Communities and Markets in Economic Development

Communities and Markets in Economic Development

This page intentionally left blank

Communities and Markets in Economic Development

Edited by Masahiko Aoki and Yujiro Hayami

Great Clarendon Street, Oxford OX2 6DP Oxford University Press is a department of the University of Oxford. It furthers the University's objective of excellence in research, scholarship, and education by publishing worldwide in Oxford New York Auckland Bangkok Buenos Aires Cape Town Chennai Dar es Salaam Delhi Hong Kong Istanbul Karachi Kolkata Kuala Lumpur Madrid Melbourne Mexico City Mumbai Nairobi São Paulo Shanghai Taipei Tokyo Toronto Oxford is a registered trade mark of Oxford University Press in the UK and certain other countries Published in the United States by Oxford University Press Inc., New York © Oxford University Press 2000 The moral rights of the author have been asserted Database right Oxford University Press (maker) First published 2000 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, or under terms agreed with the appropriate reprographics rights organizations. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above You must not circulate this book in any other binding or cover and you must impose the same condition on any acquirer British Library Cataloguing in Publication Data Data available Library of Congress Cataloging in Publication Data Communities and markets in economic development/edited by Masahiko Aoki and Yujiro Hayami. p. cm. Includes bibliographical references and index. 1. Economic development–Social aspects. 2. Community organization. I. Aoki, Masahiko, 1938-II. Hayami, Yujiro, 1932HD75 .C646 2000 306.3–dc21 00-061120 ISBN 0-19-924101-5

Contents List of Figures vii List of Tables ix List of Abbreviations xii List of Contributors xiii Introduction: Communities and Markets in Economic DevelopmentMasahiko Aoki and Yujiro Hayami xv Part I.Historical and Theoretical Perspectives 1. Impersonal Exchange and the Origin of Markets: From the Community Responsibility System to Individual Legal Responsibility in Pre-modern EuropeAvner Greif 3 2. Community and Market in England: Open Fields and Enclosures RevisitedRobert C. Allen 42 3. The Two Paths of Agrarian System Evolution in the Philippine Rice BowlYujiro Hayami and Masao Kikuchi 70 4. Community Norms and Embeddedness: A Game-Theoretic ApproachMasahiko Aoki 97 Part II.Community in Market Development 5. Middlemen in a Peasant Community: Vegetable Marketing in IndonesiaYujiro Hayami and Toshihiko Kawagoe 129 6. Market Integrators for Rural-based Industrialization: The Case of the Hand-Weaving Industry in LaosAkihiko Ohno 153

vi

CONTENTS

7. The Role of Business Networks in Market Development in Sub-Saharan AfricaMarcel Fafchamps 8. Risk and Insurance in Transition: Perspectives from Zouping County, ChinaJonathan Morduch and Terry Sicular Part III.Governance of Local Commons 9. Water Community: An Empirical Analysis of Cooperation on Irrigation in South IndiaPranab Bardhan 10. State, Community, and Market in the Deterioration of a National Irrigation System in the PhilippinesMasao Kikuchi, Masako Fujita, and Yujiro Hayami 11. Evolution and Consequences of Community Forest Management in the Hill Region of NepalKeijiro Otsuka and Towa Tachibana 12. Liberal Reforms and Community Responses in MexicoAlain de Janvry, Céline Dutilly, Carlos Muñoz-Piña, and Elisabeth Sadoulet 13. Community Arrangements to Overcome Market Failures: Pooling Groups in Japanese Fisheries Jean-Philippe Platteau and Erika Seki CommentsDouglass C. North Index

186 215 247 265 295 318 344 403 409

List of Figures 2.1. 2.2. 3.1. 3.2. 3.3. 3.4. 3.5. 5.1. 5.2. 5.3. 5.4. 5.5. 6.1. 6.2. 6.3. 6.4. 8.1. 10.1. 10.2. 10.3. 10.4. 10.5.

Real open field rent, South Midlands (five year moving average) Output per worker in agriculture, 1300–1800 Map of the central plain of Luzon, indicating provinces and municipalities covered by the field surveys Regional distribution of different harvesting systems commonly used by municipalities in the central plain of Luzon, 1968 Regional distribution of different harvesting systems commonly used by municipalities in the central plain of Luzon, 1978 Relationship between the year the harvesting arrangements changed and the year in which land tenure status changed across municipalities in inner central Luzon Regional distribution of different types of permanent labor arrangements commonly used by municipalities in the central plain of Luzon, 1987 Location of the study village relative to market centers Traditional cropping systems New cropping systems Marketing channels for vegetables Seasonal fluctuation of cucumber prices and marketing margin for inter-village collectors Map The marketing process of cloth Gross sales of a retailer in Talaat Sao Contract choice Hierarchy of agricultural insurance provision by the People's Insurance Company, 1993 Location of the NIA irrigation system Canal layout of the NIA irrigation system Organization of the NIA as of December 1995 Irrigation ratio per year for the entire system, upper reaches, and the lower reaches of Lateral C Operation and maintenance expenditures per hectare in the NIA system excluding NIA staff salaries, deflated by GDP deflator (1995=100), three-year moving averages, 1980–1993

50 54 71 77 78 79 84 130 131 132 135 145 155 158 165 167 236 267 268 269 271 272

viii

LIST OF FIGURES

10.6. Rice price and NIA project expenditures, 1965–1995 10.7. Changes in the ratio of irrigated area per year to service area in the NIA system by lateral, three-year moving averages, 1985–1994 12.1. Equilibrium attendance at assemblies 12.2. Cooperation and community size: percentage of members who assist at assemblies and number of ejidatarios 13.1. Sophisticated fishermen's groups as clusters of interlinked attributes 13.2. Choosing to pool or not to pool when there are profitability and insurance advantages associated with pooling, and when group heterogeneity and local status effect are present 13.3a. The payoff matrix of a 2×2 game where two boats have to choose between two fishing locations in a decentralized manner 13.3b. The payoff matrix of a 2×2 game where two boats have to choose between two fishing locations in a coordinated manner

273 275 330 333 350 383 398 398

List of Tables 2.1. 3.1.

Crop Yields and Enclosure, c. 1800 Imputation of Wage Rates for Rice Harvesting Work under Different Contracts in the Lowland Rice Area of Laguna, Philippines 4.1. The Payoff Consequences of the Exchange Game 5.1. Credit Cost for Vegetable Producers under Alternative Credit Arrangements in the Majalengka District, 1990 5.2. Trade Practice of Farmers and Number of Collectors in the Majalengka Village, 1989/90 6.1. Major Features of Traders 6.2. Frequency of Contractual Arrangements 6.3. Possibility of Agent's Opportunism 6.4. Social Characteristics of the Large-scale Master Weavers 7.1. Loyalty to Suppliers in African Manufacturing 7.2. Percentage Ethnic Composition of the Ownership of African Manufacturing Firms 7.3. Estimate of Ethnic Bias in Firm Entry 8.1. Average Real Income Growth, 1990–1993 8.2. Contributions of Agricultural and Non-Agricultural Income to Income Growth, by Quartile, Percentage Points 8.3. The Potential for Risk Sharing: The Average Coefficient of Variation of Household Consumption under Hypothetical Scenarios 8.4. Inter-household Transfers, 1990–1993 8.5. Transfers Received from Inside versus Outside the Village 8.6. In-Kind Transfers, 1990–1993 8.7. Determinants of Transfers Received, 1990–1993, Random Effects Estimates 8.8. Determinants of Transfers Received, 1990–1993, Random Effects Tobit Estimates 8.9. Net Financial Savings, 1990–1993 8.10. Net Savings Rates by Average Household Income 8.11. Determinants of Net Saving Rates, 1990–1993, Random Effects Estimates 9.1. Dependent and Independent Variables 9.2. Ordered Logit Estimates 9.3. Logit Estimates 9.4. Logit Estimates 9.5. First-Stage Logit Estimates

47 90 100 140 143 159 161 161 174 188 196 197 222 223 224 225 226 228 229 230 232 233 234 251 257 258 258 259

x

LIST OF TABLES

10.1. Length of Laterals, Service Areas, and Number of Sub-laterals, Villages and Farms in the Five Laterals of the NIA system, 1996 10.2. Benefits and Costs to Farmers of Switching from the Surface to Pump Irrigation, and Social Benefits and Costs of NIA's investment in the System 10.3. Social Benefits and Costs of NIA's Institutional Development vs. Increased O&M Expenditure for Irrigators' Associations (IAs) in the System 10.4. Private Benefits and Costs to NIA of Institutional Development vs. Increased O&M Expenditure for Irrigators' Associations (IAs) in the system 11.1. Basic Characteristics of Sample Forests 11.2. The Number of User Groups Initiating Informal and Formal Management by Period 11.3. Forest Conditions in 1978, 1992–96, and 1997 by Management Regime in 1997 11.4. Frequency of Sample Forests Adopting Regulations of Forest Resource Extraction/Use by Management Regime in 1997 11.5. Determinants of the Emergence of Forest User Group Management: Probit and Ordered-Probit Regressions 11.6. Marginal Effects for Ordered Probit Regression 11.7. Determinants of Regeneration per Hectare in 1997 12.1. Access to Land in the ejido, 1997 12.2. Importance of Common Property Resources (CPR) across ejidos, 1997 12.3. Use of CPR Land in ejidos, 1997 12.4. Population Pressure for Access to Land in the ejido 12.5. Indicators of Cooperation at the ejido Level, 1997 12.6. Determinants of Cooperation in 1994 and 1997: Assistance to Assemblies (Robust Ordinary Least Squares) 12.7. Effect of Group Size on the Percentage of Members who Assist in Assemblies (partial results from robust ordinary least squares regression) 12.8. Determinants of Change in Cooperation between 1994 and 1997: Change in Assistance at Assemblies (Robust Ordinary Least Squares) 12.9. Appropriation of Common Property Land, 1994–1997 12.10. Consequences of Individual Appropriation of CPR Land 12.11. Incorporation of New Members in the ejido, 1994–1997 13.1. Functioning Characteristics or Attributes of the Three Sample Groups

279 283 286 290 300 301 303 305 310 310 313 324 325 325 326 327 332 333 335 337 339 341 348

LIST OF TABLES

13.2. The Rationales of Pooling as Perceived by the Sample Skippers (frequencies of answers) 13.3. Covariances between the Gross Incomes (per haul) of each Individual Fishing Unit and the Aggregate Incomes of the Rest of the Group to which it Belongs (for the three sample groups) 13.4. Revealed Preferences for Fishing in Heta and Oki and Observed Performances Therein—A Comparison between Groups A and B (based on data collected during June–August 1997) 13.5. Frequencies of Net Damage and Labor Time Spent for the Search and Repair of Nets Damaged Over the Period June–end of August 1997, for Groups A and B in Shinminato 13.6. Estimation Results of Marginal Product Functions (based on production data collected during June–end of August 1997) 13.7. Probabilities of Net Damage and Losses According to Fishing Locations and Groups 13.8. Fishing Intensities and Estimated Marginal Products of Fishing Effort in Heta and Oki, for Groups A and B 13.9. Comparison of Daily Landings of Group A between the Two Fishing Periods (April–May and June–August 1997) 13.10. Comparative Catch Performances of the Three Sample Groups over the Two Fishing Seasons and Minimum Prices Obtained on a Day (1997) 13.11. Comparative Degrees of Heterogeneity in the Three Sample Groups, as Measured by Output Performances

xi 362 363 366 367 372 373 373 376 377 388

List of Abbreviations CNC CPR CRRA CRS DFO FASID FCA GDP GPS IA IDO NIA O&M OECD PAP PIC PROCEDE PWD RAN SGPE SRA TFP TSA VDC

National Confederation of Peasants (Mexico) common property resources constant relative risk aversion Community Responsibility System district forest officer (Nepal) Foundation for Advanced Studies on International Development Fishery Cooperative Association (Japan) gross domestic product geostatic positioning system irrigators' association (the Philippines) institutional development officers National Irrigation Administration (the Philippines) operation and maintenance Organisation for Economic Co-operation and Development Parambikulam Aliyar Project People's Insurance Company (China) Program to Certify Ejidatario Rights and Land Plots (Mexico) Public Works Department National Land Registry (Mexico) sub-game perfect equilibrium Secretariat of Agrarian Reform (Mexico) total factor productivity turn-out service area (group) (the Philippines) village development council (Nepal)

List of Contributors Robert C. Allen University of British Columbia, Canada Masahiko Aoki Stanford University Pranab Bardhan University of California at Berkeley Alain de Janvry University of California at Berkeley Céline Dutilly University of California at Berkeley Marcel Fafchamps Oxford University, England Masako Fujita Takushoku University, Japan Avner Greif Stanford University Yujiro Hayami Foundation for Advanced Studies on International Development (FASID), Japan Toshihiko Kawagoe Seikei University, Japan Masao Kikuchi Chiba University, Japan Jonathan Morduch Princeton University Carlos Muñoz-Piña University of California at Berkeley Douglass C. North Washington University in St. Louis Akihiko Ohno Aoyama-Gakuin University, Japan Keijiro Otsuka Tokyo Metropolitan University Jean-Philippe Platteau University of Namur, Belgium Elisabeth Sadoulet University of California at Berkeley Erika Seki University of Namur, Belgium Terry Sicular University of Western Ontario, Canada Towa Tachibana Hokkaido University, Japan

This page intentionally left blank

Introduction: Communities and Markets in Economic Development Masahiko Aoki and Yujiro Hayami There is an increasing awareness among economists that the ways in which economic systems are organized in developing economies make a difference for their developmental performance and potential. We conceptualize an economic system as a comprehensive coherent arrangement of institutions through which competition among, and expectations of, people are coordinated for the use of resources. In the past it was customary to classify economic systems in terms of varying combinations of the state and the market in their control of resource allocation. At one end of the spectrum we find the central planning system in the Soviet model characterized by the dominant role of the state over markets and, at the other end, we find the liberal market system in the tradition of Adam Smith, with the socalled ‘developmental state,’ ‘welfare state,’ ‘corporatist state,’ etc. somewhere in between. Our advocacy in this book is to recognize the community as the third major element constituting an economic system. Further, we depart from the conventional view that markets and community norms are to be treated as two alternative ways of organizing economic activity, and that economic development is to be viewed as movement away from community norms to markets and contracts. We argue that markets and community norms may sometimes be complementary rather than substitutes, in that the presence of the latter may facilitate the development of markets, just as the state may sometimes enhance the ability of the private sector to coordinate through markets. We also ask, under what conditions does this occur? In our definition a ‘community’ is a group of people characterized by intense social interactions among themselves. The common usage can encompass the range of communities from the family to the ‘national community’ and even further to the ‘international community.’ However, the communities with which this book is concerned are limited to those characterized by personal relationships among mutually identifiable agents. In developing economies they are typically observed as tribes and villages characterized by non-voluntary membership based on kinship or territoriality. However, in developed economies also, community relationships, which are formed in various spheres, such as the workplace, alma mater, church, sport and other hobby clubs, have significant influences on business transactions and political activities (Hayami 1997: 241–2; Putnam 1993; Ellickson 1991).

xvi

MASAHIKO AOKI AND YUJIRO HAYAMI

What role does the community play in economic development, along with the market and the state? The market is the institution that coordinates the activities of profit-seeking individuals through autonomous exchanges, under the guidance of prices toward efficiency-improving production and consumption of private goods. The state can be conceived of as the institution that supplies public goods (alternatively ‘public bads’)—including the protection of property rights—by means of the legitimate coercive power under its monopoly. On the other hand, through the repeated interactions of members, the community is capable of endogenously generating rules (norms) for coordinating the activities of members and thus making them self-enforcing without third-party involvement. In other words, community norms are informal institutions that can guide community members toward achieving voluntary cooperation (or alternatively failing to do so) in the supply of local public goods in a broad sense. This role of the community has increasingly been recognized in recent years with respect to the conservation and use of common property resources (local public goods in a narrow sense), such as forests, grazing land, fishing grounds, and irrigation, commonly called the ‘local commons’ (Feeny et al.1990; Ostrom 1990; Ostrom and Walker 1994; Bardhan 1993; Baland and Platteau 1996). Since local commons are characterized by the difficulty of charging users cost-prices because of technological non-excludability, it cannot be expected that sufficient effort for their conservation will be elicited through the profit-motive of agents in the market domain. The state may try to force the people to conserve the local commons through its coercive power. The state management, however, is subject to inefficiency arising from the weak incentives of bureaucrats for serving the needs of local people, as well as the high costs involved in accurately grasping the local needs and in monitoring whether the conservation rules are duly observed at the grassroots level. In this regard a high potential exists in rural communities in developing economies to organize collective action for their common goods based on cooperation among the users constrained by sanctions, such as social opprobrium and ostracism against free riders. An emerging new paradigm among development practitioners is to advocate for delegating the management of the commons to the initiative of local communities (e.g. World Bank 1996). The role of rural communities in the management of local commons cannot be over-emphasized. How to enhance their capacity to fulfill this role is the high-priority question on development research agendas. In fact, a major proportion of this book, particularly Part III, addresses this question. Also the greater focus of this volume is placed on the possible role of the community in enhancing the development of a market economy. In the past community norms and markets have been considered two rival institutions in suppressing and/or substituting for each other. A popular view, from Thomas More's Utopia to narodniki in Czarist Russia, and from U.S.

COMMUNITIES AND MARKETS IN ECONOMIC DEVELOPMENT

xvii

populists in the nineteenth century to Mahatma Gandhi in India, asserts that the intrusion of market activities into subsistence-oriented communities is bound to destroy the traditional norms of mutual help and subsistence guarantees, resulting in the destitution and misery of the poor. The opposing model, held by enlightenment philosophers such as Montesqieu, identifies the traditional customs and norms of communities as oppressors of human minds and conduct, and identifies market development as a liberator (Hayami 1989). These two rival models share the common view, however, that the community and market are mutually exclusive institutions from the normative perspective. In an influential economic anthropological study, Polanyi (1944) identified pre-modern economic systems as arrangements in which exchange terms are governed by customs, commands, etc., sharply distinguished from market economies which are governed by the law of supply and demand. A theoretical treatise of economic history by Hicks (1969) follows the same track. However, recent developments in institutional economics involving the theories of games, transaction costs and principal–agent contracts, suggest the possibility that community relationships can be an important basis for market development, especially in its early stage. The major barrier against the development of markets in developing economies is the absence of an effective mechanism to protect property rights and enforce contracts. For example, when it is uncertain if a supplier will honestly deliver supplies to a manufacturer at an agreed-upon quality and time, or if the manufacturer will honor the contract of paying an agreed-upon price, potentially mutually beneficial production and trading may fail to take place. Theoretically this market failure can be corrected by contract enforcement through legal procedures. In practice, however, the judicial costs involved in formal court procedures may be large, often exceeding the expected gains from dispute settlements on the small transactions typical of less-developed economies. Moreover, when judges and police are not motivated to be the faithful agents of citizens, market failures cannot only go uncorrected but can even become larger through government failures (e.g. bribery, predation). In this situation where legal resources and competence are underdeveloped, community norms based on intense social interactions can become a low-cost substitute for the legal enforcement mechanism. The incidence of contract violation should be smaller among people whose personal interactions are intense because each can identify and accurately predict the behavior of others. Also, mutual trust and community norms, nurtured through close personal relations within a community involving credible social sanctions, should work as an effective brake on moral hazards. It is therefore no surprise to find historically thriving trade and financing activities among certain ethnic groups, such as the Maghreb Jews in the medieval Mediterranean and the Chinese in Southeast Asia. Cultural beliefs leading to the conviction that default on a contractual agreement

xviii

MASAHIKO AOKI AND YUJIRO HAYAMI

with any member of the community should be collectively retaliated against by exclusion from further trading relationships and are able to prevent moral hazards from occurring in transactions among group members as well as outside agents (Greif 1989, 1993; Hayami and Kawagoe 1993). The possibility of honest exchanges based on reputation, however, may not be limited to an occurrence within ethnic groups. Imagine that there is a village community that has developed a dense information network through the collective management of a traditional local commons, or has otherwise faced the possibility of becoming engaged in exchanges with outside markets through intermediary traders possibly indigenous to the community. Then the reputation building of the members of the community (e.g. regarding the quality and timing of supply) vis-à-vis external markets, and that of the intermediary traders vis-à-vis the community (e.g. regarding the due and fair payment) may be regulated through a self-enforcing sanction mechanism replicating community norms under certain technological, social, and political conditions (Hayami and Kawagoe 1993; Tanimoto 1998; Hayami 1998). Also, communities that have developed their own internal contract enforcement mechanisms may mutually enhance impersonal trade across communities by effectively attributing collective responsibility for any breach of a contract by a trader to his community (Greif, Chapter 1 of this volume). Thus the initial development of markets can be embedded by community relationships. However, there can also be limits and disadvantages to the community mechanism as well. The community mechanism of contract enforcement is based on intensive social interactions. In general, the smaller the size of a community, the more intensive the interactions are and hence, the stronger its enforcement power. Therefore, trade based on community relationships tends to be small in range and characterized by a strong entry barrier. An equilibrium outcome from competition within the small group is naturally inferior to that evolving in a trade domain with a larger number of participants. Moreover, mutual trust and cooperation among community members often provide a basis for collusion to suppress competition even within the group. Indeed, it is not unusual to find that a trading institution based on community relationships which had initially served for the expansion of exchanges later became a barrier to further competition. A typical example is the guilds that were created in medieval Europe, Japan, and elsewhere. The guild's aim was to reduce transaction costs by controlling the quality standards of commodities and regulating transaction procedures among its members. Also, in Europe the capacity of the guild to mobilize collective penal actions against rulers who reneged on their promises to protect merchants' property rights served as a source of credible commitment on the side of the rulers for the expansion of trade (Greif et al.1994). The guild can be regarded as a kind of community

COMMUNITIES AND MARKETS IN ECONOMIC DEVELOPMENT

xix

with a strong entry barrier and an autonomous self-regulating mechanism. However, the success of guilds in reducing transaction costs eventually entailed the formation of a monopoly and became a major hindrance to market expansion in the late medieval and early modern eras. Whether the merit of community-based trade relationships outweighs their drawbacks may depend on the stage of economic development. This book represents an effort to shed some light on the question of what types of community relationships promote or deter market development. The role of the community in the conservation of local commons is essentially the same as its role in market development. In both cases, it enforces rules at a modest cost. The cost can be modest so long as the community is small in size and relatively closed. Just as the community mechanism of rule enforcement is more efficient in controlling trade in a small local market than in wide regional or national markets, it is more efficient in the conservation of a forest within a village than in a national park. Therefore, the conventional theory dictates that the operation of broad markets, as well as the conservation of large commons, must be supported by the state's coercive power in the stipulation and enforcement of formal laws. Yet, the theory cannot tell precisely which sizes of markets and commons need to be controlled by communities and which sizes by the state. This is partly an empirical question whose answer may differ, depending on development stages and cultural traditions. However, fully relevant answers are unlikely to be found through trying to demarcate between what activities should be born by state bureaucracy and what should be left to community-level collective action. Rather, an optimum mix arrangement may be found in their complementary roles. For example, a large national forest can be conserved more efficiently if the cooperation of several villages across the forest are organized through education and training as well as by providing them with adequate incentives, compared to the case of management by state forest rangers alone. Thus, similar to the advocacy for the ‘market-enhancing’ role of government in market development (Aoki et al.1997), government activities geared for strengthening the communities' capacity to organize collective actions for their common goods may be desirable. This book aims at drawing policy implications from this perspective. Let us now introduce the chapters to follow. The whole volume is divided into three Parts. The first Part, entitled ‘Historical and Theoretical Perspectives,’ consists of four chapters. Chapter 1, entitled ‘Impersonal Exchange and the Origin of Markets: From the Community Responsibility System to Individual Legal Responsibility in Pre-modern Europe,’ by Avner Greif challenges the view that communities and markets are alternative modes of governing transactions by providing game-theoretical and historical analysis of the community responsibility system. This institution relied on intra-community contract enforcement mechanisms to support

xx

MASAHIKO AOKI AND YUJIRO HAYAMI

inter-community exchange that has been impersonal, but up to trader's community affiliation. In the late thirteenth century attempts were made throughout Europe to abolish this system and establish legal systems administered by the state. This transition reflects the nature of institutional dynamics as a historical process since the former did not collapse because of exogenous shocks. It decayed because it contained the seeds of its own destruction: the growth of traders' communities in size and number, and economic and social heterogeneity. While the initial development of markets may be embedded by community relationships, there can also be developmental paths toward market economies evolving on the destruction of peasant community relationships. Robert Allen in Chapter 2, entitled ‘Community and Market in England: Open Fields and Enclosures Revisited’ makes such an argument about the evolution of English agriculture from communal open fields to enclosed farms. Allen challenges the traditional analysis of English agricultural organization in which enclosure triggered economic progress by abolishing community management of the open fields. On the contrary, Allen provides evidence that open field organization achieved high productivity and that enclosure gave only a small boost to total factor productivity. However, with the emergence of great estates, landlords substituted individual tenants for unified communities in order to increase land rents. Thus, the communal mechanism of open fields was uprooted by market capitalism rather than enclosure overthrew communities and created markets. However, Allen argues that a new type of social capital conducive to the efficient management of these new, large estates emerged between landlords and large-scale tenants and replaced the old community norm among the yeomen. Chapter 3, entitled ‘The Two Paths of Agrarian System Evolution in the Philippine Rice Bowl,’ by Yujiro Hayami and Masao Kikuchi focuses on the importance of historical factors in determining the evolutionary paths of economic systems despite basic similarities in environmental conditions. Early Spanish settlement of the Coastal Regions in the Philippines led to a close-knit system of small landlords and tenants while a hacienda system of large plantations developed in Inner Central Luzon as a result of the later settlement. Later events, such as land reform and technological advance, failed to homogenize the systems in these two regions: the different communal norms adapted to these exogenous changes in respective ways and still produced different farmer–laborer relations. Chapter 4, entitled ‘Community Norms and Embeddedness: A Game-Theoretic Approach’ by Masahiko Aoki, provides a unified game-theoretic framework for conceptualizing community norms and the sociological concept of ‘social embeddedness.’ It examines conditions necessary for norms to become self-enforceable in the domain of community as well as among quasi-community groups formed endogenously in the domain of initially

COMMUNITIES AND MARKETS IN ECONOMIC DEVELOPMENT

xxi

impersonal exchanges. It tries to relate some of the empirical findings in other chapters of this volume to this framework and provides illustrative examples from Japanese historical experiences from the Tokugawa period to the post-war growth period. Part II, entitled ‘Community in Market Development,’ contains four chapters dealing with the role of the community in the process of market development in contemporary developing and transitional economies. Chapter 5 entitled ‘Middlemen in a Peasant Community: Vegetable Marketing in Indonesia,’ challenges a famous anthropological thesis by Clifford Geertz that the entrepreneurship for modernizing, non-farm business activities cannot emerge endogenously from within the village. They looked at the emerging Indonesian vegetable market in which village-based traders act as the intermediaries, delivering the produce to towns. In order for this operation to be effective, credit must be advanced; this induces moral hazard problems. However, community norms—and not any legal system—enforce these contracts and countervail against the temptation of farmers to cheat. Further, traders are compelled to give farmers a fair price, since there is symmetric information in the village on the market prices. Thus, the market and community norms act in concert to prevent trader cheating. Chapter 6, entitled ‘Market Integrators for Rural-based Industrialization: Case of the Hand-Weaving Industry in Laos,’ examines similar issues in the emerging market for hand-woven cloth from the Laos countryside. Large scale ‘master weavers’ bridge the information gap between the demands of the city market and the suppliers in the village; they are able to take advantage of community-based trust to avoid moral hazard problems. Macroeconomic stability allows for this relationship to flourish. Ohno shows that the currency crisis in Laos increased the gains from opportunism; as a result, the trust-based contracts were in some cases replaced by spot transactions. Again, community norms and market conditions reinforce each other in economic development. The previous two chapters examine how community-based trust aids in early market development by lowering transaction costs. Chapter 7 by Marcel Fafchamps, entitled ‘The Role of Business Networks in Market Development in Sub-Saharan Africa,’ extends this analysis to Sub-Saharan Africa. Much like the vegetable markets in Indonesia or hand-weaving in Laos, Sub-Saharan African markets use business networks to establish relationships and trust, which allow for forward contracting, contract enforcement, information sharing, and other processes which lower the cost of trading. However, while the networks in Chapters 5 and 6 are centered around the village, the membership in the SubSaharan business networks is likely to be determined by common religion or ethnic background, but it cannot be considered as entirely exogenous to the market formation process. Referral by family and friends is the most likely channel through which

xxii

MASAHIKO AOKI AND YUJIRO HAYAMI

ethnic concentration arises. The author also discusses alternative institutions that perform functions similar to those of business networks. A case where there is a lack of community–market interaction is addressed by Jonathan Morduch and Terry Sicular in Chapter 8 entitled ‘Risk and Insurance in Transition: Perspectives from Zhouping County, China’. In China, the development of markets has led to impressive growth over the past two decades. But with the de-collectivization of Chinese agriculture in the 1980s came the volatility of the markets, without the development of formal insurance mechanisms to buffer households against these shocks. Community-based transfers do not fill this need; they are too small and homogenous to be effective. Households would normally then rely on their extended family mechanisms to smooth consumption; however, shrinking family size has led to an increase in savings, but not enough to eliminate vulnerability. In light of these needs, Morduch and Sicular put forth a proposal for insurance coverage that combines community and market. Part III, entitled the ‘Governance of Local Commons,’ shifts the focus from community norms interacting in emerging markets to communal control over local public goods. In contrast to Garett Hardin's solution to the ‘tragedy of commons’ problem, i.e. state control or the establishment of private property rights rules, the first four chapters deal with complementary roles of the state and the community in the governance of local public goods and the last chapter with collective private control of fishing. In Chapter 9 entitled ‘Water Community: An Empirical Analysis of Cooperation on Irrigation in South India,’ Pranab Bardhan examines 48 irrigation communities in south India, using multiple regressions to come to conclusions about the interaction of government and community in regulating water usage. In general, both village elites and average farmers will consider the rules fair if they have input in framing the rules, instead of being imposed rigidly by the government. Further, Bardhan finds that cooperation is hindered by inequality in landholding but helped by social homogeneity, small group size, longer duration of access to water, monitoring by guards, and other factors in some cases. Chapter 10 by Masao Kikuchi, Masako Fujita, and Yujiro Hayami, entitled ‘State and Community in the Deterioration of a National Irrigation System,’ cautions against putting too much faith into the community management paradigm. Specifically, organization at the rural community level may not be sufficient to manage large complicated systems because of monitoring difficulties. In the Philippines, the national agency which had regulatory control over the irrigation system was able to function rather effectively with infusions of external funds. However, when the management of the irrigation system was turned over to the communities in light of budgetary pressures, some communities succeeded, but many others failed. The authors conclude that rural communities lacked the sufficiently

COMMUNITIES AND MARKETS IN ECONOMIC DEVELOPMENT

xxiii

strong organizational capacity to control free-riding problems. Also, the existence of a market exit option deprived the farmers of an opportunity to effectively organize their voice in irrigation management. Thus, it may be advisable for the government to play a complementary, facilitating role for local communities to develop their own capabilities to regulate their local commons, rather than leaving them on their own. To follow up those conclusions, Chapter 11 by Keijiro Otsuka and Towa Tachibana, entitled ‘Evolution and Consequences of Community Forest Management in the Hill Region of Nepal,’ demonstrates the effectiveness of community management regimes when supported by the government. When deforestation occurred in Nepal, informal user groups began to emerge; these groups voluntarily began to manage their local forest resources. The government began to actively support these groups, handing over forest rights to user groups. These formal user groups—who work with the government's forest officers—have been able to achieve reforestation at levels above those in areas managed by informal user groups or without any management. For the most part, the communities studied in Chapters 9 through 11 were formed endogenously over a period of time. In Chapter 12 entitled ‘Liberal Reforms and Community Responses in Mexico,’ Alain de Janvry, Céline Dutilly, Carlos Muñoz-Piña, and Elisabeth Sadoulet examine a set of communities called ejidos that were exogenously formed by the Mexican government to serve political ends. Though the community owned all land, some of it was parceled out in individual plots of land and the rest became common property with regulated access for all members. Though regulated by community leaders, the state had effective control over their management. Reforms in 1992 eliminated governmental oversight of the ejidos: the economic costs outweighed the political benefits. There was a varied response to this hand-over: some ejidos experienced increased cooperation, others preserved the status quo and some saw a decrease in cooperation and increased privatization. Much like Chapters 10 and 11, Chapter 12 emphasizes the importance of government redesigning of the ejido communities, particularly in their size. In order to take advantage of the gains from cooperation, the Mexican government should aid the ejidos in their transition. Chapter 13 turns to another classic case of ‘tragedy of the commons’: fishing. Jean-Philippe Platteau and Erika Seki argue in ‘Community Arrangements to Overcome Market Failures: Pooling Groups in Japanese Fisheries’ that in this sector there are considerable gains to pooling resources. Economies of scale can be captured by working together including: work coordination such as collective lost-net retrieval and repair, information sharing, regulation of fishing activity and sales price stability. In return for these collective efforts, income pooling can distribute these gains. Platteau and Seki observe that in a Japanese village practicing such

xxiv

MASAHIKO AOKI AND YUJIRO HAYAMI

an egalitarian sharing scheme, fishermen of high capability are compensated by high social esteem paid by other village members. Finally the book concludes with the record of Douglass North's comments from a transaction cost perspective on drafts of the chapters discussed at the final conference leading to the present volume.

References Aoki, M., Kim, H., and Okuno-Fujiwara, M. (eds.) (1997) The Role of Government in East Asian Economic Development, Oxford, Oxford University Press. Bardhan, P. (1993) ‘Symposium on Management of Local Commons’, Journal of Economic Perspectives, 7(Fall): 87–92. Baland, J.-M. and Platteau, J. P. (1996) Halting Degradation of Natural Resources: Is there a Role for Rural Communities? Oxford, Oxford University Press. Ellickson, R. C. (1991) Orders Without Law, Cambridge, MA, Harvard University Press. Feeny, D., Berkes, F., McCay, B. J., and Acheson, J. M. (1990) ‘Tragedy of the Commons: Twenty-two Years Later’, Human Ecology, 18(1): 1–19. Grief, A. (1989) ‘Reputation and Coalitions in Medieval Trade: Evidence on the Maghribi Traders’, Journal of Economic History, 49(Dec): 857–82. —— (1993) ‘Contract Enforceability and Economic Institutions in Early Trade: The Maghribi Traders' Coalition,’ American Economic Review, 83(June): 525–48. —— , Milgrom, P., and Weingast, B. R. (1994) ‘Coordination, Commitment, and Enforcement: The Case of the Merchant Guild’, Journal of Political Economy, 102(Aug.): 745–76. Hardin, G. (1968) ‘The Tragedy of the Commons,’ Science, 162: 1243–8. Hayami, Y. (1989) ‘Community, Market, and State’, The Elmhirst Memorial Lecture at the 20th International Agricultural Conference of Agricultural Economists, in A. Maunder and A. Valdes (eds.), Agriculture and Government in an Interdependent World, Aldershot, Gower, pp. 3–14. —— (1997) Development Economics: From the Poverty to the Wealth of Nations. Oxford, Oxford University Press. —— (ed.) (1998) Toward the Rural-based Development of Commerce and Industry: Selected Experiences from East Asia, EDI Learning series, Washington DC, World Bank. —— and Kawagoe, T. (1993) The Agrarian Origins of Commerce and Industry: A Study of Peasant Marketing in Indonesia, New York, St. Martin's Press. Hicks, J. (1969) A Theory of Economic History, Oxford, Oxford University Press. Ostrom, E. (1990) Governing the Commons, New York, Cambridge University Press. —— and Walker, J. (1994) Rules, Games, and Common-Pool Resources, Ann Arbor, MI, University of Michigan Press. Putnam, R. D. (1993) Making Democracy Work, Princeton, Princeton University Press. Polanyi, M. (1944) The Great Transformation, New York, Rinehart. Tanimoto, M. (1998) Nihon ni okeru Zairaiteki Keizai Hatten to Orimonogyo (Indigenous Economic Development and the Textile Industry in Japan), Nagoya, University of Nagoya Press. World Bank (1996) World Bank Participation Book, Washington, DC, World Bank.

Part I Historical and Theoretical Perspectives

This page intentionally left blank

1 Impersonal Exchange and the Origin of Markets: From the Community Responsibility System to Individual Legal Responsibility in Pre-Modern 1

Avner Greif

1.1 Introduction Communities and markets are considered by economists as alternative modes of governing transactions. In contrast, this paper calls attention to the importance of considering them as complements. In economic development and economic history, the view that communities and markets are substitutes has long been characterized by postulating that the community is an inferior mode of governance. It entails personal, as opposed to impersonal, exchange and hence limited division of labor. Accordingly, it has been held that the historical process of economic development requires substituting away from community based governance to market based governance. The feasibility of this transformation, in turn, depends on the nature of a society's contract enforcement institutions. It necessitates a transition from contract enforcement relying on personal, repeated relations within a community to one based on formal, state-mandated legal contract enforcement among market's participants. This view has been based, to a large extent, on our understanding of the process of market expansion in pre-modern Europe. North (1991: 101), for example, invokes the experience of the Western world to argue that a legal system administered by the state is a necessary condition for an advanced division of labor and a market economy. ‘Undergirding … markets are secure property rights, which entail a polity and judicial system to permit low cost contracting’. Hence, market societies ‘need effective, impersonal contract enforcement [provided by the state], because

1

This research was supported by the National Science Foundation Grants 9009598-01 and 9223974. Masahiko Aoki, Gregory Besharov, Brent Daniel Goldfarb, and participants in Workshops in the University of Chicago, Oxford University, Cambridge University, the Hebrew University, and Stanford University provided useful comments. Yadira Gonzalez de Lara provided indispensable research assistance and comments.

4

AVNER GREIF

personal ties, voluntaristic constraints, and ostracism are no longer effective’ (p. 100).2 In recent years, however, inspired by Coase's analysis of the relationships between the firm and the market, this view of the nature of market economies has been altered: it has been recognized that even in modern, developed economies contract enforcement based on personal and repeated relations is important for economic efficiency. It is important because even impartial legal contract enforcement entails transaction costs due to asymmetric information, incomplete contracts, and the costs required to verify past actions and states of the world by the court. Within communities, however, informal contract enforcement mechanisms can, at times, economize on these costs. Relevant information may be more readily available to one's community members, incomplete contracts can be replaced by tacit knowledge of a code of conduct, and a third party with the ability to verify past actions is not required for informal contract enforcement to function. This Coasian view led to a different understanding of the relations between markets and communities (surveyed in Greif 1997): markets and communities are substitutes in providing contract enforcement, each of which entails distinct cost and benefit for the organization of exchange. The benefit of community-governed exchange is the low cost of contract enforcement but its cost is the efficiency loss due to the confinement of exchange to a relatively small number of individuals. The benefit of market exchange is that transactions are not confined to the relatively small number of the community's members but the exchange within the market entails a relatively high cost of contract enforcement. Hence, communities and markets interrelate in a manner equivalent to the Coasian vision of how firms and markets interrelate. An efficient organization of an economy entails that transactions are allocated between market and communities in a manner that minimizes transaction costs. Guided by this view of communities and markets as substitutes, theoretical and empirical studies examined the factors determining the cost and benefit associated with each mode of transacting and the nature of intra-community contract enforcement mechanisms.3 The related theoretical and empirical studies of the dynamics of the relationships between markets and communities have integrated this Coasian view of communities and markets as

2

See also: Weber (1981 [1927]: 277) who argued that for the European ‘capitalistic form of industrial organization’ to emerge, it ‘must be able to depend upon calculable adjudication and administration’ of the law. Similarly, Rosenberg and Birdzell (1986: 116) argued that the establishment of a legal system able to facilitate exchange in the 18th century provided (1986: 117) ‘an important element in the development of capitalist institutions’.

3

e.g. Macauly (1963), Granovetter (1985), Macneil (1985), Greif (1989, 1993, 1994), Montgomery (1991), Bernstein (1992), Klein (1992), Guinnane (1994), Landa (1994), Fafchamps (1995).

PRE-MODERN EUROPE

5

substitutes with Durkheim's analysis of economic development. This analysis maintains that the relationships between communities and markets are unidirectional. Increases in the efficiency and extent of markets undermines the governance of exchange by communities since it provides outside options to a community's members. Theoretical considerations and the historical experience of pre-modern Europe, however, indicate the limitations of these static and dynamic views of the relationship between market and communities. Theoretically, (either formal or informal) intra-community contract enforcement mechanisms can provide a basis for the functioning of intercommunity impersonal exchange. More specifically, communities (rather than an impartial legal system with authority over both transacting individuals) can provide the basis for exchange that is impersonal up to one's community affiliation. Furthermore, this exchange is not necessarily spot exchange. Rather, similar to the exchange made possible by the court, as long as past actions can be verified at relatively low cost, this exchange can be with respect to transactions (such as credit and contracts for future delivery) characterized by separation over time and space between the quid and the quo. In other words, markets and communities are not necessarily substitutes. They can be complements. Communities enable the operation of markets and (relatively) impersonal exchange. Indeed, the emergence of markets can crucially depend on the existence of appropriate community structures rather than on an impartial court. It is in the absence of such community structures to provide contract enforcement mechanisms that the operation of the market requires a well functioning legal system. Conversely, a reduction in the effectiveness of the intra-community contract enforcement mechanism may lead to a reduction in the extent to which impersonal, market relationships are possible. Furthermore, such a reduction in the communities' ability to support market relations may not reflect an exogenous alternative provided by the market to members of a community. Rather, it reflects a process through which the same market relationships that communities enable gradually undermine these communities' ability to provide the appropriate institutional foundations for the market. Perhaps more important than the above theoretical argument, the experience of pre-modern Europe indicates the important complementarity between communities and markets in the process of development. Indeed, a particular system of inter-community contract enforcement based on intra-community contract enforcement supported market expansion in Europe during the late medieval commercial revolution (of the eleventh to the fourteenth century). It enabled impersonal exchange characterized by separation between the quid and the quo despite the absence of an impartial legal system with authority over the transacting individuals. It was indeed during this period that western Europe witnessed trade expansion and market integration as well as a widespread reliance on arrangements

6

AVNER GREIF

such as credit, contract for future delivery, negotiable securities, and maritime insurance. Furthermore, the same processes that this community responsibility system fostered—processes through which trade expanded and merchants' communities grew in size, number, and economic and social heterogeneity—reduced its economic efficiency and the intra-community political viability. Yet, the extent to which communities could have replaced the community responsibility system with an alternative institution based on Individual Legal Responsibility depended on their political environment. The political system in England and France was conducive to such a transition while that in the politically fragmented Italy was not. The paper elaborates on the above theoretical argument and presents the community responsibility system that prevailed in Europe prior to the rise of contract enforcement provided by the state.4 Furthermore, it elaborates on the reasons for the system's decline and how the nature of the distinct political systems in England and Italy directed the details of the system that replaced the community responsibility system. Specifically, Section 1.2 discusses the issue of impersonal exchange during the late medieval period. Section 1.3 presents a theory of the community responsibility system. Section 1.4 utilizes the insights of the theoretical analysis, as well as information from contracts, court records, charters, and royal and community regulations, to evaluate the extent to which a community responsibility system indeed functioned in Europe during that time. Section 1.5 examines, theoretically and historically, the deficiency of the community responsibility and attempted transitions to an alternative system based on individual responsibility. Conclusions follow.5

4

An earlier generation of scholars have noted the operation of this system (e.g. Santini 1886; Arias 1901; Maitland and Bateson 1901; Patourel 1937) and this paper extensively relies on their important contributions. By and large, these scholars did not analyze the system but concentrated on the inter-community disputes which this system entailed viewing them as reflecting archaic and barbaric behavior. For a more recent presentation of some of the system's elements, see Moore (1985).

5

This paper relates to several recent studies in economic history which combined game theory and historical analysis to examine the institutional infrastructure underlying the late medieval Commercial Revolution. Greif (1989, 1993) analyzed the institution that governed the relations between merchants and their overseas agents among the Maghribi traders who operated mainly in the Muslim Mediterranean during the eleventh century. Greif (1994) provided a comparative study of eleventh- and twelfthcentury institutions among the Maghribi and Genoese traders and these institutions' inter-relations with their cultures and societal organizations. Greif et al. (1994) examined how the institution of the Merchant Guild secured alien merchants' property rights despite the ability of rulers to abuse them using their coercive power. Two papers, Benson (1989) and Milgrom et al. (1990), are directly concerned with institutions that supported exchange characterized by separation between the quid and the quo among merchants who did not expect to interact frequently. In particular, the latter paper proposed that during the late medieval period a Merchant Law System enabled impersonal exchange by providing the appropriate information and generating the appropriate incentives. Their explicit game-theoretical analysis indicated how a local legal system, without any coercive power but with a monopoly over the ability to verify merchants' past actions and store and communicate information, could ensure contract enforcement in impersonal exchange. It did so by controlling the information required for a multilateral reputation mechanism among the merchants. Thus, they argued that impersonal exchange at the Champagne Fairs—the central trading fairs in Europe during the Commercial Revolution—was possible because the local court functioned as a Merchant Law System.

PRE-MODERN EUROPE

7

1.2 On the Nature of Exchange Characterized by Separation Between the Quid and the Quo During the Commercial Revolution The historical records indicate that exchange characterized by a separation between the quid and the quo over time and space was common in western Europe during the late medieval Commercial Revolution, perhaps for the first time since the fall of the Roman Empire. In towns, fairs, and market-places, merchants provided and received credit, bought and sold through contracts for future delivery, and insured the cargo they shipped over the sea.6 While we cannot qualitatively measure the efficiency contribution of such exchange relations, their contributions were arguably great. Lopez (1976), the eminent historian of the Commercial Revolution, has viewed credit as necessary for commercial expansion to occur with a monetary system based upon a limited supply of precious metal. The ‘take-off [of the Commercial Revolution] was fueled not by a massive input of cash, but by a closer collaboration of people using [commercial] credit’ (p. 72). The historical records also reflect the identity of the individuals who enter into exchange characterized by a separation between the quid and the quo over time and space during the Commercial Revolution. Such exchange was often conducted among individuals who lived in close proximity to each other. Herman Van der Vee (1977), for example, has noted that exchange relations characterized by a separation between the quid and the quo were frequently established among merchants and their local consumers in Europe during the later middle ages. ‘Consumer credit underwent great expansion in the local economy’ and ‘merchants from the town or country regularly bought up agricultural produce from the farmers before it had been harvested against partial or complete cash payment.’ At the same time, ‘farmers and workers received credit guaranteed by their future harvest or future labor output’ (p. 300). It is more intriguing to note, however, that exchange characterized by a separation between the quid and the quo was also established from as early as the twelfth century among merchants who did not live in close proximity to each other. For example, around the middle of the century traders from Asti regularly sold on credit northern textiles imported from the Champagne Fairs to Genoese traders (Reynolds 1929, 1930, 1931; Face 1958). Credit arrangements among individuals from other localities are frequently mentioned in Genoa's historical records. In 1190, for example, two Genoese traders, Bonifacius della Volta and Nicola Mallonus,

6

For a description of these developments, see, for example, Lopez and Raymond (1955: 157–238); de Roover (1963: 42–118); and Postan (1973).

8

AVNER GREIF

bought goods for a Piacenzan merchant for 120 lira with one year to pay. On the 28th of March, 1210, Rubeus de Campo of Genoa paid a debt of one hundred marks sterling in London on behalf of Vivianus Jordanus from Lucca.7 Credit transactions among individuals from distant localities were not confined to Italy during this time. Such transactions were common in England during the twelfth and thirteenth centuries among English merchants from different localities and among English merchants and French, Flemish, and German traders.8 Similarly, contracts for future delivery among individuals from distant localities were common in Italy, England and France. For example, in 1191, a Genoese merchant named Ugo Mallonus bought, from a Pavian and a Roman, 5 bales of fustian of Pavia at 40 pieces per bale, including 13 vermilion, 6 green, the rest brown, and contracted to buy 3 more bales at mid-Lent and 2 more bales at Easter. At the Fairs of Champagne, where much of the trade between northern and southern Europe was conducted during the twelfth and the thirteenth centuries, merchants from different localities frequently entered into contracts for future delivery.9 Contract enforceability is necessary for any exchange but the necessity of enforcement is particularly important in exchange characterized by separation between the quid and the quo. In the absence of appropriate institutions, a borrower, for example, can make himself better off after obtaining a loan by not repaying his debt. Expecting such behavior ex post, a borrower would not lend ex ante. Similarly, a merchant who is paid to deliver goods in the future will find it optimal to retain possession over these goods, implying that the buyer would not be willing to pay ex ante. Hence, exchange characterized by separation over time and space between the quid and the quo requires contract enforcement institutions that enable the transacting parties to ex ante commit to carry out their contractual obligations ex post. What were the institutions that enabled such exchange during the Commercial Revolution? How could a twelfthcentury borrower from Lucca, for example, commit himself ex ante to repay ex post a debt to a lender from London? Did late medieval Europe develop contract enforcement institutions that enabled impersonal exchange? Or was exchange confined to personal exchange in which repeated interactions or family relationships mitigated the commitment problem? Given the available

7

Obertus Scriba (1190), no. 669 and see also nos. 138, 139. Lanfranco Scriba (1952), vol. 1, no. 524, p. 234.

8

e.g. Calendar of the Patent Rolls Preserved in the Public Record Office, 1266–72, p. 20. English Historical Documents, vol. II, pp. 1012–3. Postan (1973).

9

Ugo: Guglielmo Cassinese (1190–2), no. 250. With respect to England and France, see Moore (1985), and Verlinden (1979). Toward the end of the thirteenth century Bruges became the main center for trade in products.

PRE-MODERN EUROPE

9

historical evidence, we cannot address this question by tracing the exchange relations of individual merchants over time. Hence, to examine the extent to which impersonal exchange was feasible in pre-modern Europe one has to determine whether an institution that enabled impersonal exchange functioned during this time. In the early days of the period under consideration there was no legal system in Europe that could have effectively supported impersonal exchange among individuals from distant localities. Even within a relatively well-organized political unit (such as England), there was no legal system that could support, at low cost, the enforcement required for such exchange.10 Local courts that could supervise and enforce contracts executed in the areas under their authority existed throughout Europe. They had the ability to provide contract enforcement required for exchange among individuals present in the (limited) territorial area over which they had legal jurisdiction. Yet, such local courts were not, by and large, unbiased agents of a central legal authority. Rather, they were the embodiment of local interests and are known to have been biased in their judgments against foreigners.11 Similarly, late medieval communities probably had informal contract enforcement institutions of the kind that often emerge among individuals who live in close proximity.12 Such local formal and informal contract enforcement institutions, however, could not govern impersonal exchange characterized by separation between the quid and the quo among individuals who lived far away from each other. The lack of a legal system that could have supported impersonal exchange characterized by separation between the quid and the quo among merchants from distant localities led to the perception common within economic history that, prior to the rise of the state, such exchange was not feasible. Yet, this assertion ignores the observation that late medieval trade actually occurred in a particular social context, namely, that of communities. Trade was not conducted among individuals without any social affiliation. On the contrary, it transpired among individuals with particular community affiliations. Evidence of the fact that long-distance traders were identified as members of a particular community can be found in the many commercial contracts that survived from the late medieval period. Since during this period last names were not yet common among non-nobles, individuals often had

10

Ashburner (1909) ; Plucknett (1949: 142); Postan (1973) ; and the information contained in Hall (1930) Select Cases Concerning the Law Merchant, A.D. 1239–1633, 2: Central Courts.

11

e.g. Hanawalt (1974).

12

Ellickson (1991) has argued, and provided some empirical support, that communities develop welfare-maximizing norms supported by informal contract enforcement institutions. For analysis of such informal contact enforcement institutions among merchants, see Greif (1989, 1993), Bernstein (1992), Clay (1997).

10

AVNER GREIF

surnames or nicknames indicating their profession, place of origin, particular features, etc. In contracts related to longdistance trade it is common to find merchants whose surnames or nicknames reflect their places of origin.13 Furthermore, such contracts were signed in the presence of witnesses whose names were registered on the contract. It is common to find that merchants from one community witnessed a contract that members of their community entered into when they were abroad, indicating close association among them. The recognition of ones' community of origin was probably facilitated by the relatively low level of social and cultural integration that prevailed during the medieval period. National languages, for example, were not yet developed and distinct cities and small regions still had their own peculiar linguistic characteristics. Furthermore, communities provided an important part of the organizational framework associated with long-distance trade. While this reflects their importance in trade it also indicates a mechanism that facilitated learning about one's community. For example, at the Champagne Fairs various merchant communities had permanent representation, scribes, and even courts of law. Merchants from a particular community often traveled abroad together, living and trading from their own compound. The organization of the German merchants who frequented the large international emporium in the city of Bruges was based on the merchants' home towns and regions. Genoa and Venice had their own quarters in Constantinople, while the Flemish merchants at the Boston Fair lived in the same quarter.14 As further discussed below, the communities with which merchants were identified and which provided the organizational framework within which long-distance trade was conducted assumed various forms. The most common ones were a home town, a borough, and merchant guild. In any case, all these communities seem to have shared a particular feature, namely, that the nature of their members' (social, economic, or legal) relations were such that the community had the ability to impose punishment on each of them. This ability, in and by itself, reflects the broader social and legal environment of the period. The economic and social cost of leaving one's community were relatively high. Is it theoretically possible that the identification of individual merchants with particular communities and the associated intra-community contract enforcement institutions were utilized in late medieval Europe to support inter-community impersonal exchange? Methodologically, to address this question the next section develops a repeated game model of exchange characterized by separation between the quid and the quo. It enables

13

Emery (1952) and Lopez (1954).

14

e.g. Verlinden (1979: 131) (the Champagne Fairs), Greif et al. (1994).

PRE-MODERN EUROPE

11

exploring the factors constraining impersonal exchange and the role of communities in overcoming these constraints.

1.3 A Theory of the Community Responsibility System Theoretically, communities with intra-community contract enforcement institutions could have supported intercommunity impersonal exchange. To highlight the function of the community, the analysis begins by examining the assumptions required to enable impersonal exchange when there are no communities or legal contract enforcement mechanisms. It then examines the more realistic model satisfying only the sufficient conditions required for a legal system to support impersonal exchange. The examination reveals that these conditions are also sufficient for communities to support impersonal exchange. Without loss of generality, the discussion concentrates on exchange in which credit is provided for trading, and for ease of exposition, technical details are not presented. The theory of repeated games suggests that exchange characterized by separation between the quid and the quo may be possible even in the absence of legal contract enforceability among individuals who expect to interact even only once with each other.15 When profitable exchange can be repeated, the fear of retaliation—the fear of losing future gains from exchange—may prevent cheating. To examine the extent to which expectations for retaliation can support exchange, consider the following simple model. Consider an economy in which there are lenders and borrowers. Each period or stage game is divided into two subperiods. In the first sub-period each borrower has to decide whether to be randomly matched with a lender or not. Borrowers who choose to be randomly matched will never be matched again with the same lender.16 A lender who was matched with a borrower can decide whether to lend (a finite amount) or not.17 A borrower to whom a loan was made travels away to trade and in the second sub-period can decide whether to repay his loan or not. One's gain from exchange or defaulting in a particular period is consumed only after the following period. Lending is efficient and profitable to both parties conditional on the borrower paying his debt. In other words (ignoring the process through which the expected gain from each loan is divided), the division of gains from lending and trading is such that it is profitable for a

15

For a survey of the relevant literature, see Fudenberg and Tirole (1991).

16

See Townsend (1981) for the appropriate specification.

17

For simplicity's sake, it is assumed that no matter how many borrowers decide to travel, each is matched with a lender. Random matching is assumed to capture the impersonal aspect of exchange.

12

AVNER GREIF

lender to lend and a borrower to borrow if the loan will be repaid and the gains divided as agreed upon. If a loan was made and not repaid, however, the lender is worse off than he would have been had he not made the loan. Furthermore, since this lender loses his capital he is unable to furnish loans in the future. Note that this (one-sided prisoners' dilemma game) is characterized by separation over time and space between the quid and the quo. If this game is played once, there is no sub-game prefect equilibrium in which lending occurs. Hence, efficient exchange is not feasible. This is the case because the best a borrower can do ex post if a loan was given is to not repay it. Anticipating ex ante that this would be the case, a lender would not furnish a loan. However, the following are sufficient conditions for exchange to be possible: (1) The game is repeated for an infinite number of periods, (2) borrowers have an identity, a ‘label’ that is known to all the lenders, and (3) the history of the game is common knowledge, namely, each player's past actions are known to all. When these conditions are satisfied, there is sub-game prefect equilibrium (SGPE) which enables exchange if a borrower is sufficiently patient or, for a given time discount factor, if his (per-period) gain from exchange is high enough. Specifically, consider the following strategy combination: a borrower chooses to be randomly matched but repays if and only if he has never cheated a lender before, while a lender provides a loan only to a borrower who has never defaulted on a loan. A borrower is induced not to default ex post by the lender's threat that if he defaults he will not be able to borrow in the future. The lender's threat is credible since he expects a borrower who cheated once to cheat again, and this indeed is a cheater's best response given the lender's strategy. Expecting ex ante that this will be the case, the lender will lend.18 Does the above analysis indicate that impersonal exchange is feasible given the above conditions? While the answer should seemingly be to the affirmative, this is not the case. It basically places identity at the center of the analysis. Exchange with a borrower who cheated in the past is not feasible despite the inherent profitability of the exchange because of the identity of that particular borrower. In other words, it indicates that personal exchange—in which a borrower's identity as captured by his past actions—is feasible. It does not indicate, however, that impersonal exchange is possible.19 Could communities have relaxed this constraint

18

For a formal exposition of such a multilateral reputation mechanism, see Milgrom et al. (1990) and Greif et al. (1994) for the case when the history of the game is known to all players.

19

Kandori (1992) and Ellison (1994) examined the operation of such mechanisms (in PD games) when each player knows only his private history. Exchange can be supported under rather restrictive conditions by imposing a general punishment, namely, on those who defaulted and those who did not. In any case, their analysis does not hold when players have a finite life-span as is appropriate to assume and is assumed below.

PRE-MODERN EUROPE

13

and enabled exchange without the need to link one's identity to one's ability to exchange? Furthermore, the above model is an inappropriate departing point for examining the role of communities in fostering impersonal exchange since it is based on the above three unrealistic assumptions. Medieval traders did not have an infinite life-span. Yet, if borrowers have a finite horizon, it is optimal for them to default in their last period. Anticipating this, the lenders would not lend in the last period, implying that borrowers would find it optimal to default in the period before the last. Yet, expecting default, the lenders would not lend in this period either. Carrying this logic further indicates that lending would not be provided in any period.20 Similarly, a merchant's identity may have been known to his exchange partners in each particular period but could not have been easily transmitted to other merchants in the age prior to last names, passports, and drivers' licenses. Finally, past conduct was private information although it may have been verified ex post by a third party. Hence, to examine the potential role of communities in fostering impersonal exchange it may be appropriate to alter the above model by dropping these unrealistic assumptions and replacing them with the assumptions that are sufficient for a court to be able to support impersonal exchange. Accordingly, assume that while the number of borrowers each period is constant, each of them lives for only a finite number of periods. Each borrower plays the above stage game each period up to his last period when he ‘retires’ and does not trade. Furthermore, assume that a borrower's identity is known to the parties to the exchange, and that past actions are private information although they can be verifiable by a third party. Note that in this model there is no equilibrium in which lending occurs. Lenders' finite life-spans are sufficient for this result. Yet, as discussed below, expanding the model to include communities implies that an equilibrium with impersonal lending does exist. To introduce communities in the analysis assume that all borrowers are members of community B while all the lenders are members of community L. (The analysis holds if there are many borrowers' communities but introducing many lenders' communities is more problematic as discussed below.) Apart from providing a community label to each borrower, a community is conceptualized as having two features: a ‘territory’ and an intra-community contract enforcement institution that can impose a punishment

20

Telser (1980) pointed out that infinite horizon games are conceptually identical to finite horizon games with an uncertain number of periods. For this to be the case, however, the probability that the game will end in each period should be constant over time. Yet, this condition does not hold with respect to the uncertainty associated with human lives. For how the finite horizon problem was mitigated in intra-community agency relations during the late medieval period, see Greif (1989, 1993).

14

AVNER GREIF

or abuse the property rights of both a community member and any individual present in the community's territory. While the following analysis holds whether a community's enforcement institution is formal or informal, this institution is referred to below as a ‘court.’ Denote the lenders' court by LC and the borrowers' court by BC. The second feature of a community is that its court has a payoff function which is the sum of the payoff functions of its (living) members. In other words, a community aims at maximizing the sum of the lifetime utilities of its constituting members.21 While the two latter aspects of the above conceptualization of a community may be intuitively appealing, they are, in a sense, an equilibrium outcome.22 Given the lenders' equilibrium strategies that enable impersonal exchange (discussed below), it is optimal for every (non-retired) borrower that an intra-community contract enforcement institution with the above payoff be established by the community. In other words, had the game been extended to enable, in the first period of the game, members of each community to decide whether to establish such a court or not, the only equilibrium entailing impersonal exchange is the one in which the communities establish courts. In addition to the actions available to borrowers and lenders in the previous game, in this game borrowers now have to travel to the lenders' community to be matched and must return there to pay their debts. Furthermore, in each second sub-period each of the lenders can complain if he was not paid; the LC can impound or confiscate the goods of any and all borrowers present in the lenders' community, verify whether or not a default occurred, and apply to the BC to demand compensation.23 If compensation is received, the LC can release the impounded goods and distribute the total proceedings in any manner among the lenders. In each first sub-period, the BC can verify any complaint about defaulting, and can decide whether to impose a punishment on any borrower, to impound any borrower's property, and to furnish a particular compensation to the LC.

21

This model relates to two lines of work in game theory. Works on games between a long-lived player and a sequence of short-lived players, and works on games with overlapping generations of players. For a survey, see Fudenberg and Tirole (1991: 168–72).

22

Examining the case in which there are multiple lenders' communities is more complicated and is left for future work.

23

Both the terms ‘impound’ (namely, to take legal or formal possession of goods to be held in custody of the law) and ‘confiscation’ (namely, seizure under public authority) seem appropriate here. In the historical reality of the late medieval period the goods were at times confiscated directly by the lender to become his or her property or were impounded by the authority and held as a pledge.

PRE-MODERN EUROPE

15

Under mild conditions, impersonal exchange can be sustained in this game as an SGPE outcome.24 The conditions are that (1) the borrowers' time discount factor and the amount the LC can impound are sufficiently high relative to the amount defaulted upon and the gains to the borrowers from future trade, and (2) that the gains for the LC from impounding the goods of all the borrowers present in the lenders' community are lower relative to the present value of future gains of providing loans.25 When these conditions hold, there is an equilibrium in which each borrower is motivated not to default by the expectation of punishment by the BC. The BC is motivated to punish one who defaulted because a failure to do so implies the loss of impounded goods and all future gains from exchange. The LC verifies any complaints, impounds goods, demands compensation if default occurred and distributes the proceedings to the lender who was cheated. The LC is motivated to do so because otherwise a cheated lender would not complain, leading borrowers to default and hence reducing the LC's payoff. Thus, each lender, knowing that the best a borrower can do is to pay, finds it optimal to lend. Specifically, if the above conditions hold, exchange that is impersonal up to one's community affiliation can be supported as an SGPE outcome by the following strategy combination: a borrower travels if and only if the LC never impounded or confiscated goods when a default did not happen, borrows if given a loan, and returns to pay his debt if and only if the lenders' strategy calls for complaint following a default and as long as the BC has not refused to pay a compensation. A lender lends as long as the BC has not refused to pay a compensation and complains if cheated. The LC verifies any complaint and, if and only if it is a valid complaint, impounds the goods of the borrowers present in the lenders' community and demands a compensation from the BC equal to the present value of the cost imposed on the lender due to the default. If compensation was given, the LC then advances it to the lender who was cheated, otherwise it impounds the goods of borrowers until equivalent compensation is obtained. Finally, the LC impounds the goods of all borrowers in its territory if it ever impounded goods when no default occurred. The BC verifies any complaint and, if and only if the complaint is valid, confiscates the defaulter's goods and pays a compensation not higher than the present value of the cost imposed on the lender due to the default. To finance the compensation the BC utilizes the defaulters' confiscated goods and any additional sum it receives from confiscating goods in an equal manner from all the other borrowers. Finally, it imposes punishment

24

Note that if the borrowers were members of several communities, exchange would have been impersonal up to one's community label.

25

Note that the game specification was such that the punishment the BC could impose on a borrower is at least as high as his gains from defaulting.

16

AVNER GREIF

on a borrower who defaulted which erases all the utility gains from defaulting. Hence, theoretically, the social context within which late medieval trade was conducted could have enabled intercommunity impersonal exchange despite the absence of a state supported, centralized legal system. Community affiliation that is known to lenders and community ability to inflict punishment based on limited inter-community social mobility provides the foundations for the operation of an economic institution, the Community Responsibility System. The CRS—by holding the entire community responsible for a default of one of its members—provides the incentives required to conduct exchange that is impersonal up to one's community affiliation. When exchange is governed by the CRS, in considering whether to enter a profitable exchange relation one does not have to examine the private history of the other party. As long as that party's community affiliation is known and the above conditions hold, exchange can be conducted. Hence, exchange is impersonal in the sense that one's past actions do not influence a decision whether to enter into exchange relations with him. In addition, the CRS is effective despite the borrowers' finite life-span and requires a relatively small amount of information to support impersonal exchange. The community internalizes the cost inflicted by the default of a borrower who is about to retire on the other borrowers thereby credibly committing to punish the one who defaulted. Although the community aggregates the payoffs only of its living members, each of whom has a finite life-span, it becomes, de facto, a substitute for a single infinite horizon player. Furthermore, in terms of information the lenders neither have to know the identity of each borrower nor his past actions. Knowing a lender's community label and having the courts able to verify past actions only in cases of default is sufficient for exchange. Finally, note the symbiosis implied by the analysis between social structures and economic institutions. The CRS is an economic institution whose functioning is based on pre-existing social structure. The affiliation of individuals with particular communities and the knowledge of such affiliation enables the operation of the CRS. At the same time, this economic institution reinforces the underlining social structure. It provides communities with additional incentives to define the boundaries between members and non-members, to devise mechanisms for collective decision-making and collective action, and to establish an intra-community contract enforcement institution. Although the CRS could have enabled inter-community impersonal exchange during the late medieval period, this does not imply that such an institution for contract enforcement indeed functioned then. Did the CRS facilitate intercommunity impersonal exchange in western Europe during the late medieval Commercial Revolution? Do the historical records provide evidence consistent with its operation?

PRE-MODERN EUROPE

17

1.4 The Community Responsibility System: History A systematic examination of the extent to which a Community Responsibility System, whose essence is captured by the above model, functioned in pre-modern Europe is yet to be undertaken. Yet, a preliminary investigation indicates it was widespread. In late medieval Europe, a CRS supported inter-community, impersonal exchange and its use and explicit recognition by contemporaries is well reflected in cartularies, charters, court cases, and other sources from Italy, England, France, and Flanders. The origin of the CRS has not yet been revealed by the historical records. It may have emerged during the ninth and tenth centuries when merchants from the same locality traveled together in armed bands to trade abroad. Historical evidence indicates, however, that by the twelfth century it prevailed in Europe. The earliest comprehensive cartulary available from late medieval Italy is that of a Genoese scribe named Giovanni Scriba. One of its entries, registered on July 22, 1164, reveals the operation of a CRS. This entry indicates that shortly prior to 1164 a Genoese trader, Amicus Zostro, received a loan from Xecha Bohadie, a Muslim trader from Tripoli. While Amicus had evidently already arranged to pay Xecha's brother or son in Sicily, Xecha claimed that no such payment had been made. In July 1164, following Xecha's assertion that payment had not been made, Amicus sent an agent named Baldezonus from Genoa to Tripoli carrying six cantras of copper. Baldezonus was instructed to sell the copper and pay Xecha, if the latter would swear in the presence of reliable witnesses that he would hold neither Amicus nor any other Genoese merchant for ransom.26 Another mention of the CRS in Italian sources of the twelfth century, however, indicates that the operation of the CRS was confined neither to the relations between Christian and Muslim traders nor to Genoa at that time. When the Emperor of the Holy Roman Empire, Frederick Barbarossa, visited Bologna in 1155, the students of Bologna's famous law school expressed to him their dissatisfaction with the CRS because the city of Bologna was holding them liable for debts incurred by members of their original communities.27 In twelfth-century England the functioning of a CRS is reflected in a charter given to London sometime between 1130 and 1133. The King, Henry I, announced that ‘all debtors to the citizens of London discharge these debts, or prove in London that they do not owe them; and if they

26

Giovanni Scriba, no. 1245. By the fifteenth century this procedure through which a European community provided a bond for the conduct of its members in the Muslim world had already been institutionalized. As an Arabic writer noted, ‘the consuls are the chiefs of the Franks and are hostages for each community. If anything happens in any community dishonoring to Islam, the consul is answerable.’ Lewis (1988), p. 76.

27

Munz (1969), p. 77.

18

AVNER GREIF

refuse either to pay or to come and make such proof, then the citizens to whom the debts are due may take pledges within the city either from the borough or from the village or from the county in which the debtor lives.’28 Such community responsibility could have even been invoked in response to illegal taxation imposed on Londoners. ‘If anyone has taken toll or custom from the citizens of London, then the citizens of London may take from the borough or village where toll or custom has been levied as much as the man of London gave for toll, and more also may be taken for a penalty.’29 The above charter focuses on the relations between members of different communities within the same political unit, namely, England. It explicitly defines one's relevant community as either his borough, village, or county. A distinct definition of one's community is found in the English legal documents inspected by Maitland (1889). They indicate that at times one's guild was considered to be the relevant community for the operation of the CRS. He noted that in England, in general, the ruling was that members of the same merchant guild, rather than residents of a particular borough, were held responsible for each other.30 ‘Every member of the guild … guaranteed the debts contracted by every member in the way of his trade—is subsidiary liable for those debts. You are a member of the commonality of X:—it is a course of action for me against you that A, who is your “peer and parcener,” your “fellow commoner,” [or] “at scot and lot” with you, has contracted a trading debt with me and has not paid it’ (Maitland 1889: 134). The governance by the CRS of impersonal exchange between merchants from England and those from other political units is reflected in a statement made by King Henry III in 1266. The king granted ‘to his burgesses and merchants of Lubeck, that during the king's life, they or their goods within the king's power shall not be arrested for any debt whereof they are not sureties or principal debtors; unless the debtors are of their commune and power and have failed to pay in whole or part and the said burgesses of Lubeck, by whom the said town is governed fail in justice to the men of the king's land and power, and this can reasonably be proved.’31 In some English boroughs, the process was simpler. Once a foreign creditor could establish that a member of the borough had failed to repay his debt the authorities paid him out of the borough's funds. Later they sought double indemnity from the debtor.32

28

English Historical Documents, vol. II, pp. 1012–13, and see discussion by Stubbs (1913: 128–30) Selected Charters and Other Illustrations of English Institutional History from the Earliest Times to the Reign of Edward the First. In this charter Henry exempted the Londoners from the Community Responsibility System, an issue which is discussed below.

29

English Historical Documents, vol. II, pp. 1012–13.

30

Notwithstanding the fact that in many towns the mercantile and municipal organizations were identical, since the merchant guild was the governing body of the borough.

31

Calendar of the Patent Rolls Preserved in the Public Record Office. 1266–72: 20.

32

Plucknett (1949: 137).

PRE-MODERN EUROPE

19

The operation of the CRS in Flanders around 1252 is reflected in statutes drawn up by Countess Margaret regarding the operation of foreign merchants visiting the Flanders Fairs. Such foreign merchants were held liable for debts assumed by their peers. Yet, consistent with the interpretation that such liability was utilized to ensure punishing the one who defaulted, only a principal debtor or his guarantor could be imprisoned for debt. Other members of the defaulter's community could be punished only by impounding their goods.33 The historical records also indicate changes in the CRS that are consistent with the above theoretical framework. Although most of the above evidence on the operation of the CRS is explicit in stating that goods will be impounded following default, theory indicates that confiscation is not always required for the CRS to enable impersonal exchange. If the value of future trade in the borrower's community is sufficiently high, confiscation or impounding goods is not required to support exchange. To see why this is the case, suppose for the moment that the LC cannot impound goods or that the BC would have to make a transfer. In this case, for the strategy combination discussed in the previous section to be an equilibrium strategy, the present value for the BC from future trade must be more than the gain from not compensating and retaining the amount defaulted upon. (Since otherwise the BC would prefer for trade to cease and hence forgo compensation.) This condition is more likely to hold if the number of borrowers is large or value of trade high, since then the present value of future trade will be higher as well, ceteris paribus. Conversely, this condition is not likely to hold when communities are small and the trade value low. When this condition does not hold, impounding goods may be required to support exchange since it increases the cost to the BC of not compensating. In deciding whether to compensate or not, the BC takes into account that a failure to compensate will also imply the loss of the impounded goods. Hence impounding goods enables exchange at a relatively low level of trade. Thus, theory implies that in the early stages of the Commercial Revolution when communities and their trade were relatively small and growing, the CRS was likely to require actual confiscation. As time passed and trade increased, however, the value of future trade was likely to eliminate the need for actual confiscations. Consistent with this theoretical prediction, thirteenth-century historical records from Italy and England reflect this transition. While the evidence regarding the operation of the CRS in Italy during the twelfth century mentioned above reflects the threat of impounding goods, by the thirteenth century this was no longer the case. A treaty signed between Pisa and Florence in 1214, for example, contains the provision that if one community refused to compensate the other, members of the latter would be allowed 40 days to leave town.34

33

Verlinden (1979: 135).

34

Santini (1886: 165).

20

AVNER GREIF

More than a century later, in 1325, a similar clause appears in a Florentine statute. It required the podestà to wait one month between declaring and acting upon any impounding of goods under the CRS.35 A similar transition to CRS, based on using the value of future trade rather than impounding is reflected in the history of the Champagne Fairs which were the most important trade fairs in Europe during the thirteenth century. Consistent with the above theoretical analysis, the operation of a CRS there relied on the value of future trade rather than impounding. In 1260 the wardens of the Champagne Fairs had the right to pronounce a sentence of exclusion from the fairs following a default, and this exclusion extended to the compatriots of the defaulters if the judicial authorities of their own town or principality did not compel them to fulfil their obligations.36 Interestingly, and consistent with the operation of the CRS, all the major trading communities had representation at the fairs by this time. The discussion so far has assumed perfect information and verifiable actions. Reality, however, is characterized by commercial disputes in which both parties disagreed about whether contractual obligations were fulfilled or not. Furthermore, different courts (and even judges and juries in the same trial) can reach different conclusions based on the same evidence. Thus an empirical study of inter-community contract enforcement institutions can be potentially facilitated by examining implications more realistically. Accordingly, consider the following alteration of the model. Assume that lender–borrower relations are characterized by imperfect monitoring—the lender receives a signal which is a random variable that depends on the action taken by the borrower. Particularly, even if cheating has not occurred, the lender's signal may indicate that he was cheated.37 Further assume that each court also has an independent imperfect monitoring ability—if a dispute occurs each court receives a signal indicating whether cheating has occurred. Each court's signal is private, non-verifiable information and the signals are not perfectly correlated.38 In other words, courts can sincerely disagree about whether cheating took place. The observable implication of this imperfect monitoring is the most efficient equilibrium in which impersonal exchange is conducted will be characterized by periods of inefficient inter-community ‘retaliations’ followed

35

Santini (1886: 168–72).

36

Verlinden (1979: 131). Most likely this right was held previously by the count and the ecclesiastical jurisdictions of the area.

37

The historical records suggest that disputes were more likely to occur when one of the contracting parties passed away, the debt was old, the contract was not clearly defined, or the contracting obligations were allegedly fulfilled by the agents of one of the parties rather than one of the parties themselves.

38

It is assumed, for simplicity, that if there is no dispute, the courts have perfect monitoring ability.

PRE-MODERN EUROPE

21

by ‘suspension’ of the retaliation.39 During retaliation, impounding would occur and exchange would cease, imposing costs on both communities. Such retaliations, however, would last for a finite number of periods following which retaliation would be suspended and exchange would resume. Despite the fact that no cheating occurred (in the sense that a borrower chose not to pay), these costly periods of inefficient retaliations are required to provide the communities and the contracting individuals with the appropriate incentives. Specifically, the only possible equilibrium strategies (that enable exchange) specify confiscation and period(s) during which inter-community exchange ceases when the LC concludes that cheating occurred and the BC concludes that cheating did not occur. Such periods of retaliation are required since if the BC's strategy calls for compensating the lender although it concludes that cheating did not occur, the LC's best response is to claim that a dispute occurred even if it did not. Similarly, if the LC's strategy calls for not continuing confiscation when it maintains that cheating occurred, the BC's best response is not to furnish compensation even if its signal indicates that cheating occurred thereby motivating borrowers' to cheat. Thus, for intercommunity impersonal exchange to be feasible despite imperfect monitoring, inefficiencies (in the form of forgone exchange) must be incurred by the communities.40 Consistent with the theory, impounding (or confiscation) and retaliations following disputes and disagreements among courts are well reflected in the Italian historical records. Since disputes are more likely to occur when trade volume is high, it may not be surprising to find that retaliation became common as trade expanded (Arias 1901). In 1238, Beatrice, wife of Marcovaldo of Florence, requested a retaliation against the properties of the people and the Commune of Pisa, for a sum of 2,000 and interest of

39

This conclusion qualitatively holds even if disputes do not occur, but courts have imperfect monitoring ability and there is a probability that each court would reach a distinct conclusion following a complaint. This puts a limit on the penalty the LC can place on a lender whose complaint was found invalid. If the penalty is too high relative to the court's probability of making a mistake, lenders would not complain even if cheated, and hence exchange cannot take place. Similarly, the BC cannot place a penalty that is too high relative to the probability of a mistake because it would encourage lenders to file false claims. Hence, if borrowers have different time discount factors or distinct probabilities of being detected cheating, the resulting equilibrium can be characterized by a temporary breakdown of cooperation.

40

If retaliations reflect the inability of two communities to objectively verify the conflicting claims made by their citizens, rather than the communities' desire to gain compensation or to take revenge, costly retaliations are unavoidable. Even arbitration will not resolve the dispute since unless it is costly enough, it will not provide the appropriate incentives. The LC, for example, would be induced to submit claims regarding disputes even if it was aware that they were groundless. Indeed, the only case known to me in which communities attempted to employ an arbitrator failed. In 1234, Volterra and San Gimignano submitted themselves to arbitration by the council of Florence in a futile attempt to end a state of retaliation between them. Santini (1886: 168). Note, however, that this arbitration was ex post and hence does not fit well the theoretical reasoning for arbitration's failure.

22

AVNER GREIF

750 dinar piccoli of Genoa, to be paid by the heirs of Ubaldo Viscount and Torritano, the son of the late Lamberto, and by two Pisans who had posted a guarantee for them. The retaliation was granted by the podestà after the Commune of Pisa, which had been asked for restitution (according to the Statute), denied cooperation.41 Such denial, according to the above model, should occur when the two courts differed in their assessment of the situation. That contemporaries considered retaliation to be unavoidable in cases of disagreement among courts is reflected in various commercial treaties. As a matter of fact, in the treaty between Pisa and Florence signed in 1214, it is specified that retaliations would follow if the judges were unable to settle the dispute.42 Disputes were not unique to Italy. A dispute may have been the reason in 1270 that ‘Gottschalk of Almain, burgher of Lynn, complains [in the court of St Ives Fair in England] of the communities of Ghent, Ppoeringen, Douai, Ypres, and Lisle as subjects of the countess of Flanders, for that whereas the said Gottschalk caused 14 sacks of wool worth seven score marks to be brought from the realm of England to Flanders to trade with it there and hosted this wool at the house of a certain Henry Thurold on Sunday.’ The wool, however, was detained in Flanders and the loss amounted to about 200 marks. Yet, the countess of Flanders refused to provide justice. Accordingly, Gottschalk requested that the court impound the goods of members of the above communities present at the fair.43 That retaliations were a calculated response aimed at fostering exchange rather than acts of revenge is suggested by attempts to confine them to only inter-community commercial matters. A 1325 statute from Florence, for example, explicitly enumerated the cases in which it was appropriate to grant retaliation. Retaliation could be granted in cases in which there were losses in currency or goods, damage to properties, extorted taxes, or personal detention. No retaliation was allowed in cases involving personal bodily offenses.44 Further evidence that retaliations were a means to ensure proper incentives rather than compensation per se is suggested by the observation that they indeed lasted for a finite number of periods and communities terminated a retaliation period by announcing a ‘suspension’ without making it conditional on full compensation. Retaliations were not necessarily carried out until full compensation was achieved, but lasted long enough to inflict the appropriate cost to the other side exactly as predicted by the theory.45

41

Santini (1886: 165).

42

Santini (1886: 166–8).

43

Pro SC 2/178/93: 14 May 1270 published in Select Cases Concerning the Law Merchant: A.D. 1270–1638. 1: Local Courts, pp. 9–10.

44

Santini (1886).

45

For a discussion of suspension, see Arias (1901: 177–88). See also Santini (1886: 165).

PRE-MODERN EUROPE

23

Theory also suggests that if the per-period cost of retaliation to two communities is high, they will invest more resources to resolve disputes without retaliation by acquiring better information and devoting more attention to the details. Indeed, as the trade of the Italian cities increased, the process through which retaliation could have been initiated became more circumspect and better regulated. This was accomplished by involving several authorities in the decision-making process. Initiating retaliation in Florence prior to 1250 was basically in the domain of the city administrator and his council. By 1325, however, in order to initiate a retaliation, the city administrator had to make two requests to the Commune to approve it and had to wait one month before enacting the retaliation. In 1415 the process was made even more difficult: a statute detailing the rules for retaliations was listed as being under the authority of consuls responsible for crafts and trade and no longer under the authority of the city's administrator. Yet, for these consuls to grant retaliations they had to be approved by two additional bodies, the Consuls of the Popolo and the Consuls of the Commune.46 The above theoretical and historical discussion emphasizes the association between communities and courts. Yet, fair courts which were not affiliated with a particular community also applied the principal of community responsibility during the late medieval period. An examination of fair court cases indicates that fair courts acted like the LC, holding one member of a community responsible for contractual obligations assumed by another member.47 As mentioned above, the Champagne Fairs held a community responsible for the contractual obligation of each of its members. English fairs acted similarly. For example, sometime in the thirteenth century in the fair of St Botulph in England, a certain James complained that several merchants of Brussels had cheated him. After verifying his complaint, the fair bailiff impounded wool belonging to the merchants from Brussels who were present at the fair.48 How were fair courts motivated to apply the strategy of the LC and demand compensation, or the strategy of the BC and punish a person who defaulted on his obligations? The events in the fair of St Botulph discussed above indicate that some motivation was provided by the fairs' courts legal obligations to the king or another lord. The bailiff of the fair neglected to adhere to the CRS. After a while the ‘bailiff of the fair aforesaid, had wrongfully delivered sacks of the wool aforesaid to the aforesaid merchants [of Brussels] to the grave damage and manifest loss of James himself, inasmuch as the same commune has not yet satisfied him in respect to

46

Santini (1886: 168–72).

47

For many examples of the applications of community responsibility in courts at the English fairs, see Moore (1985). See above for the Champagne Fairs.

48

Selected Cases Concerning the Law Merchant, vol. II, no. 7, pp. 11–12.

24

AVNER GREIF

the debt aforesaid.’ Accordingly, James entered a plea against the bailiff before the ‘lord king’ of England. Yet, economic incentives may have provided an even more important motivation. The right to hold a fair court was granted by the lord who established the fair, and running a successful fair and its court was a profitable business. (Courts were paid by the case.) Hence, those who held the rights to the fair and its court stood to gain from increasing the volume of trade at the fair under their jurisdiction. Guaranteeing inter-community impersonal contract enforcement increased the fair's attractiveness to alien merchants. Note that such incentives might be diluted if the fair court was also a community court since, in that case, the community's own merchants might have to bear the cost of retaliation in case of dispute. Fairs which did not have an affiliated merchant community might have been better able to promote impersonal exchange. Interestingly, a salient feature of Europe's late medieval fairs is that the most important fairs did not have an affiliated community. This is true, at least with respect to the Champagne Fairs, the important trade center of Bruges which replaced the Champagne Fair as northern Europe's trade center, and one of England's most important fairs, the fair of St Ives. A CRS enabled, during the Commercial Revolution, inter-community exchange that was impersonal up to one's community affiliation. Its novelty is that it took advantage of the existing social structure and the repeated nature of the interactions among members of a community to support inter-community impersonal exchange. Although information was limited and each community member had a finite life-span, one's identification by others as a member of a particular community and the strategy this recognition enabled motivated each community member to establish and employ the intra-community enforcement institution to retaliate following a community member's default. Similar to an exchange supported by a centralized legal system, however, the extent to which the CRS could support impersonal exchange was limited by the ability to verify the past actions of people at a relatively low cost. Exchange in which conduct could not be verified by the court, as was the case, for example, in agency relations, could not be governed by the CRS.49 Yet, the CRS enabled inter-community impersonal exchange despite the absence of any legal authority with jurisdiction over the whole scope of the transaction. It was this institution that facilitated the Commercial Revolution by enabling contracts for future delivery and credit arrangements between individuals from various corners of Europe.

49

Indeed, agency relations during the Commercial Revolution were governed by institutions other than the CRS. See, for example, Greif (1989, 1993, 1994).

PRE-MODERN EUROPE

25

1.5 Institutional Decline and Transition: From CRS to Individual Legal Responsibility During the late medieval period the CRS enhanced efficiency by supporting impersonal exchange. Yet, the historical records of the late thirteenth century reflect attempts to abolish the system in Italy, France, and England. Indeed, at that time it was replaced, at least in England, by contract enforcement provided by the state and based on the concept of individual legal responsibility. What led to the decline of the CRS? The common explanation to similar organizational and institutional changes that occurred in the late medieval period is that they reflect either economies of scale associated with increasing population and trade (North and Thomas 1973) or an attempt by the state to dominate communities and fill its own coffers (Benson 1989). Yet, there is not much evidence to support either of these hypotheses with respect to the CRS. There is no measure of the economies of scale associated with either system. Furthermore, the historical records do not reflect a return to the CRS following the great contraction of population and trade that occurred in first half of the fourteenth century and lasted for generations. Similarly, the mere fact that transition away from the CRS was repeatedly attempted also in politically fragmented Italy indicates that the rising centralized state was not necessarily the driving force behind this transition. Hence, it is appropriate to consider whether the decline of the CRS reflects intra- and inter-community economic, social, and political processes that reduced the system's efficiency and undermined its intra-community political viability. Addressing these questions requires examining the cost and limitations of the CRS and its inter-relations with economic, social, and political processes inside and outside communities. The results of a preliminary examination point to the centrality of intra- and inter-community economic and social processes in leading to the decline of the CRS. These processes increased the system's inefficiency and decreased its intra-community political viability. Ironically, the process of economic growth, community expansion, and social integration which was facilitated by the CRS also led to its decline. Theoretically, the first line of causation between the CRS's economic efficiency and social processes is due to the CRS's dependency on the extent to which one's community affiliation can be established. While the above analysis assumes that one's community affiliation is common knowledge, the extent to which this assumption is appropriate depends on the social context within which exchange occurred. If this social context is such that it is relatively easy to falsify and difficult to verify one's community, the less efficient is the CRS. In the early days of the Commercial Revolution, cities may have been sufficiently small and merchants' communities sufficiently few that falsification of one's community was difficult and verification

26

AVNER GREIF

of alleged affiliation was easy. Yet, the late medieval period was a time in which communities grew in size and number. For example, between 1200 and 1300 the population of Genoa increased from about 30,000 to 100,000, while that of Venice increased from about 70,000 to about 110,000.50 In England there were a little more than 200 boroughs at the turn of the thirteenth century, but there were about 500 at its end.51 By the second half of the thirteenth century the ease of falsification and the difficulty of verification seem to have hindered the operation of the CRS in England. As Moore (1985: 119) has noted, ‘this procedure [of the CRS] apparently worked well enough in many cases, but it could be cumbersome and time consuming both for the creditor and the court: it usually seems to have involved long disputes over whether or not the original debtor and/or the men actually being sued for the debt were truly members of their town community or gild, with everyone scurrying to disclaim responsibility for the obligation’. Similarly, Plucknett (1949: 137–8) noted that ‘there seems to have been much trafficking between foreign merchants and natives whose mercantile status was doubtful, and whose assets and persons were by no means entirely within the territorial jurisdiction of a local court’. The ability of individuals to falsify their identity and the strategic use of this ability is well reflected in a case brought before the court of the St Ives Fair ‘on Wednesday next after the feast of St John before the Latin Gate in the eight year of Abbot William,’ which happened to be the year 1275.52 On that day William and Amice of Fleetbridge brought a complaint against Thomas Coventry of Leicester. But since he was not present at the fair, several of Thomas's other ‘peers and parceners,’ namely other merchant members of the community of Leicester, were summoned to the court. William and Amice claimed that Thomas owed Amice money for a sack of wool he bought three years ago from her (late) father, who lived in Leicester, and for which Thomas had been supposed to pay in the following year but had not. To prove their case, William and Amice produced a tally from the court of Leicester.53 Leicester's merchants, who were present at the court and held responsible for the debt, denied, however, ‘any breach of the peace of the lord Abbot and the bailiffs or the fair and the damage of the said William and Amice’ and were ‘ready to verify in such manner as the court shall award that the said Thomas Coventry was never peer or parcener of theirs or at scot and lot with them or a member of the commonality of Leicester.’ The court

50

Bairoch et al. (1988).

51

Beresford and Finberg (1973).

52

The case: pro. SC 2/178/94: 8 May 1275. Parts of the document appeared in Select Pleas in Manorial and Other Seigniorial Courts, Reign of Henry III and Edward I. pp. 145–6, 155.

53

On the tally and its use, see Select Pleas in Manorial and Other Seigniorial Courts, Reigns of Henry III and Edward I. p. 133.

PRE-MODERN EUROPE

27

refused to accept their claim and judged in favor of William and Amice. Yet, shortly after these proceedings, Thomas of Coventry appeared at the fair and did not deny being from the commonality of Leicester. He did claim, however, that William and Amice had brought a false accusation against him, causing him ‘no small damage,’ most likely by the response of the merchants of Leicester whose goods were impounded by the fair court. William and Amice could not defend themselves but claimed not to be under the jurisdiction of the court since they were from London.54 Another link between social and economic processes and the efficiency of the CRS is due to the costs implied by the periods of costly retaliation. Such retaliations are unavoidable transaction costs required to support informal contract enforcement in a world characterized by limited ex post verifiability of actions. Yet, as communities and their trade expand, so does the probability of dispute and hence retaliations. This is the case because the possibility of retaliation increases the uncertainty associated with trade to all merchants. The above model abstracted away from this uncertainty, yet, the CRS implies that there is a positive probability that the goods of a (risk averse) merchant will be confiscated or impounded, he himself will be taken hostage, and his trade will be interrupted, even if he fulfilled all his contractual obligations. The courts' ability to impound goods only from some merchants—those present in their jurisdiction—but not from others, further increases the cost implied by this uncertainty. Whenever retaliation looms on the horizon, merchants are likely to attempt to ensure that their own goods will be saved by leaving the city that is likely to impose sanctions. Hence, theoretically, the relationship between trade expansion and the cost implied by possible disputes is not linear. Rather, the nature of the uncertainty implied that cost would rise faster than trade volume. Indeed, Florence's historical records indicate that once a retaliation was expected, merchants of the communities involved would refrain from trading in each others' cities, while merchants who permanently lived in the other communities would leave their residences with their merchandise.55 Communities attempted to reduce the inefficiency associated with these costs by agreeing to restrict the penalty that could be imposed on a particular individual or allocate the costs more evenly. In 1251, Genoa contracted with Florence to warn its merchants at least two months prior to a retaliation so that Florentine merchants could leave Genoa in an orderly manner.56 After trade expanded sufficiently to enable the CRS to function without impounding or confiscation, taxes on members of one community

54

See below regarding the exception of London from the CRS.

55

For this behavior and its costs in Italy, see Arias (1901: 156–8).

56

Arias (1901: 52).

28

AVNER GREIF

were used by another to better allocate the costs of retaliation.57 Statutes issued by Countess Margaret in 1252 regarding foreign merchants visiting the fairs of Flanders ruled that only a debtor or his guarantors could be imprisoned for debt. Only goods could be confiscated from other members of the debtor's community.58 Yet, the uncertainty implied by the CRS could not be completely avoided and as size of communities increased, the cost of retaliation and uncertainty increased by more. There is, however, another link between a community's size and the inefficiency associated with the CRS that causes the CRS's inefficiency to increase by more than a community's size. The CRS intensifies the moral hazard problem associated with credit financing. It provides insufficient incentives to lenders to examine the creditworthiness of borrowers and the extent to which incentives were insufficient increased with a community's size. Thus, a consequence of the CRS (which is not captured in the above simple model) is that it increased the probability that a borrower would not be able to pay his debt. Since retaliation often occurred following a borrower's bankruptcy, the CRS also increases the probability that a retaliation will occur. Under credit financing, the lender assumes all downside risks while the borrower retains all the gains above a certain amount. Hence, credit financing encourages borrowing for high risk ventures. Appropriate incentives to lenders to evaluate ex ante the creditworthiness of the borrowers mitigates this problem, but the CRS undermined such incentives. Under the CRS the future trade of all members of the borrower's community was the de facto collateral for the loan, and hence a lender had a relatively weak incentive to verify the borrower's ability to repay it.59 In other words, credit financing under the CRS attracted borrowers with ‘bad’ projects which were likely to fail, and thus were more likely than other projects to lead to retaliation. While the above discussion indicates a theoretical possibility, in order to influence the lenders' behavior during the late medieval period they had to be aware that this possibility existed. Interestingly, lenders during this period clearly understood this moral hazard problem. On February 8, 1281, several cities in Tuscany agreed not to retaliate against each other. In announcing this agreement to their merchants, the authorities stated that the merchants should start paying more attention to the personal creditworthiness of merchants from the other towns they would be dealing with. This extra precaution was required, it was argued, since from now

57

For example, on February 22, 1296, Florentine merchants petitioned their city to agree that Bologna would impose a toll (pedaggio ) on Florentine goods entering Bologna in order to settle a retaliation. Arias (1901: 165).

58

Verlinden (1979: 135).

59

Almost one hundred years prior to the invention of Information Economics, Arias (1901: 166), noticed this adverse effect of the Community Responsibility System.

PRE-MODERN EUROPE

29

on ‘a chui dato, a colui rechesto’ (that is, ‘to whom it is given, from him it will be asked’) because retaliation would not be permitted.60 Theoretically, communities could have attempted to mitigate this moral hazard problem by either increasing the cost of defaulting to a borrower who took a risky venture or increasing the cost of default to a lender.61 Indeed, late medieval communities took such actions but with a limited success. The increasing ease by which one could move from one city to another in the thirteenth century implies that, following bankruptcy, one could have escaped to another city. The extent to which this option was exercised and an attempt to curtail it is reflected in a treaty signed on April 7, 1279, among Florence, Genoa, and most of the other towns of Tuscany, Lombardia, Romana, and March Trevigiana. It established that merchants fleeing with money belonging to other people could be imprisoned in the territories of the towns who signed the treaty and there they would be kept until brought to justice.62 In England, the authorities were usually unable to locate an individual who escaped from his place of residence.63 Furthermore, one did not even have to escape in order to avoid paying a debt. During this period English law precluded selling one's house or real estate to repay a loan, or even punishing a borrower who defaulted with imprisonment.64 Another way to mitigate the moral hazard problem is to provide a lender with better incentives to screen prospective borrowers. Had the LC been able to perfectly verify the claim that a particular borrower indeed went bankrupt and not demand compensation in that case, the lender could have been provided with the appropriate incentives. Yet, medieval courts had limited ability to do this. Indeed, retaliation often occurred due to bankruptcy and disputes over the distribution of the remaining funds among various borrowers. Hence, in Italy and England alike, the authorities gradually resorted to increasing the cost of default to the lender by demanding that prior to requesting justice from the legal authorities of his own community, a lender had to travel to the other community and make his case there. Only if justice were not provided, could he apply to his own community court requesting retaliation.

60

Arias (1901: 166–7).

61

Note, however, that if these costs were raised too high, they would undermine the operation of the CRS and the incentive to trade.

62

Arias (1901: 100).

63

Plucknett (1949: 142). As late as the 17th and 18th century, ‘a felon could consider himself distinctly unlucky if he was captured by the authorities. Policing was left largely in the hands of the local community. The maxim was not efficiency, but financial economy, making the system of public order pay for itself. A king with no proper permanent army, who could not pay the members of his own household with regularity, was not likely to visualize or finance a proper police system’ (Bellamy 1973: 201).

64

Jones (1979).

30

AVNER GREIF

For example, the city of Cambridge received a charter and the right to establish a merchant guild as early as the middle of the twelfth century. Yet, only a charter given to the city by King Henry III in 1256 states that ‘our beloved burgesses of Cambridge … may forever throughout the whole of our land and dominion have this franchise, namely that they themselves or their goods, wheresoever found in our dominion, shall not be arrested for any debt of which they shall not be the sureties of principal debtors, unless perchance the debtors shall be of their commonality and power and shall have to make satisfaction for their debts in whole or in part and the said burgesses shall have made default in justice to the creditors of the same debts and this can be reasonable proven.’65 The cost implied by the moral hazard problem depended, similar to the cost associated with falsification, on social factors. First, it depended on community size—since the larger a community is the more likely it is that a lender will be able to recover, at relatively low cost, a debt in case of default. Second, it depended on inter-community ease of mobility. The more advanced is the social and economic integration of a society and the more secure are life and property for aliens abroad, the more difficult it is for a community to discipline its own members. Finally, to the extent that community growth implies increasing intra-community social and economic heterogeneity, it is likely to reduce the intra-community political viability of the CRS. Community growth and heterogeneity imply that some segment of the community bears the cost, but not the benefit, of the CRS. This segment of the community will then be motivated to act within the commune to abolish the CRS. For example, rich, well established merchants gain less from the CRS than others since they already have the connections, reputation, and wealth required to conduct trade, based on personal rather than impersonal exchange, and with little reliance on credit. Yet, the CRS implies that they will have to bear the cost of the system, and perhaps an even higher cost, than those less-well-to-do merchants whose trade depends on the CRS. Similarly, the non-mercantile part of a community is likely to bears the cost implied by the absence of alien merchants in the community during retaliations without directly gaining much, or anything, from the CRS. Hence, while the CRS fostered inter-community economic interactions and facilitated the growth of trade, as well as growth in size, number, and heterogeneity of merchants' communities, theory suggests that these processes diminished the system's effectiveness, increased its economic costs, and undermined its political viability. Unfortunately, establishing quantitatively the extent to which this was the case seems impossible, but there is an indication that these costs were rising in the second half of the thirteenth century in England according to the Close Rolls. Debts could be

65

Maitland and Bateson (1901: 14–15).

PRE-MODERN EUROPE

31

registered in these chancery rolls, thereby placing the transaction under the jurisdiction of the Common Law. This implies that property and goods could have been placed as bond for repaying debts.66 Registration, however, was costly and prior to 1271 few debts, if any, were enrolled each year. Their numbers rose substantially by 1271, however, just before (as discussed below) the CRS was abolished in England.67 Yet, the historical records do reflect, consistent with the above theoretical considerations, that wealthy traders and large communities attempted, in England and Italy alike, to be exempted or to abolish the CRS and different segments of cities had distinct incentives with respect to the CRS. Moore (1985) examined cases brought before the court of St Ives Fair in England and noticed that in the second half of the thirteenth century there ‘was an increasing numbers of individuals … able to respond to … suits [related to the CRS] by producing royal licenses of immunity from prosecution for any debts except those for which they were principal debtors or pledges’ (p. 119). In England large communities sought an exemption from the CRS from the king. With about 25,000 residents, twelfth-century London was the largest city in England and a large city by western European standards of that time.68 In 1133, Henry I declared that the citizens of London ‘shall appoint as sheriff from themselves whomever they may choose, and shall appoint from among themselves justice whomsoever they choose’ and ‘no other shall be justice over the men of London.’ Indeed, as mentioned above, William and Amice claimed that they could not be judged at the St Ives court. Yet, as cited above, Londoners had the right to apply the principle of community responsibility to residents of other cities.69 Later, the king provided other large communities with various exemptions from the CRS. In Flanders, Ypres was the largest city and about twice the size of London. Indeed, King Henry III, sometime between 1225 and 1232, assured the merchants of Ypres that none of them ‘will be detained in England…nor will they be partitions for another's debts. ’ Only the debtor, or those who made a pledge for him, would be liable for the debt. Yet, the King seemed to be aware of the need to induce the community of Ypres to ensure any merchant that he would receive justice if a complaint was brought against him. It was agreed that ‘if any aforesaid burger or merchant of Ypres were to offend the King or other men or merchants from

66

See discussion in Moore (1985: 120), no. 105.

67

Close Rolls of the Reign of Henry III, 1227–72. 14 vols. London: His Majesty's Stationery Office, years 1256–1272. There is one entry for 1257; four for 1269; and 43 for 1271. For the high cost of a merchant using the common law court, see Plucknett (1949: 137).

68

Bairoch et al. (1988: 33).

69

English Historical Documents, vol. II, no. 270, pp. 1012–13.

32

AVNER GREIF

England, or if a dispute were to arise between another man of his and a man from England, it will stand by law in the courts of the king by the king, or by his judges, or by his bailiffs in that place where the offense was committed for the purpose of amending the mistake and making proper payments.’70 The historical records from Italy reflect the reduction in the CRS intra-community political viability. Distinct incentives among various segments of Florence with respect to the CRS are reflected in an appeal made on 22 February, 1296, by some Florentine merchants to the authorities of the city regarding a retaliation with Bologna. These were merchants whose livelihood depended on being able to pass through Bologna. They proposed setting up a toll (pedaggio) to be levied almost exclusively on their goods, just to settle the retaliations in which they were probably not directly involved. It seems that the rest of Florence did not care about settling the matter.71 Similar distinct incentives are reflected in Florence's fifteenth-century regulations. In 1415 it was forbidden to retaliate against foreign rectors, officials, or against traders coming to Florence to sell edibles.72 The mercatores of Florence were the city's affluent merchants whose business during the thirteenth century was carried out over most of Europe, as far north as Sweden. While they may have had the ability to exchange based on their own reputation, they had a great deal to lose from retaliations. Indeed, once they secured political control over the city, they entered into a sequence of treaties aimed at moving Florence away from the CRS.73 On April 9, 1279, the cities of Florence, Venice, Genoa, as well as most of the cities of Tuscany, Lombardy, Romagna, and Marca Trivigiana, agreed ‘that from this day forth nobody of the said city-states is able to be or should be, on behalf of another, detained or taken captive or disturbed, in person or goods, but it should be demanded of him alone to whom it should be given, or of him who by justice should be held.’ To enable impersonal inter-community exchange, however, it was also agreed that each town would imprison any merchant in its territory who was fleeing with other people's money and that his creditors would receive justice.74 That most of Italy's large city-states also sought an end to the CRS in the second half of the thirteenth century suggests that by that time, the most important Italian communities sought to abolish the system. The examination of the costs associated with the CRS, and the relationship between these costs, and social, economic, and political processes within and between communities, suggests that these processes were important factors in the decline of the CRS. In England, the state may have

70

Calendar of the Patent Rolls Preserved in the Public Records Office, 1232–1339: 460.

71

Arias (1901: 165).

72

Santini (1886: 168–72).

73

Arias (1901: 170–6).

74

The Latin version of this treaty is contained in Arias (1901: 400–4).

PRE-MODERN EUROPE

33

contributed to an increase in the ineffectiveness of the CRS by granting individuals and communities legal immunity from impounding or confiscating their goods. In doing so, however, the state seems to have been responding to demands reflecting processes within and between communities. The observation that the Italian city-states also attempted to move away from the CRS toward the end of the thirteenth century, strongly supports the view that the state was not the originator of this transition. Yet, the extent to which communities in Italy and England could have abolished the CRS did depend on the existence of a central state. In Italy, communities could not rely on a third party, a King, to provide them with an alternative institution. Indeed, while the extent to which inter-community treaties to abolish the CRS in Italy achieved their aim should still be examined empirically, retaliations did continue after the thirteenth century. This was not the case in England, where the state enabled the communities to abolish the CRS. In 1275, King Edward I issued the Statute of Westminster I that forbade applying community responsibility to debts, and established the following ruling with respect to any ‘stranger who is of this kingdom [namely, an Englishman from one locality present in another].’ Such an individual should not ‘be distraint in a city, borough, vill, fair or market for what he is neither debtor nor pledge for, and he who does this is to be severely punished and the distress is to be released without delay by the bailiffs of the place or by the other, the king's bailiffs if need be.’75 The Statute of Westminster I did not establish any alternative institution for contract enforcement and the results—in terms of inability to contract—were apparently severe.76 A later statute in 1283 declared that ‘merchants who in the past have lent their substance to various people are impoverished because there was no speedy law provided by which they could readily recover their debts on the day fixed for payment, and for that reason many merchants are put off from coming to this land with their merchandise to the detriment of merchants and of the whole kingdom.’77 Edward seems to have abolished the system, recognizing its cost, only to later realize its benefits. The alternative contract enforcement institution established by Edward I may have been influenced by a similar institution that had been enacted in France. This conjecture gains some support from the observation that in England it was first established in the Channel Islands in 1279, and only later, in 1283, in the rest of England.78 The Statute of Acton Burnell of 1283 established this alternative contract enforcement institution by

75

English Historical Documents, vol. III, p. 404.

76

Although the merchants could have used, as mentioned above, the costly and time consuming common law court.

77

English Historical Documents, vol. III, p. 420.

78

See discussion in Patourel (1937: 97) who also refers to the work of Giry, Manuel de Diplomatique Paris (1925: 649–52, 835–54) who discusses this institution in France.

34

AVNER GREIF

enabling, though not requiring, a creditor and debtor to appear before the mayor of either London, York, or Bristol, acknowledge the debt, and have it registered in a roll. The creditor would then be given the debtor's bond sealed by a special royal seal kept by these mayors for this purpose. In case of default the creditor did not need to bring an action of debt, but could resort to immediate confiscation and selling of the chattels and divisible property that were placed as bond. Recovering the loan through a forced selling of the bond, if necessary, was the responsibility of the mayor or sheriff in whose jurisdiction the bond was to be found. If the proceedings from the bond were not sufficient to cover the debt, the debtor would be imprisoned.79 Despite its intent, the Statute of Acton Burnell failed to provide an appropriate contract enforcement mechanism, and two years later, in 1285, the Statute of Merchants had altered some of its provisions because ‘merchant complained to the king that sheriffs misinterpreted his statute and sometimes by malice and by misinterpretation delayed the execution of the statute to the great detriment of the merchants.’80 The provisions of this statute indicate the difficulties in providing the appropriate incentives to those who were supposed to administer the legal procedures established by the Statute of Acton Burnell. Among the provisions were the following: instead of one administrator, two were required to produce a roll specifying a debt, one nominated by the king and the other by the local authorities. Two rolls specifying the bond had to be made, and after being sealed by the seal of the debtor, each would be held by another administrator. Forced sales were abolished and the debtor was imprisoned until the matter was settled but he was given three months to enter any contract necessary to raise the money required for paying his debt. If he failed to do that, the Statute of Merchants and the Statute of Westminster II established that his chattels and his land could be handed over to the creditor in compensation. The land, however, could not be alienated, but the proceedings from the land could be used to pay the debt.81 Finally, the ability to register and seal contracts subject to this procedure would be made available to merchants in towns other than London, York, and Bristol, and in every fair. The procedures established in England between 1283 and 1285 provided the basis for a contract enforcement mechanism that enabled impersonal exchange based on a central legal system and individual responsibility. The system did not mature overnight, and several improvements were made over the years to enhance its functioning. For example, in 1352, the

79

The Statute is published in English Historical Documents, vol. III, no. 54, pp. 420–2. For a discussion of this Statute and other relevant developments, see Plucknett (1949), pp. 138–50; Moore (1985), p. 120.

80

This Statute appears in English Historical Documents, vol. III, no. 58, pp. 457–60.

81

For the Statute of Westminster II of 1285, see English Historical Documents, vol. III, no. 57, pp. 428–57 (and see in particular c. 18).

PRE-MODERN EUROPE

35

common creditors were ranked with the crown's creditors insofar as imprisonment of the defaulted debtors were concerned, and outlawry was extended to debt and actions of account.82 Yet, even as late as 1543, the authorities could not break into the locked house of a debtor who defaulted on his debt.83 A contract enforcement institution based on individual responsibility, similar to the contract enforcement institution based on collective responsibility which it replaced, developed slowly.

1.6 Conclusions A Community Responsibility System enabled exchange that was impersonal up to one's community affiliation in late medieval Europe. It took advantage of the existing social context, namely, the existence of communities, the relatively easy identification of one's community, and intra-community contract enforcement institutions that were effective despite their local nature because of the social and economic cost of inter-community mobility. These features of the medieval society were supplemented by the strategies that provided individuals and communities with the appropriate incentives. It enabled impersonal exchange despite the absence of the legal contract enforceability provided by a state and human's finite lifetime. Furthermore, at least in England, the state itself attempted to further enhance the institution based on community responsibility by making its strategies the legal obligation and right of courts in communities and fairs. While the preliminary historical examination presented above indicates that this contract enforcement institution facilitated impersonal exchange throughout Europe, it came with a price. It restricted exchange to members of identifiable communities and required local courts. Yet, the biggest cost it imposed was due to courts' imperfect monitoring ability. It led to periods of retaliation, increased the uncertainty of trade, and amplified the moral hazard problem associated with credit financing. Over time, various communities and legal authorities became aware of the origin of these costs and labored to contain them by increasing the accuracy of information, smoothing the uncertainty to individual merchants due to retaliation, and increasing the cost of bankruptcy to borrowers and lenders alike. The growth of long-distance trade and communities' size, number, and heterogeneity fostered by the CRS also reduced the system's economic efficiency. These economic and social processes made it easier to falsify community affiliation, hindered verification of one's community, reduced the

82

Plucknett (1949: 324–6, 343). Neither did the system enhance the ability to recover debt when the debtor was an alien and landless in England. See discussion in Moore (1985: 121).

83

Jones (1979: 13–16).

36

AVNER GREIF

cost of escaping to another community for those who defaulted or went bankrupt, and increased the severity of the moral hazard problem. Furthermore, increasing diversity within communities implied that the cost of the CRS to communities or some segments in the community outweighed the benefit. By the late thirteenth century, communities were laboring to abolish the CRS and, where possible, the state stepped in to provide an alternative. The process through which individual legal responsibility came to dominate the modern legal system moved forward. This study of a particular impersonal contract enforcement institution during the late medieval period reveals the importance of understanding the inter-relations between social, economic, and political factors in determining the set of feasible and actual economic institutions. It indicates that economic institutions can be based on and provide a positive reinforcement to a particular social structure. While the CRS was based on the existing, community-based social structure, it reinforced it by motivating the community members to clearly define their communal membership, to communicate it to the rest of the society, and to strengthen their intra-community contract enforcement institutions. At the same time, the dependency of the economic institution on its social foundations implies that social, as well as economic, processes can erode the effectiveness and political viability of the economic institutions. While similar conclusions had been reached in studies that examined contract enforcement within communities (e.g. Ellickson 1991; Greif 1989, 1993, 1994), the study of impersonal exchange reveals the importance of a particular aspect of social context. Sociologists have long emphasized the cognitive aspects of communities—namely, that a community exists when individuals consider themselves and are considered by others as members of that community. The study of the CRS indicates the importance of incorporating this cognitive aspect in game-theoretical analysis for comprehending the inter-relations between the social context and economic institutions. For the CRS to function, the social affiliation of individuals with communities had to be known. Unless this is the case, impersonal, intercommunity exchange is not feasible even if communities do exist. There is much evidence for the operation of economic institutions that built on the existing social context to enable impersonal exchange in past and present non-European economies.84 Yet, the extent of their contribution to past and present economic development, the exact forms of these institutions, and their evolution and interrelationships with political and legal institutions are yet to be explored. We do not know, for example,

84

See Varian (1990), Besley and Coate (1995), Bouman (1995) regarding joint liability in credit markets in developing economies and Fairbank (1992: 177) regarding exchange in imperial China.

PRE-MODERN EUROPE

37

what implications distinct social processes associated with development in pre-modern Europe and developing countries have had on the ability of community enforcement institutions to enhance economic efficiency. Analyzing these institutions and their inter-relations with social structures in different historical episodes will lead the way to a comparative institutional investigation that may shed additional light on the evolutionary process of economic institutions and development. Be the conclusion of this analysis as it may, the broad policy conclusion that this paper yields is clear: one can attempt to utilize a society's existing social structures to foster contract enforcement required for advancing development. At the same time, one has to realize that social structures can be taken as exogenous to economic processes only in the short or medium run. The same economic processes that economic development brings about can erode the underlying social structures. Unless an economy is able to respond by developing contract enforcement institutions that can replace those based on social structure, the economy may be stifled.

References Arias, G. (1901) I Trattati Commerciali della Repubblica Fiorentina, Firenze, Successori le Monnier. Ashburner, M. A. W. (1909) The Rodian Sea-Law, Oxford, Clarendon Press. Bairoch, P., Batou, J. and Chêvre, P. (eds.) (1988) The Population of European Cities from 800 to 1850, Genève, Center of International Economic History (in French and English). Bellamy, J. (1973) Crime and the Courts in England, 1660–1800, Princeton, Princeton University Press. Ben-Porath, Y. (1980) ‘The F-Connection: Families, Friends, and Firms and the Organization of Exchange,’ Population and Development Review, 6: 1–30. Benson, B. L. (1989) ‘The Spontaneous Evolution of Commercial Law,’ Southern Economic Journal, 55(January): 644–61. Beresford, M. and Finberg, H. P. R. (1973) English Medieval Boroughs: A Handlist, Newton Abbott, David and Charles. Berman, H. J. (1983) Law and Revolution: The Formation of Western Legal Tradition, Cambridge, MA, Harvard University Press. Bernstein, L. (1992) ‘Opting Out the Legal System: Extralegal Contractual Relations in the Diamond Industry,’ Journal of Legal Studies, XXI (January): 115–57. Besley, T. and Coate, S. (1995) ‘Group Lending, Repayment Incentives and Social Collateral,’ Journal of Development Economics, 46(Feb.): 1–18. Bouman, F. J. A. (1995) ‘Rotating and Accumulating Savings and Credit Associations: A Development Perspective,’ World Development, 23(3): 371–84. Calendar of the Patent Rolls Preserved in the Public Record Office (1893–1910) 14 Vols. London, His Majesty's Stationery Office.

38

AVNER GREIF

Cassinese, Guglielmo. Cartolare di. 1190–1192. Edited by Hall, M. W., H. G. Krueger, and R. L. Reynolds 1938. (Documenti, XII–XIII), Turin. Chiaudano, M. (ed.) (1935) Giovanni Scriba, Cartolare di, 1154–1164, vols. I, II, Turin. —— and Morozzo Della Rocca, R. (eds.) (n.d.) Obertus Scriba de Mercato, 1186–1190, (Documenti, XI and XVI). Turin. Clay, K. (1997) ‘Trade Without Law: Private-Order Institutions in Mexican California,’ Journal of Law, Economics, & Organization, 13(1): 202–31. Close Rolls of the Reign of Henry III, 1227–72 (1902–1938) 14 vols. London, His Majesty's Stationery Office. de Roover, R. (1963) ‘The Organization of Trade,’ in The Cambridge Economic History of Europe. vol. III. Cambridge, Cambridge University Press. Ellickson, R. C. (1991) Order Without Law. Cambridge, MA, Harvard University Press. Ellison, G. (1994) ‘Cooperation in the Prisoner's Dilemma with Anonymous Random Matching,’ Review of Economic Studies, 61(3): 567–88. Emery, R. (1952) ‘The Use of the Surname in the Study of Medieval Economic History,’ Medievalia et Humanistica, VII: 43–50. English Historical Documents (1968) Edited by David C. Douglas and George W. Greenaway, London, Eyre and Spottiswoode. vol. II, pp. 1042–189. English Historical Documents (1975) Edited by Harry Rothwell, London, Eyre and Spottiswoode. vol. III, pp. 1189–327. Face, R. D. (1958) ‘Techniques of Business in the Trade Between the Fairs of Champagne and the South of Europe in the Twelfth and Thirteenth Centuries,’ The Economic History Review, 10: 427–38. Fafchamps, M. (1995) ‘Trade Credit in Zimbabwe,’ in Spring, A. and McDade, B. (eds.), African Entrepreneurship, Indiana University Press. Fairbank, J. K. (1992) China, A New History, Cambridge, MA, Harvard University Press. Fudenberg, D. and Tirole, J. (1991) Game Theory, Cambridge, MA, MIT Press. Granovetter, M. (1985) ‘Economic Action and Social Structure: The Problem of Embeddedness,’ American Journal of Sociology, 91(3): 481–510. Greif, A. (1989) ‘Reputation and Coalitions in Medieval Trade: Evidence on the Maghribi Traders,’ Journal of Economic History, XLIX(Dec.): 857–82. —— (1993) ‘Contract Enforceability and Economic Institutions in Early Trade: The Maghribi Traders' Coalition,’ American Economic Review, 83(3). —— (1994) ‘Cultural Beliefs and the Organization of Society: A Historical and Theoretical Reflection on Collectivist and Individualist Societies,’ The Journal of Political Economy, 102(5): 912–50. —— (1997) ‘Contracting, Enforcement, and Efficiency: Economics Beyond the Law,’ in Bruno, M. and Pleskovic, B. (eds.), Annual World Bank Conference on Development Economics, Washington, DC, The World Bank. —— (Forthcoming) Culture and the Institutional Foundations of States and Markets. Historical and Comparative Institutional Analysis of Genoa and the Maghribi Traders, Cambridge University Press. Grief, A., Milgrom, P., and Weingast, B. (1994) ‘Coordination, Commitment and Enforcement: The Case of the Merchant Guild,’ The Journal of Political Economy, 102(4) (August).

PRE-MODERN EUROPE

39

Gross, C. (ed.) (1908) Select Cases Concerning the Law Merchant: A.D. 1270–1638, 1: Local Courts, Selden Society Publications, 23, London, B. Quaritch. Guinnane, T. W. (1994) ‘A Failed Institutional Transplant: Raiffeisen's Credit Cooperatives in Ireland, 1894–1914,’ Explorations in Economic History, 31(Jan.): 38–61. Hall, H. (ed.) (1930) Select Cases Concerning the Law Merchant, A.D. 1239–1633. 2: Central Courts, Selden Society Publications, 46, London, B. Quaritch. Hanawalt, B. (1974) ‘The Peasant Family and Crime in Fourteenth-Century England,’ Journal of British Studies, 13: 1–18. Hirschman, A. O. (1982) ‘Rival Interpretations of Market Society: Civilizing, Destructive, or Feeble?’ Journal of Economic Literature, XX (4, Dec.): 1463–84. Jones, W. J. (1979) The Foundations of English Bankruptcy: Statutes and Commissions in the Early Modern Period, Philadelphia, Transactions of the American Philosophical Society. Kandori, M. (1992) ‘Social Norms and Community Enforcement,’ Review of Economic Studies, 59. Klein, D. B. (1992) ‘Promise Keeping in the Great Society: A Model of Credit Information Sharing,’ Economics and Politics, 4(2): 117–36. Krueger, H. C. (1957) ‘Genoese Merchants, Their Partnerships and Investments, 1155 to 1164,’ in Studi in Onore di Armando Sapori, Milano, Institudo Editoriale Cisalpino. —— (1962) ‘Genoese Merchants, Their Associations and Investments, 1155 to 1230,’ in Studi in Onore di Amintore Fanfani, Milano, Multa Paucis. —— and Reynolds, R. L. (eds.) (1952–4) Lanfranco Scriba, 1202–1226, (Notari Liguri del secolo XII e del XIII, I–III), Genoa. Landa, J. T. (1994) Trust, Ethnicity, and Identity, Ann Arbor, University of Michigan Press. Lewis, B. (1988) The Political Language of Islam, Chicago, Chicago University Press, p. 76. Lopez, R. (1954) ‘Concerning Surnames and Places of Origin,’ Medievalia et Humanistica VIII: 6–16. —— (1976) The Commercial Revolution of the Middle Ages, 950–1350, London, Cambridge University Press. Lopez, R. S. and Raymond, I. W. (1955) Medieval Trade in the Mediterranean World, New York, Columbia University Press. Macaulay, S. (1963) ‘Noncontractual Relations in Business: A Preliminary Study,’ American Sociological Review, 23: 55–70. Macneil, I. R. (1985) ‘Relational Contract: What We Do and Do Not Know,’ Wisconsin Law Review, 3: 482–524. Maitland, F. W. (ed.) (1889) Select Pleas in Manorial and Other Seigniorial Courts, Reigns of Henry III and Edward I, London, Seldon Society Publications, 2. —— and Bateson, M. (1901) The Charters of the Borough of Cambridge, Cambridge, Cambridge University Press. Milgrom, P. R., North, D. C., and Weingast, B. R. (1990) ‘The Role of Institutions in the Revival of Trade: The Law Merchant, Private Judges, and the Champagne Fairs,’ Economics and Politics, 2(1): 1–23.

40

AVNER GREIF

Montgomery, J. (1991) ‘Social Networks and Labor-Market Outcomes: Toward an Economic Analysis,’ American Economic Review, 81(5): 1408–18. Moore, E. W. (1985) The Fairs of Medieval England, Toronto, Pontifical Institute of Medieval Studies. Munz, P. (1969) Frederick Barbarossa, Ithaca and London, Cornell University Press. North, D. C. (1991) ‘Institutions,’ Journal of Economic Perspectives, 5(1): 97–112. —— and Robert P. T. (1973) The Rise of the Western World, Cambridge, Cambridge University Press. Patourel, J. H. le (1937) Medieval Administration of the Channel Islands, 1199–1399, Oxford, Oxford University Press. Pirenne, H. (1956 [1936]) A History of Europe, New York, University Books. Plucknett, T. F. T. (1949) Legislation of Edward I, Oxford, Clarendon Press. Postan, M. M. (1973) Medieval Trade and Finance, Cambridge, Columbia University Press. Reynolds, R. L. (1929) ‘The Market for Northern Textiles in Genoa 1179–1200,’ Revue Belge Philologie et d'Historie, 8: 831–51. —— (1930) ‘Merchants of Arras and the Overland Trade with Genoa Twelfth Century,’ Revue Belge Philologie et d'Historie, 9: 495–533. —— (1931) ‘Genoese Trade in the Late Twelfth Century, Particularly in Cloth from the Fair of Champagne,’ Journal of Economic and Business History, 3: 362–81. Rosenberg, N. and Birdzell, L. E. Jr (1986) How the West Grew Rich, New York, Basic Books. Santini, P. (1886) ‘Appunti sulla Vendetta Privata e sulle Rappresaglie,’ Archivio Strico Italiano, XVIII: 162–76. Stubbs, W. (1913) Selected Charters and Other Illustrations of English Institutional History from the Earliest Times to the Reign of Edward the First, 9th edn. Oxford, Clarendon Press. Sugden, R. (1989) ‘Spontaneous Order,’ Journal of Economic Perspective, 3(4): 85–97 Telser, L. G. (1980) ‘A Theory of Self-enforcing Agreements,’ Journal of Business, 53: 27–43. Tirole, J. (1996) ‘A Theory of Collective Reputation (with applications to the persistence of corruption and to firm quality),’ The Review of Economic Studies, 63(1): 1–22. Townsend, R. M. (1981) ‘Models of Money with Spatially Separated Agents,’ in Kareken, J. H. and Wallace, N. (eds.), Models of Monetary Economies, Minneapolis, Federal Reserve Bank. Van der Vee, H. (1977) ‘Monetary, Credit, and Banking Systems,’ in Rich, E. E. and Wilson, C. H. (eds.), The Cambridge Economic History of Europe, vol. V, Cambridge, Cambridge University Press. Varian, H. R. (1990) ‘Monitoring Agents with Other Agents,’ Journal of Institutional and Theoretical Economics, 146 (1): 153–74.

PRE-MODERN EUROPE

41

Verlinden, C. (1979) Markets and Fairs. Vol. 3. The Cambridge Economic History of Europe, Edited by M. M. Postan, E. E. Rick, and M. Miltey. Cambridge, Cambridge University Press. Weber, M. (1927) General Economic History, translated by F. H. Knight, New York, An Adelphi Publication, Greenberg. Williamson, O. E. (1985) The Economic Institutions of Capitalism, New York, The Free Press.

2 Community and Market in England: Open Fields and Enclosures 85

Robert C. Allen Opposition between ‘community’ and ‘market’ originated with nineteenth-century social theorists like Marx (1867) and Tonnies (1887), who contrasted the modern, capitalist world with a preceding feudal or communal world. The modern world was seen as rational, efficient, and progressive. Economic activities were coordinated by the market and economic agents responded to prices. In contrast, the earlier communal world was seen as irrational and stagnant. Tradition governed economic behavior and thwarted progress. Redistribution rather than growth was the goal. The replacement of ‘community’ by ‘market’ was essential for economic progress. These views have been elaborated in the twentieth century by Polanyi (1944), Geertz (1963), and Scott (1976). Agrarian organization was the prime example of this dichotomy. In this case, ‘community’ meant the open field village. Open field agriculture was widespread (but not universal) in England and on the continent. Settlement in open field areas was in nucleated villages rather than dispersed farms and hamlets. Each farmer's land consisted of strips scattered uniformly around the village. The strips were grouped in great fields (usually two or three), which served as crop rotation units. In the case of the three field system, for instance, all the land in one field was planted with a winter grain (wheat or rye), another was planted with a spring crop (oats, barley, beans, or peas), while the third was left fallow. In addition to the arable, there was some meadow and rough pasture. The latter was owned collectively and managed as a common on which the village livestock was grazed. The village herd was also grazed on the fallow field and the cropped fields after the harvest. In this way, the land was fertilized and weeded. While there were elements of private ownership in this system—each farmer owned particular strips in the fields—there were significant communal aspects: The common pasture was collectively owned and managed, and the village herd was also grazed in common on the fallow field. For that to happen, all of the

85

I would like to thank Masahiko Aoki, Hans Binswanger, Marcel Fafchamps, and Yujiro Hayami and other participants at the World Bank Conferences in Paris (1998) and Stanford (1999) for their comments and discussion.

COMMUNITY AND MARKET IN ENGLAND

43

farmers had to follow the same crop rotation, and that was regulated by the village community often under the aegis of the manorial court. A long tradition regards the open field village as hidebound and resistant to progress. In Ernle's (1912: 199) memorable phrase, ‘open-field farmers were impervious to new methods.’ In this view, agricultural progress required the elimination of village control through enclosure. In an enclosure, commons were abolished, land ownership was consolidated, and the institutions of communal management were abandoned. Each landowner or his tenant farmer could use the land as he saw fit. Enterprising farmers could no longer be held back by a hidebound peasant majority opposed to progress. The overstocking of commons was also eliminated. When communalism ended, progress began. This essay presents an alternative interpretation of the open fields based on the historical research of the last thirty years. The guiding idea of this research is that the open fields were an efficient institution for meeting the needs of small scale, grain growing farmers. These needs included diversification against the risk involved in grain production and increasing agricultural productivity. McCloskey (1972) has championed the view that the scattered landholdings in the open fields can be explained as diversification. Land was not uniform, so the productivity of different parts of a village's land responded differentially to variation in the weather. In years of high rain fall, low lying land might have been waterlogged and given low yields, while higher land might have been productive. Conversely, when rainfall was light, the upland might have been too dry to produce well, while yields might have been high in the low lands. Scattering is a feature of low income agriculture around that world, so a simple explanation applicable everywhere commends itself. Diversification is the usual explanation (Hayami 1997: 261). The second need of poor farmers was higher productivity, and the open fields were an efficient organization for achieving that goal. Since this conclusion is so important and so inconsistent with the standard, negative assessment of open fields, the first purpose of this paper is to present the case for their progressiveness. The conclusion that open field agriculture was conducive to technical progress raises two further questions that have not been intensively studied. The first relates to decision making in the open fields. The negative assessment of open fields presumes that technical change required widespread agreement, which was hard to obtain. How broad a consensus was, in fact, required, and, how was it achieved? The second relates to the enclosure movement. If the open fields were so productive, why were they enclosed? This paper addresses these questions through a case study of open field performance in Spelsbury, Oxfordshire. While history never repeats itself exactly, it can suggest likely patterns of development. While the Ernle view of open fields as backward and

44

ROBERT C. ALLEN

unprogressive raised the presumption that communal institutions were a poor basis of economic organization, the more positive vision of the open field community developed here leads to the presumption that communal institutions can effectively manage economic resources and secure rising productivity.

2.1 Efciency of Open Fields—Micro Evidence The economic performance of open field agriculture has been extensively researched since the World War II. A variety of indicators—crop choice, yield per acre, output per worker and various measures of total factor productivity—have been used. This research all points to a far more favorable assessment of open fields than is found in the social science literature that contrasts the open field ‘community’ with the enclosed ‘market’ economy. Before considering the evidence on agricultural efficiency, it should be emphasized that the Polanyi view that open field agriculture was ‘precommercial’ is not seriously entertained by most historians today. Quite the contrary. MacFarlane's highly influential The Origins of English Individualism (1978) argued that English farmers were commercial as far back as the middle ages. This has become the modern view, as evidenced, for instance, by the title of Britnell and Campbell's recent collection A Commercializing Economy: England 1086 to c. 1300 (1994). The Polanyi depiction of open fields as premarket organizations—a view asserted without evidence—can be tested with probate inventories commencing in the early sixteenth century. Most people in this period—including, in particular, small scale farmers—made wills and when they were proved, the executor of the estate was required to produce a list and valuation of the deceased's property. Most farmers had unsold grain in the barn, and one can compute the value per bushel placed on this wheat and barley by the executors. These values were similar to wholesale market prices both in level and movement, as a comparison of the two series shows (Allen 1988). Evidently, the executors, who were themselves often farmers, were attuned to market prices and used them for valuation. It is difficult to see how these findings can be squared with a ‘precommercial’ view of the open fields. If open field farmers had a commercial outlook, it would be surprising if they did not try to raise efficiency. This issue has been explored with data ranging from the middle ages to the nineteenth century, and the consensus is that open field agriculture was progressive. Our knowledge of open fields in the middle ages (1000–1500) is sketchy, but even for that period there is overwhelming evidence for advances in efficiency. The two field system was the norm in many open fields in the early middle ages, but many of these systems were reorganized as three

COMMUNITY AND MARKET IN ENGLAND

45

(sometimes four) field systems by the early modern period (Gray 1915; Thirsk 1973: 255–62). This change reduced the fallow from one half to one-third of the farm land, leading to a substantial increase in output. Such a reorganization would have affected everyone and so is considerably important. A second important change in medieval agriculture was the substitution of horses for oxen. While this change did not directly involve the village community, there may have been a link in that the shift to three field farming probably increased the production of oats, which were eaten by the horses. In any event, the adoption of the horse shows that small scale open field farmers were willing to take up innovations (Langdon 1986). Several important changes in land management were made in the early modern period (1500–1750). First, peas and beans were added to open field rotations. Since the new crops replaced other spring crops, the change was not too far reaching but still shows the innovativeness of open field farmers (Hoskins 1950, 1951). Second, crop yields rose substantially in the open fields. For wheat, the premier crop, the increase was particularly large and equalled most of the gain achieved by enclosed farmers c. 1800 compared to their medieval, open field predecessors. For barley and oats, the gains were less substantial but still impressive (Allen 1999). Why yields rose is an ongoing subject of research, but that they rose is further testimony to the innovativeness of open field farmers. Third, open field villages in areas well suited for pasture converted much of their arable to grass and greatly expanded livestock production. These changes emulated the pastoral husbandry followed in enclosures in the same districts. Open field farmers seem to have converted less arable to pasture than did enclosed farmers, but the change was far reaching for the open field villagers since it required a reorganization of the main aspects of field management (Allen 1992: 121–5). All told the changes in farm methods in the early modern period were far reaching in their importance. Finally, the innovativeness of open field farmers can be assessed with village by village surveys conducted around 1800. Systematic comparisons of cropping in open and enclosed villages was first undertaken by Yelling (1977) and, particularly, by Turner (1982, 1986). Studies of this sort require stratifying the data by soil type so that the profitability of the new method is the same for the open and enclosed villages being compared. My own research (Allen 1992) has concentrated on the south midlands—a region extending roughly from Oxford to Cambridge to Leicester. The region includes 2,850,866 acres of land and 1,568 villages. The south midlands comprises three natural environments—the pasture, light arable, and heavy arable districts. In each district there was an improvement or cluster of improvements that raised productivity in the eighteenth century. In the pasture district, conversion of most (sometimes all) of the land to grass was its most profitable use. Convertible husbandry was sometimes

46

ROBERT C. ALLEN

practiced here. In the light arable district, most land was kept as arable, and the soil was light enough so that the Norfolk rotation (turnips–barley–clover–wheat) could be profitably used. New Leicester sheep, an improved breed, were often introduced to eat the clover and turnips. In the heavy arable district, there was little change in cropping, but subsurface drains were introduced to deal with standing water. The comparisons show that Ernle clearly overstated his case, for there is considerable evidence that open field villages adopted the rent maximizing techniques. Nonetheless, the pace of innovation was slower than it was among the enclosed villages. In the pasture district, for instance, open villages allocated an unusually large fraction of land to grass, but the proportion was less than in enclosed villages, and there seems to have been no alteration of land between pasture and arable—the hallmark of convertible husbandry. In the light arable district, open villages certainly cultivated clover and turnips, but to a lesser extent than did enclosed villages. The latter were also more likely to have raised improved sheep. Finally, in the heavy arable district, enclosed villages were much more likely to have drained their land. So there is evidence that open field villages were more rigid than enclosed villages although not as hidebound as Ernle would have us believe. A more favorable assessment of open field villages is reached, however, by using other indicators of performance. Crop yields have received considerable attention. Table 2.1 compares yields in the three districts of the south midlands. In most cases the yields were highest in the enclosed villages, but the percentage gain was often small and erratic. Yields had approximately doubled between the middle ages and the nineteenth century, and that increase provides another yardstick for judging the gap between open and enclosed villages. The difference between open and enclosed yields was often small compared to the growth of yields that had occurred between the middle ages and enclosed yields c. 1800. This conclusion applies with great force in the case of wheat, for the difference between open and enclosed yields was small for this—the preeminent—crop. The difference between open and enclosed yields was somewhat larger in the case of oats and barley, especially in the heavy arable district. Overall, the yield comparisons lend support to the conclusion that enclosed villages were more efficient than open, but indicate that the advantage was small. Moreover, open field villages had accomplished most of the gain that had occurred between medieval levels and the standard of efficiency reached by enclosed farms c. 1800. These results carry over to comparisons of total factor productivity (TFP). These comparisons are often developed in the context of the rental value of the land. Enclosures were pushed forward by large scale land-owners with the aim of raising the rent they received from their tenants. McCloskey (1972) first pointed out that rent was an indicator of efficiency since an

47

COMMUNITY AND MARKET IN ENGLAND

Table 2.1 Crop yields and enclosure, c. 1800

Heavy arable district Wheat Barley Oats Beans Average Light arable district Wheat Barley Oats Beans Average Pasture district Wheat Barley Oats Beans Average

Open bushels/acre

Enclosed bushels/acre

Enclosed relative to open Enclosure gain relative to 19th century enclosed—medieval

19.7 26.5 23.5 18.8 21.2

20.2 31.8 33.0 22.2 24.1

2.2 20.0 40.4 18.1 14.7

5.3 35.3 44.6 27.9 23.8

20.0 27.0 26.5 19.9 23.4

19.7 29.3 32.5 18.1 24.7

−1.5 8.5 22.6 −9.0 5.6

**** 18.4 28.8 **** 10.9

20.9 28.0 36.9 22.4 24.7

21.9 32.2 38.1 23.4 26.7

4.8 15.0 3.3 4.5 8.1

8.9 27.3 4.5 7.5 14.2

Source: Allen (1992: 136).

enclosed farmer could afford to pay a higher rent only if he increased his output per acre or reduced his (nonland) costs per acre or some combination. Any such combination implies a rise of output with respect to total inputs, that is, higher total factor productivity. The relationship between Ricardian surplus per acre and total factor productivity can be shown easily for a Cobb–Douglas technology. Let the production function be

(1) where Q is output, K, L, and T are the quantities of capital, labour, and land, and A is the index of total factor productivity. The exponents are positive fractions and sum to one. In that case, the dual cost function is:

(2) where r, w, and s are the prices of capital, labor, and land, while P equals the cost (assumed to equal the price) of agricultural products. Solving equation 2 for A shows how efficiency can be inferred from land values:

(3)

48

ROBERT C. ALLEN

The variables in equation 3 can be interpreted as relative values in the instances being compared. The denominator involving P, r, and w is a price index that deflates changes in land values for inflation. If P, r, and w are constants, then the rise in total factor productivity can be calculated by raising the relative change in Ricardian surplus to a power equal to the exponent of land in the Cobb–Douglas production function. For instance, if the exponent of land was one-third and if rents rose eight-fold, then total factor productivity doubled (two is the cube root of eight). Allen (1982) generalizes this result to production functions besides the Cobb–Douglas. The most direct way to infer total factor productivity from land values is direct calculation of the Ricardian surplus, that is, the revenue minus the opportunity cost of the labor, capital, and materials applied to the land. Calculations for the south midlands show an erratic pattern. Enclosure in the heavy arable district came closest to the standard story, for enclosed farms realized revenues about one half pound per acre more than open farms with the same cost structure. As a result, the rental value of the land rose by the rise in the value of output. This corresponds to a 13 percent increase in TFP. In the light arable district, however, the gain in TFP was only 3 percent. In the pasture district, the results were mixed with some old enclosures, apparently on the most productive lands, showing significant increases in farm revenue per acre due to high stocking densities of beef cattle. These output increases translated into a 13 percent rise in TFP. In other districts, enclosure did not result in higher stocking densities, and TFP changes ranged from −4 to +0.2 percent. On average, enclosure gave only a small boost to TFP but in some districts the gains were significant (Allen 1992: 176–7). Calculations of Ricardian surplus have also been used to test the hypothesis that common pastures were overgrazed. If access were unlimited, then profit maximizing farmers would have increased herds until the cost of an extra animal equaled the average product of animals on the common. In that case, rent would have been zero. While this case is theoretically alluring (Hardin 1977), it has always been doubtful on historical grounds since many villages were aware of the common property problem and set stints limiting grazing. In addition, it has been possible to test the hypothesis econometrically with a sample of 231 farms described by Arthur Young in his tours of England c. 1770. Ricardian surplus was regressed on a variety of variables including the number of sheep grazed on commons. The coefficient of sheep was positive, statistically significant, and equal to the value of the annual ‘profit of a sheep’ reported in contemporary literature (Allen 1982). This statistical evidence, thus, runs counter to the overgrazing hypothesis. Direct calculations of Ricardian surplus are the exception rather than the rule. Rents actually paid are the usual measure of land value. These figures are readily available in estate records, and, indeed, McCloskey

COMMUNITY AND MARKET IN ENGLAND

49

originally proposed that productivity be inferred from rent in order to take advantage of this information. There are several limitations to this approach, however. First, poor rates and land taxes were also paid out of Ricardian surplus, and they should be added to rent to ascertain the surplus. Second, and probably more importantly, land markets were often out of equilibrium. This has been shown for two data sets in the late eighteenth and early nineteenth centuries (Allen 1982, 1992: 176–9). Such a disequilibrium could arise since mobility among tenants was low, rents were adjusted only at long intervals, and accounting was so primitive—certainly with respect to family labor and capital costs (livestock and implements) that bulked large in total costs. Even surveyors relied on rules of thumb and used prevailing rents, rather than Ricardian calculations, as the basis of valuation. In the long run, no doubt, rents tended to follow Ricardian valuations, but significant discrepancies could arise. Despite these caveats, many economic historians (myself included) have succumbed to the easy availability of rents and used them to infer productivity. Clark (1998b), for instance, has compiled about 20,000 quotations of rents received for lands owned by English charities. Many (on the order of 93 percent) of these reports applied to atypically small and highly valued properties, which are unrepresentative of the normal course of rents. Nonetheless, Clark reports only small rent gains at enclosure. Rents can also be used to measure productivity growth over time, although in this application they must be deflated by an index of farm input and output prices to purge them of inflation. Such calculations point to significant productivity gains in the open fields. Figure 2.1 plots real rents for open field villages in the south midlands from 1450 to 1800. These curves show considerable rent lag during periods of inflation (the price revolution of the sixteenth century and the late eighteenth and early nineteenth century). Nonetheless, the overall rise in rent suggests productivity growth in the open fields. Such a result is not surprising. Since there was significant productivity growth in English agriculture between the middle ages and the nineteenth century—a fact attested by the rise in crop yields and one supported by other evidence to be considered—a corollary of the finding that enclosure made only a small contribution to productivity growth is that productivity rose sharply in open field villages in the early modern period. This implication is borne out by Fig. 2.1. The evidence reviewed here runs strongly counter to the depiction of open fields in the standard story. Open field farmers achieved significant productivity growth in the early modern period, and they went part way in adopting the famous improvements of the eighteenth century. Nonetheless, enclosed farming usually had an advantage over open field farming whether the indicator be new methods, crop yields, or rent paid. There is thus some evidence in favor of the view that enclosure raised efficiency. The main point, however, is that the efficiency advantage of enclosed farming

50

ROBERT C. ALLEN

Fig. 2.1 Real Open Field Rent, South Midlands (Five Year Moving Average)

was small, and that open field farmers had achieved almost all of the productivity gains available between 1500 and 1800. Enclosure explains neither the productivity advantage that England enjoyed over other countries c. 1800, nor the rise in efficiency that had occurred since the middle ages.

2.2 England: Macro Evidence The focus of research in English agricultural history has shifted to macro evidence about output and productivity. This research provides further insight into the performance of open fields and enclosures by testing whether output increased during major enclosure waves as the proponents of enclosure claimed. Overton (1996a, b), who is the only economic historian in recent years to have reaffirmed the view that eighteenth-century enclosures were critical for agricultural modernization, rested his case mainly on these

COMMUNITY AND MARKET IN ENGLAND

51

reconstructions. However, his use of the evidence is idiosyncratic for reasons to be discussed. More plausible reconstructions of the aggregate series support the conclusions derived from the micro evidence. All of the reconstructions start with the identity that agricultural output (Q) equals the population (N) multiplied by per capita consumption of agricultural products (c) and by the ratio of agricultural production to agricultural consumption (t):

(4) The factor t incorporates international trade in agricultural goods into the analysis. Before the nineteenth century, it played only a small role in most cases, but it is included for completeness. Overton used equation 4 to reconstruct aggregate farm output by making the additional assumption that per capita consumption of agricultural products (c) was constant. Since t was always close to one before 1800, this procedure effectively interpolated agricultural output from the population. The English population expanded from 1500 to 1650, was then static to 1750, and after that began to grow rapidly in the world's first population explosion. Assuming that farm output followed the same pattern implies a 115 percent increase in output between 1500 and 1650, a 13 percent rise from 1650 to 1740, and a 39 percent surge in ouput from 1740 to 1800 at the time of the parliamentary enclosure movement. From this, Overton concluded that there was no agricultural revolution in the open fields between 1650 and 1750, and that the parliamentary enclosures were critical to agricultural productivity growth. While the assumption of constant per capita food consumption was once commonly made, it has been rejected by most investigators for over twenty years. Crafts (1976) first pointed out that consumption depended on income and prices, and he and subsequent investigators (Jackson 1985; Allen 1999) have used that fact to estimate agricultural output by specifying a demand curve for farm goods. The simplest formulation is to specify per capita consumption of agricultural goods (c) as:

(5) where p is its price, i is income per head, m is the price of other consumer goods, and a sets the units of measurement.86 Calculation of agricultural

86

The calculations are explained in Allen (1998b), and the data sources are described in detail in Allen (1998a). Briefly, p is a retail price index of foodstuffs, i is the wage of building craftsmen, and m is a retail price index of nonfood goods. All of these variables are expressed in real terms by deflating them by an international, intertemporal consumer price index made up of the food and nonfood prices used in p and m.

52

ROBERT C. ALLEN

consumption requires values for e, g, and b. These are taken from analyses of English budget studies and from demand studies for developing countries. The usual assumptions in the historical literature (Crafts 1976; Jackson 1985; Clark et al.1995; Allen 1999) are that the income elasticity of demand was 0.5, the own price elasticity was [0.6, and the cross price elasticity between food and manufactures was 0.1. (Other reasonable values give similar results.) Substituting equation 5 into equation 4 gives the following equation for estimating the growth in agricultural output:

(6) Making demand a function of income and prices implies a very different pattern of output growth from extrapolations from the population. Far from being a period of stagnation, the years between 1650 and 1740 stand out has a period of significant advance when output increased by 26 percent. This finding supports the view that output grew rapidly before the parliamentary enclosures when small scale family farming (the yeoman mode of production) was predominant in the open fields. Furthermore, the period from 1740 to 1800, which includes major waves of parliamentary enclosures, was characterized by a negligible increase (only 3 percent) in farm production. While the population was increasing, which tended to raise demand, the prices of agricultural goods were rising rapidly in the last half of the eighteenth century, and that inflation indicates that supply was not keeping pace with demand. The demand curve simulations do indicate a resumption of agricultural output growth after 1800, but the main period of enclosures is remarkable for its stagnation. The aggregate evidence, thus, points to substantial productivity and output growth in the seventeenth and early eighteenth centuries. Growth slackened in the second half of the eighteenth century. The progressive period is the same as that in which the open field, yeomen farmers were ascendant, while the period of stagnation coincided with the parliamentary enclosures and the elimination of peasant proprietorship. Micro and macro evidence point to similar stories of the agricultural revolution.

2.3 Agricultural Labor Productivity: A European-Wide Perspective Further insight into the causes of agricultural productivity growth and of the role of enclosure can be gained by extending the macro evidence to a European-wide perspective and by combining the output estimates with estimates of agricultural employment to gauge labor productivity.

COMMUNITY AND MARKET IN ENGLAND

53

The point of departure is Wrigley's (1985) well-known paper on economic structure and labor productivity. He presumed that the population could be split into agricultural and nonagricultural components. Dividing agricultural ouput (as defined by equation 4) by the agricultural population (A) gives an expression for agricultural labor productivity:

(7) Like Overton, Wrigley assumed that c was a constant and t was one (in most cases). With those assumptions, output per worker is inversely proportional to the share of the population in agriculture (A/N), and the problem of measuring agricultural productivity is reduced to demography. N and A are required to implement Wrigley's procedure. Historical demographers have estimated the total populations of European countries back to the middle ages. Estimates have also been made of the urban population. The difference is, of course, the rural population. Wrigley split it into agricultural and nonagricultural components as follows: the division in 1800 was obtained from nineteenth-century censuses. Values ranged from about 0.5 to 0.8. Wrigley guessed that 80 percent of the rural population was agricultural in 1500—a plausible value in view of Dutch evidence (de Vries and van der Woude 1977: 233). The increase in the rural, nonagricultural share from 20 percent in 1500 to as much as 50 percent in 1800 corresponds to the ‘proto-industrial’ revolution that occurred in that period. The nonagricultural fraction of the rural population was interpolated for intervening dates. Wrigley undertook these calculations for England and France from 1500 to 1800. Output per worker rose in both countries, but the pace of advance was higher in England. The results provide support for the traditional view that enclosure raised productivity since they show England ahead of France and since ‘labor productivity’ measured in this way rose sharply between 1750 and 1800, i.e. during the intense phase of parliamentary enclosing. But that interpretation cannot be sustained for the same reason that Overton's reconstructions of English farm output were abandoned: Wrigley's calculations do not provide an accurate depiction of labor productivity since per capita consumption depended on income and prices. The solution is to compute per capita consumption c with equation 5. In that case, output per worker becomes:

(8) Output per worker depends on the fraction of the population in agriculture A/N, the trade balance t, and per capita consumption, which is a function of income and prices.

54

ROBERT C. ALLEN

Fig. 2.2. Output Per Worker In Agriculture, 1300–1800

Figure 2.2 plots the results for England, Belgium, Italy, Spain, the Netherlands, and Austria/Hungary, which cover the main possibilities. Most of the European continent—Spain, Italy, France, Germany, Austria/ Hungary, Poland—shared a common agricultural productivity history, as exemplified by the curves for Spain, Italy, and Austria/Hungary in Fig. 2.2. Italy, whose data run back to 1300, shows a rise in labor productivity from 1300 to 1400 due to the population losses associated with the Black Death. Between 1400 and 1800, population grew steadily, and output per worker in agriculture declined. There is no evidence of an agricultural revolution on the continent. The history of labor productivity in present day Belgium was quite different. Output per worker was very high in the middle ages—Flanders was renowned for its agriculture—but did not increase thereafter. However, it set a standard showing what might be achieved. The Netherlands and England had agricultural revolutions in the early modern period. In 1500, output per worker in Dutch agriculture was similar to that on most of the continent, but productivity rose sharply to 1700

COMMUNITY AND MARKET IN ENGLAND

55

when the Belgian level was surpassed. This rise was achieved despite a rapidly growing population. From 1300 to 1600, the productivity history of England was like that of most of the continent. Output per worker rose after the Black Death and remained high to 1500 due to the continued stagnation of the English population. Demographic growth accelerated in the sixteenth century and labor productivity fell due to diminishing returns. This typical continental pattern was broken after 1600 when labor productivity surged upward reaching the highest level in Europe in 1750. The patterns in Fig. 2.2 have two important implications for agrarian history. First, the rise in English labor productivity was confined to the period 1600–1750, that is the period when open field, yeoman agriculture was ascendant. The last half of the eighteenth century again stands out as a period of stagnation despite the enclosure movement. Second, the figure requires us to reformulate the search for the causes of agricultural productivity growth. Three countries were highly productive—England, Belgium, and the Netherlands. What did they have in common that distinguished them from the less successful countries? They certainly did not share enclosures—the notion that these institutions were the recipe for rapidly rising agricultural productivity dies a fast death when subjected to the test of comparative history. It is more accurate to say that they shared systems of owner-occupying family farms, for they were very common in England between 1600 and 1750 and were widespread in the Low Countries. Open fields were also common there. Indeed, their agricultures have been criticized for the prevalence of just such institutions. However, Fig. 2.2 shows that they were the bases of progress in the early modern period—not poverty. While open fields and family farms were a good base for agricultural advance, they are not a complete explanation for these institutions were found elsewhere as well. However, there was one thing that farmers in England and the Low Countries had in common and which distinguished them from farmers elsewhere in Europe, and that was rising urbanization. In medieval Belgium where high levels of productivity were first achieved, 39 per cent of the population was urban. From 1500 to 1750, the share of the population living in cities in the Netherlands rose from 30 percent to 36 percent, and, in England, the share rose from 7 percent to 23 percent over the same period. These percentages are higher than elsewhere or rising more (Bairoch et al.1988). Urbanization affected agriculture through the demand for farm products—a theme running back to von Thuenen—and it also affected agriculture through the labor market. Early modern cities had extremely high mortality rates and could not exist, let along expand, without high rates of rural–urban migration. Wrigley estimated that half of the natural increase of the English population was required by the growth of

56

ROBERT C. ALLEN

London, and the same was true of Dutch cities (Wrigley 1967: van Zanden 1993). Wages were very high in the cities of northwestern Europe (Allen 1998a). Agriculture responded to the high urban demand for labor in ways that raised agricultural labor productivity. In England, where the property law favored the landlords, agriculture was eventually reorganized by enclosure and large farms—changes which generally reduced employment per acre. Throughout the region, peasants responded to the high demand for labor by increasing their output since that raised their incomes without reducing employment.

2.4 A Case Study of Decision Making in the Open Fields The conclusion that open fields were highly innovative makes sense on one level but not on another. The finding makes sense in that early modern England was a poor, backward economy, and so it would be a surprise if its farmers did not want a higher standard of living. From that perspective, it is not surprising that their basic institutions supported productivity growth. On the other hand, the progressiveness of the open fields is hard to square with the ‘inflexibility’ usually imputed to them in the historical literature. The place to begin is the presumption that unanimity—or, at least, widespread agreement—was required for most changes in cropping to occur, so a few hidebound traditionalists could block progress. Testing this view requires looking inside the open fields to see how decisions were made and, in particular, whether unanimity was required and how it was achieved. This can only be done with a case study, and I review evidence for the parish of Spelsbury in Oxfordshire. For this parish, one can trace the adoption of new crops in the eighteenth century. In one case, unanimity was required at the outset, and that requirement raised significant problems that were only overcome by coercion on the part of the manorial lord. In other important cases, the need for unanimity was avoided, and the changes in field management were structured so that farmers had a individual choice as to whether to take up the new crops. After the initial phase of voluntarism, the extension of the cultivation of the new crops was done on the basis of majority rule; unanimity was never required. There were, thus, at least two approaches to organizing innovation in the open fields. Flexibility was a desirable strategy from an economic, as well as a social, point of view. Technical progress involved experimentation. There was a risk that new crops might fail, and, even if they were suitable, their culture had to be adapted to local conditions. Demonstration or experimental farms are a twentieth-century solution to these problems, but there were no state agricultural services before the industrial revolution. The home

COMMUNITY AND MARKET IN ENGLAND

57

farms of great estates rarely served this function. The task was left to the peasants. In the open fields, experimentation was accomplished by letting people do what they wanted. In that case, enterprising individuals could try out new methods and adapt them to local conditions. Their more risk averse neighbors could follow suit when they were convinced the new techniques paid off. This view of open fields is, of course, the antithesis of the standard straight jacket model, which presumed that entrepreneurial farmers could try new crops only after enclosure. The records of Spelsbury give some insight into the way experimentation and innovation were carried out in an open field community. The adoption of sainfoin early in the eighteenth century was one of the first examples of the innovativeness of open field farmers adduced by historians (Havinden 1961). I will review those agreements, and then carry the story forward to the enclosure of the village in the early nineteenth century. Spelsbury was an unusually large parish (3,900 acres according to the 1851 census) containing several settlements with their own field systems. I am concerned here with the townships of Taston, Fulwell, and Spelsbury proper. Their fields amounted to 2,016 acres divided into 53 yardlands of 38 acres each. A yardland was a standard early modern farm. It could have been cultivated by a family without much hired labor. In the early eighteenth century, most holdings were held as copyholds. Fifteen holdings were one yardland, and twelve amounted to half a yardland in size. These were yeomen farms. In addition, there were a few two yardland holdings, which required considerable hired labor, and a single four yardland (152 acres) holding, which was a fully fledged capitalist farm. The demesne of eight yardlands (304 acres) was also operated as a single farm. With the exception of the four and eight yardland holdings, family farming was the predominant mode of production in Spelsbury.

2.5 The Adoption of Sainfoin Three new crops—sainfoin, turnips, and clover—became available to Spelsbury farmers in the eighteenth century. They were extensively cultivated in nineteenth-century enclosed farms, and open field farmers adopted all of them in the eighteenth century. Sainfoin was grown in meadows for hay. It was the first to be adopted. Special agreements were made to reorganize the land for sainfoin cultivation. The first agreement was made in Taston in 1701, and it created a special sainfoin meadow. The agreement provided for withdrawing several furlongs from the fields of Taston and fencing them with walls or hedges and ditches. Sainfoin was to be planted and mowed for hay and then grazed. The agreement calls this ‘inclosing’ and it was in that the land was

58

ROBERT C. ALLEN

withdrawn from the common fields, fenced, and managed separately. But many features of the open fields remained: landholdings were not consolidated; each copyholder retained his original, scattered strips within the enclosure. While each copyholder now mowed his own strips for hay, the aftermath was grazed with a common herd. Each proprietor was required to maintain the perimeter fence where it crossed his strips. Three fieldmen were chosen on the first of each year to establish the dates when sainfoin would be planted, when animals could graze, and to enforce the maintenance provisions. Violations of the agreement (like withdrawing land or allowing stock to graze at unauthorized times) were punishable with fines paid to the lord of the manor. The Taston sainfoin agreement created an improved meadow that was operated in the traditional, open field manner. The Taston sainfoin agreement was signed by all of the copyholders in Taston.87 It was thus the result of a unanimous decision. The ‘inflexibility’ of communal decision making could be transcended. Achieving unanimity was not effortless, however. William Canning, the steward of the estate, regularly wrote to the Earl of Litchfield about estate matters. On 3 April, 1703, Canning recounted that ‘at the Court I found a great disturbance a Mungst them of Tastone & Fullwell about Settling their Methods of Managing the sainfoine Grass that they had Sowed.’ The management of the sainfoin enclosure had shifted to the manorial court—this was not a surprise—for the court was the arena in which rules for the fields were normally promulgated, field men appointed, and fines assessed. While Canning believed that sainfoin cultivation ‘is like to be a great Improvement if it be inCouraged, & carryed on as it should be, But I found so much Crossness … a Mongst the Mainuagers of it; that if they be not overawed, the designe will soon be destroyed.’ Canning does not tell us what they were arguing over. The dispute may have centered on the schedule of sowing, mowing, and grazing, or it may have been more fundamental like removing land from the enclosure, something which was explicitly prohibited in the agreement. In any event, dissent was confined to two copyholders. ‘Therefore, I took upon me to force Wilt. Rooke & John Hull of Tastone to a better Comployance, and the way that I took was the next day after the Court, I took John Freeman [another copyholder in Taston] with me & went to them both.’ Canning ‘told them that if they wold not Imedially Comploy to the Well ordering of all those Matters we had on foote, I wold forthwithe Report to ye Losp, their Ill Manners & misbehaviour at the Court Thursday that their should be such a course taken with them as to have them severely

87

This conclusion is based on a comparison of the Spelsbury quit rental for 1703 (DIL II/b/33) and the Taston Inclosure Agreement (DIL II/n/1). All documents referenced in this paper are deposited in the Oxfordshire County Record Office.

COMMUNITY AND MARKET IN ENGLAND

59

punished.’ In addition, Canning threatened to harass them until they complied. He proposed to punish them ‘for leting their Houses run to Ruin.’ In addition, ‘for every little offense which they should commit against ye well ordering of the Sainfoine I would have a writ on their Backs, and as for Hull if he did not Imediably pay me 20s for the Tree he cut down I wold serve him with a Writ the next day.’ These threats worked. ‘So then they bothe agreed with me that they Wold comply to any orders I should make.’88 Evidently unanimity could be obtained by manorial coercion, but that was not the only—or, indeed, most desirable—way to effect change in the fields. The alternative was to restructure the change so that it could proceed voluntarily. This was done in a 1708 agreement to enclose land in Spelsbury. The purpose of the agreement was to make ‘a certain quantity of land Every years Land,’ that is, land that was continuously cropped. As in Taston, communal grazing was practiced after the harvest, fieldmen were chosen to regulate the grazing, and fines were assessed for violations of those rules. The main difference with the Taston agreement was that each copyholder could use his Every Year's Land as he liked—‘to soo Corne or Sainfine or What they please.’89 The only sensible use of this enclosure was to cultivate sainfoin as in Taston, but that was not required of everyone at the outset. It is likely that it became the usual practice, however. Court records later in the eighteenth century refer to the ‘old sainfoin’ field90 as does the field map prepared at enclosure.91 By providing flexibility at the outset, the kind of disputes that occurred at Taston were avoided, the entrepreneurial copyholders could proceed with experiments, and the others could adopt the sainfoin culture after it was proved in their village.

2.6 The Adoption of Turnips and Clover Sainfoin was a major improvement, but clover and turnips were the most famous new crops of the eighteenth century. Their cultivation in England began in East Anglia in the seventeenth century. At first, their culture was not linked; it was only in the eighteenth century that they were combined in rotations like the famous Norfolk rotation (turnips–barley–clover– wheat). In the eighteenth century, the cultivation of clover and turnips spread north and west. The earliest reference to turnip cultivation in Oxfordshire was in 1727 (Allen 1992: 111). How receptive were Spelsbury farmers to these most famous of crops? The first reference to them was in the manorial court rolls for 1751. The management of the open fields was carried out in the manorial court, and

88

DIL I/k/1h.

89

DIL II/n/26.

90

E.g. DIL II/w/134.

91

Misc. Sta. I/1.

60

ROBERT C. ALLEN

the court rolls, the minutes of the proceedings, frequently summarized orders regulating the fields. The Spelsbury orders for 1751 state: First, we do order and agree that the Over Furlong Shooting into Chipping Norton Road in a field called Sinquefoil field in Spelsbury be sowd with turnips this next Season, we agree to abate our sheep common for every Bushel of Barley Seed that the Ground shall take for Sowing thereof, any neighbour that shall agree to Sow Turnips in any part of the Same Field shall have the same abatement, that every person shall make his own mind.92 This order is significant in several respects. First, the principle of voluntarism was respected: each person could decide for himself whether to grow turnips. Second, the turnips were not introduced at first in the fields but into the sainfoin enclosure, the original Every Year's Land where everyone could choose his crops. Third, the abatement of common rights was normal: If people were not growing sainfoin on the every year's land they were not contributing aftermath to feed the common herd, and so they had to reduce the number of animals they could put in the herd. (The measurement of common rights in terms of the barley seed that could be sown on the land in question was the standard formula in Spelsbury.) By the late 1750s, these procedures were modified. While the cultivation of turnips was still confined to the ‘sainfoin enclosure,’ cultivation became obligatory. Unanimity was not required, but the majority ruled. Thus, the 1758 field orders state that ‘We do order and agree to Sow White Turnips the next Season in such part of Spelsbury Field called Old Saint Foyne Field as the Landholders or Major part of them shall by next May Day Agree.’93 By 1762, a portentous change had occurred. The cultivation of turnips was shifted out of the sainfoin enclosure into the open fields themselves: We order & agree that the Barley quarter shall be sowed with Turnips and that each person shall abate three sheep for a quarter of Barley seed. Also it is ordered and agreed that if any person shall let his lambs go loose out of the fold and thereby trespass upon his Neighbours Turnips he shall forfeit ten shillings for each offense.94 The orders for 1763 were more explicit in indicating where the turnips were to be planted: Also We Order and Agree to Sow Briar Furlong Turnips and to hain [fence] them so soon as the Turnips are up And that winter beer furlong and Butt furlong shall be sowed with Turnips and hained as the others, to be sowed as low down as the owners of the Land shall think proper, to be Mounded by the Outside Proprietors and for each breach of this Order the person offending shall pay to the Lord the Sum of five shillings.95

92

DIL II/w/108.

93

DIL II/w/108.

94

DIL II/w/134.

95

Ms. summary of Spelsbury court rolls.

COMMUNITY AND MARKET IN ENGLAND

61

1762 was also the first year in which clover was mentioned in the court records. Their cultivation began in the fields: Also We agree to plant almshouse Furlong and Winter Bere with Clover and that the same shall be hained from Christmas Day to tenth of April under the penalty of ten shillings for each person who shall turn in any Cattle between Christmas and the said Tenth of April.96 Subsequent courts record a slow shifting of turnip and clover cultivation around the fields. In 1765, for instance, winter beer furlong was planted with turnips, and briar furlong with grass seed.97 This was a reversal of the practice in earlier years. Gradually, the cultivation was spread to other furlongs. By the 1780s, the cultivation of clover and turnips had been standardized: At this court it is agreed to sow Clover on Costar Hill and Dean Field side and that the same shall be hained at Michaelmas and broke the twelfth of July with Sheep and the Sheep to be stocked as usual And to sow Turnips from Jack's Brake to Slate Pits and to hain the same as soon as they are up & that they shall be mounded by the Occupiers. And that the Clover shall lie two years at Coom Road to be hained the first of January and broke the twenty sixth of April.98 While land was planted with turnips for only one year before it was shifted to another crop, the land was planted with clover for two years before being rotated. It might also be noted that the language had shifted subtly from that of the 1760s. Twenty years before, it was the owners (i.e. copyholders) who were required to mound the perimeter of the turnip field, but by the 1780s, it was the occupiers. This change in language reflects the replacement of copyholders with tenant farmers, a change to be analyzed shortly. Separate field orders were recorded for Spelsbury and for Taston, but agricultural practices developed in similar ways in both hamlets. In 1785, the court minutes record: Also it is ordered and agreed to sow Clover on West Field from caught-house to Banbury Gap & to sow Clover upon small Narnills for a Sheep Common to lye with Sharigo Hills to be paid by the yard Land. To sow Turnips upon Enwellin Quarter up to Green Ball to be hained as soon as they are up. The Clover to be hained at Michaelmas and broke the Twelfth of July. The Sheep commons to go as usual. All the moundng to be done by the yard Land. The drift Road for the Sheep to go by Shilcott Wood. Sharigo Hill & small Narnills to be hained the first of November and broke at old May Day. Also that no person shall turn their plows on the Sain Foin that shall be sown this year, next Spring under the penalty of Ten Shillings to the Lords for each year.99

96

DIL II/w/134.

97

DIL II/w/18.

98

Ms. summary of Spelsbury court rolls.

99

Ms. summary of Spelsbury court rolls.

62

ROBERT C. ALLEN

The reference in the last sentence to sowing sainfoin is significant in two respects. Organizationally, it represents a transfer of jurisdiction over the sainfoin to the manorial court. The first sainfoin enclosures were created with formal contracts independently of the court. By the middle of the eighteenth century, sainfoin cultivation was sufficiently routine that the enclosures were managed by the manorial courts along with the ordinary business of the fields. Agriculturally, the reference to sowing sainfoin is important since it suggests that the farmers of Taston were practicing a form of convertible husbandry in which land was alternated between sainfoin meadows and arable planted with corn, clover, and turnips. For instance, the court rolls for 1766 recorded an extension of the sainfoin field. Ordered … that the Field of Taston from the old St. Fine wall down to Guys Close and as far as the Landholders can agree shall be planted with St. Fine in the Spring of 1767 and mounded by the Michmas following by the Yard Land to continue for Twelve Years to be hained at Christmas and broke at Michmas but no sheep to go thereon at any time. And all Folds to be off on the one Cropp Land by Lady day and to have the Folds to be staked from Ladyday till Harvest is in under the penalty of Ten Shillings to the Lord for each offence.100 The minutes for 1788 prescribe the reverse sequence: ‘Also it is agreed and ordered that Turnips shall be sowed on part of the Old Saint Foin down Deadman Hill as far as each tenant thinks proper and that the same shall be hained as soon as they are up.’101 Converting field land to sainfoin enclosures for a dozen years, and then converting the land back to cultivation of turnips, clover, and corn amounts to convertible husbandry—hardly the practice that Lord Ernle imagined open field farmers to have been capable of. We have seen how open field farmers took their first steps in adopting sainfoin, clover, and turnips. There was an evolution in practice across the century as farmers learned the best way to grow and integrate these crops. This is not surprising. In 1700, no one knew the optimal system; it was developed by trial and error everywhere. Spelsbury was no exception. What is perhaps more surprising is that the open fields were a suitable environment for this evolution, for two reasons. First, furlongs rather than fields were the fundamental operating units, so land could be shifted to new or experimental uses in small quantities. Second, not everyone in each furlong had to do the same thing. The first Spelsbury sainfoin enclosure was set up so that each person could grow what he wanted. The aim of the enclosure was certainly to grow sainfoin, and ultimately it did, but uniformity waited until the gains became obvious to everyone. (Failure to follow this procedure resulted in disputes like those in Taston. While manorial

100

Ms. summary of Spelsbury court rolls.

101

Ms. summary of Spelsbury court rolls.

COMMUNITY AND MARKET IN ENGLAND

63

authority could force a minority to comply with a majority, voluntary procedures could achieve similar ends without coercion.) The voluntary principle was also applied again when turnip cultivation was tried. ‘Every person shall make his own mind.’ By letting those eager to try the new crops take the first steps, small scale experiments were undertaken to establish whether and how the new crop should be cultivated. Other farmers soon followed. Eventually, majority rule replaced individual decision making. Even then, however, the open fields catered to many tastes. By the end of the eighteenth century, the majority had not forced all the land to a four course Norfolk rotation. Instead, a complicated system evolved incorporating the new crops but also old practices like the use of the fallow. The late eighteenthcentury survey data discussed earlier in this paper shows that to have been the typical pattern in the open fields. The flexibility of the open fields, which had been a strength at the outset when it allowed enterprising individuals to try out new crops, became a weakness later since it continued to make way for the least enterprising.

2.7 Enclosure Reconsidered If open fields were so good, why were they enclosed? There are three explanations in the literature. First, enclosure accelerated technical progress. We have seen that there is some truth to this and, indeed, have just offered an explanation for why the open fields lagged behind enclosures in the use of the newest crops at the end of the eighteenth century. But evidence discussed earlier shows that the gains from these changes were small, and the increases in land value were frequently not large enough to justify the expense of enclosing. Second, enclosure redistributed income. Enclosure extinguished many insecurely established customary rights. It also gave the landlords an opportunity to raise rents. Redistribution may have been a more important factor in enclosure than one might expect due to the primitive nature of cost accounting and agricultural knowledge. While today's economists can computerize eighteenth-century data and compute by how much enclosure on a particular type of land should have raised its value, such calculations were impossible for people in the eighteenth century. Even in the nineteenth century, land surveyors paid lip service to Ricardo, but valued land by rules of thumb because they could not put Ricardo's concept of surplus into practice (Allen 1992: 171–87). All landlords knew was that they could usually raise rents after enclosure, and that is what improvement meant to them. Whether the rent gains came from higher efficiency or income redistribution was unclear. The concept of social capital (Coleman 1990; Putnam 1993, 1996) provides another approach to the incentive to enclose and relates enclosure to changes in the social relations of production in agriculture. Much of

64

ROBERT C. ALLEN

England's open field farm land was in the hands of small-scale farmers in the eighteenth century, who were quasi proprietors and held their land on copyholds or beneficial leases. In Spelsbury, as we noted earlier, most of the land at the beginning of the eighteenth century was divided into farms of one half or one yardland (19 or 38 acres). Three holdings were leases for lives, but the great majority were copyholds. The latter were held ‘according to custome of one life & a widdow's estate in possession & one in reversion’.102 At any time, there were at most three people—specifically named—with a right to the copyhold—a man, his wife, and a third person, usually a child. If nothing was done, the copyhold would revert to the lord of the manor at the death of the three named people. Before the eighteenth century, it was usual for parents to extend the agreement to include a child by ‘buying a reversion.’ According to a document dated 1705, a reversion cost five times the annual commercial value of the estate. A yardland was valued at five pounds per year in 1705, so the value of the reversion was 25 pounds.103 The lord of manor was under no legal obligation to sell reversions, and he stopped doing so in the first quarter of the eighteenth century. The change in policy took many decades to achieve its full effect because some copyholds had been recently renewed and the heirs lived long lives, while others were coming up for renewal and were held by old people. By the end of the eighteenth century, only three copyholds were left. The history of Spelsbury was typical of many villages in England. The concentration of landownership in great estates obviated the advantages of open fields. They had been an effective means of diversification for small farmers, but scattering land over the village was no attraction to the lord who now owned the whole village. The creation of great estates presented new management problems, and open fields frustrated their solution. Before the eighteenth century, manorial lords were not directly concerned with motivating farmers. Land was let as copyholds, and the lords realized income from them when entry fines or other levies were collected, but the management of the land, including any subtenants, was the responsibility of the copyholders. Even demesnes had usually been let as beneficial leases, so, again, the lords collected a fine occasionally but otherwise land management was limited. In both cases, the tenants farmed with fixed and low rents for long periods of time, and consequently received the full benefit of any improvements as higher income. Tenancy terms encouraged innovation but made it difficult for landlords to capture the gains. When copyholds and beneficial leases were run out, landlords had to deal directly with farmers. They had the chance to appropriate increases in land values due to rising productivity or prices, but they had to do this without destroying the farmer's incentive to improve: If rents were raised

102

DIL II/b/32.

103

DIL II/a/4e.

COMMUNITY AND MARKET IN ENGLAND

65

whenever productivity increased—i.e. if English landlords followed the (perhaps mythical) Irish model of ‘predatory landlordism’ (Mokyr 1983: 81–111)—then the incentive to improve would have been destroyed. On the other hand, leaving the tenants with fixed rents indefinitely would provide a great incentive to raise productivity but the gains would stay with the tenants. Share cropping was one possible solution, but it was not utilized in eighteenth-century England. Tithes (a 10 percent tax on grain output) were regularly denounced as a tax on improvement, and the much larger percentage of output that would have been paid as rent would have been a considerably greater disincentive to increase efficiency. Furthermore, improvement often meant the conversion of arable to grass, and that, in turn, meant that output fell with labor costs dropping more. If the landlord's income was a fixed share of output, then it would have fallen even though the commercial value of the land was rising. Share cropping did not commend itself as a simple way of managing landed property. Long-term, fixed rent contracts were another solution, and, indeed, twenty-one year leases were frequently advocated. They were not frequently used, however, because the unstable price level between 1750 and 1850 meant that a reasonable rent could not be predicted in advance. Instead, the English gentry and aristocracy developed a management system consisting of several elements. Legally, tenancies were short term; indeed, most farmers were merely tenants at will who held their land for a year beginning on Michaelmas (the 29th of September), and rents were officially determined at the outset of that year. In practice, however, these agreements were renewed repeatedly, and farmers occupied the same farms for many years. Rents were raised infrequently, so farmers had a financial incentive to raise productivity since they realized the gains. Rents might not be adjusted for decades, although the period could be shortened if inflation was rapid. Rent adjustment was done in a bureaucratic manner, i.e. by hiring a surveyor to revalue the estate, so that it appeared objective and fair. Fluctuations in farm prices and yields were accommodated by allowing tenants to postpone rent payments (run up arrears) when farm incomes were low. The object of this system was to create incentives for farmers to raise productivity while guaranteeing that eventually productivity gains were passed to the landlords in the form of higher rents. While the rental agreements were officially annual, the tenants had to regard them as long term if they were to have an incentive to raise productivity. Tenant confidence and loyalty were fostered by the infrequent and bureaucratic method of raising rents, by renewing tenancies every year, and by allowing the tenants to fall behind in rent payments. Landlords gained when tenants felt a long run attachment to the estate and believed they would be fairly treated by it. These tenancy terms and associated norms of behavior and

66

ROBERT C. ALLEN

expectations of proper treatment were forms of social capital. When this capital was in place, productivity rose. The open fields had also had social capital, but it was directed antithetically. The open field village fostered solidarity amongst the tenants. The communal management of open field land required a forum—usually the manorial court—in which the tenants could negotiate their economic affairs together. Courts were also responsible for other aspects of local government (McIntosh 1999). These continued associations and the economic benefits that derived from them encouraged solidarity among the villagers. Moreover, this solidarity was often opposed to the lord. Many times in the preceeding millennium peasant communities had had to act in concert to protect their interests vis-à-vis the lord. This too reinforced solidarity, which may have facilitated the management of the fields. Open field farming strengthened bonds between peasants and made them wary of lords. While the social capital of the open field village raised productivity in that context, it was counterproductive when the yeomen were eliminated and replaced by tenants at will. The self-governing institutions of the open fields were now manned by tenants: participation in these institutions encouraged solidarity among the tenant farmers vis-à-vis the lords. In particular, each tenant could plead that he could not improve his cropping because the village majority would not accommodate it. More generally, tenant self-government strengthened bonds among tenants whereas the new system of short-term leases put a premium on the tenants identifying with the great estate. The new system of short-term leasing worked best if the social capital of the open field village were replaced by a new kind of social capital. Once landlords had run out copyholds and beneficial leases, the days of the open field village were numbered.

2.8 Conclusion Nineteenth-century interpretations opposed ‘community’ to ‘market.’ Economic relations were noncommercial and nonprogressive in the ‘community,’ and the reverse with the ‘market.’ The research of agricultural historians in the last fifty years has demolished this paradigm. The open field community has been shown to have been innovative, effective, and progressive. An important reason that the open fields were progressive was because communal field management was far more flexible and open to experimentation than was imagined by proponents of the community–market dichotomy. The biggest puzzle about open fields is not why they persisted but why they disappeared. A variety of explanations were reviewed, and we have also suggested that another factor that was at play. In the eighteenth

COMMUNITY AND MARKET IN ENGLAND

67

century in many villages, the yeomen lost their land as their lords refused to renew their copyholds and beneficial leases. The land was then let on short-term agreements to capitalist farmers. The new system of leasing required a new psychological commitment. The open fields had their characteristic form of social capital in which an assembly of villagers governed the fields and opposed lordly claims. These associations fostered solidarity among the yeomen in opposition to the lord. In contrast, the great estates that emerged in the eighteenth century required that each farmer identify his fortunes with the estate rather than with his fellow villagers. Open field social capital was obsolete and had to be replaced with social capital appropriate to the great estate. Enclosure was the result. It was not, therefore, that enclosure overthrew community and created market capitalism. Rather, the reverse was the case: market capitalism undermined the open field community from within and precipitated enclosure.

References Allen, R. C. (1982) ‘The Efficiency and Distributional Consequences of Eighteenth Century Enclosures,’ Economic Journal, 92: 937–53. ––—(1988) ‘Inferring Yields from Probate Inventories,’ Journal of Economic History, 38: 117–25. —— (1992) Enclosure and the Yeoman, Oxford, Clarendon Press. —— (1994) ‘Agriculture during the Industrial Revolution,’ in Floud, R. and McCloskey, D. (eds.) The Economic History of Britain since 1700, Cambridge, Cambridge University Press, pp. 96–122. —— (1998a) ‘The Great Divergence: Wages and Prices in Europe from the Middle Ages to the First World War,’ U.B. C. Department of Economics Discussion Paper No. 98–12. —— (1998b) ‘Agricultural Output and Productivity in Europe, 1300–1800,’ U.B.C. Department of Economics Discussion Paper No. 98–14. —— (1999) ‘Tracking the Agricultural Revolution,’ Economic History Review, 2nd series. Bairoch, P., Batou, J., Chêvre, P. (1988) La population des villes europeennes: banque de donnees et analyse sommaire des resultats, 800–1850, Geneva, Librairie Droz. Britnell, Richard H. and Campbell, Bruce M. S. (eds.) (1994) A Commercializing Economy: England 1086 to c. 1300, Manchester, Manchester University Press. Clark, G. (1998a) ‘Commons Sense: Common Property Rights, Efficiency, and Institutional Change,’ Journal of Economic History, 58: 73–102. —— (1998b) ‘Renting the Revolution,’ Journal of Economic History, 58: 206–10. —— Huberman, M. and Lindert, P. H. (1995) ‘A British Food Puzzle, 1770–1850,’ Economic History Review, 2nd series, 48: 215–37. Coleman, James S. (1990) Foundations of Social Theory, Cambridge, MA, Harvard University Press.

68

ROBERT C. ALLEN

Crafts, N. F. R. (1976) ‘English Economic Growth in the Eighteenth Century: A Re-Examination of Deane and Cole's Estimates,’ Economic History Review, 2nd series, pp. 226–35. de Vries, Jan, and van der Woude, Ad (1977) The First Modern Economy: Success, Failure, and Perseverance of the Dutch Economy, 1500–1815, Cambridge, Cambridge University Press. Ernle, Lord (1912) English Farming: Past and Present, London, Heinemann Educational Books Ltd and Frank Cass and Company Ltd, 1961. Geertz, Clifford (1963) Agricultural Involution: The Processes of Ecological Change in Indonesia, Berkeley, University of California Press. Glennie, P. (1988) ‘Continuity and Change in Hertfordshire Agriculture, 1550– 1700: II—Trends in Crop Yields and Their Determinants,’ Agricultural History Review, 36: 145–61. —— (1991) ‘Measuring Crop Yields in Early Modern England,’ in Campbell, B. M. S. and Overton, M. (eds.), Land, Labour and Livestock, Manchester, Manchester University Press, pp. 255–83. Gray, H. L. (1915) English Field Systems, Cambridge, MA, Harvard University Press. Hardin, Garrett (1977) The Tragedy of the Commons, San Francisco, W. H. Freeman. Havinden, M. (1961) ‘Agricultural Progress in Open Field Oxfordshire,’ Agricultural History Review, 9. Hayami, Yujiro (1997) Development Economics: From the Poverty to the Wealth of Nations, Oxford, Clarendon Press. Hoffman, P. (1991) ‘Land Rents and Agricultural Productivity: The Paris Basin, 1450–1789,’ Journal of Economic History, 51: 771–805. Hoskins, W. G. (1950) ‘The Leicestershire Farmer in the Sixteenth Century,’ in Hoskins, W. G. (ed.), Essays in Leicestershire History, Liverpool, pp. 123–83. —— (1951) ‘The Leicestershire Farmer in the Seventeenth Century,’ in Hoskins, W. G. (ed.), Provincial England, London, Macmillan, pp. 149–69. Jackson, R. V. (1985) ‘Growth and Deceleration in English Agriculture, 1660–1790,’ Economic History Review, 2nd series, 38: 333–51. Langdon, John (1986) Horses, Oxen, and Technological Innovation: The Use of Draught Animals in English Farming from 1066 to 1500, Cambridge, Cambridge University Press. McCloskey, D. (1972) ‘The Enclosure of Open Fields,’ Journal of Economic History, 32: 15–35. MacFarlane, Alan (1978) The Origins of English Individualism, Oxford, Basil Blackwell. McIntosh, Marjorie K. (1999) ‘The Diversity of Social Capital in English Communities, 1300–1640 (with a Glance at Modern Nigeria), Journal of Interdisciplinary History, 29: 459–90. Marx, K. (1867) Capital. Mokyr, Joel (1983) Why Ireland Starved: A Quantitative and Analytical History of the Irish Economy, 1800–1850, London, Allen & Unwin. Overton, M. (1979) ‘Estimating Crop Yields from Probate Inventories: An Example from East Anglia, 1585–1735,’ Journal of Economic History, 39: 363–78.

COMMUNITY AND MARKET IN ENGLAND

69

Overton, M. (1991) ‘The Determinants of Crop Yields in Early Modern England,’ in Campbell, B. M. S. and Overton, M. (eds.), Land, Labour and Livestock, Manchester, Manchester University Press, pp. 284–322. —— (1996a) ‘Re–Establishing the Agricultural Revolution,’ Agricultural History Review, 44(I): 1–20. —— (1996b) Agricultural Revolution in England, Cambridge, Cambridge University Press. Polanyi, Karl (1944) The Great Transformation, Boston, Beacon Press. Putnam, Robert D. (1993) Making Democracy Work: Civic Traditions in Modern Italy, Princeton, Princeton University Press. —— (1996) ‘Tuning In, Turning Out: The Strange Disappearance of Social Capital in America, PS Political Science and Politics, 28: 644–83. Scott, James C. (1976) The Moral Economy of the Peasant: Rebellion and Subsistence in Southeast Asia, New Haven, Yale University Press. Thirsk, Joan (1973) ‘Field Systems of the East Midlands,’ in Baker, A. R. H. and Rutlin, R. A. (eds.), Studies of Field Systems in the British Isles, Cambridge, Cambridge University Press, pp. 232–80. Tonnies, Ferdinand (1887) Community & Society (Gemeinschaft und Gesellschaft), translated and edited by C. P. Loomis, East Lansing, Michigan State University Press, 1957. Turner, M. E. (1982) ‘Agricultural Productivity in England in the Eighteenth Century: Evidence from Crop Yields,’ Economic History Review, 2nd series, 35: 489–510. —— (1986) ‘English Open Fields and Enclosures: Retardation or Productivity Improvements,’ Journal of Economic History, 41: 669–92. —— Beckett, J. V. and Afton, B. (1997) Agricultural Rent in England, 1690–1914, Cambridge, Cambridge University Press. —— —— and—— (1998) ‘Renting the Revolution: A Reply to Clark,’ Journal of Economic History, 58: 211–19. van Zanden, J. L. (1993) The Rise and Decline of Holland's Economy: Merchant Capitalism and the Labour Market, Manchester, Manchester University Press. Wrigley, E. A. (1967) ‘A Simple Model of London's Importance in Changing English Society and Economy, 1650–1750’, Past and Present, xxxvii: 44–70. —— (1985) ‘Urban Growth and Agricultural Change in the Early Modern Period,’ Journal of Interdisciplinary History, 15: 683–728. Yelling, J. A. (1977) Common Field and Enclosure in England, 1450–1850, Hamden, CT, Archon Books.

3 The Two Paths of Agrarian System Evolution in the Philippine Rice Bowl Yujiro Hayami and Masao Kikuchi The evolution of economic systems is inherently dependent on historical paths. According to the perspective of evolutionary game theory, an economic system evolves as a process of adaptation to environmental changes through trials and errors by human beings, characterized by bounded ratio-nationality and myopia (Mailath 1992; Kandori et al. 1993; Matsui 1996). Once a historical accident has twisted a system's evolution path, it seldom returns to the old path even if the shock of the accident may subside. Strong inertia usually operates for the preservation of its unique evolutionary path because of scale economies in the use of any specific institution as well as complementarity among various institutions in the system (Aoki 1994, 1996). The evolution of economic systems, therefore, seldom converges into a single path, even if the economic conditions surrounding them are similar. Such a process of the system's evolution is most clearly illustrated in the evolution of agrarian systems in a major riceproducing area in the Philippines. An ‘agrarian system’ is here defined as a combination of land tenure and labor employment arrangements which coordinate and constrain rural people in their use of land and labor, the two primary factors of agricultural production. In the central plain of Luzon Island, two distinct agrarian systems emerged and evolved despite basic similarities in economic conditions, such as ecological conditions for rice production, access to market, colonial history and post-independent government policies such as land reform. The difference emerged mainly from the different community norms that were created through different settlement histories since the Spanish colonial period. Critical influences of the path-dependent community norms on economic system evolution toward multiple equilibria will be demonstrated in this paper on the basis of field research over the 1970s and 1980s.

3.1 Traditional Agrarian Systems Our investigation covers seven provinces—Laguna and Rizal in the Southern Tagalog Region, and Bulacan, Nueva Ecija, Pampanga and Tarlac in the Central Luzon Region.104 These provinces constitute the

104

This section and the next draw heavily on Kikuchi et al. (1979), Kikuchi and Hayami (1980) and Hayami and Kikuchi (1981, ch. 4).

AGRARIAN EVOLUTION IN THE PHILIPPINE RICE BOWL

71

largest rice-producing area in the Philippines, commonly called the rice bowl’ of this nation (Fig. 3.1). Despite the geographical contiguity and ecological homogeneity as well as similarities in the access to metropolitan Manila market, the region is not homogeneous in terms of agrarian Fig. 3.1 Map Of the Central Plain Of Luzon, Indicating Provinces and Municipalities Covered By the Field Surveys

72

YUJIRO HAYAMI AND MASAO KIKUCHI

structure. The region is sharply divided into two subregions characterized by different agrarian systems—the ‘Coastal Region’ and ‘Inner Central Luzon.’ The Coastal Region includes Laguna, Rizal, the southern parts of Bulacan and Pampanga, and the northern part of Pangasinan. Inner Central Luzon encompasses the remaining landlocked area.105 Most areas in the Coastal Region are easily accessible from Manila by sea or lake and have a long history of settlement since before the Spanish conquest in the sixteenth century. In contrast, the landlocked Inner Central area had largely been left uncultivated until the late nineteenth century, since when it has been developed as a rice-producing area.

Two Modes of Landlordism Corresponding to the difference in the settlement history, there emerged a major difference in land-ownership patterns between the Coastal Region and Inner Central Luzon. Both regions are characterized by pervasive landlordism. However, in the Coastal Region, the large estates of native nobility (datu) recognized by Spanish laws or established as royal grants to Spanish conquerors were subdivided gradually over a long period of time into smallholdings through inheritance and sales (Phelan 1959). At the same time, piecemeal accumulation of lands by Chinese or Chinese mestizo traders progressed through their moneylending operations by taking native peasants' lands as collateral (McLennan 1969). The landholdings were thus relatively small and scattered, typically less than ten hectares. A common landtenure arrangement was share cropping called kasamahan (a Tagalog word meaning ‘partnership’) by which both harvest and production costs were shared equally between landlord and tenant. In general, the relation between small resident landlords and their tenants in the Coastal Region was a paternalistic one. A landlord as a patron helps the tenant by advancing credit and by utilizing his connections and social influence, in return for the loyalty of the tenant (Anderson 1964; Larkin 1972). In contrast, Inner Central Luzon was characterized by the development of huge estates (haciendas) of several hundreds to thousands of hectares. The haciendas originated as royal grants and purchases of undeveloped royal domain. This process of large-scale land acquisition by Spaniards began in the eighteenth century after the removal of restrictions on Spanish residence in the provinces. Although some royal grants and large purchases

105

The two regions have similar climatic patterns, both subject to the influence of the southwest monsoon, and have the same farming systems except for harvesting and threshing. On the average, irrigation systems are better developed in the Coastal Region than in Inner Central Luzon. But there are a number of districts in Inner Central Luzon where irrigation systems are equally well established.

AGRARIAN EVOLUTION IN THE PHILIPPINE RICE BOWL

73

did occur in the Coastal Region, too, they took place more recently in a much larger scale in Inner Central Luzon, which then represented the frontier of settlement. Until the late nineteenth century, most areas in Inner Central Luzon were covered by jungle, and the large haciendas engaged primarily in cattle ranching. The raising of cattle came to an end during the 1880s when a series of rinderpest and hoof-and-mouth disease decimated cattle herds (McLennan 1969). Subsequently, the haciendas located in the lowland areas developed a system of rice monoculture carried out by tenants who had migrated from Ilocos and Southern Tagalog, whereas the haciendas located in upland areas were converted into sugar plantations. In these large rice haciendas with several hundreds or thousands of tenants, the landlord–tenant relationship was inevitably less paternalistic than in the Coastal Region. Typically, the owner lived in Manila and the hacienda was managed by an encargado (farm manager) and a number of katiwalas (overseers) (Umehara 1974). The tenure contract was geared more strictly to economic considerations, and it was enforced more strongly by legal means than by a sense of obligation based on patron–client relationships. As a result, the haciendas became a hotbed for tenants' uprisings represented by the Hukbalahap movement (Dalisay 1937; Kerkvliet 1971; Pelzer 1945; Rivera and McMillan 1954).

Two Rice Harvesting Arrangements Corresponding to the different modes of landlordism, different labor employment relations emerged in the form of two distinct rice harvesting arrangements. Traditionally used in the Coastal Region was a labor contract called ‘hunusan’, by which, when a farmer specifies a day of harvesting, anyone can participate in harvesting and threshing and the harvesters receive a certain share of the harvest.106 The share ranges from one-fourth to one-ninth of the harvest, but traditionally one-sixth is the most common. The typical land-tenure arrangement before land reform was such that the harvest was divided into equal halves between landlord and tenant after the share of the harvesters was deducted. In this system the crop is cut manually with sickles and the threshing is done by hand beating on a wood-bamboo plate by the same worker. In contrast, the tilyadora system in Inner Central Luzon is characterized by the use of big threshing machines called tilyadora. The tilyadora threshers are of McCormic–Deering design. The first few machines imported from the United States during the 1910s were run by a steam engine. They

106

The hunusan system has an old origin. Before sickles were introduced, it was a common practice in the Philippines as well as other countries in Asia to cut the ears using hand knives with mass participation of village people. The harvested ears were bundled and the harvesters received one out of six bundles. For the traditional harvesting systems in the Philippines, see Africa (1920).

74

YUJIRO HAYAMI AND MASAO KIKUCHI

were replaced during the 1920s and 1930s by machines run by belt pulley from a tractor of 40–50 horsepower. The threshing by tilyadora is generally done in one central location on each tenant's farm where unthreshed crops are hauled and stored in large stacks. Crop cutting and threshing are two separate operations. While the threshing process is mechanized, the crop cutting is still done by hand with sickles. The labor contract under the tilyadora system is entirely different from the hunusan system. Under the tilyadora system it is difficult to determine harvest shares at the stage of crop cutting because of the difficulty of ascertaining the quantity of grain per sheaf.107 Therefore, harvesters are employed at a fixed daily wage rate (upahan) or at a fixed payment per unit of area harvested (pakyaw). The cost of harvesting before land reform was shouldered by the tenant.108 The fee of threshing by tilyadora (at a rate of 4–6 percent of the harvest) was deducted from the harvest before it was shared between landlord and tenant.109 The introduction of tilyadora in Inner Central Luzon can be partly explained by a relative labor shortage in the areas of new settlement. However, a more important reason appears to be that the tilyadora system was used by the hacienda owner to collect share rent. Under the share cropping contract the first requirement for collecting the right amount of rent was to ascertain the quantity of the harvest. In the Coastal Region, with its paternalistic relationship between landlords and tenants, the landlords could largely trust the honesty of their tenants in their report on the crop harvested. They could also check it easily as they lived together in the same locality. Thus, the transaction costs involved in the collection of share rent were relatively modest when the traditional hunusan system was used for harvesting. Share cropping (kasamahan) was the common form of tenancy in Inner Central Luzon also before the recent land reform after the Second World War. However, in the earlier stage of the development of the rice haciendas, the common tenure arrangement had been the leasehold at a fixed rent (canon). At the start of a settlement, the hacienderos gave a parcel to each settler without rent payment for the period of the land opening, and

107

It is easier to standardize the grain quantity per bundle when crops are cut at the neck of the ears by hand knives than when they are cut at the bottom of plants by sickles. Crop sharing by bundle was especially common in Indonesia, where knife cutting was still commonly used in the 1970s. However, even in Indonesia, crop sharing for harvesters has been shifting from bundle terms to threshed-grain terms as sickle cutting has been introduced. See Collier et al. (1973).

108

In the share-tenancy contract under the hunusan system the cost of harvesting as well as other costs such as seeds, fertilizers, and transplanting were commonly shared half and half by tenants and landlords. Under the tilyadora system, the harvesting cost was shouldered by tenants but the transplanting cost was shouldered by landlords, while the other costs were shared half and half.

109

In reality, wide variations existed for each category of harvesting system. There were cases where the fixed-wage contract was used together with animal treading instead of tilyadora.

AGRARIAN EVOLUTION IN THE PHILIPPINE RICE BOWL

75

requested from him only a nominal rent even after the gratis period. However, as the population and labor force grew, the canon was raised gradually, and finally shifted to share cropping.110 This shift in tenure status took place largely during the 1920s and 1930s. The national census of 1903 indicates that there were 2,215 leasehold cash tenants and only 290 share tenants in Nueva Ecija. According to the 1918 census there were 2,796 cash tenants and 1,798 share tenants. By 1939, the transition was complete with only 867 cash tenants and 50,831 share tenants. The 1920s and 1930s were also the years when the tilyadora became the common method of threshing. The data on the number of tilyadora threshers are not available. However, according to sales records of International Harvester, the number of threshers shipped to the Philippines from the USA increased from only three in 1926 to 37 in 1939.111 Such concurrent growths in the share-tenancy system and the use of tilyadora suggest the hypothesis that the major factor underlying the diffusion of the tilyadora system in Inner Central Luzon was the motivation of hacienderos to reduce the transaction costs required for the collection of share rent and that the tilyadora facilitated the conversion of the land-tenure system from leasehold to share tenancy.112Tilyadora was not adopted in the Coastal Region presumably because the possibility of reducing transaction costs expected from its introduction was much smaller with the paternalistic relationship between landlords and tenants.

3.2 Agrarian Change in Inner Central Luzon From Tilyadora to Hunusan The best data to test our hypothesis are the changes in the harvesting system in Inner Central Luzon following postwar land reform. We should expect that, if the introduction of tilyadora were mainly based on the motivation of hacienderos to reduce the transaction costs incurred in the collection of share rent, the tilyadora system should have been abandoned

110

The shift from leasehold to share tenancy was based on the preference of hacienderos to raise the average rate of land rent at the expense of shouldering a part of the risk involved in rice production. Although the shift from the canon to the kasama system was found widely in Nueva Ecija, where large haciendas were pervasive, it was also observed in other areas in Inner Central Luzon such as Northeastern Pampanga.

111

Personal communication from Mr. Greg Lennes, Corporate Archivist, International Harvester, Chicago, August 31, 1978. Data before 1926 and during 1927–38 are not available.

112

The results of our personal interviews with older people (including an ex-manager of a hacienda) who witnessed the process of tenure change are consistent with the hypothesis that the share cropping system could not have been adopted in the haciendas without the use of the tilyadora.

76

YUJIRO HAYAMI AND MASAO KIKUCHI

according to the logic of institutional complementarity when land reform programs abolished share tenancy. The post-war land reform in the Philippines began during the Magsaysay regime, but its basic structure was stipulated in the Agrarian Reform Code of 1963, specifying two major reform operations: (a) the conversion of share tenancy to fixed-rent leasehold tenancy, and (b) the subsequent transfer of ownership of tenanted land devoted to rice and corn crops in excess of the landlord's retention limit (originally 75 hectares and later reduced step-wise to 7 hectares) to cultivating tenants. The land-reform programs were strongly reinforced by Presidential Decrees No. 2 and No. 27 in 1972 under the Martial Law proclaimed by President Marcos, with the result that the rice-hacienda system was broken, and share croppers were converted into leaseholders or amortizing owners, mostly during the 1970s.113 Upon the abolition of share tenancy, major changes in harvesting systems occurred. The results of our survey covering 43 municipalities in six provinces,114 as shown in Fig. 3.1, clearly show that the regional distribution of harvesting systems in 1968 followed the traditional pattern, i.e. tilyadora was commonly used in Inner Central Luzon whereas hunusan was practiced in the Coastal Region (Fig. 3.2). However, with land reform an entirely different pattern had emerged by 1978 (Fig. 3.3). The most dramatic change was a shift from the tilyadora system to the hunusan system in Inner Central Luzon. In 1968, 96 percent of farmers in Inner Central Luzon in our sample used the tilyadora system. However, the percentage declined sharply to 33 percent in 1978, even though the custom service of tilyadora thresher with operators continued to be available as before. The majority of the farmers shifted completely to hunusan with hand threshing for both wet and dry seasons or adopted hunusan partially for the wet season while using tilyadora for the dry season (‘mixture’ in Fig. 3.3). The shift from the tilyadora to the hunusan system was not uniform within Inner Central Luzon. The shift was complete in the Eastern part of the region (Northern Bulacan and South-to-Central Nueva Ecija) where hunusan was adopted by most farmers for both wet and dry seasons. In the Western part of Inner Central Luzon (southern Pangasinan, Tarlac and Northern Pampanga), the diffusion of hunusan was slower; it was introduced for the wet season only. In Fig. 3.4, the years of shift of sample farmers from the tilyadora to the hunusan system are compared with the years of shift in their tenure status from share croppers to leaseholders or amortizing owners. The fact that no observation lies below the 45-degree line is consistent with the hypothesis that the tilyadora system was introduced

113

For more details on land reform in the Philippines, see Hayami et al. (1990).

114

The two surveys were conducted in February and July 1978 covering 100 farms in 43 municipalities. The sample consisted of farms covered by the Central Luzon/Laguna Loop. Surveys conducted by the International Rice Research Institute for 1966, 1970, and 1974. Historical information was supplemented by those earlier surveys.

AGRARIAN EVOLUTION IN THE PHILIPPINE RICE BOWL

77

Fig. 3.2 Regional Distribution Of Different Harvesting Systems Commonly Used By Municipalities In the Central Plain Of Luzon, 1968

Source: Kikuchi et al. (1979: 12); reproduced in Hayami and Kikuchi (1981: 86) in order to facilitate the management of the haciendas under the share cropping system and, therefore, that the change in the tenure status was a prerequisite to the shift in the harvesting system. In Fig. 3.4, almost all observations from Northern Bulacan and South-Central Nueva Ecija in our sample are located on the 45-degree line, implying that most farmers shifted from the tilyadora to the hunusan system in the year of their tenure change. Sample observations from the Northern Nueva Ecija and Tarlac are located above the 45-degree line, indicating that time lags were involved in the shift from the tilyadora to the hunusan system. A regression equation

78

YUJIRO HAYAMI AND MASAO KIKUCHI

Fig. 3.3 Regional Distribution Of Different Harvesting Systems Commonly Used By Municipalities In the Central Plain Of Luzon, 1978

Source: Kikuchi et al. (1979: 12); reproduced in Hayami and Kikuchi (1981: 86) was estimated for all regions with the year of the adoption of the hunusan system (H) as a dependent variable and the year of tenure change (T) and regional dummies (D for Northern Nueva Ecija and D for Tarlac) as independent n t variables. The least-squares estimation for the sample of 40 farmers for which data are available resulted in

where the numbers inside the parentheses are the student t-values, and R2 is the coefficient of determination adjusted for the degree of freedom.

AGRARIAN EVOLUTION IN THE PHILIPPINE RICE BOWL

79

Fig. 3.4 Relationship Between The Year the Harvesting Arrangements Changed and the Year In Which Land Tenure Status Changed Across Municipalities In Inner Central Luzon

Source: Kikuchi et al. (1979: 14); reproduced in Hayami and Kikuchi (1981: 88) The results of the regression analysis show that the intercept is not significantly different from zero and that the coefficient of T is not significantly different from one at conventional levels of probability. They support the hypothesis that the relation between H and T is the 45-degree line. In other words, the estimated relation between H and T implies that the shift from the tilyadora to the hunusan system occurred as soon as the tenure status changed, except for regional time lags. The coefficients of D and D indicate that the shift in the harvesting system lagged behind the tenure n t change, on the average, 1.6 years for farmers in northern Nueva Ecija and 4.6 years in Tarlac. Such time lags seem to reflect the geographical route and direction of the diffusion of hunusan. According to the information obtained from the interviews in our survey, the diffusion of hunusan in Inner Central Luzon was channeled primarily through seasonal migrant workers from Laguna and Southern Bulacan. Most farmers that we interviewed replied that the workers from the south asked the farmers to adopt the hunusan contract.

80

YUJIRO HAYAMI AND MASAO KIKUCHI

The diffusion of hunusan first began in the southeastern part of Inner Central Luzon and then moved to the north and to the west. The diffusion from the coastal area of Pangasinan to the south was not very significant, perhaps because of the lack of large scale migration of harvesting workers. The speed with which farmers in Bulacan and Nueva Ecija shifted from the tilyadora to the hunusan system after the land-tenure change reflects the inefficiency of tilyadora as a harvesting technology for farmers despite its merit in the collection of share rents by hacienda management. It must also be recognized that, although the land-tenure reform was the basic condition for the shift from the tilyadora to the hunusan system, technological developments in rice production had made the use of tilyadora more inefficient. First, during the 1960s the area under irrigation expanded greatly, especially in Nueva Ecija, with the construction of the Upper Pampanga River Irrigation System. Second, in the late 1960s there was the beginning of a dramatic diffusion of modern rice varieties with short-maturing and nonphotosensitive characteristics. Those developments made the double-cropping of rice a common practice. In the traditional single-cropping regime with the photosensitive varieties, the rice crop matures after November when the length of the day becomes shorter and the rainy season is over. Because the weather is dry, harvested crops can be stored in the fields without danger of spoilage until the tilyadoras arrive. However, in the double-cropping system, the modern varieties mature before the rainy season is finished. Since heavy tilyadoras cannot enter wet fields, the harvested crop must be hauled over a long distance to available dry spots. The risk of crop damage is large due to possible delay in the arrival of the tilyadora. Despite these developments, the tenant farmers were not able to abandon the tilyadora system as long as the regulations of the haciendas were still in force. It seems reasonable to hypothesize that the accumulated frustration of the farmers resulted in an overnight switch to hand threshing as soon as they were emancipated from the yoke of haciendas. As a test of this hypothesis, we estimated a regression equation, where the dependent variable is the time interval between the year of shift from tilyadora to hand threshing (H) and the year of change in the tenure status (T) and the independent variable is the time interval between T and the year of introduction of the system of double-cropping of rice (M). We expect that the longer the interval between T and M, the more clearly farmers recognized inefficiency in the use of tilyadora and, thereby, the shorter was the interval between H and T. The least-squares estimation based on the sample of 26 farmers for which the data were available resulted in

AGRARIAN EVOLUTION IN THE PHILIPPINE RICE BOWL

81

where the numbers in parentheses are the student t-values and R2 is the coefficient of determination adjusted for the degree of freedom. The results are consistent with our hypothesis, showing the significant effect of the time interval between the tenure change and the introduction of double-cropping in reducing the time lag between the tenure shift and the shift in the harvesting arrangement. The shift in the harvesting contract as outlined above clearly illustrates the force of institutional complementarity that makes it difficult for one institution to change independently of other institutions in an economic system, even if its change becomes economically profitable.

Emergence of Permanent Labor Contracts It is easy to enumerate the shortcomings of the land reform program in the Philippines.115 The land retention limit for a landlord (seven hectares) was still too high and it was often evaded by registering excess holdings in the names of relatives and friends. The reform applied only to tenanted land planted in rice and corn. It did not apply to land under the direct administration of landlords. Nor did it apply to land used for cash crops such as sugar. Therefore, it was a widespread practice of landlords to expand the area under their direct management by evicting tenants under the guise of voluntary submission of land from tenants to landlords or to plant sugar in paddy land. Often, collusion between landed elites and local officials invalidated the effect of the land reform program. The operation of transferring land ownership has been particularly slow and its effect has been limited (Purtzel 1992; Riedinger 1995; Takigawa 1994). Despite such apparent failures, the fact cannot be denied that large haciendas were broken down and that most tenants established their status as leaseholders or amortizing owners during the 1970s, though sizable areas remained under landlords' direct administration (Hayami et al.1990, ch. 4; Umehara 1992, 1997). The beneficiaries of land reform were able to capture a large economic surplus because rice yields increased significantly due to irrigation development and application of new varieties and fertilizers, while both rent and amortization payments were fixed. Thus, the land reform was successful in transferring much of the economic return to land from absentee landlords to ex-share croppers. On the other hand, it produced inefficiency in land use as the reform regulations made the land market inactive. Moreover, it created serious income inequality within village communities because the reform produced a major gain

115

This section draws on Hayami et al. (1987), Hayami et al. (1990, ch. 4) and Hayami and Otsuka (1993b).

82

YUJIRO HAYAMI AND MASAO KIKUCHI

to tenant farmers alone with no direct benefit accrued to agricultural laborers subsisting on casual farm work. As a typical ‘land-to-tillers’ program in Asia, Philippine land reform was geared at transferring rights on farm lands from non-cultivating landlords to cultivating tenants, while giving no right to agricultural laborers who are at the bottom of rural communities; wage laborers are not considered the ‘tillers of soil’ even if they work on a farm (Hayami et al.1990, ch. 1). Ironically, the successful income transfer from landlords to ex-share croppers induced the land reform beneficiaries to shift their status from actual tillers of land to non-tilling farmers or semi-landlords. The major income gains from rice yield increases under fixed rent/amortization fees raised the disutility of labor of land reform beneficiaries. On the other hand, landless agricultural laborers whose wage income did not rise, or even decline under strong population pressure, desperately sought employment opportunities. Correspondingly, new employment arrangements developed to facilitate the substitution of landless workers' labor for the family labor of land reform beneficiaries. Rice farming in the Philippines was known for its high dependence on hired labor (Barker and Cordova 1978). However, the hired labor contract was usually short-term, mostly for a day, or for a task which is concluded within a day. Permanent laborers employed for a year or a crop season—relatively common in India and Nepal—were seldom found in the rice sector of the Philippines as well as in other Southeast Asian countries, although they were common in plantations growing tropical cash crops—e.g. dumaan in sugar haciendas in the Island of Negros (Hayami et al.1990, chs. 5 and 6). Indeed, sociological and economic investigations into the agrarian structure of rice villages in the Philippines found an insignificant incidence of permanent labor contracts in the past (Hester and Mabbun 1924; Rivera and MacMillan 1954; Anderson 1964; Takahashi 1969; Umehara 1974; Hayami and Kikuchi 1981; Ledesma 1982). However, there is a sign that a new class of agricultural laborers employed under long-term contracts for a crop season or a year, who resemble the permanent farm laborers of India, has been emerging in Inner Central Luzon. This form of labor contract has been used as a substitute for the land tenancy contract under conditions of dramatic progress in new rice technology and land reform regulations. This development suggests the transformation of the agrarian system in the rice bowl of the Philippines into a structure characterized by bifurcation between non-working farmers and landless laborers, analogous to that of India (Bardhan and Rudra 1980). A survey conducted in August 1987 covering 36 municipalities in six provinces identified three types of permanent or attached labor contracts being practiced there. These were contracts for (a) farm servant, (b) semi-attached laborer, and (c) semi-tenant laborer. A laborer in the first type is

AGRARIAN EVOLUTION IN THE PHILIPPINE RICE BOWL

83

called kasugpon in Ilocano and kasugpong in Tagalog, literally meaning ‘farm helper.’ He is usually a young single male who lives in a farmer-employer's house, is provided with food and clothing, and works exclusively at the employer's farm. A laborer in the second type is also called kasugpong, but is not necessarily single. He may live with his family in a shanty either inside or outside the employer's residential area, usually with no meals provided. He performs certain preassigned tasks, though he is allowed to work outside his master's farm as a casual worker so he can supplement his income. In both types of arrangements, laborers receive fixed wages in paddy ranging from 10 to 30 cavans per season (1 cavan=45 kg), depending on their experience and skill and the scope of the tasks assigned to them. The third type is commonly called porsientuhan, meaning ‘share.’ This arrangement is similar to that of the ‘semiattached laborer’ type but with a more comprehensive coverage of farm tasks to be assigned and with greater autonomy in labor allocation to the assigned parcel of land. Payments are usually in a share of output (usually 10 percent). This arrangement resembles share tenancy. Usually, the semi-tenant and the semi-attached laborer types coexist in the same municipalities. In fact, these two types are difficult to distinguish. For all three types, the contract period is one crop season, but the contract is usually renewed continuously. According to a survey conducted in 1987, employment contracts of the farm servant type were clustered in a rainfed area lying from the Nueva Ecija, Tarlac border to Pangasinan's southern part (Fig. 3.5). This practice is said to be an old one, dating back to before the Second World War. Its incidence did not increase recently but rather declined somewhat because young landless laborers tended to emigrate to irrigated areas. In contrast, the semi-tenant type was commonly observed in the irrigated areas of Nueva Ecija. It is a new system which became common from the late 1970s in replacing contracts of the farm servant type. In Bulacan and Pampanga, the semi-attached laborer and the semi-tenant types seems to have coexisted. These arrangements, especially of the semi-tenant type, became common in these areas. Historical studies suggest that the original kasugpong, or farm servant arrangement, was introduced by migrants from the Ilocos region in the north to Inner Central Luzon when it opened up for rice production in the late nineteenth century (McLennan 1969). In sparsely populated frontier lands in which peace and order were still not well established (e.g. where there was high incidence of carabao theft), it would have been convenient for settlers to have young male helpers living in their farms. At the same time, this arrangement should have provided the young men with a step up from agricultural wage laborer to tenant farmer, since it enabled them to acquire farming skills and to save money needed to purchase farm implements and/or carabao.

84

YUJIRO HAYAMI AND MASAO KIKUCHI

Fig. 3.5 Regional Distribution Of Different Types Of Permanent Labor Arrangements Commonly Used By Municipalities In the Central Plain Of Luzon, 1987

Source: Hayami et al. (1987, 1988) However, as the land frontier reached its limit, it became increasingly difficult for the landless to ascend the agricultural ladder. Moreover, with the implementation of land reform, which protected tenants' cultivation rights and controlled land rent, landowners became more reluctant to rent

AGRARIAN EVOLUTION IN THE PHILIPPINE RICE BOWL

85

out their lands. Thus, the avenue for laborers to become tenants was almost completely closed, unless they were fortunate enough to inherit land cultivation rights from their parents. Today, their options are either to continue living in the same village as casual farm laborers or to emigrate in search of better opportunities elsewhere. One such opportunity is to become semi-tenant laborers in irrigated areas. Meanwhile, the disutility of labor increased among land reform beneficiaries, creating strong demands to substitute hired labor for family labor in rice-farming operations. Traditionally, hired labor performed rice transplanting and harvesting. Rice transplanting was contracted out on a daily wage basis to a crew of transplanters organized by a labor drafter called kabesilya, while harvesting was done under various employment contracts as already explained. In contrast, land preparation had traditionally been the major task that farmers were supposed to perform themselves, because it required due care in plowing and harrowing without harming their draft animals. Recently, however, plowing has increasingly been contracted out to the custom services of large tractors. The remaining tasks performed by farm operators and family members are those that require care and judgment and, therefore, are more difficult to monitor, such as water control, and fertilizer/chemical application and weeding. Unlike transplanting and harvesting, labor demands for these tasks spread thinly over a crop season in an unpredictable manner. Altogether, transaction costs associated with the use of casual labor for these tasks tend to be higher than long-term contracts with a few specific laborers with whom an employer can create a relationship akin to the cooperative-game solution (Hayami and Otsuka 1993b). It is reasonable to expect that the demands for permanent or attached laborers would have become especially large in areas where major yield gains were achieved with the progress of land reform. If a well-to-do land reform beneficiary in a well-irrigated area would want to minimize his own farm work including labor supervision, it would be most efficient to have a tenancy contract with a worker, leaving all the tasks to him for a fixed or harvest-share rent. In fact, some new leaseholders opted to have subtenancy contracts with landless laborers. However, the subtenancy arrangements are highly hazardous because, if the lessee dares to appeal to the Agrarian Reform Office, the lessor's tenancy title would be forfeited and transferred to the lessee if the lessee proves himself to be the real ‘tiller of the land’. Therefore, subtenancy arrangements are bound to be limited to a narrow circle of relatives and close friends. The labor employment arrangement of the semi-tenant type can be considered an institutional innovation to fill the demand of well-to-do land reform beneficiaries in irrigated areas, who want to withdraw

86

YUJIRO HAYAMI AND MASAO KIKUCHI

from farm work while keeping the titles on lands. It plays a role similar to tenancy contracts, while it can easily be disguised as a labor employment contract even though farming based on the employment of permanent laborers is proved to be less efficient than that based on the family labor of either owner or tenant operators (Hayami and Otsuka 1993a, ch. 9). A typical example was found in the case of a large farmer holding three hectares of leasehold land and two hectares of amortizing land in the municipality of Muñoz, Nueva Ecija. While he operated a small rice mill and a grocery (sari-sari) store, he left almost all the farm tasks, including land preparation (except transplanting, weeding, and harvesting), to a family of three working males (father and two sons). For their services, this family received 10 percent of the gross paddy output from this farm. They lived in a shanty in the farmer's residential area but neither food nor clothing was provided to them by the employer. They were entitled to receive daily wages when they engaged together with other casual laborers in transplanting, weeding, and harvesting activities. In addition, they supplemented their income by entering into a semi-tenant labor contract with another farmer. The use of permanent laborers as a substitute for tenancy contracts was not limited to land reform beneficiaries. It was also used by large landlords who still maintain large tracts of land under their direct administration, because nontenanted lands are exempt from land reform programs. Such a case was also found in Muñoz. The widow of a former owner of a rice hacienda of 200 hectares were managing a farm of 30 hectares, while the rest was taken away by reform programs. She was managing this farm with ten permanent laborers each assigned two to three hectares of land. These laborers were supervised by one overseer. While these workers received 10 percent of gross output, the overseer received a 5 percent share. In addition, the overseer was allowed to cultivate 1.5 hectares under a 50:50 sharetenancy contract. An interesting aspect of this case is that the landlord forced her workers to sign written contracts indicating that they are laborers but not tenants. In fact, the appeal of one worker to the Agrarian Reform Office to establish a tenancy title was turned down because of this document. The emerging agrarian structure in Inner Central Luzon resembles that of India where farmers of upper castes do not themselves work but only supervise the work of laborers of lower castes, with no agricultural ladder bridging them. The agrarian system of India was crafted on the social norm that prohibits the low-caste laborers' ascent to farm operators. It was further reinforced by the prohibition of share tenancy by land reform laws (Hyami and Otsuka 1993a: 85–9). Land reform in the Philippines has had the same effect. This ‘Indianization’ of the Philippine rice bowl will be

AGRARIAN EVOLUTION IN THE PHILIPPINE RICE BOWL

87

reversed only with the creation of sufficient nonfarm employment opportunities for the rural poor.

3.3 Agrarian Change in the Coastal Region An apparent anomaly compared with the case of Inner Central Luzon was the sharply different mode of agrarian system evolution in the Coastal Region represented by the diffusion of the gama harvesting system (Fig. 3.3) with little incidence of permanent labor contracts (Fig. 3.5).116 In terms of these characteristics the province of Laguna, which may be called the heartland of the Coastal Region for its old settlement history and traditionally easy access to Manila by means of safe lake lanes, was the most distinct. Henceforth, our discussions will refer mainly to the case of Laguna. However, it must be noted that this province's experience was rather typical of agrarian system changes in old-settled peasant communities in the Philippines as well as other economies in Southeast Asia (Ledesna 1982; Hayami and Kikuchi 1981, chs. 7 and 8).

From Hunusan to Gama Concurrent with a shift from the tilyadora to the hunusan system in Inner Central Luzon, a shift occurred in the Coastal Region from the hunusan to the gama system (Fig. 3.3). Gama is a harvest-sharing arrangement similar to hunusan except that the employment for harvesting and threshing is limited to workers who weeded the fields without receiving wages. In other words, in the gama system the free service of weeding is a prerequisite for laborers to be employed in harvesting and receive a certain output share, usually one-sixth, of the harvest. In 1968, 83 percent of our sample farms in the Coastal Region used the hunusan system but the ratio declined to less than 50 percent in 1978 as it was replaced by gama. This shift was most complete in Laguna where the ratio of gama adopters increased from 27 to 100 percent during the same period. It should be recalled that hunusan was an old form of contract in this area. It was ingrained with the traditional norm of ‘work and income sharing’ within a village community. This norm dictates that well-to-do members in a village community should provide income-earning opportunities to poor neighbors by retreating themselves from work. It is a kind of

116

This section draws heavily on Hayami (1998) and Kikuchi and Hayami (1983).

88

YUJIRO HAYAMI AND MASAO KIKUCHI

community principle of mutual help to guarantee minimum subsistence to the poor.117 A condition for the original establishment of the sharing principle could have been a low level of agricultural productivity with high risk. Until rather recently, Southeast Asia had been characterized by sparse population relative to available land for cultivation. Before the diffusion of modern high-yielding varieties since the late 1960s, rice farming was typically extensive with little fertilizer application and weeding practice, so that yield differences between diligent and idle farmers were not pronounced. Therefore, whether a farmer himself worked hard on his field or left the work to hired laborers did not greatly affect the level of yield. On the other hand, production risk was high, especially in the absence of irrigation and drainage systems. During drought, crops in elevated locations may be destroyed, whereas a bumper crop may be harvested from lowerlying marshy fields. The reverse is likely to be the case during a season of heavy rain and flooding. Similarly, an outbreak of pest may eliminate crop in a certain area, while other areas might be little damaged. Because it is hazardous to rely on production from a single plot for subsistence, a peasant usually holds his land in small parcels scattered over a wide area. Similarly, it greatly reduces risk, if he allows other villagers to share work and output in his farm, while he is allowed to share work and output in others' farms. This insurance mechanism of work and income sharing would be especially valuable in economies where the market is underdeveloped, providing villagers with no other means of insuring against risk in farm production, such as off-farm employment opportunities, formal insurance, and credit systems. This is one of many insurance mechanisms in traditional societies, which generally involve diversifying family members' economic activities widely across different locations (Rosenzweig 1988a, b; Stark and Lucas 1988). Thus, it is hypothesized that the community-wide sharing system as represented by hunusan emerged through experiments by peasants to secure subsistence at a low level of land productivity. It was established because it was mutually beneficial to sharing parties. As such, egoists should have found it profitable to observe sharing arrangements such as hunusan harvesting in terms of their rational economic calculations. However, the sharing system would not have been elevated to a social norm, and maintained

117

A different community norm operates in Japanese villages, which strongly dictates villagers' cooperation in collective actions for construction, operation, and conservation of common-property resources, such as irrigation systems, forest, and pasturelands. It seems that this norm was formed in Japan under strong population pressure on natural resources since the eighteenth century (Hayami 1998: 251–4). In contrast, such a norm has been rather weakly established so far in Southeast Asia, because a comparable scarcity of natural resources began to be felt there only as late as the 1950s (Hayami and Kikuchi 1981: 20–3).

AGRARIAN EVOLUTION IN THE PHILIPPINE RICE BOWL

89

as such, unless violations from this norm (e.g. receiving shares from neighbors without reciprocating to them) could be expected to receive social sanctions such as social opprobrium (which may eventually escalate to ostracism) in the small village community characterized by intensive personal interactions (Becker 1974; Akerlof 1976). Such social sanctions would not have been so credible and effective unless supported by a common cultural belief to consider the sharing arrangements fair and just. If the community principle of work and income sharing originated in rational choice under certain economic and technological conditions in the past, its practice would have changed corresponding to changes in those conditions. In the traditional hunusan system, every villager could participate in harvesting and normally receive one-sixth of harvested rice. In the past when rice farming was associated with low yields, one-sixth of output could well have been close to harvesting labor's contribution to output (or labor's marginal productivity). However, with the advent of the Green Revolution the use of modern rice varieties and chemical fertilizers was promoted, nearly doubling yields per hectare from a level of about 2 tons per hectare in the mid 1960s to over 4 tons by the early 1980s. The yield increases resulted in parallel increases in harvesters' receipts with their share fixed at one-sixth. On the other hand, market wage rates remained largely stable under the pressure of rapid labor force growth that resulted from explosive population growth of the order of 3 percent per year. As a result, the rate of return to labor under the traditional hunusan contract rose cumulatively above the market wage rate. At the same time, as in Inner Central Luzon, farmers' income positions greatly improved relative to landless agricultural laborers due to concurrent progress in land reform and the Green Revolution. Unlike Inner Central Luzon, however, landlords in Laguna were typically small, not many owning more than the land retention limit of seven hectares. The operation of land ownership transfer in land reform, therefore, had a relatively small impact in Laguna, but the operation of tenancy reform had a major impact as it converted tenant farmers from share croppers to leaseholders with their rent payments fixed at rates lower than market equilibria. Moreover, under the fixed rents new leasehold tenants were able to capture all the increments of rice yields resulting from the Green Revolution. It was under these conditions that the gama system diffused rapidly in Laguna. With gama, not only harvesting but also weeding was handed over to hired laborers. Because the gama system requests harvesting laborers to render a free weeding service for receiving the same share of output, the implied wage rate is lower in gama than in hunusan. Although this adjustment reduced the rate of compensation to labor per hour, it was accepted by villagers because it maintained the traditional sharing arrangement as well as the sharing rate, thereby remaining consonant with the norm of the village community.

90

YUJIRO HAYAMI AND MASAO KIKUCHI

With the adoption of the gama system the gap between harvesters' share and labor's marginal productivity under the traditional hunusan system was closed. This is illustrated by calculations based on periodic farm surveys in the province of Laguna conducted by the International Rice Research Institute (Table 3.1). First, labor inputs per hectare per crop season are measured in workdays for both harvesting and weeding per hectare (rows 1 and 2). Dividing harvesters' output share (row 4) by the number of work days for harvesting (row 1) produces the imputed wage rate of harvesting labor under the hunusan contract in terms of rough rice (paddy) equivalent (row 5). The imputed real wage rate under the gama contract (row 6) can be obtained by dividing harvesters' share by the number of work days for both harvesting and weeding (row 3). Those imputed wage rates are compared with the market wage rates in rough rice terms, which are obtained by dividing the nominal wage rates of daily employed laborers by the farm-gate prices of rough rice. The results show that, if hunusan had been used in the 1970s and 1980s when modern rice Table 3.1 Imputation of Wage Rates for Rice Harvesting Work Under Different Contracts in the Lowland Rice Area of Laguna, Philippines, Wet seasons for Selected Years, 1970–1990. 1970 Sample size (no. 152 of farms) Percent of gama 28 adoptors Number of work days per hectare (1) Harvesting 27.2 and threshing (2) Weeding 13.3 (3) Total 40.5 419 (4) Harvesters' output share (kg/ ha) Imputed wage rate (kg/ha) (5) Hunusan: (4)/ 15.4(154)b (1) (6) Gama: (4)/(3) 10.3(103) Market wage rate 10.0(100) (kg/ha)a

1975 67

1981 100

1984 28

1990 59

85

90

86

74

29.2

29.2

23.0

18.8

27.4 56.6 493

16.5 45.7 594

16.7 39.7 479

16.6 35.4 461

16.9(201)

20.3(185)

20.8(175)

24.5(183)

8.7(104) 8.4(100)

13.0(99) 13.1(100)

12.1(102) 11.9(100)

13.0(97) 13.4(100)

Sources: Laguna Survey, 2nd to 11th rounds, conducted by Social Science Division, International Rice Research Institute, Philippines. a b

Notes: Rough rice equivalent. Indices with the market wage rates set equal to 100 are shown in parentheses.

AGRARIAN EVOLUTION IN THE PHILIPPINE RICE BOWL

91

technology was widely diffused, the rate of compensation to harvesting labor would have been about 50 to 100 percent higher than its market wage rate. This gap was eliminated almost completely with adoption of the gama system.118

The Community Mechanism of Work Enforcement These findings are consistent with the hypothesis that the gama contract represents an institutional arrangement designed to reduce disequilibrium between labor's remuneration rate and marginal productivity within the framework of work and income sharing in the community. Relative to other methods to recover equilibrium, such as lowering harvesters' share rate or replacing the sharing contract by the fixed-wage time-rate contract, the gama contract should have entailed lower transaction costs for its greater congruence with the traditional arrangement. Moreover, the gama contract by nature tends to promote a patron–client relationship among villagers, because it links harvesting with weeding and continues to be renewed over seasons. Wages are not paid at the time of weeding. In the minds of villagers, weeding with no direct payment is considered not a part of a contract based on economic calculation but an expression of gratitude by laborers for the goodwill of a farmer patron who provides them a guarantee of stable incomes from contracted plots at a time-honored share rate, thereby reducing their subsistence risk. Such a personal relationship is further strengthened through employment in other tasks as well as exchanges of gifts, credits and personal services. Both the sense of moral obligation and the fear of losing the patron–client relationship motivates laborers to exert conscientious work effort and to refrain from committing moral hazard such as stealing of harvested rice and shirking in weeding work. Where this community mechanism of work enforcement operates, it would not be necessary for a farmer employer to bind particular laborers by permanent labor contracts in order to elicit their conscientious efforts. This appears to be the reason why permanent labor contracts failed to make inroads into Laguna where the community norm of work and income sharing was strongly established as a tradition. Underlying the relatively efficient working of the community enforcement mechanism was the traditional agrarian structure in which small/ medium landlords and share croppers were tied together in a patron–client relationship. When this relationship was broken by the land reform

118

The diffusion of gama reached a peak in the early 1980s. Thereafter, gama was gradually replaced by ‘new hunusan ’ which is the same as old hunusan except that only those who are invited by farmers can participate in harvesting. This change occurred with the adoption of portable threshing machines as well as increased urban influences. For detail, see Kikuchi and Hayami (1999).

92

YUJIRO HAYAMI AND MASAO KIKUCHI

program and, accordingly, farmer share croppers rose to the status of well-to-do leaseholders, they were expected to behave as new patrons vis-à-vis agricultural laborers whose population increased partly because of closed cultivation frontiers and partly because of inactive land rental market under land reform regulations. Indeed, the land reform beneficiaries behaved as if they were legitimate patrons according to the traditional community norm by themselves retreating from farm work. By doing so they saved the transaction costs of hiring labor. While handing over farm tasks to landless laborers at relatively low transaction cost, they were able to allocate their own labor to non-farm work or to allocate their children's time to higher education. In this way, land reform beneficiaries have increasingly moved out from the ‘tillers of soil’ and approached the urban middle class. A similar transformation of land reform beneficiaries have also taken place in Inner Central Luzon. However, the traditional relationship between large hacienderos and share croppers had been opposite to the patron–client mode that prevailed in the Coastal Region. Terms of conditions in share cropping contracts had clearly been specified in written documents and enforced by means of direct supervision by overseers and managers, often backed up by the police as well as private armies. When this system was broken by land reform and former share croppers rose to the status of amortizing owners or leasehold tenants, they were not well endowed with the tradition of the community mechanism of work enforcement. Hunusan was introduced from the Coastal Region by migrant harvesting laborers, as explained before. However, as subsequent increases in rice yield created disequilibria between harvesters' share and market wage rate, the share rate was pulled down from one-sixth with no hesitation. The traditional share rate of one-sixth in the Coastal Region, especially in Laguna, was not something to be respected as a traditional norm in Inner Central Luzon where the hunusan system was newly introduced. Consequently, weeding was performed separately by laborers employed on daily wage contracts instead of being combined with harvesting into gama. In both harvesting and weeding, direct supervision by farmers themselves or overseers was commonly practiced in Inner Central Luzon, whereas farmers in the Coastal Region often themselves refrained from visiting the fields where hired laborers were working (though they usually tried hard to monitor the work efforts of their employees indirectly by such means as informal hearing from neighbors who observed the laborers' work). In the Coastal Region, if a farmer employer would show the gesture of a benevolent patron who trusts his clients, the open exercise of opportunism by the client laborer would be penalized through social opprobrium and eventual ostracism. Under the community norm of work and income sharing that was commonly shared between farmers and agricultural laborers, transaction costs associated with hiring labor were relatively modest.

AGRARIAN EVOLUTION IN THE PHILIPPINE RICE BOWL

93

In the absence of such a tradition, farmers in Inner Central Luzon whose income position markedly improved due to land reform and new rice technology were compelled to pay the high cost of direct supervision. For the tasks for which direct supervision was difficult, farmers opted to bind certain laborers by formal permanent labor contracts which were designed to elicit laborers' work effort for the benefit (or fear of losing the benefit) of stable employment and income.119 In contrast, farmers in the Coastal Region needed no such formal contracts, because they could rely on de facto long-term contracts based on the tacit understanding that casual daily wage contracts continue to be renewed indefinitely, akin to the world of the Folk Theorem (Fudenberg and Maskin 1986; Abreu 1988).

3.4 Conclusion The comparative history of two regions in the Philippine rice bowl illustrates the mechanism of economic systems to evolve along multiple paths. Despite basic similarities in environmental conditions for rice production as well as the impacts of Spanish colonial policies, two different agrarian structures developed—the large hacienda system characterized by the management hierarchy for supervising a large number of tenants in Inner Central Luzon and the relatively small-scale landlordism in which landlords and tenants were tied by a patron–client bond involving little direct supervision in the Coastal Region. The difference originated from different settlement processes which were initially based on differential access to Manila market, i.e. the Coastal Region was settled much earlier than landlocked Inner Central Luzon because of the farmer's advantage in utilizing safe lake lanes. While this difference was later reduced by developments in inland transportation systems such as railways and highways, the hacienda system consolidated itself in Inner Central Luzon with such innovations as the introduction of large threshing machines for enforcing share cropping contracts. Meanwhile, the community-wide harvest-sharing arrangement continued to be practiced in the Coastal Region that relied solely on hand threshing. This harvest-sharing arrangement saved labor transaction costs as it was consistent with the community norm of work and income sharing. The difference between the Coastal Region and Inner Central Luzon was further reduced by land reform programs. The hacienda system was broken and many ex-share croppers in Inner Central Luzon established their status as amortizing landowners with the expectation that they will become full-fledged owners after finishing amortization payments for the lands they

119

For the mechanism of permanent labor contracts to elicit laborers' work efforts for the tasks on which direct monitoring is difficult, see Eswaran and Kotwal (1985) and Hayami and Otsuka (1993b).

94

YUJIRO HAYAMI AND MASAO KIKUCHI

received. Concurrently, most farmers in the Coastal Region were converted from share croppers to leaseholders at low fixed rents—almost the same status as amortizing owners. New agrarian structures in the two regions created from land reform were similar, both characterized by the rise of land reform beneficiaries as new elites relative to landless agricultural laborers who staked out a meager subsistence from hired farm work under mounting population pressure. The income disparity between the two classes widened further with the Green Revolution that equally affected the two regions. Despite the similarity in newly emerged agrarian structures, the modes of labor employment relations were different. In both the regions, farmers whose income position improved significantly began to retreat from farm work by handing it over to agricultural laborers. In the Coastal Region where the community norm of work and income sharing had traditionally been established, farmers tried to strengthen the community mechanism of work enforcement by such means as the use of the gama system which adds weeding to harvesting for the traditional harvest share for laborers in the same community. In Inner Central Luzon where the community had traditionally been bifurcated between hacienderos and share croppers, new elite farmers introduced permanent labor contracts to bind particular laborers for eliciting their conscientious work efforts. Such contrasts clearly illustrate the power of different social norms to diverge the evolution of economic systems in different modes, despite converging economic, social and technological environments.

References Abreu, D. (1988) ‘On the Theory of Indefinitely Repeated Games with Discounting’, Econometrica, 56: 383–96. Africa, A. A. (1920) ‘A Preliminary Survey of the Comparative Costs of Different Methods of Harvesting Rice,’ Philippine Agriculturist, 8: 272–92. Akerlof, G. A. (1976) ‘The Economics of Caste and the Rat Race and Other Woeful Tales,’ Quarterly Journal of Economics, 90: 599–617. Anderson, J. N. (1964) ‘Land and Society in a Pangasinan Society,’ in Espiritu, S. C. and Hunt, C. L. (eds.), Social Foundations of Community Development, R.M. Garcia Publishing Co., Manila, pp. 171–93. Aoki, M. (1994) ‘The Contingent Governance of Teams: Analysis of Complementarity,’ International Economic Review, 35: 657–76. —— (1996) ‘Toward a Comparative Institutional Analysis: Motivations and Some Tentative Theorizing,’ Japanese Economic Review, 47: 1–19. Bardhan, P., and Rudra, A. (1980) ‘Types of Labor Attachment in Agriculture: Results of a Survey in West Bengal, 1979,’ Economic and Political Weekly, 15 (August 30): 1477–84. Barker, R., and Cordova, V. G. (1978) ‘Labor Utilization in Rice Production,’ in Economic Consequences of New Rice Technology, International Rice Research Institute, Los Baños, Philippines, pp. 113–36.

AGRARIAN EVOLUTION IN THE PHILIPPINE RICE BOWL

95

Becker, G. S. (1974) ‘A Theory of Social Interactions,’ Journal of Political Economy, 82: 1063–93. Collier, W. L., Wiradi, G., and Soentoro (1973) ‘Recent Changes in Rice Harvesting Methods,’ Bulletin of Indonesian Economic Studies, 9: 36–45. Dalisay, A. (1937) ‘Types of Tenancy Contracts on Rice Farms in Nueva Ecija,’ Philippine Agriculturist, 26: 159–98. Eswaran, M., and Kotwal, A. (1985) ‘A Theory of Two-Tier Labor Markets in Agrarian Economies,’ American Economic Review, 75: 162–77. Fudenberg, D. and Maskin, E. (1986) ‘The Folk Theorem in Repeated Games with Discounting or with Incomplete Information,’ Econometrica, 54: 533–54. Hayami, Y. (1998) ‘Norms and Rationality in the Evolution of Economic Systems: A View from Asian Villages,’ Japanese Economic Review, 49 (February): 36–53. —— and Kikuchi, M. (1981) Asian Village Economy at the Crossroads: An Economic Approach to Institutional Change, University of Tokyo Press, Tokyo (Published for North America by Johns Hopkins University Press in 1982). —— and Otsuka, K. (1993a) The Economics of Contract Choice: An Agrarian Perspective, Oxford University Press, Oxford. —— and—— (1993b) ‘Kasugpong in the Philippine Rice Bowl: The Emergence of New Labor Institutions after the Land Reform,’ in Hoff, K., Bravermann, A., and Stiglitz, J. E. (eds.), The Economics of Rural Organization, Oxford University Press, Oxford. Hayami, Y., Marciano, E. B., and Bambo, L. M. (1987a) ‘Kasugpong in Central Luzon: “Indianization” of the Philippine Rice Bowl,’ IRRI Department of Agricultural Economics Paper, 87–22, International Rice Research Institute, Los Baños, Philippines. —— —— and—— (1987b) Rice Harvesting in Central Luzon and Laguna Revisited, IRRI Research Paper Series, 133, International Rice Research Institute, Los Baños, Philippines. Hayami, Y., Quisumbing, M. A., and Adriano, L. S. (1990) Toward an Alternative Land Reform Paradigm: A Philippine Perspective, Ateneo de Manila University Press, Quezon City, Philippines. Hester, E. D. and Mabbun, P. (1924) ‘Some Economic and Social Aspects of Philippine Tenancies,’ Philippine Agriculturist, 12: 367–444. Kandori, M., Mailath, G., and Rob, R. (1993) ‘Learning, Mutation, and Long Run Equilibria in Games,’ Econometrica, 61: 29–56. Kerkvliet, B. J. (1971) ‘Peasant Society and Unrest Prior to the Huk Revolution in the Philippines,’ Asian Studies, 9: 164–213. Kikuchi, M., Cordova, V. G., Marciano, E. B., and Hayami, Y. (1979) ‘Changes in Rice Harvesting System in Central Luzon and Laguna,’ IRRI Research Paper Series, 31, International Rice Research Institute, Los Baños, Philippines. —— and Hayami, Y. (1980) ‘Technology and Labor Contract: Two Systems of Rice Harvesting in the Philippines,’ Journal of Comparative Economics, 4: 357–77. —— and—— (1983) ‘New Rice Technology, Intrarural Migration, and Institutional Innovation in the Philippines’, Population and Development Review, 9: 247–57. —— and —— (1999) ‘Technology, Market and Community in Contract Choice: Rice Harvesting in the Philippines,’ Economic Development and Cultural Change, 47: 371–86.

96

YUJIRO HAYAMI AND MASAO KIKUCHI

Larkin, J. A. (1972) The Pampangans': Colonial Society in a Philippine Province, University of California Press, Berkeley, Los Angeles and London. Ledesma, A. (1982) Landless Workers and Rice Farmers: Peasant Subclasses under Agrarian Reform in Two Philippine Villages, International Rice Research Institute, Los Baños, Philippines. McLennan, M. S. (1969) ‘Land and Tenancy in the Central Luzon Plain,’ Philippine Studies, 17: 651–682. Mailath, G. J. (1992) ‘Introduction: Symposium on Evolutionary Game Theory,’ Journal of Economic Theory, 57: 259–77. Matsui, A. (1996) ‘On Cultural Evolution: Social Norms, Rational Behavior and Evolutionary Game Theory,’ Journal of Japanese and International Economics, 10: 262–94. Pelzer, K. J. (1945) Pioneer Settlement in the Asiatic Tropics, American Geographical Society, New York. Phelan, J. L. (1959) The Hispanization of the Philippines: Spanish and Filipino Responses, 1565–1700, University of Wisconsin Press, Madison. Purtzel, J. (1992) Captive Land: The Politics of Agrarian Reform in the Philippines, Monthly Review Press, New York. Riedinger, J. F. (1995) Agrarian Reform in the Philippines: Democratic Transitions and Redistributive Reform, Stanford, Stanford University Press. Rivera, G. E., and McMillan, R. T. (1954) An Economic and Social Survey of Rural Households in Central Luzon, Cooperative Research Project of the Philippine Council for United States Aid and the United States of America Operations Mission to the Philippines, Manila. Rosenzweig, M. R. (1988a) ‘Risk, Private Information, and the Family,’ American Economic Review, 78: 245–50. —— (1988b) ‘Risk, Implicit Contracts and the Family in Rural Areas of Low Income Countries,’ Economic Journal, 98: 1148–70. Spillman, W. J. (1919) ‘The Agricultural Ladder,’ American Economic Review, 9(1 Supplement): 170–9. Stark, O., and Lucas, E. B. (1988) ‘Migration, Remittance, and the Family,’ Economic Development and Cultural Change, 36: 465–81. Takahashi, A. (1969) Land and Peasant in Central Luzon, The Institute of Developing Economies, Tokyo. Takigawa, T. (1994) Tonan Ajia Nogyo Mondai Ron (Treatise on Agrarian Problems in Southeast Asia), Keiso Shobo, Tokyo. Ullmann-Margalit, E. (1978) The Emergence of Norms, New York, Oxford University Press. Umehara, H. (1974) A Hacienda Barrio in Central Luzon: A Case Study of a Philippine Village, The Institute of Developing Economics, Tokyo. Umehara, H. (1992) Philippine no Noson: Sono Kozo to Henka (Rural Villages in the Philippines: Their Structure and Change), Kokon Shoin, Tokyo. —— (1997) ‘Sonraku Level de Miru Philippine Nochi Kaikaku no Tenmatsu’ (Outcomes of Land Reform in the Philippines Observed at a Village Level), in Mizuno, H. and Shigetomi, S. (eds.) Tonan Ajia no Keizai Kaihatsu to Tochi Seido (Economic Development and Land Systems in Southeast Asia), Institute of Developing Economies, Tokyo, pp. 153–98.

4 Community Norms and Embeddedness: A Game-Theoretic 120

Masahiko Aoki The community is a somewhat elusive concept. In this chapter I define it tentatively as a group of agents who repeatedly interact with each other in social and economic domains and thus are mutually identifiable. In the course of discussion I will provide a more precise meaning. In any case, the members of the community may share (limited) information regarding each others' choices of action/action-plans and derive utilities from repeated interactions in a social domain. I do not assume, however, that the members of the community necessarily share the same a priori objective. That is, they can be agents seeking private interests while interacting with each other. Yet, through repeated interactions they may autonomously generate implicit rules regulating their individual choices and thus internalize their welfare. We may refer to such endogenous rules as community norms. As such, an open-ended set of traders engaged in impersonal exchanges in an anonymous market is not a community. On the other hand, a group of traders who interact repeatedly and share information regarding the history of each others' action choices may be referred to as a community of traders (e.g. Greif's chapter in this volume). Groups of peasant families who share the use of local commons, such as an irrigation system or forestry, are certainly included in a class of communities in our definition (Part III). Even an organization in the modern industrial context of which membership is relatively stable so that the coordination of activities among members may be substantially regulated by implicit rules and conventions of its own, may be conceived of having an aspect of a community. The community as defined above becomes of interest to us, since, as said, it may generate internal rules—community norms—that are endogenous and thus self-enforceable without third-party involvement. Under what conditions is such a rule likely to be generated and sustained within a community? What are the efficiency implications of such a rule? Is its existence detrimental or facilitating to the development of impersonal economic relationships (market relationships) beyond the confines of a community?

120

I am indebted to comments and suggestions by Christopher Kingston, Yasumasa Koga, Yujiro Hayami, Bins Hansberger, and Pranab Bardhan.

98

MASAHIKO AOKI

Is there any role that a community mechanism (the enforcement of a self-generating norm) can play in the context of a developed market economy? If so, under what conditions and how? These basic questions are investigated in chapters in this volume in various historical and geographical contexts from diverse perspectives. This chapter concerns some of the formalism and attempts to find some generic logic involved in the arguments in these chapters. More specifically, it aims to narrate a basic analytical logic of community-mechanisms from a game-theoretic perspective without being involved in technical rigor and sophistication. The organization of this chapter is as follows. Section 4.1 formulates the basic structure of an economic domain in game form and distinguishes between exogenous rules of the game (e.g. technology, statutory law), on the one hand, and community norms as endogenous rules of the game, on the other. It restates a familiar, generic logic of the reputation mechanism to facilitate subsequent discussions of the endogenous formation of community norms. Section 4.2 then discusses the endogenous formation of clientship or business networks (e.g. Fafchamps' chapter) that regulate exchanges in the context of the initially impersonal exchange domain (random matching). Sections 4.3–4.4 deal with a situation in which externalities in an economic transaction domain, say the local commons domain, are so severe as to be autonomously regulated in that domain, if it stands alone. However, we can envision a situation in which the economic transaction domain is embedded in a social exchange domain in which the same members repeatedly interact socially and invest in, and enjoy returns from, social capital. Then, to keep the value of social capital, the members of community may refrain from moral hazard actions in the economic transaction domain through the formation of a social norm internalizing externalities. This situation can be considered to correspond to the one which the sociologist Granovetter dealt with in his seminal paper criticizing the concepts of ‘modern man’ in neoclassical and transaction economics. We analyze a possible outcome of ‘social embeddedness’ in terms of linked games in which agents coordinate their strategies in two domains (one being the social exchange domain and another an economic transaction domain) and then provide a few interesting illustrations. Section 4.5 turns to dynamic issues by applying the idea of linked games. Specifically, we consider the possible roles of a community norm and structure in the transition of a rural community to contractual relationships with outsiders to the community, as well as those of replication of a community norm in the context of a modern industrial organization. As is well known, repeated games can generate multiple equilibria (the so-called Folk theorem). Within the present state of game theory there is no convincing theory to single out one equilibrium over the many that are possible nor is such a theory likely to develop. Thus, an understanding of why one

COMMUNITY NORMS AND EMBEDDEDNESS

99

particular institutional arrangement, such as a norm, evolved as an self-enforcing rule of the game (that is, as an equilibrium) needs to be pursued in combination with the analysis of historical and comparative information derived from specific contexts. Thus this chapter and the other chapters of this volume complement each other. Section 4.6 is a summary.

4.1 The Basic Logic of the Reputation Mechanism Imagine a group of agents who repeatedly interact in a domain of economic transaction over a number of periods. Each agent is endowed with a fixed set of physically feasible actions (e.g. produce, sell, or buy a unit of a commodity). A combination of action choices by all agents—an action profile—generates a physical consequence in the state space describing all possible physical states (e.g. final consumption of goods by the agents). A rule which assigns a consequence with every possible combination of action choices by agents is called a consequence function. It may be parametrized by various environmental factors including technology, asset endowments, as well as legal rules, if any.121 In other words, it may be thought of as representing the exogenously given rule of the game in the domain. Each agent has a preference ordering over possible consequences in the state space. Then the composite of the consequence function and an agent's preference function may be interpreted as the payoff function of that agent in the ordinary sense of game theory. The reason we decompose it into the objective consequence function and subjective preference function is to clarify the notion of exogenous (or technologically determined) rules of the game, which are to be distinguished from the notion of institutions as ‘endogenous, self-enforcing rules of the game’ that we seek. When all agents in the domain strategically choose their plans of action so as to maximize the values of their preference functions subject to the exogenous rules of the game and their own expectations regarding others' choice plans, the situation is called a game. If the set of agents is fixed in a certain domain of the game in which the membership of any agent is identifiable, and if substantive characteristics of actual actions of the agents can be observed and known to all agents in the domain and utilized as a basis for their subsequent action choices, we

121

This implies that legal rules do not directly restrict the set of feasible actions of agents, but may affect their consequences. For example, a law that regulates the traffic of drugs per se does not alter the action choice sets of agents, since they can still smuggle drugs in the physical sense. Only if an agent actually smuggles drugs and a law enforcement officer catches him/her and chooses to punish him/her according to a specification of the law, might an adverse consequence to the smuggling agent (e.g. imprisonment) result. However, to take a bribe and dismiss the smuggler may also be in the choice set of the law enforcement officer. Such setting is useful for making the phenomena of corruption susceptible to institutional analysis. See for example Kingston (1999).

100

MASAHIKO AOKI

may call the set of agents a community. Such a community evolves in domains of exchange as well as in production and use of a commons. A community can generate and uphold the endogenous, self-enforcing rules of action choices through repeated interactions in order to regulate the possible moral hazard behavior of community members, while a group of agents who fail to generate such rules may fail to sustain group coherency as a community. We can refer to an implicit rule endogenously generated within a community and enforced on its members without any third party mechanism as a community norm. The nature of the self-enforcement mechanism can be theoretically captured as an equilibrium of a reputation mechanism. Its basic logic is well known, but is recapitulated here in the context of an exchange domain as a starting point of subsequent discussions. A commodity exchange opportunity is characterized as involving a typical one-shot prisoner's dilemma situation: two agents can each benefit from mutually honest exchange, but if either of them can unilaterally and solely cheat the other, there is an even greater gain for him/her at the expense of the other. If there is no device to control such dishonesty, then exchange that could be potentially beneficial to both parties may not take place. For an example, see Table 4.1. The matrix represents payoff outcomes to two trading partners from a single exchange, depending on the combination of strategies they adopt. H denotes playing Honest and C denotes playing Cheat. The first number in each entry indicates the payoff accruable to the row player (say, the buyer), and the second the payoff accruable to the column player (say, the seller). If they mutually play H, the net gains from honest exchange are Γ, equally divided between the trading partners. If either of the parties unilaterally cheats the other, he/she can derive personal benefits measured by α > Γ/2, while causing external damage to the other party is measured by −β < γ. In this situation, there is a social loss measured by Γ − (α − β) > 0. If both players play C, payoffs to both agents are negative (−γ < 0). Each player has a no-exchange option which will yield zero payoffs to both players. Then, if either trader expects that the other will play C, he/she will opt for no exchange. This is the unique Nash equilibrium when the game is played only once and in isolation. If two particular traders, or a buyer and a seller, meet repeatedly, however, the threat of terminating exchange in the future by each partner in the event of the other's cheating may mutually deter cheating, provided that they do not discount the future benefits from exchange too heavily Table 4.1 The Payoff Consequences of the Exchange Game H C

H Γ/2, Γ/2 α, −β

C −β, α −γ, −γ

COMMUNITY NORMS AND EMBEDDEDNESS

101

(i.e. they are patient), and that the one-time gain from cheating is not too large. For example, let us assume that the above exchange game is repeated every week and that both players discount one unit of utility available after one week by a discount factor δ. Suppose that each party adopts the following contingent strategy: play H as long as the other party also plays H; if the other party plays C this week, switch to no-trade strategy forever. In order to see if such strategy profiles constitute a Nash equilibrium, let us examine the effect of unilaterally deviating from this strategy for one week. Suppose that a trader plays C, while the other trader plays H. The current net gain for the former to play C is α − Γ/2, while he/she suffers from zero payoffs forever from the next period on so that the present value sum of the costs of the deviation will be [δ + δ2 + ··· + δt + ···] × Γ/2 = δ/(1 − δ) × Γ/2. Therefore, if agents are sufficiently patient (i.e. discount factor δ is large) and the one-time gain from cheating α is not too large relative to gains from honest trading Γ/2 so that δ > (α − Γ/2)/α, it will never be beneficial for the trader to play C. This is the case of a permanent no-trade punishment. The described strategy imposes a very high cost (as high as the cost for the punished) on the punishing agent himself even if strategy C is adopted by mistake by the trading partner. Therefore, let us consider the following alternative strategy profile. Suppose that the trader who played C once will be punished for the consecutive T weeks by the trading partner playing C. If the former keeps playing H during this period, then he/she will be forgiven and mutually honest trading will resume at that time. If at any time during the punishment phase the punished ever plays C again, then a new punishment phase starts. Suppose that one trading partner plays C in one week while the other plays H. If the other party sticks to the prescribed punishment strategy, then the cheater will not only lose gains from trade Γ/2 but will also suffer from penalty cost β from the next period on for T periods. The present value sum of the costs of the deviation is δ(1 − δT)(Γ/2 + β)/(1 − δ). If δ and T are large enough so that this sum is larger than the one-time benefit from deviation, α − Γ/2, then it will never be beneficial for the trader to play C. On the other hand, it is necessary for T not to be so large that it is incentive compatible for the defector to accept the punishment by playing H while being retaliated against by his opponent. This condition is given by:

These two conditions are simultaneously satisfied for some positive T, if δ > (α − Γ/2)/α as before. On the other hand, if a trader plays strategy C, the best response of his/her trading partner will be to retaliate for T* weeks, where T* is the maximum T that satisfies the above conditions (assuming that only pure strategies are allowed). When mutually beneficial exchanges take place every week under this bilateral reputation mechanism, traders'

102

MASAHIKO AOKI

mutual beliefs that cheating will never occur, because it will credibly invite costly retaliation by the other partner, are referred to as ‘personal trusts’. Let us now imagine that a particular group of traders gather to exchange goods at a local market that opens every week. Each trader is matched with another trader each week and they play the same trading game, referred to as a stage game specified in Table 4.1. If two traders can meet every week repeatedly then their personal trust can support mutual cooperation. However, suppose that traders randomly change their trading partners every week. Suppose that they play the honest strategy as long as their trading experiences have been satisfactory, but that if a cheating incident occurs in the local market place, then the news spreads very quickly and all traders stop going to the local market forever (no exchange option). Then the incentives of each player are identical with those in the case of a permanent non-exchange solution in the two-person repeated trading game. However, this solution is extreme in that any defection from honest trading by a single trader, even if by mistake, will lead to the complete closure of the entire local market and impose a very high cost on everybody. But if cheaters can be correctly identified and made known to every trader in the local market, then the limited punishment strategy described before can be selectively applied to a cheater (Kandori 1992: Proposition 2). That is, suppose that when traders meet randomly they can identify through rumor whether the other party has ever cheated, and if that is indeed the case they can then refuse to exchange honestly with the partner and demand that the partner play H (if any other player cheats while a player is being punished, the latter can be forgiven and only the most recent cheater is punished). This selective punishment requires more information about cheating (it is not sufficient for every trader to know that cheating had occurred but necessary to know who cheated). However, its merit is that much less cost is imposed on honest traders for punishing cheaters when cheating occurs, while cheaters are potentially exposed to the same degree of punishment as under the mechanism of personal trust. It is tempting to refer to this autonomous mechanism as the traders' community norm. However, how does the group of traders gain possession of information about cheaters? In the above abstract construct, there is a tension between the assumption of random matching (impersonal exchange) and that of perfect information dissemination. Is there any way to reconcile these seemingly inconsistent conditions in a situation where no centralized information clearing house has emerged?122

122

A familiar solution to this problem is the introduction of a centralized information clearing house, such as the Law Merchants (Milgrom, North, and Weingast), the credit bureau (Fafchamps in this volume), and the state as a third-party legal enforcement mechanism.

COMMUNITY NORMS AND EMBEDDEDNESS

103

4.2 Clientship and the Endogenous Formation of a Trader's Community For innocent observers, the bazaar appears to be a chaotic exchange domain, warranting the assumption of random matching and lacking a centralized information disseminating agent. How then can the temptation for dishonest trading by agents be regulated if not perfectly? The acute economic anthropologist, Clifford Geertz, once observed a bazaar in Morocco and recognized that stable clientship ties between buyers and sellers constituted one of the most important elements of a bazaar's institutional structure.123 He noted: Clientalization is the tendency, marked in Sefrou, for repetitive purchasers of particular goods and services to establish continuing relationships with particular purveyors of them, rather than search widely through the market at each occasion of need. The apparent Brownian motion of randomly colliding bazaaris conceals a resilient pattern of informal personal connections (Geertz 1978: 30). Let us now consider how clientship can evolve in exchange domains where impersonal exchanges (random matching) have initially prevailed. Although motivated by the question raised above, the purpose of the following exercise is not to reproduce the historical process of clientalization as such, but to understand the basic logic involved in the ways in which norms evolve among an endogenously formed group of traders and can help them respond to system-level information deficiencies in an enlarged exchange domain. Let us assume that the market (stage) game in the bazaar (now the generic name for impersonal exchange domains) is represented by an unstructured bundle of exchange opportunities providing the same payoff structures as represented by Table 4.1. The set of agents is now composed of many traders who come from different localities and are randomly paired. They do not know the past history of their potential trading partners who they happen to meet at the bazaar and cannot be sure whether they can trust these strangers. However, we assume that matched traders can meet again in the future if they so choose. In this sense the bazaar is not an anonymous market. If the matching continues to be random, the only possible Nash equilibrium could be a no trade option. In the expectation of possible gains from long-term cooperation, traders may exchange words indicating their intention to play H and to continue to play H if the relationship extends into the future. They may also try to infer the intention of their trading partners from languages they use or signals they send, such as appearance, hobbies, habits, accents, names of friends and relatives, etc., the meaning of which may not necessarily be obvious to others. If words and/or types

123

Another important element is bargaining.

104

MASAHIKO AOKI

match, traders may then be tempted to enter actual trading—the matching that Robson refers to as ‘secret handshaking.’ However, there is the risk that some traders could promise to cooperate in this way, but then play C and never meet with the same partner in the future. Let us call them ‘parasites.’ Because of their existence, ‘cheap talk’ with limited language cannot support honest trading. Eventually Nash equilibrium evolves in which every player defects from honest trading. Suppose then that there is a successive invasion of ‘mutants’ who use a new language or signal to circumvent parasites and are engaged in secret handshaking among themselves.124 However, parasites corresponding to the new generation of mutants will soon emerge. There will be no Nash equilibrium in this process in which every player cooperates, although there may be one in which some cooperate and some defect (Robson 1990). Imagine now that traders mutually exchange gifts while handshaking prior to exchange, or sink some relation-building costs (such as time for frequent visits, favors) during initial preliminary trading to substantiate their intention.125 The gift should be costly to the giver, but be of no value to others (e.g. fancy wrapping) or should be consumed quickly (within a week) by the receiver. Otherwise parasites could receive the gifts, then cheat and recycle the gifts to deceive their next trading partners—Carmichael and MacLeod (1997) refer to such players as ‘recycling parasites.’ Suppose that there are players who adopt the following strategy: if their messages coincide with those signaled by a randomly matched trading partner, they exchange gifts or mutually sink relation-building costs, play H with that partner, and, as long as the other party also plays H, stay matched until they die. Otherwise, they separate and seek new matches in the bazaar. Call those players who adopt this strategy ‘clienteles.’ In order for this strategy to be incentive compatible, the following two conditions need to be satisfied:

where G is the cost of gift-giving or relation-building. The first inequality says that the present value of net gains from continued cooperation realized by a one-time gift-giving/relation-building cost this week are strictly greater than those from the parasite strategy, i.e. renewed gift-giving/relation-building

124

Robson describes the requirement of a mutation, whereby inefficient evolutionarily stable equilibrium (ESS) can be destroyed, as follows: ‘[The] mutants must involve more than simply a different choice from the original set of strategies. Indeed, the mutants here entail the possession of a signal, that is an observable characteristic which can be taken to have zero inherent cost.’ Robson makes an analogy with a game among the Harris sparrows which mutually differentiate their behavior against opponents according to the color of plumage.

125

The following mechanism draws on Carmichael and MacLeod (1997).

COMMUNITY NORMS AND EMBEDDEDNESS

105

every week and cheating. The second inequality says that the present value gains from continued cooperation are greater than the cost of one-time gift-giving/relation-building. Combining these two conditions, if it holds that

it is expected that a clientship tie can be sustained to the mutual advantage of traders. Suppose that a small proportion of the population dies every year and is replaced by a new generation. Most of them simply mimic the strategies of their parents, but a very small fraction of them experiment with randomly chosen strategies. For this kind of game, Carmichael and MacLeod (1997) showed that for strategies of ‘clienteles’ with the same sized gift/cost satisfying the above inequalities, there is what they call a weakly evolutionarily stable equilibrium characterized by the following two properties: (1) it is a Nash equilibrium and (2) mutants can never realize a higher utility than ‘clienteles.’ Specifically, neither of the following is better off: a secret group of parasites who use a new message, exchange gifts/sink costs, but defect against everybody else, nor a group of agents who use a new message, do not exchange gifts/sink costs, but cooperate among themselves. When an agent with the clientele strategy meets another with the same strategy, he/she leaves the matching market and stays with the other until one of them dies. Parasites remain in the market forever. The proportion of parasites in the bazaar therefore gets higher, which reduces the returns of being a parasite. The social cost of the gift exchange/relation-building for enhancing an honest market transaction can be low if traders are sufficiently patient. If the discount factor is close to one, G can be very close to α − Γ/2, which is far less than the gains from continued honest exchange. Thus, the clientalization can be considered an efficient response by agents to the information deficiencies of the bazaar.126 It is important to note that the clientalization can emerge as an endogenous outcome in a context where there is no prior social contacts among agents and can be maintained without any external reinforcement, social or legal, once it is formed (in that sense this mechanism should be theoretically distinguished from social embeddedness, discussed below).

126

Consistent with the game-theoretic prediction, Geertz submits: ‘Clientalization represents an actor-level attempt to counteract, and profit from, the system-level deficiencies of the bazaar as a communication network—its structural intricacy and irregularity, the absence of signaling systems and the underdeveloped state of others, and the imprecision, scattering, and uneven distribution of knowledge concerning economic matters of fact—by improving the richness and reliability of information carried over elementary links within it (Geertz 1978: 31).

106

MASAHIKO AOKI

The above mechanism was formulated as emerging in the bazaar as a set of quasi-permanent bilateral relationships between pairs of players who initially exchange the same signals and gifts (alternatively sink costs). However, just as ‘personal trust’ (a bilateral reputation mechanism) can be replicated within a group of traders who recognize each other, it may be conjectured that the above mechanism can also be replicable by a group of players. For example, a group of traders in a population of many players may begin to identify each other by using the same language (signal) and sink initial ‘fees’ to form a club or a relational association. The fees paid should be spent in a form that is useless for outsiders or not individually portable, for example, on ritualistic ceremonies, member certificates, club monuments, etc. After its formation, each club constitutes a separate sub-domain of the initial market game in which club-specific norms can regulate mutual transaction relationships in a manner replicating the community mechanism described in the previous section. Namely, any violation of the standard of behavior (honest trading) within the club can be punished by the termination of club membership. In this way, traders' community norms can operate, not only in domains where its membership is exogenously fixed because of the technological nature of transactions or to geographical confines, but also in domains endogenously created by a particular group of players who can identify each other as possibly trustworthy, exclusive clientele. We call such quasi-community norms operating in an endogenous group of agents club norms. One potential problem with club norms as a market governance institution is the potential loss that may result from inertia. In the above parable, the technological environment (payoffs) of the exchange game are assumed to be constant for indefinite periods of time. Then, it makes sense that trading partners who identify each other as honest partners maintain those relationships until they die. However, suppose new merchants with new products arrive at the bazaars every year so that old relationships quickly lose their value. Nevertheless, ‘mistrust of strangers’ may prevent traders from exploiting the potential gains from such new opportunities. We will come back to a situation like this in Section 4.5 below.

4.3 Linking the Social Exchange Game and the Commons Game So far we have considered the case in which a sufficient amount of surplus is created by honest trading (i.e. Γ − (α − β) is large) and the future discount factor of the agent is sufficiently high so that the autonomous norm in the exchange domain can prevent cheating. However, there may be cases in which such conditions do not hold. For example, consider a commons domain in which the members of the community share the construction, maintenance, and use of public goods (e.g. irrigation system, forestry),

COMMUNITY NORMS AND EMBEDDEDNESS

107

but it is not technologically possible to exclude a member from using them. Then it will be difficult to punish a member who does not bear the costs of their maintenance by denying him/her an access to their benefits. Suppose, however, that the same community members are also repeatedly engaged in social exchanges and derive benefits from them. The present value of future benefits from such exchanges for an agent are considered his/her individual assets. Then, shirking in the commons domain may become controllable from agents' concerns over sustaining such assets. Let us explore this possibility in this section, starting out with one motivating historical example. Example 1. The period from 1600 to 1867 is known as the Tokugawa period in Japan and it is estimated that between the years 1600 and 1721, the population grew from 8–10 million to 26 million, implying a 1.0 percent annual growth. This period can be characterized as the Period of Great Reclamation which realized a 40 percent increase in taxable farming lands. An expansion of such magnitude was largely realized by the conversion of alluvial plains into rice paddy fields in the relatively underdeveloped Eastern regions. In spite of the remarkable expansion of arable farmlands, however, it did not match the growth of the population, and the feasibility of productive reclamation became increasingly difficult. Therefore around the end of the first sub-period, greater effort began to be directed toward improvements in irrigation, an increase in the use of commercial fertilizers (e.g. dried fish), and more intensive use of human labor. With the development of the irrigation system, the dominant form of agricultural production shifted from dry field farming to wet farming. The village cultivating system in the Eastern region was characterized as the ‘scattered strip system’ in which rice paddy fields possessed by individual peasant families for their own cultivation were scattered all over the village farming land and mutually intermeshed. This system presumably evolved partly because of the incremental increase in farmland through the collective efforts of reclamation over many years. Another reason might have been the egalitarian distribution of advantageous/disadvantageous access to the water supply and diversification of the natural risk associated with the location of paddy fields. It is not yet a settled issue (among irrigation historians) when the arrangement of separate supplies of water to the scattered plots (paddy fields) of each family became the prevailing practice. Some irrigation specialists argue that it evolved at the latest in the second half of the Tokugawa period (e.g. Nagata 1971: 34–9). Others take the position that by the time of farm land re-zoning in the Meiji, there was not necessarily a separate water intake point belonging to each paddy field, and irrigation at the time of seedling planting took the form of ‘across-paddy field irrigation’ (tagoshi kansui) over multiple families (Tamaki and Hatade 1975: 234–5; Tamaki et al. 1980: 20). This was a gravity irrigation system in which

108

MASAHIKO AOKI

water drawn from a canal was successively supplied from one paddy field to the next using the natural slope. In any case individual farms, constituted as integrated sets of paddy fields, developed incompletely and slowly in the late Edo period. Regardless of whether across-paddy field irrigation was practiced beyond or limited to individual farmers, there were substantial needs for collective coordination and cooperative work among peasant families in the village community. For example, the maintenance and productive use of the irrigation system required the hard work of removing dirt and weeds regularly from the water channel, cleaning the drainage, keeping equipment in shape, preventing destruction of the system during floods, and actively participating in sometimes violent ‘water-disputes’ (mizuarasoi) with neighboring villages located either upstream or downstream of the water supply system in times of scarcity, etc. Also, aggressive and individualistic claims over water rights needed to be restrained. Coordination in the timing of seedling-planting, as well as drainage for preventing the excessive growth of stems in summer, needed to be made collectively. In the typhoon season just before harvest, flood control required the collective effort of village members on an ad hoc basis. How was the compliance of villagers in the collective control of construction, maintenance, and use of the irrigation system enforced? One consequence of the separation of military-political power from the rural community was the absence of an external enforcement mechanism. Once large scale civil works encompassing multiple villages were completed, the construction, maintenance, and use of the local irrigation system was entrusted to the autonomous control of the village community. But a hidden implication of the technological–ecological characteristics of the irrigation system just described was that there was the potential hazard of free riding among villagers in the collective work, as well as aggressive egocentric behavior exerting severe external diseconomies on others. However, once the peasant's farm land became an integral part of the water supply system, even if he shirked collective maintenance tasks or harmed others by aggressive water use, it would have been technologically difficult to exclude him from using the water. However, in spite of technological non-excludability, there was the credible threat of ostracizing him from participating in other spheres of the social, political and economic life of the village: the practice known as mura hachibu, literally meaning ‘80 percent separation from the village.’ Other villagers could refuse to cooperate with a shirker by denying him mutual aid when necessary (e.g. roof thatching, helping with the sick) and excluding him from participation in social events such as ritualistic eating-together, seasonal festivals, etc.—with the exceptions of fire extinguishing and funeral services. This threat was effective in eliciting a high degree of cooperative effort in constructing, maintaining, and using the irrigation system without the intervention of an external enforcement mechanism.

COMMUNITY NORMS AND EMBEDDEDNESS

109

Let us turn to modeling inspired by the above example. Imagine the village community composed of a number, say N, of homogenous families playing an irrigation game and a community social exchange game simultaneously over infinite periods of time. In the domain of the irrigation game, families have an option to choose an action from the same choice set {Cooperate, Shirk cooperation} in the collective tasks of construction, maintenance, and use of the irrigation system. Suppose the effort cost to each family of cooperation per period in this game is C , and the benefit to i each individual family from using the irrigation system per period is B when everybody cooperates, and B − nd when n i i i families shirk. We assume that

The first inequality implies that there is incentive for each individual to shirk, while the second shows that such shirking will impose external diseconomies on the community as a whole, resulting in net welfare loss. However, we assume that because of a technological reason, it is difficult to punish any shirking family by excluding it from the future use of the irrigation system at a reasonable cost. In the community social exchange game each family can contribute to the production of social goods with some costs C , and enjoy the benefit from the consumption of social goods (perhaps tacit goods measured by utility unit), the s quantity of which is a non-decreasing function of the number of other contributing members, B (n), where n denotes s the number of other families contributing to social goods. We assume that there exists such that B '(n)=0 for all n s satisfying , i.e. there is a saturation point in the productivity of social goods. The community social exchange game is played repeatedly, and at the beginning of each stage game any family can be excluded from participating in the production and consumption of social goods by the other families. Imagine the path of full cooperation in the community social exchange game in which all N families cooperate without any exclusion and see if there is any incentive for a family to deviate from this path when non-cooperative social behavior is punishable by permanent ostracism from community social relationships. If this game is played separately from the irrigation game, the incentive compatibility condition for a family not to shirk cooperation is given by

where δ is the time discount factor of the family. That is, the saving of the cost of effort (the left-hand side of the first inequality) should be less than the present value sum of the sacrifice of future benefits arising from

110

MASAHIKO AOKI

ostracism (the right-hand side). The right-hand side term of the first inequality may be identified as a village family's social capital to be lost by social ostracism.127 We assume that this condition holds since δ is sufficiently large, that is, the family is sufficiently patient. Let z denote slack, defined as z=δB (N)−C , which may be identified as the value of social s s capital in flow terms. Now suppose that the irrigation game is played repeatedly by the families of the same community every spring, while the community social exchange game is played by them every autumn for an indefinite number of years. Village families can coordinate their strategies in both games contingent on the outcomes in the preceding games. Suppose that families adhere to the following strategy combinations, depending on the specified contingencies: (1) each family plays shirk in the irrigation game and do not participate in the community social exchange game if it has played shirk in any previous irrigation game. Otherwise it cooperates in both the irrigation and community games; and (2) families exclude any family, and only that family, who has ever shirked in the irrigation game from enjoying the social goods in all future years. The belief of each family is that all other families have played, and will play in the future, the strategy combination prescribed above except in the case when they actually observe a deviation from that strategy. To show the strategy combination as an equilibrium, we only need to check if any one-time deviation from it is beneficial to any family.128 First, note that if any family has ever shirked in the irrigation game before, cooperating in both games from then on will not improve its future payoffs. On the other hand, if a family has always cooperated in both games before, it is not worth it to shirk in the irrigation game. The benefits of doing so are the saving of labor cost C + C in each of the current and future periods, and the costs are the sacrifice of benefits from cooperation i s δB (N) + d in each of the current and all future periods. Therefore, the incentive constraint for each individual family s i not to shirk is given by C + C < δB (N) + d , or i s s i

127

The concept of ‘social capital’ originates with Coleman. He defines it as follows: ‘I will conceive of these social-structural resources as a capital asset for the individual, that is, as social capital. Social capital is defined by its function. It is not a single entity, but a variety of different entities having two characteristics in common: They all consists of some aspects of social structure, and they facilitate certain actions of individuals who are within the structure …. Unlike other forms of capital, social capital inheres in the structure of relationships between persons and among persons. It is lodged neither in individuals nor in physical implements of production’ (Coleman 1990: 302). There is no explicit reference to a game situation, although he invoked analogous argument elsewhere (op. cit. p. 11). However, there is no explicit contradiction between his and my definitions.

128

We are concerned here with the subgame perfectness of the described strategy profile.

COMMUNITY NORMS AND EMBEDDEDNESS

111

It is clear that even if the incentive compatibility condition in the irrigation game (C < d ) is not satisfied, if that of the i i community social exchange game is satisfied with a sufficiently large slack (z > C − d ), then this inequality can hold. i i Thus, linking the two games relaxes the incentive constraints. To check if the second portion of the strategy combination is in equilibrium, note that there is no gain at all for other families to cooperate in the community social exchange game with shirkers in the irrigation game if the number of shirkers is less than . On the other hand, there is no gain from ostracizing any family who has never shirked either. Therefore the threat of social ostracism is credible and cooperation among at least families can become an equilibrium outcome, even if there is a strong incentive for free-riding in the stand-alone irrigation game. Thus cooperation in the irrigation game can become the norm of behavior at least among families in the village community. The second part of the equilibrium strategy combination represents the rational response of village families toward a family who has shirked cooperation in the irrigation game once such shirking has occurred. Therefore it can also represent rational beliefs that can be shared in a self-sustaining manner among village families regarding what would happen to them if they shirked. Such beliefs can deter village families from actually shirking and under normal circumstances only cooperative behavior among village families can be observed as a norm. We refer to such a standard of cooperative behavior, supported by credible beliefs of the occurrence of ostracism in the event of deviant behavior, as a community norm. Although the construct is simple, the above model has a sufficiently generic structure for us to draw two general implications from it: one about the nature of norms, and the other regarding the mechanism of public goods provision. First, by using the idea of linked games, we have tried to make the notion of a community norm susceptible to economic analysis. In the pre-game theory stage, economists tended to regard norms as more or less irrelevant to economic analysis, or detrimental to the development of efficient markets (a notable exception was Kenneth Arrow). In traditional sociology before the surge of the phenomenological school (P. Berger and T. Luckman), norms were regarded as an a priori social entity waiting to be internalized by individuals through the socialization process rather than as something socially constructed and reconstructed through ‘everyday experiences.’ We have conceptualized a community norm as an endogenous outcome of linked games rather than seeing it as an exogenous constraint (rules of the game) given from outside the social system. We have identified generic conditions for a community norm to evolve and become self-enforcing—such as the accumulation of social capital through social exchange among members of the community and the near-unanimity of interests in sanctioning deviants in the use of the

112

MASAHIKO AOKI

commons. Also, in a specific context, political and social events making community members more or less homogenous to fulfill those conditions needs to be identified (homogeneity issues are discussed by Bardhan and Platteau and Seki in this volume). Doing so may help clarify the role of norms in sustaining a certain economic order but not others, while providing a clue for understanding why other solutions, such as the integration of ownership to internalize externalities (Demsetz), occurred elsewhere. Second, the logical structure of the linked game may suggest one generic way to cope with the so-called ‘free riding’ problem in public goods provision (or the problem of the tragedy of the commons). Seeking self-interest in some domain of economic transactions need not necessarily imply that, for the purpose of fulfilling it, agents are limited, willingly or not, to collect and utilize information available only in that domain. However, that was exactly what economists' arguments of market failure in the provision of public goods assumed. Private agents were supposed to try to free ride on public goods without paying the costs, since they were not technologically excludable from the use of public goods even if they did not bear the cost. As a consequence, the possibility of efficient provision of public goods was sought in the domain of governmental services. However, in that domain there is an equally troublesome problem of government failure. The government may not be responsive to the peoples' real needs because of its inability to collect the necessary information, political bias, the difficulty of offering a social choice that is democratic, informationally efficient and consistent at the same time, or due to administrative inefficiency, etc. In making decisions concerning their own contributions to public goods provision, agents may actually collect information from outside the relevant commons domain and be constrained in their choices by what that information implies. As an example, let us refer to an interesting article by Daniel Klein on the provision of turnpikes in early America (Klein 1990). In late eighteenth to early nineteenth-century America, the construction and maintenance of turnpikes was not managed by governments nor financed by taxes, but commissioned to turnpike companies chartered by the states to issue stock. Turnpikes provided large external benefits to nearby farmers and businesses. However, subscriptions to stock in these companies as a means of paying for roads was not expected to be a profitable investment. Meanwhile, it was costly to exclude ‘free riders’ from using turnpikes partly because of legal restrictions on toll collection. Thus people could have benefited directly and indirectly from the road without buying stock. In spite of this free riding problem, numerous citizens of the towns' communities indeed subscribed to the stock. Why? Klein argues that they did so from a sense of social obligation sustained by ‘negative selective incentives … [They] are particularly effective in closed, homogenous groups. The failure of agents to cooperate will attract attention’ (Klein 1990: 803, emphasis added). Negative selective incentives were

COMMUNITY NORMS AND EMBEDDEDNESS

113

institutionalized above all as social pressures exercised through morally obligatory participation in the town meeting where solicitation for stock subscriptions occurred. Formalizing such a situation might lead to a model which would exhibit a striking isomorphic relationship to our irrigation parable. In both cases, it is a credible threat of sanction, or ‘negative selective incentives,’ exercised in the domain of community social exchange, that regulated the free riding of community members in the domain of the commons (public goods) game. Only, the social exchange domains in North American town communities were more open-ended and internally diverse so that possible sanctions were more subtle in manner, less austere in substance, and more permissive of deviance than those in the closed rural community in Tokugawa Japan. Technically speaking, however, the essence of the analytical results could be isomorphic in that some strategies that are not an equilibrium (and thus not self-enforceable) in an isolated commons domain (say, cooperation in the irrigation domain or the turnpike provision domain) could become profitable strategies for agents when linked to a community social exchange domain. An important general implication could be that social context needs to be taken into consideration when analyzing issues of public goods.

4.4 Social Embeddedness and Social Status We have seen in the previous section that even if the cooperative standard of behavior in a commons game, such as in the irrigation game and turnpike provision game, does not appear to be self-enforceable in isolation, if the same agents are engaged at the same time in a social exchange domain which can produce a sufficiently large amount of social capital for them, it may become enforceable. The same idea can be extended to a linkage between another type of domain, such as trade, organizational and even polity, on the one hand, and a social exchange domain, on the other. Agents may refrain from actions imposing external diseconomies on other agents belonging to the same community from the fear of losing social capital in the event of misconduct. Although not explicitly formulated in game form, what sociologist Mark Granovetter deals with in his seminal paper, subtitled ‘The Problem of Embeddedness’ (1985), is considered to imply such a situation. Thus we call this type of linkage social embeddedness. In that paper, Granovetter criticizes the ‘oversocialized’ conception of man in modern sociology as represented by Talcott Parsons, as well as the ‘undersocialized’ conception of man of neoclassical and transaction cost economics operating in the utilitarian tradition. On the one hand, he argues that systems of values and norms are not a once-and-for-all influence as Parsons appeared to assume, but an ongoing process, continuously constructed and

114

MASAHIKO AOKI

reconstructed through interactions. In other words, the process may be perceived by individuals as exogenously given, but actually be endogenously shaped by them, ‘in part for their own strategic reasons’ (Granovetter 1985: 57). On the other hand, he argues that agents in exchange and organizational domains in the modern society generate trust and discourage malfeasance by being embedded in ‘concrete personal relations and structures (networks).’ Granovetter does not commit himself to the view that the construction and reconstruction of norms can be exclusively explained by strategic motives but rather he emphasizes the need for analyzing the historical uniqueness of the formation and nature of each concrete network. However, some essential aspects of his argument are captured in the notion of linked games between the social exchange and other transaction domains.129 The domain characteristics of linked games in the previous section, both irrigation and social exchange, were assumed to be symmetrical with respect to the agents. In particular, the amount of social capital each agent obtained from following the community norm was assumed to be identical. It is considered that the relative symmetry of the members of the community is in general a facilitating prerequisite for the community norm to operate for the ease of peer monitoring, the effectiveness of standard incentive provision, as well as the equalization of outside option values (this symmetric issue is empirically investigated in the chapter by Bardhan in this volume). However, in some cases the agents in a commons domain may be differentiated by ability and other endowments. Then, an implication of a community norm may imply that ‘each contributes according to his ability, all equally enjoy common goods.’ How can such a norm become self-enforceable? Example 1. A comparative fieldwork study of Japanese fishery communities by Platteau and Seki in a chapter in this volume offers an interesting perspective on this issue. They found variations among the fishermen's groups they studied. The most productive group was characterized by various cooperative norms and practices, such as the active sharing of expertise and information regarding the best fishing locations, collective control over access to fishing space, collective retrieval and repair of lost nets, synchronization of fishing hours and the number of hauls. This group also practiced the equal sharing of net incomes among fishing units. On the other hand, among other groups, work coordination was limited and the sharing of knowledge minimal, and there was no or limited pooling of incomes and input costs. The authors collected evidence to show that the former group intended to, and indeed did, enhance aggregate profits through the reduction of congestion costs and risk of net damage, better

129

For a formalization of the concept of social embeddedness in terms of linked games and examples, see Aoki (2001), chapter 8.

COMMUNITY NORMS AND EMBEDDEDNESS

115

exploit the gains from specialization, as well as have oligopolistic control of sales prices for fish. However, the cooperative norms regarding effort expenditure and knowledge sharing, combined with income and material input cost pooling, may imply that the more able the fishermen were, the less favored they were materialistically in comparison to their counterparts in other, less cooperative groups. Why were the able fishermen willing to comply with such an arrangement instead of defecting from it like their counterparts in another group did some years ago, dissolving the cooperative group? The authors submitted interview evidence supporting the idea that the able fishermen were able to gain higher status and esteem among the cooperative group by making a best effort in contributing to collective incomes. In terms of our jargon, the opportunity costs of cooperation incurred by the more able fishermen were compensated for by having a greater amount of social capital allotted to them. Their study suggests that, given the asymmetry in the characteristics of agents in the commons game (fishery game), there can be multiple equilibria: socially embedded cooperative fishing with income pooling, and competitive fishing based on individuated skill (that may be potentially detrimental to resource conservation, although their interviewees did not see the situation that way with some exceptions). Example 2. A structure of social status differentiation isomorphic to the one observed above can be observed in Japanese shop floor practices where informal individual cooperation is the norm, ranging from the voluntary teaching of green workers by veteran workers and the sharing of skills for doing things not found in manuals, to mutually helping solve problems, such as machine breakdowns, defective output, tool repair, and the like. It is important to note that workers assiduous in such activities are generating a substantial degree of externality. Although there are elements of reciprocity, the disparity among workers in the workshop team in their ability to provide externalities is inevitable. Are they fully compensated by wage differentials reflecting this disparity? The workers at the Japanese factory have been traditionally paid according to ranks unrelated to job classification, and wage differentials across ranks are rather compressed.130 Further, promotion along the rank hierarchy is not a tournament type where only the winner gains all the prizes associated with the highest rank. The general expectation is that everybody will sooner or later reach the highest rank in their lifetime (this is the crux of the so-called ‘seniority wage’), but differentiated by the speed of promotion (the high flyer may be promoted to a foreman with which a different rank hierarchy may be associated). Differentials in the speed of rank promotion, say,

130

See Aoki (1988), chapter 3, for a systemic description of rank hierarchy in the Japanese firm as an incentive device and an analysis of its working.

116

MASAHIKO AOKI

whether promotion to the fourth rank in ten can be achieved by age 35 or not, is of great concern to the workers. It seems doubtful, however, that lifetime income differentials associated with the relatively faster promotion awarded to more able and helpful workers are sufficient to compensate for the complete value of the externalities they are generating, because wage differentials across ranks are more compressed and the speed of promotion is in general only gradual. However, it is apparent that more able workers obtain non-pecuniary satisfaction from the recognition that faster promotion symbolizes, and other workers have high esteem, and sometimes are jealous as well, for those workers. Usually, the individual contributions of workers to collective productivity and other's capacity for development are well recognized within the work group on the shop floor and room for managerial discretion on promotion is severely limited. Whether the described convention will be able to persist or not is currently an issue of great debate in Japan. I personally think that it will not completely disappear, particularly in the traditional work group in the manufacturing industry where fine horizontal coordination is still valuable, although it may be modified to some extent in the direction of rewarding the able workers more favorably in pecuniary terms as well.

4.5 Dynamic Embeddedness We have seen that the mechanism of (social) embeddedness can deter non-cooperative action choices in the commons, exchanges, and organization domains under certain conditions (e.g. the existence of enough social capital, relative homogeneity of agents). The basic nature of equilibrium choices in the social exchange domain may remain relatively stable over time, which may roughly correspond to what is normally referred to as a cultural pattern (note, however, that we do not take it as exogenously given and fixed). In contrast, parameters defining the game form of the economic transactions domain may change at a relatively faster speed. The competence, capacities, and other human assets of agents accumulate or depreciate through learning and demographic change. Their information structure relevant to economic transactions, as well as the range of their economic activities, are also refined and expanded by the development of communications and transportation technology. Production technology, statutory rules, etc. that parameterize the consequence functions of economic transaction domains also change. Even new domains that are characterized by novel sets of actions and consequence functions may be created. Then what will happen to the possible linkages of those economic transaction domains with the social exchange domain? Will the former be emancipated from the latter's governance or does an embedded structure recur, possibly with adaptive modification?

COMMUNITY NORMS AND EMBEDDEDNESS

117

These questions have hardly been answered in a systematic manner in economics. Some chapters in this volume examine these unexplored issues in concrete developmental and historical contexts. Inspired by those enquiries, this section tries to identify a couple of generic mechanisms in which community norms and/or embeddedness may play a role in the process of dynamic institutional transformation. This attempt is still at a conjectural and extremely primitive level, but I will present it to stimulate further dialogue between historical and empirical development studies, on the one hand, and (game-theoretic) studies in institutions, on the other.131

Overlapping Embeddedness In Section 4.3, we built a model of the rural community in which the members were engaged in agricultural production by the collective use of an irrigation system and generated social capital through mutual interaction in the social exchange domain. Such a situation can be found in a relatively closed rural community at the premarket stage of development. Indeed, as noted, the model was inspired by the historical experience of Tokugawa Japan between the seventeenth to mid-nineteenth centuries. However, the rising penetration of extra-community market relationships will lessen the coherency of the social life of village families and accordingly erode the threat of community sanctions by ostracism. Then, are community norms, characterized by such practices as collective cooperation, reciprocity, and sharing, bound to become dysfunctional? Will they need to be replaced by entirely new market mores? Traditional views, whether those of economists (e.g. Hicks) or scholars in other social science disciplines (e.g. Polanyi), have drawn a sharp line between the market economy and pre-modern economy, entertaining such a view. However, the rural community bound by community norms could play a positive role in facilitating the gradual transition of pre-modern rural economies to market economies depending on circumstances. For example, the traditional community norm may constrain the community members' opportunistic behavior vis-à-vis the outsiders or vice versa, which may reduce contract enforcement costs in the initial absence of a third party governance mechanism encompassing the two. Or, the distribution of stocks of knowledge and information channels that had been established among the community members may facilitate the emergence of a new governance mechanism over new types of transactions among the insiders in a way that accommodates the new exchange opportunities. If such cases are indeed possible, the complete destruction of rural communities (as in the case of eighteenth- to nineteenth-century England described in the

131

For a formal definition of the following two characterizations of dynamic embeddedness and their examples, see Aoki (2000) chapter 9.2.

118

MASAHIKO AOKI

chapter by Allen in this volume) may not be a necessary condition for market development. In some cases, the presence of a community norm may even be complementary to the transition to a market economy (Chapter 5 by Hayami and Kawagoe on the development of vegetable trading by peasant families in the Indonesian rural communities with middlemen appears to tell such a story). By exposing themselves to outside exchange opportunities, community members' stakes in the community socialexchange, and thus their social capital within the community, will eventually decline. Thus, the role of community in the process of involvement in outer market relationships may be a transitional one, gradually yielding the primary role of governance to a formal mechanism. Nevertheless, the overlapping of community embeddedness of a traditional transaction domain and an emergent open exchange domain can provide one important source of the pathdependency of institutional evolution. Example 3. When a relatively closed rural community started to produce surplus products of relative homogeneity, yet initially lacked its own marketing competence, the opening of substantial trade with the outside could be initiated through an agreement between an outside merchant and the community as a whole. In the beginning of the second half of the Tokugawa period, it was indeed widely observed that the village headmen (shoya) traded cash crops with specific outside merchants as an agent for village families (Shinpo and Hasegawa 1998: 257). Since village families initially offered more or less homogenous goods (e.g. surplus rice, rape seeds, cotton) they could be substitutes for each other as trading partners to outside merchants. Therefore, if an outside merchant were able to trade with them individually, he could possibly prey upon some families and then switch to other families to gain from further trading. However, under the collective arrangement he was presumably compelled to believe that if he had actually behaved opportunistically, he would be shunned from any further trading in the village and lose the value of his initial relationbuilding cost. This belief could be based on the presumed ability of the village families to punish any family within the village who would defect from boycotting trade with a dishonest merchant.132 On the other hand, the merchant could threaten that if the villagers behaved dishonestly he would terminate trade with them and tell his story to all his fellow members of the merchant guild community in the city. If the future value of trading was assessed as better than a no-trade option by

132

The village headmen emerged as embryonic indigenous merchants by mediating an agreement with outside merchants. However, being cultivating peasants themselves, they were more like agents, rather than rulers, of village communities and their conduct was also under the social control of the village.

COMMUNITY NORMS AND EMBEDDEDNESS

119

the village families, even if individual temptations to cheat vis-à-vis the merchant were high, peer pressure could persuade them not to jeopardize future trading opportunities. For example, if any family delivered defective products to the headman's garden, the defector could easily be spotted by other village members and accused in a manner reminiscent of the community norm.133 Thus, the vesting of social capital with village families may have provided a foundation for them to initiate trade with outside merchants and enforce honest trading on both sides as an equilibrium outcome. As productivity differentials in cash crops or craft production widened and their products became gradually specialized, community cohesion among the villagers started to erode. By then the outside merchants also became more knowledgeable about the traits and capabilities of some individual families in the village. Individual contracting began between those families and outside merchants. Further, more successful entrepreneurial families started organizing intra-village subcontracting relationships with less entrepreneurial peasant families. In order to implement mutually beneficial contracts without the support of effective third party mechanisms, however, mutual trust (reputations) had to develop between the indigenous merchant and peasant subcontractors. As we have seen, theoretically this required two things: the effective threat of terminating relational contracting in the event of a partner's default, and the dissemination of information regarding any contract default throughout the relevant trade domain. Evidence from Japanese experiences toward the end of the nineteenth century indicates that there were multiple indigenous merchants competing in the same locality and it was possible for peasant families to switch putting out contracts among them, if necessary. On the other hand, there were peasant families serving only as buffers against fluctuating demands but who were potentially capable of replacing relational contractors. This situation might have assured that the first condition was satisfied.

133

The collective control of the quality of goods was instrumental for the village families to strengthen their bargaining power vis-à-vis outside merchants and therefore was in their mutual interest. In a situation where the adverse selection problem of product quality was severe, resourceful merchants could have taken advantage of it and claimed that goods supplied to them were of lower quality, even if it was not really the case, to abuse innocent village families. To prevent merchants from adopting such an opportunistic strategy, village families needed to be able to unambiguously control and prove the high quality of the goods they supplied. In the Tokugawa period, peasants' tax obligations were defined and fulfilled in terms of the bulk quantity of paddy rice first and processed brown rice later, rather than using paddy panicles. The ruler's purpose was to leave as little room as possible for peasants to manipulate the delivered bulk quantity by compromising on quality (such as not removing immature grains, straw, or increasing the moisture content). An unintended consequence of this obligation was that they accumulated the ability to control and prove the quality of supplies (Koga 1982, 1997). If trade in rice had been made in terms of paddy panicles as in most monsoon Asian economies, merchants might have been able to adopt the opportunistic strategy just mentioned. Village families continued to cooperate in improving post-harvest processing technology through the collective ownership of equipment and exchange of information, etc. to strengthen their collective bargaining power.

120

MASAHIKO AOKI

Since the merchants had indigenous roots, information regarding any misconduct on either side must have been easily disseminated among them as well as the villagers. Thus the intra-village information infrastructure that supported the preceding community norm might also have served as grounds for the second condition to be fulfilled. However, the way in which contracts were made self-enforceable differed substantially from the traditional community norms; the expected penalty imposed on deviants was exclusion from contractual relationships rather than from the traditional network of social exchanges per se.134

Revoked Embeddedness Suppose that a social exchange domain embeds an economic transaction domain. Subsequently, another economic transaction domain emerges outside the latter, with the partial immigration of agents from the old to the new. Suppose that an autonomous governance mechanism is used as an experiment in this new transaction domain, but it never gets firmly institutionalized or it malfunctions after its initial attempt at enforcement. Meanwhile, the agents in the new transactions domain self-organize a corresponding social exchange domain through their continued participation in the former. In response to a shock or crisis, a mechanism to regulate agents' choices in the new transactions domain may eventually evolve in a manner isomorphic to the old embedding. Since the set of agents of the old social exchange domain and those of the new are not identical, albeit partially overlapped, it is not precisely a transplant of the old community norm. However, the partial overlapping of agents between the two may make the isomorphic embedding a natural focal point. In that sense,

134

Around the 1980s the putting-out system in fabric manufacturing spread widely in Japan and in spite of the parallel development of the factory system it remained as a major constituent of the Japanese textile industry until as late as the mid-1920s. Between 1902–11, the textile industry produced about one quarter of the total industrial output and generated about 50 percent of exports. According to a recent estimate by M. Tanimoto, the proportion of textile workers employed in the putting-out system was more than 70 percent of those in the industry in 1905, and the corresponding figure employed by the factory system was only a little more than 10 percent. Further, the large factory system was more concentrated in the spinning of cotton yarn and production of standard fabrics, while smaller scale factories and producers under putting-out contracts were more specialized in differentiated fabric manufacturing. The contractors were mostly peasant families who kept farming as their major production activity yet allocated their family resources (labor power, time, and attention) among farming of various kinds, contract work, and domestic work, as an integrative decision-making unit. The integration of farming and domestic manufacturing within peasant families was a major feature of the Japanese system, distinct from the ‘proto-industrialization’ observed in eighteenth-century Flanders where grain producing farming and rural-based industry became bifurcated into separately managed activities. This transitory feature induced by the overlapping embeddedness had path-dependent impacts on the ways in which manufacturing industries became organized in later years (see the last section of the chapter by Hayami and Kawagoe).

COMMUNITY NORMS AND EMBEDDEDNESS

121

we can call this dynamic mechanism ‘revoked embeddedness.’ (As for an example, see Chapter 3 by Hayami and Kikuchi in this volume regarding the emergence of the gama harvest arrangement.) Example 4. A claim is often made that the group norm as observed on the shop floor or in the business office of a Japanese firm—or more widely in East Asian economies during the period of the Asian miracle—is a replication of the community norm that originally evolved among peasants collectively using the irrigation system. However, the preceding analysis pointed to one condition crucial for a community norm to evolve and govern the use of the commons under the severe free riding problem. This is the presence of a social-exchange domain that embeds the domain of the economic commons and can potentially generate social capital for participants in both domains. It would be difficult for such conditions to immediately prevail in the emergent industrial economy. It is more likely that the coherence of the social exchange domain would be eroded by the transition from the community based rural economy to the market-based industrial economy. It is also not self-evident how an alternative social-exchange domain with a functional equivalence can be reconstructed around the organizational domains in the context of the modern factory or administrative organization. There was an attempt by the government in the early Meiji period to transplant Westernstyle textile factories and employ the daughters of urban ex-samurai. But their management was a total failure and the factories were quickly privatized by fire sales of the assets to private capitalists. Newly established private factories recruited female workers in groups from the same rural localities and put them in dormitories. Some paternalistic capitalists even provided middle-school education, flower arrangement, and sewing classes to the employed workers at the dormitories to promote a community-like atmosphere. Likewise, industrial factories in emergent heavy industry relied on contractors for the recruitment of workers in groups from the rural sector. This practice reflected the competition from the ruralbased putting-out system and indicated that the residue of social capital held by peasants constrained their massive and individual mobility to the urban industrial sector. In any case, industrial workers, particularly skilled ones, remained scarce even after the turn of the century, and they were able to reap premium wages by responding to competition in recruiting among employers rather than fearing exclusion from yet unformed social capital at the workplace. Lifetime employment and seniority wages as incentive devices for workers to invest in (organization-specific) skills and/or settle with single firms in spite of competitive lures were introduced around the time of the First World War by advanced government-run armories and shipbuilding yards, but were mostly limited to them. It is interesting to note that a

122

MASAHIKO AOKI

major demand of militant workers at privately owned factories in the Keihin industrial district in the late 1920s was to be recognized as ‘an equal member of the company’ together with white collar workers rather than to assert their class interests like contemporary European workers (Gordon 1985). Yet their demands were repressed by authoritarian management with the help of police intervention. Labor immigration from the rural to the modern industrial sector started to rise, both from a supply push and demand pull, in the 1930s when the rural economy suffered from a prolonged severe depression, and the government started to intervene in every aspect of the economy to accelerate heavy industrialization. In the 1940s when the war economy deteriorated, the mobilization of the workforce into the industrial sector, in quantity as well as incentive-wise, became one of the keen issues the government faced. Further, the emergent problems on the shop floor, such as machine breakdowns, material and labor shortages, wide-spread absenteeism of the workers, etc., needed to be dealt with by ad hoc, collective adaptation on site. However, there was a shortage of skilled workers and in spite of wage control over new hires their mobility remained high. To control these problems, the wartime government tried to place ‘social pressure’ on the workers through formation of an enterprise branch of the Industrial Patriotic Society. This organization encompassed blue-collar as well as white-collar workers, and tried to reduce long-standing status differentiation among them. In the midst of scarcity and the need to support the families of drafted workers, the enterprise branch of the Society evolved to partially function as a mutual aid organization. It also became increasingly difficult for the workers to voluntarily quit war production factories because of intensified monitoring of the government to curb mobility. ‘Non collaborators’ in production were threatened with exclusion from the benefits of mutual aid (e.g. distribution of rationed products like food and cloth) or punished morally and physically (labeled as a ‘non-patriot,’ victimized in military drills, etc.). This may appear reminiscent of the social ostracization in the Tokugawa village. However, the workers at this time were coerced into participating in the punishment of ‘non-collaborators’ out of fear that they themselves might be pointed out as non-compliant by the military representative in the establishment and punished accordingly. The situation may be more appropriately characterized as ‘government-led, quasicommunity control’ rather than autonomous community control. An unexpected twist of history is that in spite of the ordeal of military repression, the mobilization of workers in the industrial sector during the war ushered in a new era in retrospect. Partly embedded in quasi-community control and partly because of the traditional value placed on craft skills, the practice of collective problem-solving started to spontaneously evolve on the shop floor, where an exercise of the foreman's authority that was

COMMUNITY NORMS AND EMBEDDEDNESS

123

not based on knowledge and skills was perceived to be illegitimate, and relatively skilled workers enhanced moral authority by assuming problem-solving leadership. A fundamental change in the power structure on the shop floor thus began to take place under the shock of war. The Patriotic Society was dissolved after the war as a part of post-war Democratization, but its all inclusive organizational experiences at the factory level provided an organizational basis for fast growing independent unionization. When the Japanese management tried to learn the techniques of American management, such as quality control, time motion studies, etc., to make up the existing productivity gaps during and after the Korean war boom, they had to face the changed power structure on the shop floor, although some workplaces had gone through divisive union rivalries in the interim period. Thus the management was constrained to implement such methods with the participation of the workers as a group rather than by imposing the analytical results of industrial engineers on workers. Efficient production and inventory control methods, which would later become widely known as the kanban system, evolved from this type of worker participation. The quasi-community control during the war may thus have had a limited, unintended impact on subsequent development by making the firm organization a focal point of social exchange on a mass basis. However, post-war political democratization and mobility of workers across factories during the chaotic post-war period made the sustenance of the government-led quasi-community control obviously impossible. Instead, small workgroup norms evolved in the workplace during the high growth period through workers' voluntary long-term associations. The provision of incentives for the workers to identify their interests with those of the employing firm was consciously pursued by the management during this period. However, the formation of the social-exchange domain at the workplace by itself was not necessarily efficiency-enhancing from an over-all organizational perspective, because it could become fragmentary, cliquey-forming, and collusion-prone. Reliance on information and effort sharing as an organizational coordination device required the development of a corresponding, organization-wide governance mechanism. Elsewhere, I have systematically analyzed complementarity relationships within a cluster of Japanese institutions supporting high post-war growth, which contained small group work norms, the bank's involvement in corporate governance contingent on relatively poor organizational performance (the contingent governance), relational banking, and the political– economy framework for bureaucratic interest mediation through industrial associations and other interest groups (bureaupluralism), among possible others (see Aoki 1992, 1995). When such a system is exposed to dramatic changes in the environment and some elements of it fail to adapt, other elements are also placed in a strained situation. Institutional dynamics then enter a juncture point of punctuated equilibria (Aoki 2001, chs. 9, 11).

124

MASAHIKO AOKI

In my judgement, this is the present state of the Japanese economy. The somewhat loose notion of the Japanese firm as a community is thus being tested. When the Japanese system undergoes the process of adaptation, norms governing organizational and other activities are likely to be modified, but in a path-dependent way. This subject matter is far beyond the scope of this paper, however.

4.6 Concluding Remarks In this chapter we began by examining the theoretical conditions under which an autonomous norm can regulate honest trading within a given community without a third-party contract enforcement mechanism. Then we discussed a case in which clientship ties and endogenous clubs of traders could self-organize to control the opportunism of traders in an initially anonymous exchange domain. Then we turned to cases in which, even though an autonomous selfregulatory mechanism might not be viable in a stand-alone economic transactions domain, if they were embedded by a social exchange domain of the community in which significant amounts of social capital were sunk by its members, external diseconomies and opportunism in the former domain could be regulated. Finally we discussed the possible roles of social embeddedness in institutional dynamics. A basic presumption of this chapter was to regard norms (and other institutions) as endogenously formed rules of the game rather than exogenously given rules. An institutional change could then occur only as a qualitative change in the equilibrium patterns of the strategic interplay of agents. As noted in the last section, norms supported by mechanisms of social embeddedness are most robust to environmental change and have the strongest inertia among institutional arrangements. Thus, it may have the most important path-dependent impact on the historical path of institutional evolution. This process is still only partially understood by economists as well as other social scientists. However, it is hoped that the approach attempted in this chapter for understanding norms (informal institutions) as equilibrium phenomena will stimulate further comparative, historical, and analytical enquiries into the process.

References Aoki, M. (1988) Information, Incentives, and Bargaining in the Japanese Economy, Cambridge University Press. —— (1992) ‘The Japanese Firm as a System of Attributes: A Survey and Research Agenda,’ in M. Aoki and R. Dore (eds.), The Japanese Firm: Sources of Strength, New York, Oxford University Press, pp. 11–40.

COMMUNITY NORMS AND EMBEDDEDNESS

125

Aoki, M. (1996) ‘Unintended Fit: Organizational Evolution and Government Design of Institutions in Japan,’ in Aoki, M. Kim, H., and Okuno-Fujiwara, M. (eds.), The Role of Government in East Asian Economic Development, Oxford, Oxford University Press, pp. 233–53. —— (2001) Toward a Comparative Institutional Analysis, MIT Press, forthcoming. Arrow, K. J. (1974) The Limits of Organization, New York, W.W. Norton & Company. Beger, P. L. and Luckmann, T. (1966) The Social Construction of Reality, New York, Bantam Doubleday Dell Publishing Group, Inc. Carmichael, H. L. and MacLeod, W. B. (1997) ‘Gift Giving and the Evolution of Cooperation,’ International Economic Review, 38: 485–509. Coleman, J. (1990) Foundations of Social Theory, Cambridge, MA, Harvard University Press. Demsetz, H. (1967) ‘Towards a Theory of Property Rights,’ American Economic Review, 57: 347–59. Geertz, C. (1978) ‘The Bazaar Economy: Information and Search in Peasant Marketing,’ American Economic Review, 68(2): 29–32. Gordon, A. (1985) The Evolution of Labor Relations in Japan: Heavy Industry, 1853–1955, Cambridge, MA, Harvard University Press. Granovetter, M. (1985) ‘Economic Action and Social Structure: The Problem of Embeddedness,’ American Journal of Sociology, 91: 480–510. Hicks, J. (1969) The Theory of Economic History, Oxford, Oxford University Press. Kandori, M. (1992) ‘Social Norms and Community Enforcement,’ Review of Economic Studies, 59: 63–80. Kingston, C. (1999). ‘On Inter-Society Adoption of Institutions: Informal Networks and Market Corruption in India,’ Stanford University, mimeo. Klein, D. B. (1990) ‘The Voluntary Provision of Public Goods? The Turnpike Companies of Early America,’ Economic Inquiry, 28: 788–812. Koga, Y. (1982) ‘Problems in the Post-harvest Processing of Rice in Southeast Asian Countries,’ in Scrimshaw, N. S. et al. (eds.), Nutrition Policy Implementation—Issues and Experience, New York, Plenum Press, pp. 385–417. Nagata, K. (1971) Nihon no Suiri Kozo (The Structure of Irrigation in Japanese Agriculture), Iwanami Publishers. Polanyi, M. (1944) The Great Transformation, Rinehart. Robson, A. (1990) ‘Efficiency in Evolutionary Games: Darwin, Nash and the Secret Handshake,’ Journal of Theoretical Biology, 144: 379–96. Shinpo, H. and Hasegawa, A. (1988) ‘Syouhin Seisan, Ryuutu no Dainamikkusu (Dynamics of Commodity Production and Distribution)’ in Hayami, S. and Miyamoto, M. (eds.), Keizai Shakai no Seiritsu: 17–19 Seiki (The Establishment of the Economic Society: 17-18 Centuries), Japanese Economic History, I, Iwanami Publishers. Tamaki, A. and Hatade, I. (1975) Fudo: Taich to Ningen no Rekishi (Cultured Landscape: A History of Land and HumanBeing), Heibonsha. Tanimoto, M. (1998) Nihon ni okeru Zairaiteki Keizai Hatten to Orimonogyo (Indigenous Economic Development and the Textile Industry in Japan), University of Nagoya Press.

This page intentionally left blank

Part II Community in Market Development

This page intentionally left blank

5 Middlemen in a Peasant Community: Vegetable Marketing in 135

Yujiro Hayami and Toshihiko Kawagoe A major controversy surrounds the question of how peasants are affected by the penetration of a market into hitherto isolated village communities. In a deep-rooted popular belief ever since Thomas More's Utopia, peasants have been considered to be passive to market forces and to have neither the desire nor the ability to participate in commercial and industrial activities. Traders' usury and monopsony has widely been believed. It has been assumed that middlemen involving trading activities belong to a social class alienated from the peasantry. This view found a typical expression in Clifford Geertz's anthropological study in Indonesia. Based on observations in East Java and Bali, he concluded that the entrepreneurship for non-farm business activities that may induce social modernization cannot emerge from ‘the immediate purview of village social structure’ but is limited to the population with ‘extra-village status,’ such as ethnic Arab traders in the East Javanese town and traditional rulers in the Balinese town (Geertz 1963: 148–9). The Geertz thesis has found a support from the theory of ‘traders’ dilemma' recently elaborated by anthropologists. This theory argues that the social norms of peasant communities urging well-to-do villagers to redistribute their income and wealth to poor neighbors prevent village-born traders from accumulating capital and thereby advancing to modern commercial entrepreneurs (Evers and Schrader 1994). However, what we observed in a field study of agricultural product marketing in Indonesia was contrary to the image developed by Geertz. In both long-settled Java and newly-opened Sumatra, indigenous entrepreneurs belonging to the upper peasantry actively engaged in the marketing and processing business. It does not appear that they were socially compelled to remain as petty traders. In fact, it was not rare to find these village-based traders overtaking town-based ethnic Chinese traders who had dominated local trades in Indonesia since the colonial era. Their strength, as compared with the ethnic Chinese, is their ability to enforce contracts with farmers and lower-rank collectors by means of common norms and personal ties within village community. This paper tries to illustrate how the

135

This paper draws heavily on Hayami et al. (1991) and Hayami and Kawagoe (1993: ch. 4). The authors also thank Masahiko Aoki, Alain de Janvry, Mirga De Silva, Gershon Feder, Douglass C. North, Jean-Philippe Platteau, and Julie Schaffner for their comments.

130

YUJIRO HAYAMI AND TOSHIHIKO KAWAGOE

peasant entrepreneurs develop their trade by taking advantage of the village community relationships, with respect to the case of marketing vegetables from an upland village in West Java to the metropolitan market of Jakarta. The field study on which this paper is based was conducted in 1990. We first conducted a survey of farmers in the study village to identify how much of their products was sold and to whom. Thereafter, we traced the marketing chain through a sequence of marketing agents to end-users, noting prices, transportation and processing costs, trade practices and contracts. When inconsistency was found in data and explanations among transacting parties, we visited them again for verification. This ‘pedestrian approach’ based on our own observations and hearings gathered through walking around rural villages and towns is indispensable for grasping the elusive behaviors and organizations of informal marketing agents in developing economies.

5.1 Commercial Vegetable Production in an Upland Village Our study village is located in the district of Majalengka in West Java, about 300 kilometers and 100 kilometers east of Jakarta and Bandung, respectively (Fig. 5.1). It is a peasant village in which more than 80 percent Fig. 5.1 Location Of the Study Village Relative To Market Centers

MIDDLEMEN IN A PEASANT COMMUNITY IN INDONESIA

131

of households stake out a living from small-scale family farming with an average operational size of only about 0.4 hectare, either as owner farmers or as tenants, whereas very few make a living in waged employment in both farm and non-farm activities. No large farm or plantation exists which is operated with wage laborers under the command of a manager–foremen hierarchy. The village lies on a moderately hilly plateau 70 to 100 meters above sea level. About a half of its farm land consists of rain-fed rice terraces and a half consists of dry upland fields. Agriculturally, the village represents a rather unfavorable or marginal area as compared with irrigated rice areas in flat lowland plains. The typical farming systems traditionally used in this village were as shown in Fig. 5.2: (a) a mixed cropping of traditional upland crops in upland fields and (b) a sequence of rice followed by soybean in rainfed paddy fields terraced along hill slopes. The first system inter-cropped corn and upland rice for the first crop season (October–January), followed by soybean for the second crop season (FebruaryMay), while cassava was grown in the edge of farm plots. This farming system was commonly practiced in rain-fed farming areas in Java. The third system (c) of rice Fig. 5.2 Traditional Cropping Systems

132

YUJIRO HAYAMI AND TOSHIHIKO KAWAGOE

double cropping was practiced in irrigated paddy fields located outside the boundary of this village. These farming systems underwent major change owing to the introduction of commercial vegetable production since the mid-1970s. Before then, indigenous low-quality vegetables were grown in gardens mainly for home consumption. The introduction of vegetables of marketable quality was initiated by a villager who used to be a vendor of kitchen utensils. Through his traveling sales, he observed commercial vegetable growing in some other areas. With seeds brought home, he himself tried to grow vegetables and, also, taught neighbors how to grow them. At the same time, he began acting as a collector to market the produced vegetables. The new farming systems that became common after the introduction of commercial vegetables are shown in Fig. 5.3. In upland fields the mixed cropping of upland rice and traditional upland crops (system a) was replaced by the combination of various vegetables (system A). In addition Fig. 5.3 New Cropping Systems

MIDDLEMEN IN A PEASANT COMMUNITY IN INDONESIA

133

to the cucumber, eggplant, and string beans shown in the figure, many other vegetables such as bitter gourd and chili were also produced. They were grown in various combinations. For example, cucumber, string beans, and chili are often inter-cropped. In many cases, traditional crops such as corn and cassava were also planted underneath vegetable plants so that they could grow tall after harvesting of the vegetables. In rain-fed paddy fields, the rotation of rice and soybean (system b) was replaced by the rotation of rice and vegetables (system B). A remarkable development is that some irrigated paddy fields which used to be left fallow during the third rice-crop season (June–September) because of insufficient water supply, were leased by vegetable growers for this dry season alone. A significant area of irrigated paddy fields was leased from other villages to vegetable growers in this rain-fed village. Through this seasonal leasing of irrigated paddy fields for vegetable production, system c in lowland villages surrounding the study village was replaced by system C. The successful introduction of commercial vegetable production was a major innovation, resulting in as much as about a three-fold increase in the value of farm output in this village (Hayami and Kawagoe 1993: 159).

5.2 Organization of Vegetable Marketing A key to the success of the innovation in farm-level production was an innovation in marketing. Vegetables produced from the new farming system are characterized by high degrees of commercialization with almost 100 percent of output in this village being sold, whereas the ratios of market sales of traditional crops were less than 10 percent for rice and 40 percent for corn, for example. Another distinguishing feature of vegetables is high perishability. Because of the need to distribute the large volume of perishable commodities from producers to consumers, a different organization of marketing was required for vegetables from that of storable subsistence crops such as rice and corn.

Marketing of Traditional Subsistence Crops Before identifying the unique nature of commercial vegetable marketing, it is necessary to understand the marketing organization for traditional subsistence crops in this and other peasant villages in Indonesia. According to our field surveys over Java and Sumatra (Hayami et al.1987, 1989), a common characteristic in the marketing of peasants' products is the division of labor between small traders who live in villages and collect farm products little-by-little from neighboring farmers, and larger

134

YUJIRO HAYAMI AND TOSHIHIKO KAWAGOE

traders who engage in the shipment to other districts of the commodities collected by the small traders. We call the former ‘village collectors’ and the latter ‘inter-village collectors.’ It is common that the inter-village collectors advance credit to the village collectors to ensure delivery of the collected commodities for transshipment. The hierarchical division of labor stems from (1) peasant producers' small marketable surpluses, (2) scale economies in transportation, (3) differences in labor opportunity costs and (4) differences in financial positions. Large traders need to collect commodities into a sufficiently large lot to exploit the scale economies for long-distance transportation of bulk commodities such as corn and soybean. For example, the transportation cost per ton of corn with a truck is smaller by one half when the full truck load of corn is collected than when only one-half load is collected. It is economical for them to let small traders having lower labor opportunity costs collect from small farmers in small amounts. Indeed, small collectors are themselves small farmers or agricultural laborers who trade these commodities in their spare time. On the other hand, large traders own real assets such as land and buildings that can be used as collateral. Therefore, their credit cost is substantially lower than for small collectors with no collateral. Usually, interest is not explicitly charged on the trade credit. Also, the practice is not used of reducing purchase prices by large traders from small collectors for advanced credits. The interest foregone is considered a premium for the assured delivery of collected goods that enables large traders to schedule shipment in large lots. This flow of trade credit is limited mainly within the circle of village-based traders. Bazaar (Pasar) traders in town, who are mostly ethnic Chinese, seldom advance credit to village-based traders because they do not have a common community tie to enforce credit repayment, while the cost of legal enforcement through a court is prohibitively high relative to small values of transactions in the countryside. By taking advantage of the close community tie with farmers and small collectors, large village-based traders who are relatively large land-owners are able to compete with townbased traders in the interregional transshipment of locally-produced farm commodities. Despite the hierarchical organization among traders tied by trade credit, there is no sign of monopolistic pricing. Price differences among various points in marketing chains are explained largely by transportation costs. There are many options open for farmers to sell their products, including direct delivery to rural based agro-processing factories and direct sale at town bazaars. However, there is a strong tendency for each farmer to continue to sell his product to the same collector. This fixity in trade relations does not reflect any monopsony power exercised by collectors. Instead, the terms and conditions offered by alternative buyers are very similar so that it is advantageous for farmers to save transaction costs by trading with one

MIDDLEMEN IN A PEASANT COMMUNITY IN INDONESIA

135

buyer continuously. In general, farmers have good knowledge of prices and other conditions prevailing in the market. Under the competitive market traders try to increase profit by enlarging their trade volume and maintaining the thin but positive margin through the exploitation of scale economies in transportation and inter-regional price differences.

Two Channels of Vegetable Marketing How different is the marketing organization for commercial vegetables from that of traditional upland crops such as soybean and corn as described above? A distinct feature of vegetable marketing in this village and surroundings is the clear separation of the channel for local consumption from the channel for shipment to other regions, as shown in Fig. 5.4. In the case of corn and soybean, a part of commodities assembled by village collectors is sold to inter-village collectors or bazaar traders who transship them to other regions, but a part is sold by village collectors directly to local processors such as feed mills and cottage manufactures of tempe (fermented soybean cake) and tofu (soybean curd). Also, bazaar traders engage in both local retail and interregional transshipment business. Fig. 5.4 Marketing Channels For Vegetables

136

YUJIRO HAYAMI AND TOSHIHIKO KAWAGOE

In contrast, vegetables destined for metropolitan markets are assembled by village collectors (pengepak), all of which are transshipped through inter-village collectors. Normally, no vegetable is diverted from this channel to local consumption. Unlike corn and soybean, bazaar traders, especially ethnic Chinese traders holding permanent shops (toko), play no role in the business of transshipping vegetables. In the marketing of vegetables for local town consumers the most important agent is what we call the ‘bazaar vendor’ (pedagang kaki lima). Bazaar vendors are mostly women in villages, who collect vegetables in quantities of about 100 to 200 kg from farmers in their villages or surroundings and bring the cargoes by minibus to the bazaar for sale in open space. Buyers from the bazaar vendors are housewives as well as peddlers and keepers of small grocery stores (warung) in town. Retailing of vegetables in the bazaar is also carried out in green groceries with roofed stalls (kios). These stall shops buy their supply from an informal wholesale market to which both farmers and small collectors bring a few bags of vegetables by bicycle, motorcycle, and sometimes small truck. These small collectors (pingepit) are distinguished from village collectors (pengepak) who assemble vegetables for inter-village collectors. According to the estimation of village collectors, about 70 percent of vegetables produced in this village are assembled by them for inter-village collectors, 10 percent to the informal wholesale market by small collectors, and 20 percent carried by bazaar vendors to local towns. Inter-village collectors play a key role in the shipment of locally-produced vegetables to metropolitan markets. A typical operation is that an inter-village collector charters a truck and sends it around village collectors' houses for collecting the full load of vegetables in bags and, then, lets the truck proceed directly to wholesale markets in Jakarta. The truck carrying the morning's harvests reaches either Pasar Induk, the major vegetable (and fruit) wholesale market in the southern part of Jakarta City, or Pasar Cibitung located in the eastern outskirt of the Jakarta metropolitan area. Upon arrival at the market, cargo on the truck is first received by a consignee. By contract, the consignee unloads the cargo and sells it in bags (typically weighing 50 kg) to ‘bazaar resalers’ (centeng or tukang tembak). No formal auction is made but, since many buyers compete with each other at the same place, the procedure is de facto bidding. For this operation the consignee receives a certain commission customarily agreed on and gives the rest of the proceeds to an agent of the inter-village collector, who rides with the cargo on the truck. Some bazaar resalers who buy bags of vegetables from the consignee are based within the same wholesale market. In stall or open space they take out vegetables from the bags and sort them into various grades for sale to retailers in smaller lots. Other resalers bring the purchased cargoes back to the bazaars of their home base where they make similar resale operations.

MIDDLEMEN IN A PEASANT COMMUNITY IN INDONESIA

137

Thus, vegetables produced by small peasant producers, which are bulked into lager lots for interregional transport through a hierarchy of village to inter-village collectors, are de-bulked at the destination into smaller lots from consignees to bazaar resalers and further to retailers before reaching urban consumers.

5.3 Peasant Entrepreneurs in Interregional Trade A unique aspect of vegetable marketing is identified as separation of the channel for local consumption from that of interregional trade. This characteristic is explained by the high perishability of vegetables. A vital consideration is how to shorten the time of marketing so as to deliver vegetables to consumers while being fresh. This requirement seems to explain the dominance of the bazaar vendor in vegetable marketing for local consumption, who links producers in her home village directly with consumers in a nearby town. Low time costs for the village women whose alternative employment opportunity is hired farm work, as well as relatively easy access to market from village in rural Java, should also explain the bazaar vendor's domination.

Conditions for the Interregional Trades of Vegetables This short circuit is not effective, however, for delivering fresh vegetables to consumers in distant markets. The cost for a vendor to travel nearly 300 kilometers to Jakarta is prohibitive relative to the value of the few bags of vegetables she can carry on the bus. In order to exploit strong scale economies inherent in long-distance shipment of bulk commodities, one full truck load of vegetables must be collected and shipped out within the day of harvest. This contrasts the case of non-perishable commodities such as corn and soybean for which a collector can wait several days or weeks before one truck load is assembled. Moreover, an occasional collection and shipment of one truck load is not a viable operation. In the case of nonperishable commodities like soybean, an inter-village collector typically brings his cargo himself with a truck to a metropolitan market center where he identifies the best buyer after several searches and bargains. The metropolitan wholesaler who accumulates his stock in this manner sells it in small lots to retailers over an extended period. In the case of vegetables, the cargo must be de-bulked immediately when it arrives in the market. The trader cannot afford to waste time for search and for a bargain. The system of vegetable distribution in the Jakarta wholesale markets consisting of consignees and resalers, as described in the previous section, is considered an institution consistent with this need. A key to the working of this system is the conscientious efforts of the consignee to handle the consignor's cargo. The consignee is

138

YUJIRO HAYAMI AND TOSHIHIKO KAWAGOE

not a large formal agent. He receives cargoes from only a few consignors. He receives and sells the cargoes in open space or in roofed floor leased from the bazaar office. Entry and exit to the business are entirely free. In the absence of the official license to consignee as well as of formal rules of auctioning, the consignee's conscientious efforts can only be secured by mutual trust to be established through long-term continuous transactions. Therefore, a local trader in the producing area must send his truck load regularly to a consignee in one wholesale market. Furthermore, more than one truck load is needed for the trader to reduce risk. As is common with perishable commodities, the price of vegetables in one wholesale market fluctuates widely corresponding to variations in the amount of the cargoes received. Even in Jakarta, prices differ day by day between Pasar Induk and Pasar Cibitung. Therefore, in order to reduce risk, it is necessary for the trader to diversify the destinations of his cargoes between the two wholesale markets of Jakarta as well as between Jakarta and other metropolitan cities. Because of both high risk and high transaction costs associated with marketing of perishable commodities, together with scale economies in transportation, the optimum size of vegetable collection and shipment becomes large. This large collection and shipment must be done regularly within the day of harvest. For this purpose, harvesting, collection and shipment must be very tightly coordinated, so that by the evening of everyday an inter-village collector can accurately estimate the quantity of vegetables farmers will deliver to their agent (village collector) for determining the right number of trucks for him to charter. Specialization of inter-village collectors in interregional trades for vegetables is considered a response to the strong demand for entrepreneurial and managerial ability to perform this coordinating role. An example of such an entrepreneur is a large inter-village collector who, during the peak harvesting season, collects as much as 60 tons or 12 truck loads of vegetables through some 20 village collectors. He ships these cargoes to both Pasar Induk and Pasar Cibitung as well as other cities such as Semarang and Yogyakarta. He, as well as his competitors, is a native Sundanese living in a village and solidly rooted in the village community. His consignees in the metropolis are also native Indonesian. Contrary to the implication of Geertz's study in East Java and Bali, this vegetable marketing case in West Java demonstrates that entrepreneurship for large business activities in Indonesia is not limited to the population of ‘extra-village status’ but can also emerge from ‘the immediate purview of village social structure’.

Credit-Tying Contract A device for the inter-village collector to promote more exact scheduling and coordination for assembling vegetables into a large bulk for shipment

MIDDLEMEN IN A PEASANT COMMUNITY IN INDONESIA

139

to the metropolis is to tie village collectors by advancing credits in return for their assurance to deliver their collected cargoes exclusively to him. This credit tying of lower-ranking traders by upper-ranking traders is not uncommon for assembling storable commodities such as corn and soybean, too. We frequently observed that inter-village collectors advance short-term credits to village or hamlet collectors for payments to farm producers for the purchase of the storable commodities (Hayami et al.1987, 1989). However, no case of credit tying of farmers by traders was found for these commodities. A unique aspect of vegetable marketing is that the credit tying is practiced between collectors and farmers as a device for the former to secure delivery of products from the latter. Typically, a village collector advances to vegetable growers credits in kind in the form of fertilizers and chemicals in the planting season with the guarantee that their harvest will be delivered to him. Although no written agreement is exchanged, a common clause in this contract is that 10 rupiah per kg will be charged to the growers if their products are sold to other collectors. Farmers' repayments for the credits are deducted from proceeds of vegetables delivered to the collector over the harvest season. The repayment schedule is flexible but, on the average, farmers are supposed to pay back the credits in about two months. The tying of farmers for vegetable collection, which is not practiced for other commodities, reflects the need for better coordination and more exact scheduling in the marketing of perishable commodities rather than farmers' need for fertilizers and chemicals. The credit tying was not observed for tobacco, for which farmers' credit need for fertilizers and chemicals is as high as for vegetables. While farmers are tied by village collectors, village collectors are also tied by credits from inter-village collectors. In order to finance credits in kind to farmers, village collectors usually receive cash credits from an inter-village collector under the same terms of conditions as for the contract that they have with farmers. For both collectors' credits to farmers and inter-village collectors' credits to village collectors, interests are not explicitly charged. The trade credits from inter-village to village collectors for the purchase of storable commodities such as soybean are very short term, typically less than a week, so that interest charges are not significant. It is an anomaly, however, if interest is not charged for the production credits extended for as long as two months in the economies characterized by severe capital scarcity and high interest rates. One possible device for collectors to charge the interests is to pay lower prices to credit-receiving farmers than to non-credit receivers. In our vegetable case, this price difference occurs very rarely, only a few days in a season, when supply is especially short relative to demand so that some inter-village collectors need to offer higher prices for marginal purchase from non-contracting farmers and village collectors for filling up their trucks.

140

YUJIRO HAYAMI AND TOSHIHIKO KAWAGOE

A more significant device is the use of differential prices of fertilizer and chemicals between farmers and traders. Collectors can purchase these inputs in a large lot by advancing their credits in kind to many farmers. Their purchase price is lower than the farmers' purchase price in a small lot. For example, collectors buy urea in the large lots of several tons from fertilizer dealers in town at the cost of 185 rupiah per kg including transportation costs, and charge to farmers at 200 rupiah which is the price that farmers have to pay if they are to buy one bag (150 kg) or two at village grocery stores. As estimated in Table 5.1, the average cost per farm of current inputs advanced as credit in kind would total to Rp (Rupiah) 70,500 per bata (0.18 ha) if they themselves would buy in cash, whereas the same inputs could be purchased by collectors at the cost of Rp65,550. In the credit tying operation, collectors charge to farmers Rp70,750 for these inputs. If this credit is paid back in two months, collectors earn, in effect, interest at the rate of 3.9 percent per month (58 percent per year). While this is a lucrative credit operation for collectors, it is also advantageous for farmers. The input cost in farmers' own cash purchase (Rp70,500), compared with their payment to collectors' credit in kind (Rp70,750), implies that the effective interest rate for farmers is as low as 0.2 percent per month, which is considered not significantly different from zero. This rate is much lower than the interest rates that they should pay if the farmers purchased the inputs on credit from fertilizer dealers Table 5.1 Credit Cost for Vegetable Producers Under Alternative Credit Arrangements in the Majalengka District, 1990 Input cost per farma (Rp) Cash purchase: Farmer (in small lot) Collector (in large lot) Credit purchase: Collectors' trade credit Fertilizer dealers' sale on credit Bank loan

Effective interest rate for Farmer (%/monthb)

Collector (%/monthb)

70,500 65,550 70,750 73,250

0.2 (2.2) 1.9 (25.8)

75,950

3.8 (56.3)

3.9 (58.1)

Source: Hayami and Kawagoe (1993: 129).

b

Notes: a Cost for 150 kg of urea, 50 kg of triple super-phosphate, 100 kg of ammonium sulphate, and one liter of Azodrin per 125 bata (0.18 ha). For the calculation procedures, see Hayami and Kawagoe (1993: Appendix B). Interest rates per year are shown in parentheses.

MIDDLEMEN IN A PEASANT COMMUNITY IN INDONESIA

141

(1.9 percent) or if they purchased them based on institutional loans from the government bank (3.8 percent). The official interest rate of collateral loans from the government bank (Bank Rakyat Indonesia) is said to be as low as 1.5 percent per month (20 percent per year) but high transaction costs, including the costs of paper work and travel to town for credit application, raise the effective rate for farmers to a large extent as high as 3.8 percent per month (Hayami et al.1991: 61–3). Thus, this credit-tying contract achieves a Pareto improvement as it benefits both creditor and debtor. Of course, the lucrative credit operation of the village collector to farmers is not really so lucrative unless he can receive an interest-free loan from the inter-village collector. If he has to mobilize the fund himself for credit-tying contracts with farmers, his credit costs may barely match his interest earnings. Then, how is interest charged on credits advanced from inter-village collectors to village collectors? The credit cost for inter-village collectors is much lower than for village collectors, also, partly because inter-village collectors have better collateral and partly because the size of their loan is much larger so that the unit financial transaction cost is much lower. Further, credit terms are, in effect, shorter for inter-village collectors than for village collectors. Usually one inter-village collector has the credit-tying contracts with many village collectors in different areas with different environmental conditions for which planting and harvesting seasons differ widely. Therefore, the inter-village collector can recoup credits from village collectors in various areas one by one continuously as he deducts credit repayments from the proceeds of vegetable cargoes delivered to him over various seasons in a year. From the credit repayment that he recoups from one village collector in the area in the harvesting season, he can finance credit to another village collector in the area in the planting season. In this way, long-term credits can be transformed into short-term revolving credits with a significant saving of interest. Even though the credit cost is thus reduced for inter-village collectors, it should still be positive and significant. It appears that they would shoulder this cost for the profit from better coordination between farm production and marketing of the perishable commodities for distant markets.

Community Mechanism of Contract Enforcement It is important to recognize that this rather complicated and elusive credit-tying contract is purely self-enforced, involving no formal rule or inter-mediation of formal or semi-formal organizations such as village assembly and officials or ‘vegetable growers’ associations' (which simply did not exist in this study site). We gather, nevertheless, that contract violations, i.e. farmers' selling harvested vegetables to other than the collectors who advanced them credits without paying the penalty of 10 rupiah per kg, are

142

YUJIRO HAYAMI AND TOSHIHIKO KAWAGOE

very few. Why would they not dare to do so even when it is apparently not difficult to conceal their moral hazards from the collectors' eyes. To this question, the farmers' answer was either that they are ashamed of doing so or that they fear a bad reputation which might arise from gossiping by some neighbors who happen to see their moral hazards. Both the shame and the fear of social opprobrium imply the existence of a social norm to penalize a villager who may cheat a fellow villager. Furthermore, an enforcement mechanism based on the social norm is enhanced especially when villagers and collectors belong to the same natural hamlet, as often observed in the study site. In a hamlet, comprised of 100 to 200 households, community members are tightly bound in locational affinity and kinship. Most spontaneous collective actions are organized within a hamlet while avoiding free rider problems (Kawagoe et al.1992). Obviously, a strong enforcement mechanism is working in this small community. In the presence of such a norm the expected cost of contract violation should be rather large, since not only may the violator lose benefits from the present contract but the resulting bad reputation may deprive him of future opportunities to other contracts with other villagers as well (Hayami and Kikuchi 1981: 11–23). This community mechanism of contract enforcement binds not only farmers but also collectors. Although not explicitly specified, it is assumed that farmers receive ‘fair prices’ for their products, which are conceptualized as collectors' market sale prices minus standard margins. This means that the market renders a standard of fairness to members in the village community—a perspective opposite to the theory of traders' dilemma (Evers and Schrader 1994). If collectors would significantly violate this implicit contract clause to pay market-based fair prices to farmers, they would risk being socially sanctioned. In fact, a collector who once operated in the study village (though he himself was not a villager but his second wife was from this village) was cut off supply from farmers as he developed the reputation of paying ‘unfair prices’. A de facto embargo was organized by farmers based on both the social norm and the dense information flows within a village. It is important to recognize that the competitive market made this sanction more credible. A vegetable farmer usually continues selling his product at prices offered by a collector with whom he has the credit-tying contract. This does not mean, however, the existence of monopsonistic power on the side of a middleman. If the offered prices are low relative to market prices, the farmer can easily shift to another collector in the next season. To our questionnaire, nearly 100 percent of farmers in the study village replied that they can potentially sell to more than three traders who operate in the village even though 85 percent of them continued to sell to the same traders without bothering to check alternative offers (Table 5.2). Also, more than 70 percent of our respondents replied that they can sell

143

MIDDLEMEN IN A PEASANT COMMUNITY IN INDONESIA

Table 5.2 Trade Practice of Farmers and Number of Collectors in the Majalengka Village, 1989/90 Number of collectors 1 2 21 3

1 Distribution of sample farmers by the number of traders that the farmers contacted before selling their products. 1 2 Distribution of sample farmers by the number of traders to whom the farmers consider they can possibly sell their products.

0

3 plus 1

Total 25

24

25

Source: Hayami and Kawagoe (1993: 131).

their products directly to neighboring town markets by themselves. They are well aware of prevailing market prices through informal communications with neighbors and friends. In such a situation it is difficult for a collector to exercise monopsonistic pricing. The village community norm and the competitive market structure reinforce each other to operate a strong countervailing power against middlemen's monopsony exercises. In this village community the ‘exit’ (individual market response) and the ‘voice’ (collective political action) in the Hirschman (1970) sense work together inseparably as a countervailing force against middlemen's malpractice or violation of implicit contracts. Farmers' ‘exit’ from a particular middleman is not their individual action but is a kind of collective action typically organized by the ‘voice’ in the form of gossiping among villagers. The mechanism of developing and enforcing contracts between farmers and middlemen for Pareto improvement, as described above, depends on the social structure of this village. Typical of a peasant village in Java, this village consists predominantly of small-scale farms based on family labor with the average operational holding of only about 0.4 hectares. While significant inequality exists among villagers in both land ownership and cultivation with the Gini coefficients of 0.57 and 0.58, respectively, no one owns and cultivates more than two hectares or 5 percent of village farm land. As such, no one has overwhelming economic power to coerce other villagers socially and politically. Even the richest farmer can hardly subject the poorest to servility, as commonly practiced in some rural areas dominated by large estate farms or plantations (Newby 1979; Schaffner 1995). Indeed, no permanent or attached labor contract is observed in the village, though casual daily-wage employment is very common among villagers. Also, unlike villages in South Asia, especially India, the community is not socially segmented by castes. Villagers in this Javanese village are homogenous socially if not economically.

144

YUJIRO HAYAMI AND TOSHIHIKO KAWAGOE

Since no well-to-do villager has power to coerce over the poorest, he has to purchase the goodwill and cooperation of other villagers to promote his business, protecting his properties and developing his social prestige and political position. Thus, he must search for mutually beneficial or Pareto-improving arrangements with his neighbors and rely on the enforcement of the agreements on the norms of the community, if he wants to promote his own profit. Such a social structure, which is rather typical of peasant villages in southeast Asia, especially Java, seems to underlie the efficient working of vegetable marketing or the absence of exploitation by middlemen, as observed in this village. In addition, the efficiency is critically supported by the availability of exit options or easy access for farmers to alternative marketing outlets based on a relatively well-developed transportation infrastructure which is common in Java, characterized by high population density. It is quite possible that corresponding to a similar development in commercial vegetable production a highly exploitative marketing system for small growers and/or laborers will emerge in an environment which is isolated from the market and where a large estate farm predominates.

Price Setting and Risk Sharing While medium-term production credits flow from inter-village collectors, via village collectors, down to farmers, shortterm trade credits seem to flow in the opposite direction. It is common for a village collector to pay in cash to farmers who bring their fresh harvests to his house (which farmers call the ‘depot’) in the morning. Then, an inter-village collector sends around his chartered truck with an agent and a couple of coolies to these depots for loading vegetables. How many vegetables are expected to be collected in these depots is informed to the inter-village collector by the previous evening. As the truck is fully loaded, it immediately proceeds to a wholesale market in the metropolis so as to reach there by late afternoon of the same day. Payments to the village collectors are usually deferred until next morning when the inter-village collector's agent brings back the proceeds of sales from the metropolitan markets. Sometimes, though not so often, the payments are deferred for several days. In effect, short-term trade credits flow from village to inter-village collectors. However, net flows of credits are hard to ascertain, since village collectors' credit repayments through deductions from the proceeds of sale to inter-village collectors are often postponed upon the request of village collectors. First of all, no fixed schedule of repayments is specified in advance, although a broad schedule is customarily determined. As explained in the previous section, cargoes brought to a metropolitan wholesale market are sold immediately by a consignee in de facto auctioning.

MIDDLEMEN IN A PEASANT COMMUNITY IN INDONESIA

145

The cargoes are sold out within an hour to 5 hours. Normally, the consignee's commission is 20 rupiah per kg for ordinary vegetables (except high valued ones such as small chili), but the commission increases (or decreases) step-wise as the sale price increases (or decreases) with the effect that his commission remains largely at 15 percent of the auction sale price. To the extent that the commission is a percentage of the market price, the consignee shares a part of the risk in market price fluctuations. Compared with the consignees, bazaar resalers who redistribute the auctioned cargoes to urban retailers bear a much higher risk as they pay in cash for the auction purchase before their sale prices and timing are determined. As a result, both price risk and ‘commodity risk’ defined as loss due to spoilage in the process of marketing are high to the resalers. The same applies to retailers. The prices that inter-village collectors pay to village collectors next morning seem to be determined essentially by the wholesale prices inJakarta. This relation is illustrated by weekly cucumber prices recorded by an inter-village collector, as plotted in Fig. 5.5. The correlation coefficient Fig. 5.5 Seasonal Fluctuation Of Cucumber Prices and Marketing Margin For Inter-Village Collectors

146

YUJIRO HAYAMI AND TOSHIHIKO KAWAGOE

is as high as 0.906 between the Jakarta wholesale price received by this trader and his payment to village collectors. A tendency is indicated that the inter-village collector's trade margin increases absolutely as the Jakarta price increases with the correlation coefficient of 0.651, while no significant correlation is observed between the Jakarta price and the margin rate (ratio of absolute margin to the Jakarta price) with the correlation coefficient of −0.08. These statistics are consistent with the hypothesis that, similar to his consignee in the Jakarta market, the inter-village collector shares price risk within the limit of a fixed percentage (or a fixed range of percentage) of market price fluctuations. However, the ways by which the inter-village collector passes price fluctuations in Jakarta to the local market in Majalengka are asymmetrical in the sense that the passing of price changes in Jakarta prices to local market is fast when the prices are falling, whereas the prices are transmitted with a significant time lag when the prices are rising. This relation is indicated by the following regression:

where M is the inter-village collector's trade margin; P is the wholesale price in Jakarta in the current week; PI is the rate of change in the Jakarta price from the previous to the current week when the price rises, but it is zero when the price does not rise; PD is the rate of change in the Jakarta price when the price decreases, but it is zero when the price does not decrease; student t-ratios are shown in brackets; DW is Durbin–Watson statistics; and adj. R2 is the coefficient of determination adjusted for the degree of freedom. The regression is estimated by the ordinary least-squares method based on 42 weekly observations from 16 October 1989 to 6 August 1990. The coefficients of P and PI are positive and significant at the 1 and 5 percent levels, respectively, whereas the coefficient of PD is not statistically significant at conventional levels. This result implies that, while the trade margin is largely a proportional share of the Jakarta price, it increases when the price is rising but does not change when the price is falling. These findings indicate that the inter-village collector increases his margin by delaying the transfer of price rise in the metropolis to the local market, while he maintains his margin through the immediate transfer of price fall. Such behavior of collectors is consistent with the findings of Alexander (1987) with respect to chili marketing. This practice seems to reflect a monopsonistic power in the hand of the inter-village collector owing to his faster access to price information in the metropolitan market. With this power he can pass a large portion of price

MIDDLEMEN IN A PEASANT COMMUNITY IN INDONESIA

147

risk to village collectors. But, this monopsonistic power is transitory and will be lost quickly as information on the price change will be made available to the local market through competition among inter-village collectors. Yet, it is likely that the high risk associated with the long-distance trade of perishable commodities is reduced, to a significant extent, by inter-village collectors' faster access to price information in the national market centre.

Village Community Tie versus Ethnic Community Tie A peculiar aspect is that town-based Chinese traders who are dominant is both the local and the inter-regional trades of other peasant crops play no role in vegetable marketing. Not only the local marketing was handled mainly by native villagers' wives but also the inter-regional trades of vegetables was managed exclusively by inter-village collectors who mostly come from the upper peasantry of the village community. The reason is the advantage of the village-based entrepreneurs to utilize the community mechanism in enforcing contracts with peasants on the delivery of their harvests at agreed-upon quantities and timing, which is vital for the success of vegetable marketing, as discussed before. Town-based traders, especially ethnic Chinese traders, have no such community mechanism of contract enforcement. It is, therefore, not surprising to find them seldom entering into the long-term contracts with village collectors and farmers on commodity delivery and credit advancement. Their purchase is usually on the spot; they sit in their shops and, as village collectors or farmers bring cargoes in, they examine the commodities, offer prices, pay in cash if the customers agree and otherwise let them go, with little bargaining. Farmers tend to admit that these ethnic Chinese are trustworthy in the sense that they are largely accurate in offering the prices prevailing in the market—no favors given but no cheating. Their retail operations with local customers is similar except that they sell on credit to some tofu factories operated by members of the same ethnic group. However, ethnic Chinese traders have a major advantage in interregional arbitrating since the ethnic tie binding them over different regions reduces the costs of information collection and contract enforcement. On the other hand, village-based indigenous Indonesian traders have an advantage in commodity assemblage through their community tie with farmers and low-level collectors. This advantage of village-based traders is especially large for perishable commodities like vegetables, for which closer coordination between farm-level collection and bulk shipment to other regions is required. The credit-tying contracts which are easily susceptible to moral hazard and opportunism cannot be enforced without strong community ties in the absence of effective third-party enforcement mechanisms typical of the rural sector of developing economies.

148

YUJIRO HAYAMI AND TOSHIHIKO KAWAGOE

It must be cautioned that, even if the village community tie may be effective in reducing the cost of enforcing contracts among villagers, it is far from being so effective as to eliminate the cost. The fact that the contract enforcement cost is significant is illustrated by the observation that the trade-credit inter-linked contract is extended to farmers only for vegetables, the returns on which are especially high. Expected gains from the contracts for storable commodities such as corn and soybean should be too small to pay for the cost. The lack of inter-ethnic competition and the stronger contract binding in the case of vegetable marketing might partly explain inter-village collectors' occasional earning of excess profits by delaying the transmission of price increases in metropolitan markets to village collectors and farmers. However, the exercise of this monopolistic practice is limited to a very short period. If an inter-village collector continued to use it, he would have to face the problem of his village collectors shifting their supply to other inter-village collectors in the next crop season. The situation is similar to the relationship between the parent company and subcontractors in Japan under strong competition among corporate groups.

5.4 Peasant Entrepreneurs in Modernization The striking finding in this study is active entrepreneurship for commerce and industry rooted in Indonesian peasant communities. Contrary to Geertz's (1963) pessimism, significantly large commercial and industrial enterprises were promoted by indigenous entrepreneurs belonging to the upper peasantry. Not only did their activities emerge out of but also were not confined to the ‘immediate purview of village social structure’. In the highly complicated and largescale operation of commercial vegetable marketing for the metropolis, peasant entrepreneurs became dominant as they properly exploited the advantage of their close community tie with village collectors and farm producers. In fact, this is only one of many examples showing active entrepreneurship among peasants in Indonesia (Hayami and Kawagoe 1993). It does not appear that the redistributive norm of peasant communities represents a particularly serious barrier against capital accumulation and business expansion by village-based traders, unlike the prediction of the traders' dilemma theory (Evers and Schrader 1994). The active entrepreneurship that we observed among peasants in Indonesia resemble those of ‘rural capitalists’ emerged from the gono (wealthy peasant) class, who supported Japan's emergence as a modern industrial state from the late Tokugawa to the Meiji period. In the initial stage of modern economic growth in late nineteenth-century Japan, wealthy peasants called gono, who typically cultivate a part of their land by themselves and rent out the other, actively participated in trading and

MIDDLEMEN IN A PEASANT COMMUNITY IN INDONESIA

149

manufacturing. This process had begun already in the feudal Tokugawa period in response to the gradual development of the market economy, and it accelerated with Japan's opening to foreign trade and national unification by the Meiji Restoration in 1868 (Smith 1956, 1959, 1988). Contributions of these small-scale, rural-based enterprises to the national economy were no less significant than those of large modern corporations developed by urban entrepreneurs who emerged from the pre-industrial merchant class (such as Mitsui and Sumitomo) and the ex-warrior class (such as Yataro Iwasaki, the founder of Mitsubishi). In fact, rural and national economic developments until the turn of the century or even later until about World War I are considered to have been supported, to a large extent, by the commercial and industrial activities of those local entrepreneurs (Rosovsky and Ohkawa 1961; Tussing 1966). This development of rural-based enterprises prepared a situation in modern Japan whereby industrialization could be less capital intensive than was the case for other latecomers to modern economic growth (Smith 1988, chapter 1). Moreover, the peasantry continued to be a major source of supply of entrepreneurs and managers to modern business even after World War I, as evidenced by the data that as many as 40 percent of presidents of the top 154 companies in Japan in 1956 came from peasant families (Smith 1960). Similar developments in commerce and industry by peasant entrepreneurs have now been widely observed in southeast Asian economies (Hayami 1998). Their active entrepreneurship is considered an important support to the ‘economic miracle’ in this region for the past three decades. The question arises as to what will be their fate in the event of successful economic modernization. Their marketing organization is a decentralized hierarchy of many self-employed informal agents tied by customary trade practices and informal contracts, where vertical integration is typically absent. At the present stage of economic development in Indonesia, the system is highly efficient in economizing the use of scarce capital and management input while making intensive use of labor which has a low opportunity cost. This system's efficiency depends, to a large extent, on the dualistic structure characterized by differentials in wage and interest rates across firm sizes. Will this system become inefficient and be replaced by vertically-integrated large corporations when the economy advances to a stage at which the factor of market dualism will be eliminated? The experience of the Japanese economy in the past three decades may shed some light on this question. Within a decade since the beginning of so-called ‘high economic growth’ (kodo keizai seicho) in the mid-1950s, the dualistic structure was largely eliminated (Minami 1973). Yet, small-scale family enterprises have survived (Kiyonari 1980; Patrick and Rohlen 1987). Many of them, especially in retail trade, have survived because of government protection and subsidies. At the same time, there have been many cases in which they maintained and strengthened their niches by

150

YUJIRO HAYAMI AND TOSHIHIKO KAWAGOE

exploiting the advantages of small- and medium-scale enterprises, such as high incentives for the self-employed and flexibility in employment and staffing. This advantage increased to the extent that the need for small-lot production of differentiated products increased, as the Japanese economy has advanced to a stage characterized by high per-capita income and people's diversified desires. Large corporations in Japan tend to prefer subcontracting via vertical integration; coordination between a parent company and small/medium firms (shitauke) has been developed to such a level of precision that the subcontracting system is integrated inseparably in Toyota's famous Kanban (just-in-time) system. In this system the subcontractors deliver their parts and materials to the assemblers just in time, i.e. the parts are used, so that no inventory accumulates in the assembling plant even for small lot-production of many differentiated products (Abegglen and Stalk 1985; Asanuma 1985, 1988). By nature, this system is very similar to the organization of vegetable marketing in Indonesia that we have studied; i.e. precise coordination is developed between the collection of vegetables by village collectors from farmers and the transshipment by an inter-village collector so as to leave no room for inventory to accumulate. This coordination is supported by mutual trust among those sharing the same community norms, similar to the relationship between an automobile assembler and its part suppliers in Japan. Indeed, building up of the efficient subcontracting system by Japanese automobile manufactures through the application of traditional community relationships to modern industrial organization and management can be traced through the history of Toyota since its foundation in 1933. According to Kazuo Wada (1998), Kiichiro Toyota, the founder of the Toyota Motor Company, intended to build ‘a pastoral factory’ in Koromo (today's Toyota City) which was then a typical rural town with few manufacturing activities. Kiichiro's idea was not only to locate the factory in a pastoral environment but also to build a production organization in which Toyota's assembly plant would be surrounded by trustworthy parts suppliers tied together by a common community spirit. To this end, Toyota tried as much as possible to enter contracts with manufacturers located near the Koromo factory, who were receptive to Toyota's guidance, even if they had no previous experience in the sophisticated precision work needed for automobile production. It appears that Kiichiro's idea of the ‘pastoral factory’ stemmed from his knowledge of the applicability of community relationship to industrial production, which he learned presumably through experience as a supplier of automatic looms for rural-based weaving entrepreneurs before he began automobile manufacturing. It appears that the traditional community relationship thus became a guiding principle for a modern big business such as Toyota. It represents an example of drawing from tradition an appropriate organizational form to accommodate modern production relationships in a new economic environment (Hayami 1998).

MIDDLEMEN IN A PEASANT COMMUNITY IN INDONESIA

151

Is there any strong reason to doubt whether entrepreneurial and managerial skills now being learned in the countryside of Indonesia through the marketing of peasants' products will provide the basis of modern subcontracting systems and, thereby, support the advancement of the Indonesian economy to a higher level of development? Are there insuperable cultural and social barriers for peasant entrepreneurs in Indonesia to follow the development of Japanese counterparts? If the emergence of peasant entrepreneurs in Indonesia followed the same pattern as in the early stage of development in Japan, is there any reason to suspect them to be incapable of becoming a major carrier of modern commerce and industry when Indonesia will advance to a regime of higher economic development? Does not the same question apply to other developing economies? It was T. W. Schultz (1964) who established a perspective that peasants in developing economies are not traditionbound, irrational creatures but are highly responsive to economic opportunities and capable of carrying out modern agriculture. This study of peasants' marketing activities in Indonesia today, as compared with the historical peasants' marketing activities in Indonesia today, as compared with the historical experience in Japan, suggests the possibility that they can also be a basis of modern commerce and industry with their effective use of community norms and relationships. We should recognize that the mechanism of development is often latent in apparent backwardness.

References Abegglen, J. C. and Stalk, G., Jr. (1985) Kaisha: The Japanese Corporation. New York, Basic Books. Alexander, J. (1987) Trade, Traders and Trading in Rural Java, Singapore, Oxford University Press. Asanuma, B. (1985) ‘Organization of Parts Purchases in the Japanese Automobile Industry’, Japanese Economic Studies, 13(1): 32–53. Asanuma, B. (1988) ‘Manufacturer–Supplier Relationships in Japan in the Concept of Relation-Specific Skill’, Journal of the Japanese and International Economies, 3(1): 1–30. Evers, H-D. and Schrader, H. (eds.) (1994) The Moral Economy of Trade: Ethnicity and Developing Markets, London, Routledge. Geertz, C. (1963) Peddlers and Princes, Chicago, The University of Chicago Press. Hayami, Y. (1998) ‘Norms and Rationality in the Evolution of Economic Systems: A View from Asian Village’, Japanese Economic Review, 49(1): 36–53. Hayami, Y. (ed.) (1998) Toward the Rural-Based Development of Commerce and Industry: Selected Experiences from East Asia, Washington, DC, World Bank. Hayami, Y. and Kawagoe, T. (1993) The Agrarian Origins of Commerce and Industry: A Study of Peasants Marketing in Indonesia, New York, St. Martin's Press.

152

YUJIRO HAYAMI AND TOSHIHIKO KAWAGOE

Hayami, Y., Kawagoe, T., Morooka, Y., and Siregar, Masdjidin (1987) Agricultural Marketing and Processing in Upland Java: A Perspective from a Sunda Village, Bogor, Indonesia, ESCAP-CGPRT Centre. Hayami, Y., Kawagoe, T., Morooka, Y., Mayrowani, H., and Syukur, M. (1989) Agricultural Marketing in a Transmigration Area in Sumatra, Bogor, Indonesia, ESCAP-CGPRT Centre. Hayami, Y., Kawagoe, T., Yokoyama, S., Bagyo, Al Sri and Zakaria, A. K. (1991) Marketing Innovation for Vegetables: Conditions of Diversification in Upland Farming, Bogor, Indonesia, ESCAP-CGPRT Centre. Hayami, Y. and Kikuchi, M. (1981) Asian Village Economy at the Crossroads, Tokyo, University of Tokyo Press; Baltimore, Johns Hopkins University Press. Hirschman, A. O. (1970) Exit, Voice and Royalty: Responses to Decline in Firms, Organizations and States, Cambridge, MA, Harvard University Press. Kawagoe, T., Ohkama, K., and Bagyo, A. S. (1992) ‘Collective Actions and Rural Organizations in a Peasant Economy in Indonesia’, Developing Economies, 30(3): 215–35. Kiyonari, T. (1980) Chusho Kigyo Dokuhon (Textbook of Small and Medium-Scale Industries), Tokyo, Toyokeizaishimposha. Minami, R. (1973) The Turning Point in Economic Development: The Japanese Experience, Tokyo, Kinokuniya. Newby, H. (1979) The Differential Worker: A Study of Farm Workers in East Anglia, Madison, University of Wisconsin Press. Patrick, H. T. and Rohlen, T. P. (1987) ‘Small-scale Family Enterprises’, in Yamamura, K. and Yasuba, Y. (eds.), The Political Economy of Japan, Vol. 1. The Domestic Transformation, Stanford, Stanford University Press, pp. 331–84. Rosovsky, H. and Ohkawa, K. (1961) ‘The Indigenous Components in the Modern Japanese Economy’, Economic Development and Cultural Change, 9(3): 476–501. Schaffner, J. A. (1995) ‘Attached Farm Labor, Limited Horizons and Servility,’ Journal of Development Economics, 47(2): 241–70. Schultz, T. W. (1964) Transforming Traditional Agriculture, New Haven, Yale University Press. Smith, T. C. (1956) ‘Landlords and Rural Capitalists in the Modernization of Japan’, Journal of Economic History, 16(2): 165–8. —— (1959) The Agrarian Origins of Modern Japan, Stanford, CA, Stanford University Press. —— (1960) ‘Landlords’ Sons in the Business Elites', Economic Development and Cultural Change, 9(1): 93–107. —— (1988) Native Sources of Japanese Industrialization, 1750–1920, Berkeley and Los Angeles, University of California Press. Tussing, A. (1966) ‘The Labour Force in Meiji Economic Growth: A Quantitative Study of Yamanashi Prefecture’, Journal of Economic History, 26(1): 59–92. Wada, K. (1998) ‘The Formation of Toyota's Relationship with Suppliers: A Modern Application of Community Mechanism.’ in Hayami, Y. (ed.), Toward the Rural-Based Development of Commerce and Industry: Selected Experiences from East Asia, Washington, DC, World Bank.

6 Market Integrators for Rural-based Industrialization: The Case of the Hand-Weaving Industry in 136

Akihiko Ohno

6.1 Introduction Great expectations are placed on the development of rural household non-farm industries in developing countries, which are more labor intensive and make fewer demands on scarce capital resources than modern large-scale counterparts. The isolation of a local market from national and international markets is often claimed as a major impediment for the development of rural industries in developing countries. The questions then arise as to who integrates local and urban markets and under what contractual arrangements. Though market integration generally involves the alignment of insiders and outsiders to the village community, the central question in this chapter is who takes the initiative in facilitating market integration. The researches on the emergence of traders in peasant societies show diverse results. They can be classified by the following three distinct hypotheses: the insider, the outsider and the culture broker hypotheses. 1.

2.

136

The insider hypothesis (Hayami and Kawagoe 1999; Landa 1994) argues that the difficulties of performing business transactions under imperfect information, such as adverse selection and moral hazard, can be mitigated when trading partners are fellow villagers, because information among villagers is relatively symmetrical, and informal sanction mechanisms embodied in rural communities ensure contract enforcement. Thus, those insiders who are entitled to exploit these advantages have the better chance of success in trading than outsiders. The outsider hypothesis suggests that traders in peasant societies should be outsiders to the village community, such as the overseas Chinese

This chapter has benefited considerably from the comments and suggestions of Yujoro Hayami (Aoyama-Gakuin University), Julie Schaffner (Stanford University) and other participants at the World Bank workshop held at Stanford. I am responsible for any errors or infelicities that remain despite all this good service.

154

AKIHIKO OHNO

3.

(Geertz 1963; Weber 1930). This thesis assumes peasants in traditional societies are too passive to be entrepreneurial traders. In addition, they stay in a disadvantageous position in having access to urban markets. Thus, innovators emerge from the outside world. The notion of the traders' dilemma also endorses the outsider hypothesis (Evers and Schrader 1994). This notion argues that village-based traders are often required to share incomes with fellow villagers according to the social norms of peasant communities. Thus, insiders are subject to social constraints on private profit seeking, which prevent insiders from accumulating capital. In contrast, staying as outsiders in terms of their social characteristics allows traders to operate businesses outside the moral norms of the community.137 The culture broker hypothesis stresses that in developing economies, where traditional values and practices coexist with modernity, neither urban traders nor village producers are familiar with the rules and norms of the other world. Thus, interpreters familiar with both worlds play an important role for commercial integration of rural and urban economies. Such an interpreter is referred to as the culture broker (Neal 1984).

Note that these hypotheses are not mutually exclusive. Actually what we observed in a field study on the hand-weaving industry of Laos (the Lao People's Democratic Republic) is that a variety of traders endowed with different socioeconomic characteristics are vigorously conducting their business by taking their respective advantages. Some are marked as insiders, some are outsiders, and others are culture brokers. It should also be noted that our field study observed only a few disorders in a ‘state of nature’ (Hobbes 1651) in transactions, despite in Laos, especially in rural areas, formal sanction mechanisms being de facto absent. This is assumed partly because trading parties have created contractual arrangements to cope with the problems caused by contract uncertainty, and partly because informal sanction mechanisms deter opportunistic traders from a breach of contract. This chapter tries to illustrate how the traders emerge as market integrators in the Lao hand-weaving industry and what types of contractual arrangements are adopted to solve the problems associated with contract uncertainty. Another concern is to explore if trust based on community relations works as a substitute for formal sanction mechanisms. The study area includes Capital City of Vientiane (hereafter Vientiane) as well as the provinces of Vientiane, Luang Prabang and Xieng Khouang provinces (see Fig. 6.1).138 Our research was conducted in August and September 1996, August and December 1997, and March 1998.

137

This does not imply that insiders cannot take up trading business with fellow villagers. There are several solutions to the traders' dilemma as will be discussed later.

138

Note that Vientiane is the capital of Laos, while Vientiane province extends to the north of Vientiane.

RURAL-BASED INDUSTRIALIZATION IN LAOS

155

Fig. 6.1 Map

6.2 The Prole of Lao Handloom Industry Turbulence Leading to a Weaving Renaissance Since ancient times, hand-weaving has been an important task for Lao women. Until recently Lao weavers spun either cotton or reeled silk, dyed their yarns and wove mainly for personal use rather than for commercial gain. When the country opened its doors in the late 1980s, the evolution of the hand-weaving industry gained momentum. Weaving workshops have been set up in the villages, and their products come on to urban as well as overseas markets. The recent history of Laos has been turbulent. A large movement of northern Lao groups to the Vientiane plain mainly from Houa Phan and Xieng Khouang provinces, which are known as repositories of hand-weaving, took place during the Vietnam War due to carpet bombing on the Ho Chi Minh route that passed through these provinces. At the time of the Civil War in 1975 many lowland Lao (about 400,000 people) in the Vientiane plain, who were associated with the previous government, left the country. Thereafter the people from the northern provinces, who were associated with the present government, settled down in the Vientiane plain.

156

AKIHIKO OHNO

The patterns of Lao cloth reflect the cultural, religious and ritual symbols of ethnic groups, regions and even villages. Greater contact among ethnic groups during the turbulence has, however, blurred the historical and cultural significance of patterns. Thereafter, these patterns turned out merely fashionable designs. In November 1986 after the Fourth Party Congress, Laos opened its doors under the rubric of the New Economic Mechanism. This has produced two significant impacts on the hand-weaving industry. First, new markets were brought about for the industry by overseas Lao who left the country, foreign tourists, and the Thais. Secondly, imported factory-spun yarns enabled weavers to produce high quality cloth. A weaving renaissance commenced in Laos. This is where our story on the Lao hand-weaving industry starts.

Weaving Techniques of Lao Cloth

139

We need to mention here weaving techniques so as to identify the unique nature of Lao hand-woven cloth as a commodity transacted in the market, since marketing organizations vary according to the nature of the commodity (Shiamwalla 1978). Lao cloth can be classified into two types in terms of weaving techniques: cloth woven on a handloom with or without a vertical heddle.140 The former is figured cloth that is referred to as ‘sin’, and the latter is cloth with splashed patterns known as ‘mat mi’. We confine our attention to the former, which is hereafter called simply ‘cloth’. In Laos the patterns of cloth are woven with the assistance of a vertical heddle. A vertical heddle, which functions in the same way with the Jacquard attachment, is the lacy mesh with pattern rods above the warp that reflects the number of changes required to make figured cloth of complex patterns. With a vertical heddle, less experienced weavers can produce twill and satin cloth. Put differently, a vertical heddle makes the weaving techniques for complex patterns transferable. When cloth was woven for home consumption or traded within a local community, weavers only had to produce cloth with indigenous patterns specific to their ethnic groups. With the advent of the weaving renaissance, the need arose for weavers to produce cloth with new patterns in response to consumers' demand that changes frequently. However, weavers generally know only a few indigenous patterns for knitting a vertical heddle or more often they do not

139

This part depends heavily on Connors (1996).

140

Heddle is the name of the small cords through which the warp is passed in a loom after going through the reed, and by means of which the warp yarns are separated into two sets so as to allow the passage of the shuttle bearing the weft.

RURAL-BASED INDUSTRIALIZATION IN LAOS

157

know how to knit the heddle. Furthermore, it is difficult for rural weavers to obtain information on patterns in vogue in urban markets. A vertical heddle can be knitted in two ways. Generally, pattern makers knit various types of meshes for a vertical heddle. An eminent pattern maker can even create new patterns that set fashions. Pattern makers thus play a crucial role in the hand-weaving industry of Laos.141 Secondly, weavers can knit a vertical heddle by making reference to a sample piece. They copy the information on the patterns of sample cloth by counting the number of warp yarns which appear on the face side for one weft yarn at a time, and transfer the information on to a vertical heddle through the reverse action of the weaving process. It is quite arduous work, though. With the second method an infringement of patterns is technically possible, but it wouldn't mean that it is actually possible all the time. Pattern makers lay strong claims to the patterns that they designed as intellectual property rights. Since a weaving community is not so large, an infringement of patterns is usually detectable. For instance, cloth woven in the Vientiane plain is mostly sold in Talaat Sao (a public market in Vientiane). Those who have rights on patterns often make inspection visits to retailers in Talaat Sao if cloth with their patterns is displayed in retail shops with which they have no transaction. Within villages, it is often the case that fellow villagers let the pattern maker know about the infringement. It is often the case that pattern makers who are villagers obtain market information from retailers who can know consumers' demands better as urban-based traders. In this sense the alignment between insiders (village pattern makers) and outsiders (retailers) integrates rural and urban markets. The following is a typical example. Master weaver Q, a migrant from Houa Phan province, resides in a settlement village in the vicinity of Vientiane. When she started her weaving business, she wove cloth with native patterns of Houa Phan province. The cloth did not sell well, since indigenous patterns of Houa Phan province did not suit the taste of urban consumers. Subsequently, Ms. Q modified her patterns, taking some advice from a retailer. Then, her products started to sell well. This story indicates that the transmission of market information from urban business to village producers is of crucial importance for market integration. However, note that it is master weavers that take the technical initiative in creating patterns.

141

The following is a story about a pattern maker in Luang Prabang (interviewed in September 1996). She charges 150 kip per pattern rod for a vertical heddle. As a vertical heddle has 50 to 70 pattern rods, she earns 9,000 kip by knitting a vertical heddle. The cost of materials for a heddle is born by customers. As she produces approximately 150 sets a month, her monthly earnings turn out to be 1.35 million kip (1,436 US dollar). Note that the agricultural wage rate in the area is 1,000 kip a day.

158

AKIHIKO OHNO

6.3 Marketing Process and Contractual Arrangements Marketing Process The marketing process in the weaving sector is depicted in Fig. 6.2. The major economic actors involved in the handweaving industry are weavers, master weavers, and retailers. The main marketing channel from weavers to retailers consists of two stages: from weavers to master weavers, and from master weavers to retailers. In our field study 17 master weavers and 12 retailers were interviewed. The major features of traders involved in the weaving sector are anticipated in Table 6.1. The majority of weavers are women living in villages, since hand-weaving is the side business for farm households. Master weavers live either in weaving villages or on the outskirts of consumption centers. The emergence of master weavers is a recent phenomenon due to progress in Fig. 6.2 The Marketing Process Of Cloth

Note: The thickness of the lines corresponds to the volume of transactions. The two-headed arrows denote that market information flows from the principal to an agent in a dyadic relation.

RURAL-BASED INDUSTRIALIZATION IN LAOS

159

Table 6.1 Major Features of Traders Traders Small-scale master weaver Large-scale master weaver Commission merchant Itinerant merchant Retailers

Insider/outsider Insider Culture broker Insider Outsider Mostly outsiders

Information on patterns Pattern maker, village resident Pattern maker, town resident Weak governance on patterns Weak governance on patterns A few possess antique cloth from which new patterns are created

Source: Author's data.

the commercialization of hand-woven cloth. They either employ weavers at their workshops (hereafter, workshop weavers) or deal with weavers working at home (hereafter, home weavers), or use a combination of the two groups. Master weavers furnish market information on patterns in the form of a vertical heddle. Retailers, basically town-based traders, have their own shops in urban areas. For instance, there are more than 100 sin shops, small shops run by families, in Talaat Sao and its vicinity. Other trading agents observed in our research are commission merchants who trade with fellow villagers and itinerant merchants who cover wider areas. The role of these merchants as market integrators compared to master weavers will be discussed in the last part of this chapter. Independent weavers are those who do not contract with master weavers nor merchants, but deal with retailers directly. Thus they reside in the vicinity of consumption centers. Generally, they ask village pattern makers to knit vertical heddles for the patterns that they obtained in the market. The products of indigenous rural industries, such as hand-woven cloth, often encounter difficulties entering new markets outside rural communities since the characteristics of traditional cloth such as patterns and indigenous yarns do not suit the taste of urban consumers. Therefore, rural producers have to modify their products by incorporating modern tastes and materials (factory-spun yarns) so that they can appeal to wider markets. It would be difficult, however, for individual producers residing in remote villages away from consumption centers to obtain such information on wider markets. Since most Lao villages do not have electricity and weavers are mostly illiterate, the media would not play a role in conveying market information to village producers. In contrast, urban traders are generally in an advantageous position for obtaining market information. Therefore, so as to produce high value-added cloth, market information on patterns has to be transmitted to village producers. Master weavers act as a channel for market information between retailers and weavers under contract. Some independent weavers obtain the information on patterns in vogue by seeing the display of retailers in a bazaar. They ask village

160

AKIHIKO OHNO

pattern makers to knit a vertical heddle for the patterns. Independent weavers thus have to reside near consumption centers. Export wholesalers bring cloth that sells well in Thailand to village producers (weavers as well as master weavers), and the producers knit a vertical heddle referring to the sample cloth.

Contractual Arrangements The above unique nature of Lao hand-woven cloth associated with the importance of obtaining market information creates a peculiar marketing organization. It is obvious that master weavers/retailers furnish market information on patterns, which is their private information, to rural producers on the condition that the latter deliver the products exclusively to the former. This may appear to be the condition under which vertical integration emerges. However, hand-weaving is done as a side business for farm households, and the majority of weavers are women, mostly married. The difficulties of shepherding these weavers into a centralized operational unit should need no explanation. In such a circumstance, transactions between urban and rural traders are most efficiently coordinated by the relational contracting system (Hayami 1998). Actually, in the hand-weaving industry in Laos, the relational contracting system mediates cloth transaction as a rule. Hereafter, trading parties are referred to as principals and agents. Master weavers act as the agents of retailers, and simultaneously act as the principals of weavers. The modes of the relational contracting system observed in our field study are classified into the putting-out system (verlagsystem) and the advance-order system (kaufsystem). Under the putting-out system, principals supply materials for processing to agents, and later collect the finished products on a piece-rate payment. Handlooms, reeds and vertical heddles are often lent out free of charge. Under the advance-order system, in contrast, principals only guarantee purchase of the products at the stipulated price. Agents have to purchase materials at their own expense. Note that these systems involve long-term recurrent transactions. The dominant mode of contractual arrangements is different between the two stages of the marketing channel, namely between retailers and master weavers and between master weavers and weavers (Table 6.2). While the putting-out system is a sole contract between master weavers and weavers, it is rarely practiced between retailers and master weavers. Though spot market transaction is widely practiced between retailers and village producers (either the master weaver or the independent weaver), they occupy only a small proportion of the total volume laid in by retailers. Moreover, the cloth is of low quality. The contrast between the two stages presents useful insights for the investigation of the conditions under which particular contractual arrangements emerge.

161

RURAL-BASED INDUSTRIALIZATION IN LAOS

Table 6.2 Frequency of Contractual Arrangements Master weavers (n = 17) Weavers–master weavers Master weavers–retailers Retailers (n = 13) Retailers–producers

Putting-out

Advance order

Spot market transaction

17 1

0 7

0 6

9

6

12

Source: Author's data. Notes:

1 2 3

Workshop weavers are regarded as those under the putting-out system. Of 17 master weavers, two export their product by themselves, and one has her own retail shop. Thus, the number of cases where a contract exists between retailers and master weavers becomes 14. Producers include independent weavers and master weavers. The retailers tend to adopt the putting-out system with independent weavers rather than master weavers.

Table 6.3 Possibility of Agent's Opportunism Contracting system Putting-out Advance order

Selling to other traders Weak Large

Embezzlement Large None

Agents' Opportunism The relational contracting system is subject to agents' opportunism. Two major problems related to Lao cloth are: (1) agents sell other traders the products with the patterns that principals designed; and (2) agents embezzle yarns provided from principals. Agents sell either the advanced yarns or cloth woven with the embezzled yarns. In an extreme case, they sell the third parties all the yarns provided from principals without delivering the products to principals. The putting-out and advance-order systems have different features with respect to the two post-contractual problems as is shown in Table 6.3. Note that there is a tradeoff relation as to the problems across the two relational contracting systems. Under the putting-out system agents are bound by the norm of obligation to deliver the products exclusively to principals, because principals can claim ownership on the yarns supplied to agents.142 In contrast, under the advanceorder system moral imperatives for the delivery of products to their principals is weak, because it is agents that purchase yarns in the yarn market at their own expense. Thus, agents under the advance-order system

142

In this respect, the putting-out system shares the same features with the credit-tying contract as discussed by Hayami and Kawagoe in Chapter 5 of this volume.

162

AKIHIKO OHNO

are likely to be tempted into selling the products to other traders when offered higher prices. While embezzlement of the yarns would not take place in the advance-order system since no yarns are provided from principals, it is likely to be serious in the putting-out system. Under the putting-out system principals have to assess the amount of yarn required for a unit product so as to check embezzlement. It is often the case in the rural weaving industry that master weavers weigh yarns before advancing them to weavers, and weigh delivered products again. This type of inspection cannot be applied for Lao cloth. Agents deliver only a few pieces at one time to their principals. As a piece of cloth (sin) in Lao is about 1.8 meters in length, it is too light to be weighed exactly and allow principals to check embezzlement. Furthermore, since Lao cloth is woven with several colored yarns, and the allocation of color is changed after weaving a few pieces of cloth even though the pattern is unchanged, it is difficult to calculate the exact number of hanks of each colored yarn for a unit piece of cloth. These factors give agents plenty of scope for embezzlement.

Measures Against Opportunism Deterring agents' opportunism is imperative to sustain the relational contracting system. Monitoring of agents' work performance is often deemed as an effective measure against agents' opportunism in the system. However, it frequently involves prohibitively high transaction costs, especially when contracts are made between urban-based traders and rural-based producers. In addition, formal sanction mechanisms on trade are de facto absent in Laos, especially in rural areas. Thus, traders have to resort to other measures against agents' opportunism. This chapter presents three measures against agents' opportunism: process-based trust, community-based trust and governance over the products. These measures constitute principals' control over agents. Recent literature has revealed that even without formal sanction mechanisms or direct monitoring, good conduct can be expected when trust checks agents' opportunism (Kramer and Tyler 1996; Klein 1997; Hayami 1997). Trust entails informal sanction mechanisms that range from village ostracism, bad reputation or humiliation, to simply termination of business relationships. Hence, trust can ensure transactions are conducted morally. It will be useful to make a distinction between two types of trust: process-based trust and community-based trust.143 Process-based trust arises

143

This idea is heavily dependent on Zuker (1986). She defines institution-based trust that is tied to formal social structures as the third type of trust. However, this trust is irrelevant in the early stages of market societies. Authoritative coordination can be another plausible measure against opportunism. However this measure is irrelevant in the context of Laos where social and economic cleavage among the actors of the weaving industry is not significant.

RURAL-BASED INDUSTRIALIZATION IN LAOS

163

through long-term recurrent transactions as the tit-for-tat strategy dictates. Under recurrent transactions rational egoists will establish cooperative relations as long as repeated transactions offer benefits that outweigh those obtained by one-shot opportunistic behavior and transaction costs for reaching the Nash equilibrium in the repeated prisoners' dilemma game (Axelrod 1984). This process is referred to as clientelization by Geertz (1978).144 Process-based trust takes time to emerge, since people have to ascertain if one's trading partner is trustful or not in the course of recurrent transactions. The process-based trust is dyadic by definition. When the rules and norms of this trust affect the contractual terms of the third parties, it produces community-based trust. Community-based trust rests on the norms of obligations and cooperation rooted in social similarity, such as ethnicity, being native to a place, and so forth. Community-based trust is assumed to be reliable because of the social sanctions mechanisms behind the similarity.145 Once trading parties ascertain that they share social similarity, this trust emerges immediately without recurrent transactions. Thus, both principals and agents can evade an adverse selection problem without high transaction costs. Trust is a type of expectation that alleviates the fear that one's exchange partner will act opportunistically. Hence, trust works as ‘an important lubricant of a social system’ (Arrow 1974). The two types of trust provide different sanction mechanisms. Since process-based trust is attached solely to a trading pair, the cheated partner will sanction opportunistic behavior by means of a termination of the relationship. Community-based trust, on the other hand, can rely on sanctions by other members of the community. In other words, these forms of trust entail personal enforcement and community enforcement respectively (Kandori 1992). In a community, once sanctioned, agents would lose future opportunities of contract from other principals within the community. The cheaters have

144

According to Geertz, clientelization is the tendency for repetitive purchasers of particular goods and services to establish continuing relationships with particular purveyors of them, rather than search widely through the market at each occasion of need. Adversaries is the word for clientship relations which are not dependency relations, but are competitive ones. Clientship is symmetrical, egalitarian, and oppositional. There are no patrons in the master and man sense here. Clientship is a reciprocal issue.

145

The following is a case of informal sanction that the author encountered during the survey. A master weaver, G, of Xieng Khouang province (see Table 6.4) has two workshops employing more than 30 weavers. In addition, she contracts with some contractors under the putting-out system. Her products are exclusively exported to Thailand through an export wholesaler of Vientiane. A few days before our interview she found village women wearing cloth with patterns that she had designed. The cloth should not be in the area. This means one of her contractors had embezzled yarns and sold a piece of cloth to the women secretly. The master weaver knows who is in charge of the particular pattern. The author asked how she coped with this problem. She replied that she would take back handlooms and yarns that she advanced to the contractor on credit, and she would speak ill of the cheater so that the cheater could not deal with any other master weavers and retailers in the area in the future. This implies that the cheater would be ostracized from the weaving community of the area.

164

AKIHIKO OHNO

to contract with traders with whom they do not share community-based trust. In this case, both agents and retailers have to bear high transaction costs associated with contract uncertainty. Fear of such censure works as a self-enforcing norm that deters opportunistic behavior. Community-based trust works within a group of people who can easily share information about cheats. Information about the cheats travels as gossip, which is an important form of everyday communication in pre-modern society. Since relatives and friends are effective transmission agents of gossip, the information tends to circulate rapidly within a group of people who share social similarity. It follows that informal sanction mechanisms tend to work effectively among people who share social similarity. Another measure against agents' opportunism emerges when principals can monopolize market information on patterns or skills for creating patterns. The ownership of such information is referred to as governance over the products. Since cloth with attractive patterns that suits consumers' taste obtains a stable order and fetches higher prices, agents who contract with the principals with strong governance can enjoy constant orders with higher remuneration. In addition, since the patterns are so distinctive the opportunistic behavior of selling cloth to a third party is easily detected. Hence, governance over the products will deter agents' opportunism.

Contract Choice Since vertical integration including the factory system is not a practical option in the rural hand-weaving industry, a choice has to be made among putting-out, advance-order or spot market contract. First of all, let us assume that the possibility of agents' opportunity can be neglected. On the part of rural agents, the putting-out system is usually the first choice because under this system principals advance materials on credit, and furnish market information on patterns.146 In addition, principals guarantee purchase of the finished products. These constitute the agents' benefits from the putting-out system. The more agents are faced with working capital constraints and experience difficulties in accessing yarn as well as cloth markets, the more they prefer the putting-out system. Principals can be divided into two groups according to whether they can create patterns for cloth that suit the tastes of urban consumers or not. Master weavers can create the patterns as they are pattern makers, while most of the retailers cannot do so, except for a few having antique cloth. For the principals who have private information on patterns, it is imperative to secure exclusive delivery of the products from their agents so that the principals can maintain a monopolistic position in a cloth market

146

Handlooms are often leased out free of charge under this system.

RURAL-BASED INDUSTRIALIZATION IN LAOS

165

through product differentiation. In this case, the most appropriate contract for them is the putting-out system, because the putting-out system works as a credit-tying contract that enforces exclusive delivery of the products. This is discussed in the Hayami–Kawagoe chapter in this volume. In this respect, the putting-out system is Pareto-optimal. Even for the principals who do not possess private information on patterns, securing a constant supply of cloth is crucial. There are seasonal changes of demand as is shown in Fig. 6.3. In the dry season, when the larger number of festivals and marriage ceremonies take place, demand tends to exceed supply. Many retailers insist that they keep buying from client weavers even in the off-season so that they can secure supply in the peak season. Though wholesale prices of products are lowered in the slack season, some retailers buy cloth at prices higher than wholesale prices so as to keep on good terms with client weavers. In order to secure a constant supply of cloth, principals adopt the puttingout system. Ceteris paribus, however, the agents' benefits from the exclusive contract with this type of principal will be less compared with principals having information on patterns. This reflects the difference of agents' inclination toward opportunism, as we will see later. Then, why do the advance-order system and even spot market transaction prevail in cloth transaction? Agents' opportunism is a major factor Fig. 6.3 Gross Sales Of a Retailer In Talaat Sao

166

AKIHIKO OHNO

behind the choice of these contracts. On the part of principals it can be reasonably assumed that when the loss from agents' opportunism associated with the putting-out system outweighs the profit gained from exclusive transaction, principals will abandon the putting-out system. As the other side of the same coin, from the point of view of agents, when the profit from the putting-out system is lowered, they will come to choose the advance-order system. As the benefits from one-shot opportunistic behavior become greater than those obtained from recurrent transactions, it becomes difficult for principals to enforce long-term contracts with agents so that they have no option but to use spot transactions. Two plausible situations that induce agents to behave opportunistically are: (1) the emergence of alternative market opportunities that include alternative traders and job opportunities, and (2) increases in the prices of yarns and cloth. The former attenuates the level of punishment, while the latter makes the benefits from one-shot opportunistic behavior greater than those from the relational contracting system. The former is highly associated with economic development, while the latter is caused by a sporadic shock such as the currency crisis of Thailand in 1997. In either case, contract uncertainty increases for principals. After the currency crisis, the prices of yarns increased sharply since factory spun yarns are mostly imported. This enhanced agents' opportunism critically. A piece of product (sin) is about 1.8 meter in length and 20–25 pieces of cloth are woven from a set of warp yarns which are 50–60 meters in length. After weaving several pieces of cloth, agents (mostly weavers in our case) deliver the products to their principals and receive a fixed remuneration per piece. Generally, it takes several months for weavers to finish one order. When the price of cloth increased after the currency crisis, other traders started to offer higher prices for the products. Then, agents started to sell the promised products to other traders. A high incidence of agents' opportunism increases the costs for preserving the putting-out system, and thus principals are obliged to shift to the advance-order system or even spot market transaction. Under these contracts, principals neither supply yarns on credit nor furnish information on patterns to agents. If there are no alternative principals who offer a putting-out contract, agents are faced with working capital constraints and difficulties in obtaining information on patterns in vogue in the market. The products transacted in the market are of low quality, rendering only a narrow profit to agents. Therefore, agents' opportunistic behavior jeopardizes their profits in the long run. This is the well known prisoners' dilemma game situation. Agents' impatience, namely a high future discount rate, due to their poverty is an underlying reason. From the above discussion, the following hypotheses on contract choice can be advanced. They are summarized in the field defined by principals'

RURAL-BASED INDUSTRIALIZATION IN LAOS

167

Fig. 6.4 Contract Choice

Note: Arrows denote institutional changes observed especially after the currency crisis of 1997. governance over the products and the level of community-based trust that principals can rely on (Fig. 6.4). Since the relational contracting system involves recurrent transactions in general, process-based trust is obtainable except for a spot market transaction. 1.

2.

When principals can resort to those measures simultaneously (strong control over agents in Zone A), the putting-out system is likely to be adopted since agents' opportunistic behavior can be largely deterred. That principals have governance over the products implies that they require exclusive delivery of the products. Thus, principals' first choice is the putting-out system. It is assumed that the principals who can rely on either one of the measures (Zones B or C) are more likely to suffer from agents' opportunism than the principals staying in Zone A. The putting-out system adopted by principals who rely on process-based trust only (Zone D) is vulnerable to agents' opportunism. The principals in Zones B and D may

168

AKIHIKO OHNO

3.

4.

continue to use the putting-out system so long as the profits of securing a constant supply of cloth exceeds the drawback from agents' moral hazards.147 As these measures become weaker, agents' opportunism becomes more hazardous. As a result, principals come to adopt the advance-order system, relying on process-based trust only. Eventually, spot market transaction emerges. Another plausible institutional change is the emergence of the (infant) factory system, in which principals can monitor the work performance of weavers intensively. This is more likely to be an option when principals have private information on patterns. Principals and agents are presumed to have different characteristics pertaining to the measures against agents' opportunism. This leads to the creation of multiple equilibria: an economy where distinct traders with different contractual arrangements co-exist.

6.4 Case Studies of Master Weavers Master weavers in Laos have links with both village producers (weavers) and urban-based traders (retailers and export merchants). Master weavers are mostly pattern makers. Through their links with urban-based traders, master weavers can obtain information on what patterns are in vogue in urban markets. Then, they translate the information on to a vertical heddle, and pass it to their agent weavers. The relationship between master weavers and workshop weavers is akin to a putting-out contract, because the former provide materials and a vertical heddle (market information) to the latter on a piece-rate payment. The work performance of home weavers can be easily monitored since weavers and master weavers reside near to each other in general, except in the cases of large-scale master weavers as we will see later. Accordingly, there is no underlying difference between home weavers under the putting-out system and workshop weavers. The most distinctive difference may be that the former are generally married, while the latter are single. Therefore, the demarcation between the putting-out system and employment relationship is insignificant. In this chapter workshop weavers are regarded as agents under the putting-out system in the same way as home weavers are. In our field survey 17 master weavers were interviewed. They can be grouped into two in terms of the number of weavers they contract with;

147

The principals staying in Zone B are less likely to establish the factory system, but tend to give up being master weavers when faced with large market uncertainty. This is because their weak governance over the products yields fewer profits when compared to the costs for establishing the factory system.

RURAL-BASED INDUSTRIALIZATION IN LAOS

169

12 small-scale and five large-scale master weavers. The average number of weavers that the former contract with is 11.25 (SD = 6.06), with 26 as the maximum, while the latter contract with more than 50 weavers, with 219 as the maximum. Their distinct characteristics may offer clues to exploring the process of market formation.

Small-Scale Master Weavers There are two weaving areas on the outskirts of the town of Vientiane: (1) a local weaving center (Hadxayfong district) which is renowned for generations; and (2) several settlements of migrants mainly from Houa Phan and Xieng Khouang provinces. Both are about 10–15 kilometers from Talaat Sao. Generally speaking, migrant weavers are better skilled and produce cloth of more complicated patterns than the local weavers of Vientiane. The number of pattern rods that a vertical heddle has is 15–25 in the local weaving area, while it is about 50–120 in the settlement area. As one pattern rod corresponds to one weft yarn, the higher the number of pattern rods, the more complicated the patterns woven. This difference, as we will see later, leads to different responses of respective producers to the currency crisis of Thailand in 1997. In Luang Prabang province weavers are mostly war refugees who were associated with the previous government. The following are typical examples of the small-scale master weavers. Case 1. Ms. A (interviewed in August 1996) was the first master weaver, in 1980, in the local weaving area of Vientiane to start a weaving workshop. She contracts with two workshop weavers and 11 home weavers on the putting-out system. She is a pattern maker, and sells her patterns knitted on to a vertical heddle at 6,000 to 14,000 kip. She has traded with a particular retailer in Talaat Sao for many years under the advance-order system. In the dry season 40 kg of silk is dyed per month and 20–30 kg in the rainy season at her workshop. From 4 kg of silk yarns 25 pieces of cloth are woven. Thus, she deals with 250 pieces in the dry season and 125–190 pieces in the rainy season. In the rainy season, weavers engage in agriculture as well. Her weavers are paid 4,500 kip per piece of cloth, which can be woven in one and a half days. Thus, a weaver earns about 3,200 kip per day, which is almost equal to the agricultural wage of the area of 3,000 kip per day. Case 2. Master weaver B lives in a village of a mountainous area about 80 kilometers north of Vientiane. As hardly any paddy land exists in this area, people make their living mainly by slash and burn agriculture. There are five master weavers in this village producing cotton cloth for a domestic market. Ms. B started her weaving business in 1993 and currently owns 11 handlooms of which nine are leased out to weavers under the

170

AKIHIKO OHNO

putting-out system. Ms. B and her daughter weave on the remaining two handlooms. The wholesale price of her products is 3,000 kip per piece when sold to an itinerant merchant coming to her village, but it is 3,200 kip when Ms. B brings them to Talaat Sao.148 Thus, she sells her products at Talaat Sao when she has a large quantity in stock. These weavers are paid 1,000 kip per piece, which can be woven in a day. This brings in earnings slightly better than the agricultural wage of the area of 800 kip per day. Other cases we observed also show that the earnings of weavers are more or less the same as the prevailing agricultural wage rates of their areas. The small-scale master weavers can exert the following measures against agents' opportunism. As they contract with neighboring weavers, (1) community-based trust works, and (2) they can monitor agents' performance at relatively low costs. In addition, (3) master weavers possess governance over the products as pattern makers. Consequently, those master weavers can maintain the putting-out system in Zone A of Fig. 6.4. According to our field observation weavers often shift to another master weaver after finishing one order. They feel little commitment to a particular master weaver. The frequent rearrangements of a trading pair may act against enhancing process-based trust. However, since villagers under tightly woven village relationships can share information on cheaters, community-based trust is supposed to work even under these situations. Changing trading partners or dealing among people who do not have frequent relations does not affect the validity of community-based trust, as long as information on contract breach is effectively transmitted among the community members (Milgrom et al.1990; Kandori 1992).

Decline of Trust Since trust in trade is calculative by nature, it may decline as the level of punishment decreases. Our field survey observed a decline of trust in Vientiane. We will consider two examples to illustrate the decline of trust and consequent institutional changes. Aforementioned Ms. A was re-interviewed in March 1998 after an interval of two years. She was about to give up being a master weaver. This is because, according to her, the weavers she contracted with under the putting-out system began to sell the promised products to newly emerging commission merchants. These merchants are village traders, but not pattern makers. They advance yarns to weavers on credit and later collect the finished products, thus the putting-out system is employed. However, commission merchants do not furnish information on patterns. The

148

Thus, the marketing margin is calculated at 6.5 percent for an itinerant merchant.

RURAL-BASED INDUSTRIALIZATION IN LAOS

171

weavers under contract with them have to purchase a vertical heddle with patterns from a pattern maker, such as Ms. A. The currency turmoil in southeast Asian countries since July 1997 is a major cause behind the change. The baht was depreciated from 25 baht against the US dollar to nearly 50 baht by the end of 1997. The kip was also depreciated against the baht (in 1996, one baht=37.5 kip, but in December 1997, one baht=46–47 kip). Thus, the kip was depreciated sharply against the dollar with wide fluctuations, from one dollar being equivalent to 940 kip in August 1996, 2,150 kip in December 1997 and 2,450 kip in March 1998. Consequently, the prices of imported yarns and the wholesale prices of cloth increased considerably.149 The rising price of yarns increased weavers' profit from embezzling yarns. In addition, the merchants started to offer higher prices for cloth due to an increase in the prices of cloth. Furthermore, the emergence of the merchants as alternative traders reduced the value of continued transactions with their long-trading principals, by attenuating the level of punishment on agents' opportunism. The profits from oneshot opportunistic behavior came to outweigh the future benefits obtained from recurrent transactions. Consequently, agents' opportunistic behavior in selling their products to other traders became serious. Ms. A now has to negotiate repeatedly with the cheaters about the assessment of damages caused by their opportunistic behavior. High negotiation costs drove her to introduce the new contracting system under which weavers are entitled to sell their products either to Ms. A or other merchants. The weavers, however, have to pay for vertical heddles with the patterns that Ms. A designed. Ms. A takes a fixed commission on the sales of yarns. The new system is closer to spot market transactions than the putting-out system. Ms. A now contracts five weavers on the conventional putting-out system and six weavers on the new system. She is planning to increase transactions under the latter system. This implies that Ms. A turns out to be a commission merchant or even a mere yarn merchant, though she acts as a pattern maker. When a market becomes unstable, the relational contracting system tends to be replaced by a spot market transaction. Case 3. Master weaver C has a workshop on the outskirts of the local weaving area in Vientiane. The workshop produces cloth with relatively complicated patterns. She installs 25 handlooms at her workshop and leases out five to home weavers living nearby on the putting-out system. Though she had employed villagers as workshop weavers in the past, they were gradually replaced with migrant weavers from Houa Phan province. At the time of the interview, out of 21 workshop weavers 15 were migrants. All were single and after working several years they returned home for marriage. Though Ms. C does not share social similarity with

149

This is particularly true for silk yarn for warps which have to be imported from China.

172

AKIHIKO OHNO

migrant weavers, she insists that migrant weavers are more obedient than fellow villagers. According to Ms. C village weavers now quit easily when offered jobs with higher wages by newly established factories, especially garment factories and they become less and less docile. Though Ms. C contracts with some village weavers on the putting-out system, she lets them weave scarves (pha biang) that require a less-skilled weaving technique, and hence less monitoring on quality. This implies that Ms. C has stayed in Zone C of Fig. 6.4 and is moving toward the zone of the factory system. As trust declines, agents' opportunism comes to impede smooth transactions under the putting-out system. Principals increasingly demand greater protections against the possibility of opportunism. Faced with the decline of trust, Ms. A abandoned the putting-out system and adopted the new system that is closer to a spot market transaction, while Ms. C moved in the opposite direction in a continuum from a spot market to the factory system. Ms. C no longer relies on community-based trust, but tries to internalize a production process by creating an infant form of the factory system. Increasing contract uncertainty and the peculiarity of the cloth that she produces allows her to choose the option of internalization. With this institutional change, Ms. C can sustain her status as a master weaver. As mentioned above, there are two weaving areas in Vientiane, namely the local weaving area and the settlement areas where migrants from the northern provinces engage in the weaving business. The above institutional changes are witnessed only in the local weaving area. In the settlement area where master weavers provide vertical heddles with high-standard patterns for weavers, agents' opportunism leading to the changes of contractual arrangements are rarely observed. Future transactions seem still to be heavily weighted by agents due to strong governance of master weavers over patterns. In addition, a strong community nexus of migrants seems to persist. These allow the master weavers of the settlement areas to stay in Zone A of Fig. 6.4. The putting-out system is, therefore, preserved there. In contrast, the master weavers of the local weaving area stay in Zone B since their governance over the patterns is relatively weak due to poor patterns. Faced with uncertainty and the emergence of other economic opportunities, community-based trust seems to be withering there. Thus, the environment of transaction is shifting from Zone B to a corner area where spot market transactions prevail.

Large-Scale Master Weavers Our survey observed four large-scale master weavers in Vientiane and one in Xieng Khouang Province. They reside on the outskirts of town. The following is an example. Case 4. Ms. D (interviewed in August 1996) migrated to Vientiane from Houa Phan province during the Vietnam War. She opened a small workshop

RURAL-BASED INDUSTRIALIZATION IN LAOS

173

in a village on the outskirts of Vientiane in 1986. She currently employs 15 workshop weavers who are migrants from Houa Phan province. They are single, and after working for an average of two years they return to their native villages for marriage. Ms. D also enters into a contract with a total of 204 weavers of two villages in Vientiane province. The villages were formed by the people who moved from Houa Phan province at the time of the Vietnam War. In these villages there is no contractor who works in lieu of Ms. D. The following is the situation in Village X about 35 kilometers from Vientiane. Ms. D started her business with the villager weavers in 1988 under the putting-out system. The weavers of this village were originally skilled but had some problems with the performance of their handlooms. Ms. D started to lease out quality handlooms free of charge. The number of handlooms has now reached 140 and weaving is performed at almost all households. In this village, women start weaving at the age of around five and become qualified weavers at approximately 15. Ms. D, an eminent pattern maker, provides vertical heddles with patterns she has designed to her weavers with the colored silk yarns that she has dyed. During the dry season, an average of 500–600 pieces of cloth is woven every month by the home weavers of the two villages. The number drops to about 250 pieces during the rainy season when farm work reaches its peak and weaving becomes difficult because silk yarns get thick by absorbing moisture. Her weaving business brings monthly earnings of 8.19 million kip (8,712 US dollars in August 1996) in the dry season. Note that a monthly salary of government officers is approximately 30–50 US dollars. An adjoining village, Z, shows a contrast with village X in the weaving industry, despite similar historical backgrounds. In village Z there exist no master weavers or pattern makers. The weavers themselves have to prepare a vertical heddle. Since they know only a few indigenous patterns for a vertical heddle, they cannot take in the patterns in vogue in the urban market. In addition, weavers have to purchase yarns and sell their cloth in Talaat Sao themselves. Otherwise, the weavers contract with a commission merchant of the village under the putting-out system. We will explore this merchant later. Because of working capital constraints the weavers can only afford cotton yarns. Cotton cloth brings a narrow margin when compared to silk cloth. In contrast, the weavers of village X do not have to worry about marketing. The master weaver supplies silk yarns and vertical heddles with patterns that she designs. Such products fetch higher prices in the market, assuring a larger margin. Approximate daily earnings from weaving are 2,000 kip in village X and 1,500 kip in village Z. Village X was electrified in 1996 in advance of surrounding villages.150 This comparison elucidates the

150

An electric cable is laid only up to the gate of the village. Villagers have to invest to lay electric lines up to their houses at their own cost.

174

AKIHIKO OHNO

importance of master weavers and suggests a shortage of master weavers as a market integrator. While small-scale master weavers contract with neighboring weavers, large-scale master weavers reside in or on the vicinity of towns and enter into a contract with weavers in remote areas. Therefore, constant interactions with agents on a close personal basis cannot be expected for the large-scale master weavers. What, therefore, enables them to contract with village weavers under the putting-out system and even on a larger scale? Possible explanations can be given from the viewpoint of a culture broker. The large-scale master weavers are differentiated from the small-scale counterparts in that the former have higher educational attainments and urban experiences (Table 6.4). Case 5. Ms. E and H, for example, studied abroad and can communicate in English and French. They export their products directly to foreign businesses without relying on local retailers. Ms. F is a medical doctor and Ms. D is a government officer. Ms. G in Xieng Khouang province studied at accountancy school in Vientiane. There, she got a contact with a cloth exporter. After returning to Xieng Khouang province, she started a weaving business using a link with the exporter. These large-scale master weavers can be said to have a better understanding of the rules and norms for dealing with urban-based traders and even foreign traders when compared to their small-scale counterparts. Besides being familiar with the urban world, the large-scale master weavers share a social similarity with their weavers. Ms. D, E, and F contract with migrant weavers from the same area they themselves migrated Table 6.4 Social Characteristics of the Large-scale Master Weavers Birth place

Residence

Social similarity with Number of weavers weavers under control Vientiane Yes 219

D Houa Phan Government officer and an eminent pattern maker E Houa Phan Vientiane University graduate (USSR) and can communicate in English F Xieng Khouang Vientiane University graduate and medical doctor. Runs a retail shop at Talaat Sao G Xieng Khouang Xieng Khouang After graduating from vocational school in Vientiane, worked there for some Khouang province H Savannakhet Vientiane University graduate (France) and can communicate in French and English Source: Author's data. Note: All are pattern makers.

Yes

67

Yes

120

Yes 60 years. Currently lives in the capital city of Xieng No

57

RURAL-BASED INDUSTRIALIZATION IN LAOS

175

from. Thus, transactions are segmented among groups of people who share social similarity. Thus, they can utilize community-based trust. The case of Ms. H is a good illustration of this point in a different sense. Since she is a migrant from the southern province of Savannakhet (see Fig. 6.1) where weaving is not common, she is unable to find weavers with whom she shares a social similarity in Vientiane. She contracts with 57 home weavers in four villages, who are migrants from the north. In order to monitor them, she hires three supervisors. Other master weavers do not rely on supervisors. This fact indicates that community-based trust works as a device for saving on monitoring costs. The large-scale master weavers are marked as the culture brokers more distinctively than their small-scale counterparts. Though village-based traders can rely on community-based trust, they are likely to be constrained by the moral norms of the community that demand income redistribution, that is, the traders' dilemma. There are several solutions to the dilemma (Evers and Schrader 1994). Being a culture broker offers an effective solution. Though they share social similarities with weavers, they are socially differentiated from their weavers in that they are highly educated and have urban experiences. In addition, they reside in or in the vicinity of towns apart from the weavers with whom they contract. Thus, they are less likely to be constrained by the community norms.151 Another solution is remaining as petty traders. As is often pointed out, the petty trader is a standard feature in peasant societies. Remaining petty traders denotes that they do not have enough resources to share with fellow villagers and therefore they are free from the traders' dilemma. Though staying as outsiders in terms of their social characteristics allows traders to operate business without conforming to such norms, it is highly probable that they suffer from agents' opportunism. In conclusion, the large-scale master weavers are those who can exploit an ambivalent position as outsiders as well as insiders. This allows them to engage in the weaving business on a large scale. The discussion so far reveals the twofacedness of the community norms. While the community norms often work as a lubricant for efficient transactions by producing community-based trust, community may impede capital accumulation in the hands of insider-traders.

6.5 Case Studies of Retailers We now turn to the subject of how retailers who are outsiders in our context can manage the transactions with village producers.

151

When aforementioned master weaver D introduced the author to one of her contracting villages, she brought a large amount of candies for village children. She is likely to show her generosity to evade the traders' dilemma. She told the author that she made it a rule not to visit the village often.

176

AKIHIKO OHNO

The following is based on interviews with 13 retailers. The retailers are mostly dealing with several groups of producers under different contractual arrangements (see Table 6.2): nine retailers adopt the putting-out system, and six adopt the advance-order system. In addition, most of them lay in a stock of cloth on a spot basis, but cloth of low quality. Retailers are endowed with different characteristics concerning the measures against agents' opportunism. The retailers we observed can be grouped into three types in terms of the degree on which they can utilize the measures against agents' opportunism. They are: (1) the retailers who can rely on community-based trust as well as governance over the products; (2) the retailers who possess antique cloth based on which they create new patterns but cannot rely on community-based trust; and (3) the retailers who can rely on neither community-based trust nor governance over the products. These are referred to as the retailers with strong, moderate, and weak control of agents. Process-based trust is not considered here since it is generally obtainable in the relational contracting system. By taking examples, we will illustrate the association between principals' control over agents and the types of contractual arrangement.

Retailers With Strong Control Over Agents Of the 13 retailers we interviewed only two are migrants from the northern provinces. As they contract with migrant weavers from the same provinces, they can rely on community-based trust. In addition they have governance over the products, since they are pattern makers, though not so eminent. Thus they possess strong control over agents, staying in Zone A of Fig. 6.4. Case 6. Ms. J who migrated from Xieng Khouang province contracts with two master weavers under the putting-out system. One, who controls approximately 100 weavers, resides in a village about 50 kilometers from Vientiane. The village was formed by migrants from Xieng Khouang province. Ms. J supplies the master weaver with yarns and vertical heddles with patterns she has designed under the putting-out system. That the putting-out system is adopted between the trading parties living apart can be explained by a social similarity they share. The other master weaver is her younger brother living in Xieng Khouang province. He contracts with approximately 50 weavers. Ms. J says that she has not experienced any agents' opportunism. Case 7. Another case is the aforementioned master weaver-cum-retailer, F, a migrant from Houa Phan province. She runs a retail shop at Talaat Sao and contracts with migrant weavers from the same province in the settlement area. She leases out some 80 handlooms under the putting-out

RURAL-BASED INDUSTRIALIZATION IN LAOS

177

system and employs 20 migrant weavers from Houa Phan province at her workshop in a village on the outskirts of Vientiane. She does not manage home weavers by herself, but assigns control over them to 11 contractors who are her relatives living in villages under the putting-out system.152 It is the contractors who decide the terms of contracts with weavers, and Ms. F does not concern herself with the terms between them. Since Ms. F can exert strong governance over the products as well as community-based trust, she is situated in Zone A of Fig. 6.4. She says that she has experienced no opportunistic behavior by the contractors. The above two retailers engage in business in Zone A of Fig. 6.4. Therefore, they can maintain the putting-out system without suffering from agents' opportunism.

Retailers With Moderate Control Over Agents Other retailers we interviewed are town-based traders born in Vientiane. They do not share social similarities with village weavers, especially migrant weavers. Some retailers have antique cloth from which they take motifs and create new patterns of their own.153 Their cloth fetches higher prices owing to product differentiation. With strong governance over the products but without community-based trust, these retailers can exert moderate control over agents staying in Zone C of Fig. 6.4. The following are examples. Case 8. Retailer K, born in Vientiane, sells quality cloth. She designs patterns from antique cloth she owns and furnishes them to some 20 client weavers, migrant as well as local weavers, under the putting-out system. The products need to be exclusively delivered to Ms. K. However, her weavers often sell the products to other retailers who offer higher prices, or sometimes weavers steal the patterns. Ms. K withdraws the vertical heddle after the assigned numbers of cloth are completed so that the weavers cannot copy the patterns. But some weavers produce an extra piece of cloth secretly in order to knit a vertical heddle from the piece later. The patterns of Ms. K are so distinctive that they can be easily identified when the weavers sell pieces to other retailers. She knows who is in charge of a particular pattern. She often checks the shops of Talaat Sao to see if cloth with her patterns is displayed in retail shops with which she does not deal. When this is discovered, she visits the cheater to take the rods off the vertical heddle and their business relation is broken off. Case 9. Retailer L, born in Vientiane, deals with seven migrant weavers from Houa Phan province under the puttingout system. The weavers

152

The contractors are different from master weavers in that the former are not pattern makers.

153

When retailers create new patterns, they provide sample cloth for their weavers and the weavers knit a vertical heddle referring to it.

178

AKIHIKO OHNO

reside in the same village with the retailer in the local weaving area. Thus, she can monitor weavers' performance at relatively lower costs. She provides weavers with a vertical heddle with patterns she has designed. However, embezzlement is annoying her, especially after the currency crisis in Thailand. Though she thinks she provides enough weft yarns, weavers often claim that there is a shortage. The retailer suspects the weavers of embezzling the yarns, but she cannot take effective measures against it. She is planning to increase stocking on a spot basis. Case 10. The same observation applies to Ms. M, Vietnamese-French, who runs a small garment workshop and a boutique in Vientiane. She contracts with some village weavers under the putting-out system. Ms. M creates patterns of her own and furnishes them to the weavers. According to her, the main difficulties in dealing with village weavers are the poor quality of products and the infringement of her patterns. As an urban-based trader, Ms. M cannot take the proper steps to enforce the contract. Loose supervision, for example, offers opportunities for the weavers to indulge in reducing the number of picks of weft yarn, which results in inferior quality. Such cloth shrinks easily when washed. This is what Landes referred to as the frictions inherent in putting-out (Landes 1969).154 The last measure, she mentions, is to establish a factory of her own. Though the above retailers adopt the putting-out system, they cannot check agents' opportunism efficiently because their control over agents is not strong enough.

Retailers With Weak Governance Over Agents Most of the retailers in Talaat Sao neither claim governance over the patterns nor rely on community-based trust. Though they do not furnish information on patterns to weavers, some of them adopt the putting-out system so as to secure a constant supply of quality cloth from skilled weavers. Case 11. Retailer N of Vientiane origin has been dealing with migrant weavers on a spot basis for more than 20 years. In the course of time, she has selected ‘skilled and trustful weavers (according to Ms. N)’ one by one, and entered into a putting-out contract with them. Now, she deals with some 50 client weavers under the putting-out system. Recurrent transactions have engendered process-based trust between the two parties. Clientelization, however, does not assure effective contract enforcement. An often-heard complaint from the retailers with weak control over agents is that some weavers would not deliver the promised products even if the retailers advanced yarns on credit. Embezzlement is also

154

Landes (1969) finds the reason for the transition in the usual difficulty of compelling performance by cottage workers and the poor quality of hand work.

RURAL-BASED INDUSTRIALIZATION IN LAOS

179

unavoidable. According to Retailer N (interviewed in December 1997), in 1997 seven client weavers among 50 did not deliver the products. She visited them to urge delivery, but only heard that they pleaded illness and so forth. Ms. N suspects them selling off the advanced yarns for rice. According to her estimation, the loss from this agents' opportunism amounts to 400,000 to 500,000 kip in 1997. In order to reduce such damage to a minimum, she tries to contract with weavers of different villages as far as possible. This is because, according to her, when such problems occur with one weaver, it spreads to others in the same village. This is a case of contagious equilibrium (Kandori 1992) in which community relations take a dysfunctional response against smooth transactions. Case 12. Another example is Retailer O, born in Vientiane, who sells cloth of poor quality. She contracts with seven weavers under the putting-out system. In order to secure delivery from the weavers, she provides consumption credit with no interest. Note that the retailer advances the yarns required for weaving. The weavers repay in cloth that they weave. When the retailer lends money to non-weavers, she charges at 5 percent simple interest per month. However, the weavers often sell the products to other retailers, especially after the currency crisis. According to her, though the delivery price of cloth is stipulated at 5,000 kip, other retailers offer 6,000 kip for the cloth. This is because the price of silk increased sharply. Retailer O is afraid of not being able to maintain the putting-out system any more. She is planning to lay in cloth on a spot basis. The putting-out system adopted by the retailers with weak control over agents is the most vulnerable to agents' opportunism, and thus is going to be replaced by an advance-order contract or even spot market transaction. The reason given by these retailers for the choice of the advance-order system rather than spot market transaction is to secure a constant supply of quality products. When asked why they do not adopt the putting-out system, the retailers unanimously express a fear that weavers would not deliver the products. Some retailers who adopt the advance-order system report the author that they once adopted the putting-out system in the past. The main reason for repudiating the putting-out system is that agents often embezzled the advanced yarns or sold the products to other retailers. And it is difficult to take indemnities from agents since they do not know how to force the cheater to compensate for the loss. As retailers are generally urban-based traders, they are incapable of resorting to community-based trust. When asked if they can rely on the police, they immediately say no. It became increasingly difficult to deter agents' opportunism as the number of traders in the market increased. This makes it difficult for retailers to maintain the putting-out system, and the advance-order system has been emerging. The advance-order system is, however, more fragile than the putting-out system especially when faced with uncertainty, since the moral obligation

180

AKIHIKO OHNO

of delivery is weaker under the advance-order system. According to the retailers, the incidence of agents' opportunism increased drastically after the currency crisis. The ongoing currency crisis has affected the hand-weaving industry to a greater extent. A shift of contractual arrangements from the putting-out system to spot market transactions was observed. Principals who can exert strong control over agents, such as master weavers and a few retailers having antique cloth, can maintain the putting-out system. In contrast, the retailers who are outsiders for village producers and some master weavers who have only weak governance over the products cannot cope with increasing agents' opportunism caused by contract uncertainty. The currency crisis has driven them to introduce the advance-order system from the putting-out system, and spot market transactions from the advance-order system. Retailers are basically town-based traders. An arresting fact about them as traders is that they do not actively search for trading partners by visiting weaving villages themselves, but only wait for village producers coming to the bazaar. It is left to the village producers and merchants who visit bazaars to make a contract with retailers. In this respect, insiders take the initiative in integrating the rural and urban markets.

6.6 Do Itinerant and Commission Merchants Substitute for Master Weavers? Another question is whether commission and itinerant merchants can substitute for master weavers as market integrators. The aforementioned commission merchant in village Z contracts with 56 village weavers under the putting-out system. There exist no master weavers/pattern makers in the village, and weavers know only a few indigenous patterns. The merchant shares social similarity with the weavers, but does not furnish market information on patterns. She does, therefore, stay in Zone B of Fig. 6.4. Thus, the merchant suffers from weavers' opportunism. On average, four to five weavers per month sell their cloth at Talaat Sao of Vientiane without delivering to the merchant. As the merchant knows the pattern of each weaver, such opportunism is easily detected when she visits Talaat Sao. However, she thinks there is no way to deter such opportunism. She goes on to say that retaining her wrath is imperative in doing her business. She not only advances yarns on credit but also sells daily necessities to weavers on credit. In addition, she lends money and accepts repayment in cloth. Interest is not charged. When she lends money to non-weaving villagers, she charges interest at 5 percent per month. It should be noted that money is not lent to weavers for weaving activities, but for daily consumption, since all the materials necessary for weaving are advanced on credit.

RURAL-BASED INDUSTRIALIZATION IN LAOS

181

In low-income countries, people are said to be highly vulnerable to income shocks. Money lent from traders functions as insurance for the smooth consumption of the poor. For a retailer, consumption loans are advanced to secure delivery of products from the weavers. Thus, a consumption loan achieves Pareto improvement both for the retailer and her weavers. A case of an itinerant merchant can be found in a village about 80 kilometers north of Vientiane. An itinerant merchant, Ms. P, deals with approximately 300 independent weavers of several villages under the putting-out system and collects about 50–100 pieces every day. She brings them to sell at Talaat Sao of Vientiane. The weavers know only a few patterns for knitting a vertical heddle, and thus weave cloth of indigenous patterns. There is no large demand for that line of goods. Itinerant merchants play an important role in facilitating market activities by dealing with weavers in remote areas who encounter difficulties in purchasing yarns and placing their products on the market. Itinerant merchants cover wider areas than master weavers do. However, wider coverage itself inevitably entails a drawback in that the itinerant merchant can share little social similarity with weavers. Since Ms. P comes from a village renowned for hand weaving, about 20 kilometers from the interview site, she possesses market information on patterns. However, she would not furnish that information to her weavers, because enforcing exclusive delivery from agents is difficult due to the absence of community-based trust. Thus, she stays in Zone D of Fig. 6.4. Some weavers under contract often sell their products to other itinerant merchants who offer higher prices, even though Ms. P has advanced yarns on credit. Thus, the cloth dealt with by Ms. P has to remain poor in terms of pattern. Regardless of agents' opportunism, yarns have to be provided to secure the collection of cloth because several itinerant merchants are doing business in the area. Commission and itinerant merchants are those who have no governance over the products. So long as these merchants do not furnish market information on patterns to weavers, they cannot be a full substitute for master weavers.

6.7 Conclusion We have observed that a variety of traders have created appropriate new strategies for channeling the products of small rural producers to the urban and even overseas markets. The major economic actors involved in the hand-weaving industry of Laos are weavers, master weavers, retailers, commission merchants and itinerant merchants. Retailers who are town-based traders, are passive in integrating rural and urban markets. It is rural-based traders, including the culture broker, who actively try to create the alignment of two markets. Several types of rural-based traders live in symbiosis in their respective niches in the marketing of hand-woven cloth.

182

AKIHIKO OHNO

Though itinerant and commission merchants deserve the name of market integrator, they only collect cloth from weavers in remote areas and place them on the urban market without furnishing market information on patterns to weavers. Those weavers know only a few indigenous patterns. Thus, the cloth they trade is of poor quality in terms of pattern. Retailers do not have governance over the products, except for a few retailers possessing antique cloth. Independent weavers can place their products on the market by themselves, however, because they reside near consumption centers. The pivotal market integrator for quality cloth in terms of patterns is the master weaver. They modify market information on patterns in vogue in urban markets and transfer it to weavers in the form of vertical heddles. In other words, master weavers can take technical initiatives in bridging information gaps between rural and urban actors. The emergence of master weavers is only a recent phenomenon brought about by progress in the commercialization of hand-woven cloth. Master weavers are differentiated from other traders in that they have the ability to create new patterns. In addition, they can count on community-based trust as insiders. Retailers, who are mostly outsiders, are subject to agents' opportunism since most of them have no such community mechanism of contract enforcement and do not have strong governance over the products. The institutional arrangement that facilitates market integration most effectively is the relational contracting system, since the principals who have a good access to market information can internalize the profits from it through the system. It should be noted that cloth of poor quality used to be transacted in local markets on a spot basis before the emergence of market integrators. As the local market becomes integrated in the national market spot market transaction is gradually replaced by the relational contracting system.155 This is contrary to the popular presumption that the putting-out system is archaic and that the spot market is a modern institution. The central issue in maintaining the relational contracting system is how effectively principals can check agents' opportunism. In the absence of formal sanction mechanisms, principals have to utilize informal sanction mechanisms relying on process-based trust, community-based trust and governance over the products. Principals, both master weavers and retailers, have different socio-economic capabilities in resorting to such measures. Different combinations of the three measures result in the emergence of plural trading classes with different contractual arrangements. Those who can exert the measures sufficiently to counter agents' opportunism

155

The shift from spot market transaction to the putting-out system in the early stage of economic development is often observed historically in many areas. See, Itoh and Tanimoto (1998) and Ohno and Kikuchi (1998).

RURAL-BASED INDUSTRIALIZATION IN LAOS

183

can introduce the putting-out system, while those who cannot have to conduct their business as a spot market transaction. The advance-order system lies in between. Our chapter suggests that community-based trust among people who share social similarity serves as an effective measure to deter agents' opportunism. Though the putting-out system is most appropriate to ensure exclusive delivery from the producers, the traders who are characterized as outsiders, namely most of the retailers, cannot adopt the putting-out system without suffering from agents' opportunism. The retailers who are incapable of exerting such measures have to introduce the advance-order system or even spot market transaction. Under these modes of transactions, however, principals would not furnish their private information on patterns since the exclusive delivery of the products is not secured. Therefore, unless effective market integrators such as master weavers emerge, production of the high-quality cloth that can meet the expanding demands from wide urban and foreign markets can grow only slowly. Another finding of this chapter is that the traders' dilemma often hinders the emergence of traders within village communities. Though small-scale master weavers can rely on community-based trust as insiders, they are subject to traders' dilemma. The large-scale master weavers, in contrast, enter into contracts with weavers with whom they share social similarity. They are differentiated from villagers in that they are educated and have urban experiences. Still, they can avoid agents' opportunism. Large-scale master weavers are well featured as culture brokers who exploit a rather ambivalent position as insiders as well as outsiders. It is likely that though market integrators emerge from insiders, those who have the ability to act as culture brokers will increase their relative role as the market expands. Our chapter suggests several facts about the association between the community and markets in economic development. In the Lao hand-weaving industry community relations facilitate market integration by deterring agents' opportunism. It should be noted that community relations are not all-powerful in conducting business with villagers under the relational contracting system. Community relations in combination with other measures, such as processbased trust and governance over the products, ensure efficient transactions for the relational contracting system. Economic development often destroys community since the emergence of traders and alternative employment opportunities attenuate the level of punishment. It should be noted that increasing uncertainty also jeopardizes trust between economic actors by inducing one-shot opportunistic behavior. Such a shock, which may not be brought about by economic development per se, may result in a rearrangement of a payoff matrix leading to the prisoners' dilemma. It is probable that the relational contracting system might revive if the Lao economy were stabilized. The stabilization of a macroeconomy

184

AKIHIKO OHNO

is thus important in preserving both process-based and community-based trust for economic development.

References Axelrod, R. (1984) The Evolution of Cooperation, Basic, New York. Connors, M. F. (1996) Lao Textiles and Traditions, Oxford University Press, Kuala Lumpur. Evers, H. and Schrader, H. (1994) The Moral Economy of Trade—Ethnicity and Developing Markets, Routledge, London. Geertz, C. (1963) Peddlers and Princes, Chicago, The University of Chicago Press. —— (1978) ‘The Bazaar Economy: Information and Search in Peasant Marketing,’ American Economic Review, 68(2): 28–32. Hayami, Y. (1997) Development Economics, Clarendon Press, Oxford. —— (ed.) (1998) Toward the Rural-Based Development of Commerce and Industry: Selected Experiences from East Asia, Economic Development Institute, World Bank, Washington. —— and Kawagoe, T. (1999) Middlemen in a Peasant Community: Vegetable Marketing in Indonesia, a paper presented to the workshop. Hobbes, T. (1962 [1651]) Leviathan, or the Matter, Forme, and Power of a Commonwealth Ecclesiasticall and Civil. Edited by Oakeshott, M., New York, Collier Books, Macmillan Publishing. Itoh, M. and Tanimoto, M. (1998) ‘Rural Entrepreneurs in the Cotton-Weaving Industry of Japan,’ in Hayami, Y. (ed.), Toward the Rural-Based Development of Commerce and Industry: Selected Experiences from East Asia, Economic Development Institute, World Bank, Washington, DC. Kandori, M. (1992) ‘Social Norms and Community Enforcement,’ Review of Economic Studies, 59(1): 63–80. Klein, D. B. (ed.), (1997) Reputation: Studies in the Voluntary Elicitation of Good Conduct, University of Michigan Press, Michigan. Kramer, R. M. and Tyler, T. R. (1996) Trust in Organizations, Sage Publications, California. Landa, J. T. (1994) Trust, Ethnicity, and Identity—Beyond the New Institutional Economics of Ethnic Trading Networks, Contract Law, and Gift Exchange, University of Michigan Press, Michigan. Landes, D. S. (1969) Unbound Prometheus: Technological Change and Industrial Development in Western Europe from 1750 to the Present. Cambridge, Cambridge University Press. —— (1986) ‘What Do Bosses Really Do?,’ Journal of Economic History, 46(3): 586–623. Milgrom, P. R., North, D. C., and Weingast, B. R. (1990) ‘The Role of Institutions in the Revival of Trade: The Law Merchant, Private Judges, and the Champagne Fairs,’ Economics and Politics, 2: 1–12. Neal, W. C. (1984) ‘The Role of the Broker in Rural India,’ in Robb, P. (ed.), Rural South Asia: Linkages, Change and Development, Curzon, London. Ohno, A. and Kikuchi, M. (1998) ‘Organizational Characteristics of Rural Textile Industries in East Asia,’ in Hayami, Y. (ed.), Toward the Rural-Based

RURAL-BASED INDUSTRIALIZATION IN LAOS

185

Development of Commerce and Industry: Selected Experiences from East Asia, Economic Development Institute, World Bank, Washington, DC. Siamwalla, A. (1978) ‘Farmers and Middlemen: Aspects of Agricultural Marketing in Thailand’, Economic Bulletin for Asia and Far East (UN-ESCAP), 29(1): 38–50. Thompson, P. and McHugh, D. (1990) Work Organization, Macmillan, Singapore. Weber, M. (1930) The Protestant Ethic and the Spirit of Capitalism, Allen and Unwin, London. Zuker, L. G. (1986) ‘Production of Trust: Institutional Sources of Economic Structure, 1840 to 1920’, Research in Organizational Behavior, 8: 53–111.

7 The Role of Business Networks in Market Development in Sub-Saharan Africa Marcel Fafchamps

7.1 Introduction In this chapter, we investigate the connection between business networks and community. We do so for markets that are at the early stages of their development, such as is the case in much of Sub-Saharan Africa. We begin by showing that, for firms above a minimum size, relational contracting is the rule in markets for agricultural products as well as for manufacturing inputs and outputs. Many markets in Sub-Saharan Africa can best be thought of as composed of a core network of trade relationships that are fairly stable over time. This core is surrounded by a fringe of atomistic competitive firms that operate with high transaction costs because they lack trust in each other. We document the many important roles that relationships play in facilitating market exchange, such as information sharing and the informal enforcement of contracts. These sets of trade and information sharing relationships can be thought of as business networks and we discuss the implications of this idea for our understanding of African markets. We then explore the role that community affiliation plays in the membership of business networks. We examine observed patterns of ethnic concentration and conclude that, at least in some African countries, these patterns are very unlikely to have been generated from a firm entry and growth process that grants equal access to business opportunities for all—even if we account for colonial heritage. From this observation, we conclude that entry into existing networks is biased. We discuss the possible sources and mechanism behind this bias. Based on the limited existing evidence, we argue that referral by family and friends is the most likely channel through which ethnic concentration arises. We also discuss the effect that business networks have on community formation.

7.2 Trust and Relationships Markets are normally thought of as decentralized allocation mechanisms and much has been written on the benefits to be derived from decentralization.

BUSINESS NETWORKS IN SUB-SAHARAN AFRICA

187

Markets, however, cannot exist without coordinated action if only to define and protect property rights. Whenever these services are not provided by the state, mafias and other private armies arise that thrive on protection money—and occasionally ambition to replace the state itself (North 1990). Even if property rights are properly defined and protected, market transactions also leave ample room for cheating (Fafchamps 1996). Economists typically focus on two important aspects of exchange: price and quantity. In real life, there are many other contractual dimensions that are equally if not more important to the parties: the quality of the product, the form and method of payment, the provision of credit, etc. In the absence of suitable mechanisms to deter cheating, exchange can only take a rudimentary form which elsewhere (Fafchamps 1998b) I have called a flea market economy: no placement of order, no invoicing or payment by check, no credit, and no warranty. Protection against opportunistic behavior enables agents to do business more easily, for instance by placing orders ahead of delivery, by securing product warranty, and by paying by check upon receipt of a monthly invoice.156 In theory, the legal system can ensure the enforcement of contracts. But in practice the deterrent effect of the legal system alone is insufficient when the size of the transaction is small or when the parties to the transaction have little or no assets to foreclose on. This means that legal institutions are unlikely to provide sufficient protection against opportunistic breach of contract in the overwhelming majority of market transactions that take place in poor countries. In everyday business, personal trust often is an effective substitute for the security provided by costless legal enforcement. In fact, when businessmen and women operating in poor countries are asked how they prevent opportunistic breaches of contract, they typically respond that they conduct business-like transactions only with individuals they can trust (Fafchamps 1996; Fafchamps et al.1994, 1995). With strangers, they revert to a cash-andcarry form of exchange: goods are inspected on the spot, and delivery takes place against instant payment in cash. The question then arises as to where trust comes from. In practice, it results primarily from a history of successful exchanges (Fafchamps 1996; Fafchamps et al.1994, 1995). Table 7.1 illustrates that relational contracting is the norm between African manufacturers and their suppliers: the length of the relationship exceeds seven years on average, and most supplies of a particular input come from a single supplier. The great majority of manufacturing firms places regular, e.g. monthly orders from their main suppliers. This is particularly true in manufacturing where firms have been

156

There even are commodities such as electricity for which exchange would be extremely cumbersome if not impossible if parties could not resort to periodic invoicing.

188

MARCEL FAFCHAMPS

Table 7.1 Loyalty to Suppliers in African Manufacturing Burundi 7.0

Length of relationship with supplier (years) Percent of in- 76.9 put from single supplier Firm places 62.8 regular orders

Cameroon 7.5

Ivory Coast 7.9

Kenya 9.3

Zambia 9.5

Zimbabwe 14.6

70.0

77.9

61.2

64.8

78.5

78.6

78.9

83.0

84.4

85.3

Source: Bigsten et al. (1998). Micro-enterprises excluded from samples.

shown to be extremely loyal to their main suppliers even when they have a choice among various sources of supply (Bigsten et al.1998). One reason is of course that a manufacturer's equipment and production plans are typically optimized around very specific inputs and delivery systems. If input specificity were the only reason for long-term relationships, one would not expect relational contracting to be prevalent in agricultural markets. Evidence to the contrary can be found in Gabre-Madhin (1997) and Fafchamps and Minten (1998a). Repeated exchange can be seen as a way of economizing on the costs of establishing personal trust. In certain markets—labor or credit, for example—repeated interaction is so prevalent that it has become the norm. These are typically markets for which either the potential for abuse is greatest and the screening costs are largest. Relationshipbased networks are therefore expected to arise in markets where screening costs are high and personal trust is a substitute for external enforcement through lawyers and courts (Hayami and Kikuchi 1981; Ghosh and Ray 1996; Fafchamps 1998b). Before proceeding further, we must specify what we mean by network—the term not yet having found its way into economic jargon. The theory of networks was first proposed by sociologists to transcend the concept of community, which was thought too vague, and no longer adequate (Mitchell 1969). Borrowing notation and concepts from the mathematical theory of graphs, individuals came to be regarded as nodes and the relationships between them as links. The combination of nodes (individuals) and links (relationships) constitutes a graph or network.157 Links can be of

157

Mathematicians give the term ‘network’ a more specific sense than that used in sociology (Foulds 1992). In this paper, we follow sociologists (and Standard English) and use the term ‘network’ to refer to what mathematicians call ‘graphs.’ Sociologists also distinguish between subjective networks (the graph of all links emanating or leading to a specific individual) and objective networks (the graph of all the links between a set of nodes or individuals). While subjective networks are always well defined, objective networks are often ‘open’ in the sense that links exist between members of the set and outsiders, e.g. between businessmen in Africa and Europe. Here we use one meaning or the other, depending on the context.

BUSINESS NETWORKS IN SUB-SAHARAN AFRICA

189

various kinds (e.g. trade, family, information sharing) and of various strengths (e.g. trade amount, closeness of family relationship). Links can be unidirectional, as when one individual knows another, but the reverse is not true (e.g. a movie star). They can be superposed, in which case sociologists speak of multiplex societies (Gluckman 1955; Basu 1986). They can also vary over time. Laid out in these general terms, the concept of network is nothing but a way of visualizing social relations of any kind. To give it life, one must specify a population of individuals or nodes, a particular kind of link, and a time frame. One could, for instance, construct the graph of all the trades that occurred between a set of economic agents during a particular time period, and call it a market or trade network. Alternatively, one may consider only certain types of transactions, e.g. repeated transactions, to generate the relational contracting network. In both cases, the resulting networks would just be a graphical depiction of the market itself, perhaps useful to the study of inter-industry linkages, but not substantively different. The concept of network comes to full fruition when the definition of a link is broadened to include other useful economic functions such as informal contract enforcement, information sharing, and the like. Abundant evidence indeed exists that such links are central to the functioning of labor, credit, and insurance markets (Udry 1990, 1994; Fafchamps 1992). In the words of Granovetter (1995), economic transactions are embedded in a social context. Mutual insurance among villagers, for instance, is now believed to revert primarily around long-term self-enforcing arrangements (Coate and Ravallion 1993; Fafchamps 1992, 1998c; Fafchamps and Lund 1998; Ligon et al.1997). In Sub-Saharan markets, personal relationships perform a variety of valuable functions. Fafchamps and Minten (1999), for instance, report that agricultural traders surveyed in Madagascar perceive relationships as the most important factor for success in business. The authors document the extent to which relationships facilitate the circulation of information about prices and market conditions, the provision of trade credit, the prevention and handling of contractual difficulties, the regularity of trade flows, and the mitigation of risk. Of these, the regularity of supply and demand and the sharing of risk appear particularly important. Fafchamps and Minten (1998b) further demonstrate that the productivity of traders depends on the number and type of relationships they have: controlling for all other inputs, a doubling of the number of known traders raises value-added by 28 percent; a doubling of the number of people who could help in times of trouble similarly raises it by a similar amount. Larger

190

MARCEL FAFCHAMPS

and more prosperous traders are thus those with quantitatively and qualitatively better relationships (see also Barr 1998a). Relationships can also help economize on search costs. Using a stylized model of Exchange, Kranton (1996) illustrates how trade within networks reduces search costs and can drive goods away from market exchange. Montgomery (1991) similarly shows that firms can use employee referral to identify and hire high quality workers. There is plenty of evidence that networks and relationships perform a matching and screening function in many developed as well as developing countries. Referral is a common hiring procedure, in the USA (Granovetter 1995) as well as in Africa (Velenchik 1995). Reportedly, employee referral is extensively used in the Dacca garment sector (personal communication from Junaid Ahmed). Similar practices are found in commodity markets as well. In their description of livestock markets in West Africa, for instance, Eddy (1979) and Staatz (1979) document the role that personal networks play in matching buyers and sellers. A similar function is attributed to networks in the case of long-distance African trade by Meillassoux (1971) and Amselle (1977). In a recent study of Ethiopian grain traders, Gabre-Madhin (1997) describes in detail the role of brokers in bridging surplus and deficit areas and in assisting traders screen grain quality and identify reliable buyers and sellers. Networks of personal relationships also serve to circulate information. Barr (1998a), for instance, demonstrates that Ghanaian manufacturers with better business contacts perform better than less well-connected firms. She interprets the evidence as consistent with the idea that large firms use networks to access information about new technologies and market opportunities while small firms rely on closely knit networks of mutual insurance. In a related paper, Barr (1998b) discusses other possible roles of information sharing. One such role that has received much emphasis in the theoretical literature is reputation. Kandori (1992) formalizes the idea that if information about breaches of contract circulates freely among economic agents, breach can be deterred by excluding cheaters from future trade. Milgrom et al. (1991) apply this idea to medieval trade in Europe. They argue that the Law Merchant of Champagne Fairs served the role of information repository, thereby making coordinated exclusion possible even among agents who did not know each other. Using ancient correspondence between traders, Greif (1993, 1994) extends a similar concept to medieval trade in the Mediterranean Sea area and argues that ethnic and religious networks served an important information sharing role that helped discipline trading agents residing in distant cities. Fafchamps (1996), Fafchamps et al. (1994), and Fafchamps et al. (1995) provide contemporary evidence regarding the role of reputation in enforcing contracts among African manufacturers. They show that referral by other firms is an important screening mechanism and that clients who can provide reliable references get supplier credit faster and more easily. This is

BUSINESS NETWORKS IN SUB-SAHARAN AFRICA

191

not without consequences: Fisman (1998) indeed shows that African manufacturers who get trade credit have a significantly higher level of capacity utilization. Not all firms are able to use referral to facilitate screening, however, for lack of personal contacts with other firms. This is particularly true among small firms and firms headed by individuals with loose ties to existing businesses (Fafchamps 1998a). Fafchamps and Minten (1999) find similar results among Malagasy grain traders: if anything, information sharing is even less developed there than in African manufacturing. In addition, none of these studies uncovers evidence of the systematic exclusion of cheaters from future trade. One possible explanation is that loosely organized market networks lack the coordination mechanism required to trigger expulsion. Furthermore, these studies find no evidence that firms that deal with cheaters get ostracized. Without metapunishment, it is unclear how a reputational equilibrium (as discussed by Kandori 1992) can be sustained. To explain the role of information sharing as a contract enforcement device together with the absence of a coordinated punishment mechanism, Fafchamps (1998b) proposes an imperfect information model of an exchange economy with good and bad agents. Bad agents reveal their type by cheating. Identifying good agents is therefore costly. Contracts are enforced not by the fear of exclusion but by the desire to economize on the cost of screening out bad agents, as in the models of Shapiro and Stiglitz (1984) and Ghosh and Ray (1996).158 Agents form long-term relationships to minimize screening costs (Hayami and Kikuchi 1981). Sharing information about past contractual history enables good agents to identify each other faster. Better connected agents are thus at an advantage. Fafchamps (1998b) demonstrates that, in such an environment, the exclusion of cheaters is not necessarily sustainable. For exclusion to be part of an equilibrium, agents must believe that good firms that cheat have turned bad, e.g. are going bankrupt. A sufficiently high level of ‘churning’ of firms is thus essential to the emergence of a decentralizable reputational equilibrium (as in Kandori 1992). Fafchamps (1998b) further shows that a reputational equilibrium is unlikely at early stages of market development. Long-term relationships also facilitate interlinking. Lorenz (1988), for instance, explains how relationship-based subcontracting among French manufacturers enables both buyer and seller to make long-term investments in each other's production process. Hart (1995) generalizes this concept to all incomplete contracts and argues that relationships provide a possible solution to the well-known ‘hold-up’ problem (Williamson 1975, 1985). Fafchamps et al. (1994) and Fafchamps et al. (1995) illustrate how

158

Dutta et al. (1989) use a similar disciplining device whereby workers with a long-term contract who become casual workers with lower pay and/or security of employment. (See also Kranton 1996.)

192

MARCEL FAFCHAMPS

these principles apply to African manufacturing and enable parties to place orders, pay by check, provide supplier credit, and obtain warranty. The typical supplier–client relationship is one in which clients pay suppliers for fear that supplies will not be delivered, and suppliers deliver for fear that clients will not pay. Not only do relationships enable agents to enter into multi-faceted transactions involving various forms of forward contracting (e.g. order, credit, warranty), they also facilitate insurance and finance. For instance, Barr (1998b) argues that small Ghanaian firms rely on their network of contacts to secure insurance against liquidity crises. Similar findings are reported in Fafchamps et al. (1995) and Fafchamps and Minten (1999) who argue that access to quick credit is essential for firms to survive liquidity crises. Fafchamps and Minten (1998b) further demonstrate that the performance of Malagasy grain traders depends critically on the number of people on which they can rely for assistance in financial emergencies. Relationships can also facilitate credit and equity financing. Finally, Fafchamps et al. (1995) document the practice of shared inputs and machinery among competitors who belong to the same network. One form of interlinking that is particularly important in difficult economic environments is contractual flexibility. If penalties for breach of contract are too lenient, opportunistic breach cannot be deterred and contracting cannot take place. At the same time, rational economic agents should refuse to incur contractual obligations if the penalty for breach of contract is too strict (Fafchamps 1996). The reason is that unforeseen yet unverifiable events can occur that make compliance either very costly or outright impossible. Consequently, if the cost of contractual compliance is hard to predict, it may be impossible to find penalties that deter cheating without discouraging agents from engaging their contractual responsibility. In this case trade does not occur, unless agents can use relationships to provide sufficient contractual flexibility, that is, to insure parties against extreme compliance costs. Bigsten et al. (1998) provide evidence that contract flexibility is an essential feature of industrial input and output markets in Sub-Saharan Africa: most disputes with suppliers and clients are resolved amicably through direct negotiations and trade is resumed between the parties. This finding is in agreement with the inherent riskiness of manufacturing production in poor countries (Collier and Gunning 1997). Fafchamps and Minten (1999) report similar findings for grain markets in Madagascar. Of course, contract flexibility requires economic agents to anticipate delayed payments and deliveries by building up inventories and liquidity reserves. Fafchamps et al. (1997) show, for instance, that Zimbabwean industries accumulate inventories of inputs in response to late delivery risk. There are yet other ways by which the coordinated and uncoordinated actions of groups of individuals can affect the development and functioning

BUSINESS NETWORKS IN SUB-SAHARAN AFRICA

193

of markets. They can also help individuals coordinate their actions to generate a variety of public goods such as the provision of common infrastructure and institutions, and the lobbying of government and local authorities for preferential treatment and for supportive laws and institutions.159

7.3 Business Networks and Contract Enforcement We have seen that relationships play a key role in facilitating exchange in Sub-Saharan Africa. These relationships differ from pure market exchange in that they perform economic functions other than trade itself, such as information sharing, informal enforcement of contracts, and interlinking. To discuss the role that these relationships play in market development, let us define a business network as the set of such relationships with the agents they involve. Armed with this definition, we now examine in more detail the market configurations that naturally arise from the interaction between markets and business networks. To focus the discussion, we organize the presentation around one particular function that has recently received a lot of attention in the literature, namely, contract enforcement. In the preceding section we have argued that networks can facilitate the establishment of trust by circulating essentially two distinct pieces of information: whether an agent has ever successfully completed a contract, and whether an agent has ever breached a contract. Both types of information refer to the transaction history of an agent and they have generally been treated interchangeably in the literature. But the latter is in general more demanding than the former: it is fairly easy to ascertain whether a carpenter has ever manufactured a chair, but it is harder to ascertain whether he or she has ever failed to complete an order for a chair. Similarly, it is easy to ascertain whether someone has ever paid a supplier; it is harder to assess whether the same person has ever paid a supplier late or not at all. Fafchamps (1998b) has shown that the characteristics of decentralized market exchange vary dramatically depending on whether the first or the latter type of information are circulated. Information sharing can play a positive role in facilitating exchange because the circulation of information about past contractual performances helps agents screen each other more cheaply.160

159

As has been amply demonstrated by sociologists, for agents to voluntarily contribute to the action of a group, they need to identify with it. The process of group identification is not, however, central to this chapter, so we shall abstract from it and use the terms group and community interchangeably.

160

Provided the information circulated within the community is reliable. In contrast to what is assumed by some theorists (e.g. Kandori 1992), gossip is a notoriously unreliable source of information. Not only is information distorted and imprecise, gossip can also be manipulated to hurt competitors. Gossip can be a channel of information or disinformation alike. This is why business men and women insist that they must know and trust the source of the information before giving it any credence.

194

MARCEL FAFCHAMPS

These theoretical results demonstrate that networks can facilitate market exchange in a purely decentralized manner, that is, without any collective action. It is clear, however, that networks sufficiently compact to form identifiable communities of common interest or ‘clubs’ may mobilize to achieve better outcomes for themselves (see Aoki, this volume). Clubs can, for instance, organize the joint punishment of cheaters through ridicule, ostracism, or worse. To the extent that economic agents socialize and intermarry with members of their business community, rejection by business partners may impose an additional personal burden as well (Aoki et al.1995; Spagnolo 1995): exclusion from social interaction such as marriage within a prosperous group is likely to be detrimental to the long-term prosperity of cheaters. Although these ideas have received much attention in the theoretical literature, evidence of collective punishments for opportunistic breach of contract is weak at best. In their studies of African manufacturing and grain trade, Fafchamps et al. (1994), Fafchamps et al. (1995) and Fafchamps and Minten (1999) show that non-payment of a supplier does not automatically lead to a loss of credit from other suppliers. This does not mean that ostracism is never used: in the course of a manufacturing sector survey in Kenya, the author met an Asian businessman who was shunned by his peers following a fraudulent bankruptcy. But such cases are very rare.161 Furthermore, in countless interviews with African manufacturers and traders of all ethnic origin, never did respondents articulate the implicit obligation to participate in a collective punishment. One Ghanaian trader in the Accra lumber market—a market known for its tight ethnic-based trading community—was asked whether he would oppose neighboring traders dealing with cheaters. His answer was: ‘If they want to deal with cheaters, it is their problem, not mine’—hardly the kind of answer one would expect if an explicit collective punishment strategy were in effect. Further discussions with respondents suggest that non-payment of a supplier triggers loss of credit only when it is interpreted as a signal of financial difficulty and possible bankruptcy. This idea is formalized by Fafchamps (1998b) who shows that collective punishment can arise in a fully decentralized manner without any explicit collective strategy, provided there is sufficient ‘churning’ of firms. Communities may also develop norms of behavior—a ‘business culture’—that are deemed suitable for the conduct of business (Hayami and Kikuchi 1981). Adherence to these norms of behavior may be further publicized through participation in religious rituals and membership of religious groups that reinforce respect of these norms. The role of these cultural norms is, for instance, documented for medieval trade in Greif (1993,

161

Shunning may be more frequent for other social transgressions such as female adultery, for instance.

BUSINESS NETWORKS IN SUB-SAHARAN AFRICA

195

1994). In their detailed analysis of the Sefrou market in Morocco, Geertz et al. (1979) illustrate the role that Muslim brotherhoods play in helping market communities join forces for a common purpose. In her study of market emergence in pre-colonial Kenya, Ensminger (1992) documents the fact, historically, that traders who wanted to do business with coastal Arab traders and their descendants found it easier if they converted to Islam. Similar processes have been documented elsewhere in Africa (Shillington 1989; Cohen 1969). Poewe (1989) reports comparable motivations behind the spread of evangelical churches in present day Zambia. Although there is some evidence that business communities often pride themselves in having superior ethics, there is also an inordinate amount of circumstantial and anecdotal evidence that demonstrates that morality is an elastic concept. The truth is that few if any human societies are free from the temptations of greed and instant gratification. Although a moral code can probably discourage outright fraud and treachery, social and economic pressures are a more effective deterrent against opportunistic breach of contract than ethics alone.162 Another possibility is that business communities or ‘clubs’ strive to build a reputation as a group with respect to other business communities with whom they trade. This idea has been formalized by Kandori (1992) and Ellison (1994). In this case, trade takes place across communities and the benefits from repeated interaction are captured not through individual relationships but through community relationships. The reputation of the group becomes an asset that serves as collateral to guarantee good behavior. Each community may then choose to punish its own members for opportunistic breach of contract. Such situations have been described for ancient trade across ethnic boundaries—e.g. between Europeans and Africans along the Zambezi river (e.g. Shillington 1989; Greif 1994; Geertz et al.1979). Although such processes may have been at work in ancient times (see Greif, this volume), recent work on manufacturers and traders in Africa has found no recent evidence that communities punish their members for cheating individuals in other communities (Fafchamps 1996; Fafchamps et al.1994, 1995; Fafchamps and Minten 1999). If anything, cheating is perceived to be more prevalent across ethnic or religious boundaries and trust between communities is often low. Irrespective of whether culture and ethics should be considered an important contribution to business acumen, it remains true that much social capital that is immediately relevant for market development and industrialization or modernization in poor countries is embedded within business networks and communities.

162

See, however, Platteau (1994a and b) who argues that during childhood individuals acquire, through what psychologists call ‘secondary socialization,’ what will constitute the moral fabric of their lives.

196

MARCEL FAFCHAMPS

7.4 Business Networks, Competition, and Firm Entry The existence of business networks also has some less desirable consequences. First of all, thanks to information sharing and joint punishment of deviant behavior, members of important and prosperous communities have a comparative advantage relative to others in the conduct of business. Evidence to this effect is provided by Barr (1998a), Fafchamps (1998a) and Fafchamps and Minten (1998b). Moreover, the more help and information agents receive from their community in their effort to screen each other, the less willing they become to screen individuals from outside their community. Economic agents in general prefer to deal with members of their own community (Macharia 1988). This process may explain why established Zimbabwean manufacturers, who for historical reasons are mostly of European and Asian origin, appear reluctant to deal with African firms in spite of good courts and widespread information sharing (Hoogeveen and Tekere 1994; Mumbengegwi 1994; Risseeuw 1994; Fafchamps et al. 1995; Fafchamps 1998a). As a result, prosperous communities have a tendency to reproduce themselves over time and to reinforce their grip on business—at least as long as they maintain their cohesion (Himbara 1994). To the extent that membership to these communities is restricted and that members intermarry, social mobility is likely to be impaired as well. This reduces efficiency because entrepreneurs end up being selected from a small percentage of the population. Kenyan Asians, for instance, constitute 1–2 percent of the Kenyan population, yet they own the majority of light industries (Fafchamps et al.1994). Table 7.2 shows that similar patterns are observed in other African countries (Bigsten et al.1998). The reader may wonder whether these patterns simply result from colonial heritage. While there is little doubt that colonial policies163 favored non-African firms, it is fairly clear that the ethnic concentrations shown in Table 7.2 could not have arisen from unbiased sampling since these policies were removed. Table 7.2 Percentage Ethnic Composition of the Ownership of African Manufacturing Firms Burundi African origin 81.7 European ori- 5.8 gin Asian origin 12.5

Cameroon 79.3 16.9

Ivory Coast 57.0 33.0

Kenya 41.7 3.6

Zambia 59.5 13.0

Zimbabwe 29.4 49.7

3.8

10.1

54.7

27.6

20.9

Source: Bigsten et al. (1998). Micro-enterprises excluded from samples.

163

Including policies conducted during the Unilateral Declaration of Independence era in Zimbabwe.

197

BUSINESS NETWORKS IN SUB-SAHARAN AFRICA

To see why this is the case, let the share of firms owned at time t by members of a previously favored ethnic group be denoted η . Assuming for simplicity that the number of firms is constant164 and that firms get replaced at a constant t rate θ, the law of motion of η can be written: t (1) where ρ is the proportion of the favored ethnic group in new firms. If members of all ethnic groups are equally likely to start a new firm, then ρ should be equal to the proportion of the previously favored ethnic groups in the total population. Given our assumption that the number of firms is constant, an approximation for θ is obtained as one over the average age of firms. Table 7.3 reports the value of ρ implied by the current ethnic composition of manufacturing in Africa, assuming that at independence all firms were owned by non-Africans. The results are inescapable: in all six surveyed countries, the implied sampling rate for non-Africans exceeds by a wide margin their proportion in the total population—which in all cases is not larger than 1 or 2 percent. According to these simple calculations, the inbreeding bias is largest in Kenya and Ivory Coast. In spite of their crude character, these findings make us suspect that the average entrepreneurial talent is below what it could be if all agents had an equal chance. This outcome is also inequitable in the Jeffersonian sense of equal opportunity for all. Finally, lack of social mobility is likely to be Table 7.3 Estimate of Ethnic Bias in Firm Entry Burundi Time between 34 independence and survey Average age of 11 firms Proportion of 18.3 non-African owners at survey (percent) Implied pro- 14.6 portion of entry by nonAfricans (percent)

Cameroon 34

Ivory Coast 34

Kenya 30

Zambia 31

Zimbabwe 15

12

15

19

18

25

20.7

43.1

58.3

40.6

70.6

17.0

36.6

59.5

27.6

28.5

Source: Bigsten et al. (1998). Micro-enterprises excluded from samples.

164

This assumption is not too much of an oversimplification given that very little growth has taken place in African manufacturing since independence. Moreover, if the number of firms is increasing, one should observe even less ethnic concentration, hence reinforcing our conclusion.

198

MARCEL FAFCHAMPS

reinforced by social stratification through ethnicity, language, caste, education, etc. For all these reasons, the existence of strong business communities often creates social tensions, the outcome of which can be the elimination of these communities and the dilapidation of the social capital they represent. History indeed abounds with examples of business groups earmarked for public retribution with or without the participation of the state. Given the importance of these issues, it is worth investigating in more detail the process by which business networks reinforce themselves over time and end up excluding non-members from business exchanges. One point that has not been adequately recognized until now is that the circulation of information among agents is detrimental to entry. The reason is that, when information on established agents is widely available, economic agents may rationally choose to wait until they meet an established agent about whom they can obtain information rather than spending resources screening a new, unknown agent (Fafchamps 1998b). In fact, the better information circulates, the harder it is for newcomers to be screened. Economies with well established information sharing networks are thus particularly inimical to new entrants as opposed to economies where information does not circulate and agents are forced to screen whoever shows up at their doorstep. The contrast between Ghana, Kenya, and Zimbabwe manufacturing is, in this respect, quite telling. Ghana manufacturers form a diverse group without a strong sense of community (Cuevas et al.1993; Fafchamps 1996). This is, in part, the result of government efforts to curtail the economic and political influence of certain groups, e.g. SyroLebanese and Asante business communities. As a result, information sharing is rudimentary and no single group is advantaged. This is not to say that networks of personal relationships are unimportant—Barr (1998a) shows that they are—but the intersection of these personal networks does not constitute an ethnically homogeneous community. In contrast, Kenyan manufacturing is dominated by entrepreneurs of South Asian origin—often second or third generation immigrants (Himbara 1994; Marris 1971; Fafchamps et al.1994). Although in the 1960s the Kenyatta government tried to prop up Kikuyu businesses, this policy failed to durably influence the ethnic composition of business (Himbara 1994). Kenyan Asians do not constitute a monolithic community, however. Survey respondents identified at least four distinct communities—‘the Shahs, the Patels, the Sikhs, and the Ismaelians’ as one respondent put it.165 Within these communities, information circulates rather freely and client

165

The last two of these groupings are based on religious affiliation while the first two loosely correspond to Indian castes, in spite of the fact that Kenyan Asians do not, in general, follow caste distinctions. Respondents nevertheless pointed out that it is fairly easy for someone to change his or her name to Shah or Patel. The caste nature of the first two categories is thus quite blurred.

BUSINESS NETWORKS IN SUB-SAHARAN AFRICA

199

referral is a common practice. The survey uncovered little or no evidence of similar networks of information sharing among native entrepreneurs—at least in manufacturing.166 As a result, Asian businesses are at an advantage. Because information circulates only in an informal manner and networks are somewhat segmented, firms cannot rely purely on referral to identify reliable clients and suppliers; they must also screen agents from outside their network. Consequently, it is possible for outsiders to gain acceptance in the business community. Entry then takes place through a lengthy probation process by which small amounts of credit are given to test the resolve and honesty of the prospective credit recipient (Fafchamps 1997). Zimbabwe presents yet another configuration. Except for very small businesses which are overwhelmingly in the hands of native entrepreneurs, most manufacturing firms are in the hands of people of European and Asian descent (Daniels 1994; Bigsten et al.1998, Table 7.2). Although this pattern of ethnic concentration was initiated by deliberate pro-white policies pursued during the colonial period and during the 1964–79 Unilateral Declaration of Independence era, it is unclear why it has survived to this day. One contributing factor may, paradoxically, have been the presence of an active and widely used credit reference bureau which quickly developed in the 1980s. The existence of this bureau means that information on bad payers circulates widely in the economy—well beyond the confines of personal networks (Fafchamps et al.1995). While this undoubtedly helps established businesses screen buyers and sellers, it also penalizes new businesses—especially those entrepreneurs unfortunate enough not to have been born within the existing business community. As a result, information sharing may in fact have ‘frozen’ the ethnic composition of Zimbabwean manufacturing. Fafchamps (1998b) investigates the theoretical conditions under which such an outcome may arise and shows that it is most likely to arise when there is little firm renewal. This condition is, by and large, satisfied in Zimbabwe where manufacturing firms are, on average, much older than in other African countries (Table 7.3). Setting up a credit reference bureau may thus have been detrimental to business entry. If newcomers find it difficult to enter, one must then ask the question of how networks renew themselves over time. One possibility is no renewal: membership of the network is constant; the business community is a closed group. Such an outcome is more likely when opportunities for gains from trade are stable over time and the population of potential buyers and sellers does not change—as was more or less the case in Zimbabwe

166

Why this is the case is unclear. Observations from the field suggest that African entrepreneurs commonly seek to establish personal relationships with existing businesses, which are predominantly Asian, instead of attempting to create a concurrent network. Granovetter (1995), in contrast, provides examples of the development of separate business networks among immigrant minorities in the US.

200

MARCEL FAFCHAMPS

over the 15 years following independence. By the same token, markets dominated by closed groups are more likely to arise for trade flows driven by static comparative advantage—e.g. primary commodities, agricultural staples, protected manufacturing goods. This may explain why long-distance trade in pre-industrial societies is often found in the hands of a tightly knit community (Greif 1993; Braudel 1986). In contrast, closed markets are unlikely for commodities that are subject to constant innovation and entry by Schumpeterian competitors, such as Silicon Valley. In these markets, constant reshuffling of firms and agents ensures that refusing to deal with unknown firms is uneconomical; free entry is more likely to arise in equilibrium. An immediate corollary of the above is that closed-shop markets are more likely in poor, stagnant economies where patterns of trade remain dominated by primary commodities. This is precisely what we observe in Africa. There is also room for an intermediate solution which is for network members to coopt new members. The advantage of this solution for the group is that new entry is reduced and competition minimized, thereby increasing the returns to the group's social capital while ensuring that sufficient entry takes place for the group to reproduce itself. Cooptation takes many different forms and raises a host of interesting issues. One possible form is for an established agent to screen a newcomer and then share the result of this test with others. A newcomer who successfully passes the test is then allowed to join the group—although he or she may not necessarily gain full access to information sharing. The client referral system described above for Kenya and, to a lesser extent, for Ghana essentially falls into this category (Fafchamps 1996; Fafchamps et al.1995). One difficulty of this kind of arrangement, which has been discussed in the finance literature (Lang and Nakamura 1990), is that sharing the result of the screening test may generate free riding: efforts by the testing agent to recoup screening costs from subsequent transactions may fail if the tested agent can immediately switch to another partner. In response, agents who perform the screening may seek to attach the tested agent for a minimum number of transactions. Examples of this strategy can be found in banks securing all the collateral of new borrowers to ensure they do not switch to another lender. Recourse to collateral is essentially unheard of in supplier credit, however (Fafchamps et al.1994, 1995; Bade and Chifamba 1994). Cooptation may also take place before testing has occurred. Nepotism is one such form of cooptation whereby a member of the community with no prior experience is recommended for preferential treatment, such as credit without screening, or a new job without a trial period. Although nepotism is incompatible with the principle of equal opportunity for all—and is often stigmatized for this reason—it may represent an efficient way for a network to renew itself. The precise conditions under which nepotism is

BUSINESS NETWORKS IN SUB-SAHARAN AFRICA

201

individually rational need to be ascertained but intuitively nepotism is efficient for the group whenever, thanks to network externalities, an average person from within the community generates more returns for the group than a high performance outsider. As to why this is the case may result from a variety of mechanisms, such as better exchange of information with other members of the group, easier monitoring of compliance with contractual obligations, extra sanctions for deviant behavior, and so on. Anticipation that poor performance will be harshly punished ought to discourage below average community members from seeking promotion through nepotism, thereby reducing adverse selection and false pretenses. Field observations suggest that nepotism is a reality, although it is unclear how important it is as a source of new entrepreneurs (Macharia 1988; Himbara 1994; Fafchamps et al.1994, 1995). These issues deserve further investigation. It occasionally occurs that several distinct communities compete in the same markets. Bigsten et al. (1998), for instance, reports that while ethnic concentration in manufacturing is strong in some African countries, in others several communities appear to be competing equally. Intuitively, in the absence of external intervention, the community whose social capital generates the largest private gains and cost reduction should grow at the expense of less efficient communities. Whether the long-term configuration of business involves one or several communities depends on whether the accumulation of social capital generates increasing or decreasing returns to scale. If returns to social capital are monotonically increasing with group size, then a single group should eventually dominate the market. If returns to social capital are monotonically decreasing with group size, exchange should remain atomistic; communities should eventually disappear. If returns are initially increasing then decreasing, there is room for one or several communities depending on market size. Returns to group size might eventually drop because the cost of information circulation increases exponentially with group size. It may also be that larger groups cannot impose social sanctions on to deviant members because they lack the capacity to set up meta-punishments, that is, punishments for those who refuse to ostracize past cheaters. Whatever the reason, if there are increasing returns to group size, one group should dominate. Which group dominates, however, may be indeterminate. In this case, history matters: favoritism by governments and colonial administrations can give one group a head-start, hence giving it an advantage that is subsequently difficult to shake (Himbara 1994; Shillington 1989). In other cases, historical accidents and relatively minor differences between groups can give one community a small initial advantage that gets reinforced over time. Expatriate communities seem to form a natural candidate for the formation of successful business communities, although there are many counterexamples as well. One possible explanation is that expatriate

202

MARCEL FAFCHAMPS

communities are, at least in part, the result of self-selection: only the most determined and the most ambitious migrate abroad in search of economic success. Expatriate communities are also often subject to residential and occupational restrictions that force them into certain neighborhoods and activities, thereby facilitating the circulation of information and raising the cost of exclusion from the group. Agriculture, for instance, is not a politically feasible option for Kenyan Asians (Himbara 1994). The same is probably true for Syro-Lebanese entrepreneurs in West Africa.

7.5 The Origin of Business Networks In the preceding section, we have remarked that business networks display a degree of ethnic concentration that is unlikely to have arisen from random matching of agents in the population at large. This raises the issue of the origin of business networks and the particular role that ethnicity plays in this respect. We begin by noting that, although ethnic identity is a fluid and constantly evolving concept, no one would deny that identification with a particular ethnic group is a very strong emotion that can drive behavior that is otherwise irrational and destructive, such as civil war. Ethnicity is an important—and often ignored—issue in market development because unsupervised markets often develop a strong ethnic bias. It is nobody's secret, for instance, that many business men and women in Indonesia, Malaysia, and Singapore are ethnic Chinese. In much of East Africa, business is in the hands of ethnic South Asians while in West Africa, individuals of Lebanese and Syrian origin dominate sectors of activity such as import–export trade. Although it might be individually rational, ethnic bias is socially inefficient, in the Pareto sense; it is also inequitable and is an important source of political tension. The question then is: where does ethnic bias come from. Fafchamps (1998a) investigates the issue in Kenya and Zimbabwe manufacturing and concludes that the two most convincing sources of ethnic bias in market exchange are statistical discrimination and network effects. Ethnicity and community effects tend to reinforce each other whenever membership to the dominant business group is partly determined by one's ethnic origin. The interface between ethnicity, communities, and economic exchange is perhaps the most explosive issue pertaining to markets. Governments interventions in the functioning of markets have nearly always been motivated, implicitly if not explicitly, by resentment towards particular business communities and ethnic groups (Bauer 1954; Jones 1959). The restrictions on entry by non-community members that naturally derive from the development of dominant business groups nearly always generate suspicion of collusion, discrimination, and unfair business practices. When

BUSINESS NETWORKS IN SUB-SAHARAN AFRICA

203

politicians become convinced that markets only serve the interests of a small prosperous minority, they are prompt to respond to economic and social crises by market repression and direct government intervention. The widespread use of roadblocks and other restrictions to the spatial movement of goods is perhaps the more vivid—and most stupid—manifestation of governments' mistrust of the capacity of the market to serve the interests of producers and consumers. Whatever the cause for ethnic bias in market activity, failing to address the issue adequately is likely to result in policy reversal and the abandonment of market liberalization. Finding ways of ensuring non-discriminatory markets is thus essential for sustained market-based economic development.167 To answer this question, we must understand the precise process that gives rise to ethnic bias. This raises the issue of whether business networks and communities organize around pre-existing social constructs or create their own, new groupings. It is often believed that family and lineage play an important part in the formation of business networks. This belief is in part based on the observation that Third World autocrats often favor businesses owned by relatives.168 Survey results from Africa paint a more nuanced picture. Family relationships are important in providing start-up capital and some initial business contacts (Hoogeveen and Tekere 1994; Bade and Chifamba 1994). They may also be a source of equity finance, possibly because information flows within the family facilitate monitoring (Fafchamps et al. 1995). For similar reasons, partnerships and other forms of joint ownership are most commonly based on close family ties, except in very large firms (Fafchamps et al.1995). This situation is not peculiar to Africa but seems to be pervasive in developed economies as well. There is, however, no evidence that family relations play a role in market exchange, contrary to what is often believed (Granovetter 1995). Trade with relatives and friends is extremely rare (Bigsten et al.1998). Whenever it happens, it has a negative effect on firm performance, probably because trade with family members blurs the boundaries of the firm (Fafchamps and Minten 1998b). Discussion with survey respondents further indicates that entrepreneurs find it difficult to keep business with relatives within the confines of a sales transaction. Many respondents, for instance, emphasize that it is difficult to collect payment from relatives (Fafchamps and Minten 1999). One Ghanaian carpenter put it best when he stated that ‘dealing with relatives is the surest way to go out of business’. This interpretation is reinforced by Fafchamps and Lund's (1998) work on mutual insurance in rural Philippines: the authors show that, among close relatives, risk sharing is more likely to take the form of gifts

167

The same can be said of gender bias.

168

The last example to receive public attention is that of business interests held by Suharto's family in Indonesia.

204

MARCEL FAFCHAMPS

rather than loans. Informal loans without interest and with no set repayment date are the dominant form of insurance with more distant friends and relatives. Platteau and Abraham (1987) report similar findings among Indian fishermen. Again contrary to some commonly held views (Granovetter 1995), kinship or place of origin appear to play little or no role in the formation of business networks. Individuals ‘from the same village’ (an African aphorism often used to refer to close ethnic and kinship ties) constitute a minute portion of the suppliers and clients reported by African manufacturers (Bigsten et al.1998). Proportions might be somewhat higher among micro-enterprises (Macharia 1988; Fafchamps 1994), but this could be an outcome of the apprenticeship system: trade skills acquired through apprenticeship are likely to be concentrated by place of origin if apprentices are recruited principally among family and friends. Fafchamps and Minten (1999) and Gabre-Madhin (1997) find no evidence of network formation by place of origin among grain traders in Madagascar and Ethiopia. These African findings are in contrast with those reported for Asia by Hayami (1996) and other contributors to this volume. Religion and ethnicity play a complex role in the formation of business networks (Granovetter 1985, 1995; Cohen 1969; Poewe 1989; Meillassoux 1971; Amselle 1977). A thorough coverage of the extensive literature on this issue would take us too far, so we shall limit ourselves to a few observations. As we have discussed in previous sections, business networks often— though not always (see Fafchamps and Minten 1999; Gabre-Madhin 1997; Bigsten et al. 1998)—display levels of religious and ethnic concentration that are extremely unlikely to result from random selection. What is often unclear is whether business networks arise from pre-existing ethnic and religious networks or the opposite. Regarding religion, for instance, there is some evidence that religious conversion is sometimes motivated by the desire to join a particular business community (Ensminger 1992; Shillington 1989). The same is true of strict observance of religious principles, such as the adherence to a particular dressing code or the obligation for devout Muslims to undertake a pilgrimage to Mecca if they can afford it (e.g. hadji169 traders in the Sahel), and of membership in religious organizations (e.g. Muslim brotherhoods in Sudan, marabouts in Senegal). The upshot of this is that the relationship between religion and business networks is unlikely to be unidirectional. In their description of the Moroccan market of Sefrou, for instance, Geertz et al. (1979) show the close integration of business networks and religious festival committees. Their interpretation is that both types of social structures reinforce each other: involvement in religious activities provide meeting opportunities for businessmen, which is precisely why they participate in religious activities in

169

A hadji is a person who has completed a pilgrimage to Mecca.

BUSINESS NETWORKS IN SUB-SAHARAN AFRICA

205

the first place. In addition, they argue that religious leaders use their moral authority to arbitrage certain disputes and to organize collective action (Platteau 1994a and b). Similar patterns are described for Chinese businessmen in South-East Asia in Geertz (1963). At first glance ethnicity might appear more exogenous than religion, but reality is more complex. First, what constitutes an ethnic group is a fluid concept. This fluidity is perhaps best illustrated by the recent history of Somalia. In 1976, Somalia launched a war against Ethiopia in an attempt to reunite all Somali people under the same flag. Fifteen years later, the same Somalia collapsed into separate fiefdoms on the basis of ethnicity, now defined differently. Ethnicity and, more generally, community identity can be shaped by history to suit the social preferences of the time. Second, given a particular definition of ethnicity, the ethnic affiliation of a particular individual is not always clear because it depends on various factors such as one's place of birth, mother tongue, place of residence, religion, race, and ethnic affiliation of one's parents. Whenever these factors do not coincide, the ethnicity of an individual is subject to interpretation. Fulanis living in Northern Nigeria are a good example: because they speak Hausa at home, they often can alternatively identify themselves as Hausas or Fulanis, depending on what suits them best (Cohen 1969). In many societies, it is even possible to change one's ethnicity—at least superficially—by learning another language or by changing one's name. Intermarriage is another possible avenue into ethnic conversion. Many manifestations of these processes were apparent in surveys, although collecting quantitative data on these issues proved impossible due to the extreme sensitivity of ethnic issues in all societies studied.170 Assuming that ethnicity and ethnic affiliation were defined without any ambiguity, the relationship between ethnicity and business networks remains loose even when business networks are ethnically concentrated. The reason is that ethnic groups are, by definition, very large, often numbering in millions. Even ethnic minorities such as Asians in Kenya or whites in Zimbabwe, still count over 100,000 individuals in each country, the overwhelming majority of whom are not in business. Moreover, those who are entrepreneurs often operate in unrelated sectors of activity. In contrast, business networks are small—a few hundred individuals at most, often much less (Granovetter 1995; Mitchell 1969; Barr 1998a)—and tightly knit. There is therefore no sense in which an ethnic ‘community’ can, by itself, serve as a platform for the establishment of a business network. A Kenyan Asian cannot, for instance, walk into another Asian's shop and obtain supplier credit without referral. Unless the two individuals find a

170

In some cases we had the distinct impression that raising the question of ethnic identification was sensitive precisely because respondents preferred to maintain some ambiguity as to their precise affiliation.

206

MARCEL FAFCHAMPS

common acquaintance who can vouch for them and guarantee repayment, credit will not be offered (Fafchamps et al. 1994). Because Kenyan Asians socialize principally with other Kenyan Asians, two Asians are more likely to find a common acquaintance than, say, an Asian and a Kikuyu. Consequently, there is a high probability that a referral system will result in ethnically concentrated business networks. Fafchamps (1998a) tests this proposition formally and shows that, once networks are controlled for, the measured effect of ethnicity on access to supplier credit falls dramatically. Although family, ethnicity, and religion play some role in the formation of business networks, the picture that emerges from numerous interviews with manufacturers and traders in Africa is one in which business networks for the most part result from business interaction itself. Bigsten et al. (1998), for instance, find that more than 90 percent of African manufacturers describe their suppliers and clients as simple business acquaintances. In many cases, commercial relationships are nurtured through business meetings and through socialization outside of work (Fafchamps et al.1994, 1995). In this respect, business is not very different from academia. Individuals who do not socialize with their clients and suppliers and who do not maintain regular business relationships are at a disadvantage (Fafchamps 1998a; Fafchamps and Minten 1998b). Ethnic concentration therefore seems to result from nothing other than historical accident and socialization patterns which are reinforced by the practice of business itself.

7.6 Communities and Networks Before concluding, it is worth spending a few moments discussing the various forms that networks can take so as to make the link with the main theme of this book, which is communities and markets. Empirical evidence collected in Africa indicates that there are many business networks that do not constitute communities. Malagasy traders and Ghanaian manufacturers, for instance, are embedded in various networks of interpersonal relationships, but they do not, as a rule, belong to an homogeneous business community. There may be identifiable communities regrouping subsets of traders or manufacturers, such as Malian traders in Accra's lumber market or Pakistani businessmen in Antananarivo, but these communities are small relative to the total size of the market. These simple observations raise the issue of when networks can be said to form communities. This is not an easy question. Granovetter (1995) and his followers argue that it takes six people or less to connect any two people in America. In this sense, the whole population of the USA—and even of the world—can be described as a single large network. While individuals must be connected to qualify as members of a community, the concept of network is

BUSINESS NETWORKS IN SUB-SAHARAN AFRICA

207

clearly much larger than that of community as we normally understand it. What then defines a community? The approach adopted by sociologists seems to have been to regard relationships as the building block of social constructs and to visualize society as a network of relationships (Mitchell 1969; Coleman 1988). Some of these relationships may be stronger than others—strong links vs. weak links in the parlance of Granovetter (1995), and relationships may be multi-faceted, that is, may take place at multiple levels—economic, social, symbolical (Basu 1986). Moving from the fuzzy concept of community to the more general notion of networks presents other advantages. Consider the issue of network capital, for instance. In general, someone who has more links has more options and, other things being equal, is likely to be better off. The number of one's connections can thus be taken as one measure of one's social capital: the more links, the more social capital (Barr 1998a; Fafchamps and Minten 1998b). In contrast, measures of social capital built on the notion of communities are unable to make such fine distinctions (Putnam et al. 1993; Coleman 1988). The concept of network also throws new light on the definition of dominant market position. Consider a situation in which individuals with very few links command a key position in the economy. To see how, suppose that N producers and M consumers of a particular product do not deal with each other directly but use intermediaries—i.e. traders. Further suppose that products are assembled from producers by one individual that we shall call the assembler, and that they are sold to consumers by another individual whom we call the retailer. Assembler and retailer deal with each other via a wholesaler. In this simplified example, the wholesaler has the least number of links—two—but the most market power: without the wholesaler, goods cannot move from producers to consumers. Unlike in a production monopoly, the wholesaler must buy what he or she sells. If transacting were costless, the wholesaler could not extract any monopoly rent: the threat of entry would prevent that. If transactions costs were nonconvex, however, as is the case in the presence of screening costs, trade concentration naturally arises as a way to economize on transactions costs (Fafchamps 1992). This simple example illustrates that the network structure of the market can confer market power to some agents in a way that is not adequately captured by the concept of business community. Within a network view of the world, communities or cliques can be defined as sets of agents or ‘nodes’ that are linked to each other in multiple and strong ways (Mitchell 1969). Examples of such communities include the set of businessmen and women who regularly go to the same church or temple, members of the same golf club, and individuals affiliated to the same professional organization.171 Some of these communities may restrict

171

In contrast, family does not, in general, define a community in this sense. This is because individuals by definition belong to partially overlapping networks of family relations.

208

MARCEL FAFCHAMPS

or condition entry, in which case one may want to call them clubs, e.g. a golf club. Others are open to all, e.g. churchgoers. Some communities are formal and have a legal status with an internal constitution and rules of procedure, e.g. a business association. Others remain informal, e.g. the individuals who meet in a particular bar or restaurant. The available empirical evidence suggests that it would be perilous to restrict the definition of communities to some of these categories only. For instance, Hendley et al. (1998) shows that, in the case of Russia, business communities can largely be identified with professional associations. Narayan and Pritchett (1996) similarly use membership in community associations as their concept of network capital. In contrast, discussions with African manufacturers indicate that Kenyan Asians meet at weddings and funerals and Zimbabwean whites meet at sport events and business conferences. Studying which form of organization is optimal is beyond the scope of this paper but, intuitively, open communities should be better than closed ones and formal organizations with clear rules and procedures should be better than informal organizations. Communities are not always free to choose their internal organization, however: the form they take is partly influenced by the state's attitude toward business. Discussions with survey respondents, for instance, revealed that Ghanaian lumber mill owners used to meet at their professional association in Kumasi. In the late 1980s, several of them found themselves indicted by the government in front of tribunals of exception. Although the exact reasons for these events are unclear, respondents expressed a strong suspicion that they were politically motivated and aimed at curbing the power of the association. In these circumstances, it is not too surprising if Ghanaian manufacturers now prefer to meet informally.

7.7 Conclusions The legal system alone cannot develop fully fledged markets. Personal trust and relationships are important, especially in the early stages of development when firms—and therefore transactions—are small, product quality is not standardized, and economic agents have no forecloseable assets. In these circumstances, business networks help to reduce transactions costs by circulating information on contractual performance and by coordinating the punishment of cheaters. Networks thus play a important positive role in market development. The early development of markets cannot be understood without investigating the role played by business networks. In the presence of non-convex transactions costs, we also showed that we must worry about the market power that business networks confer to certain individuals who occupy key positions in an exchange system.

BUSINESS NETWORKS IN SUB-SAHARAN AFRICA

209

Regarding market development policy, we argued that the emergence of business networks can generate various forms of discrimination and exclusion which operate to the disadvantage of non-network members. These effects get compounded by statistical discrimination when business networks are built around particular religious, ethnic, or racial groups. Efforts to develop markets should take into account the potential for a political backlash when resentment against successful business networks reaches a breaking point. Having spent time and effort investigating the role of business networks in market development, is it possible to imagine markets without them? The answer is a guarded yes. There are indeed ways by which institutions or technology can substitute for networks. Brand recognition is a good example. In a world where production is undertaken by a myriad of small producers and products are highly heterogeneous, such as agricultural products and crafts in poor countries, assessment of product quality is costly. In contrast, developed economies are characterized by the presence of a small number of producers offering products which are highly standardized and homogeneous over time. This enables consumers to economize on quality assessment by relying on a brand name instead. Knowing this, producers may invest in the reputation of their products, which then becomes an additional guarantee of quality. A similar process takes place when traditional seeds, which often result from centuries of informal breeding by farmers and are extremely variable across space, are replaced by standardized modern varieties (Hayami and Ruttan 1985). Even in the absence of returns to scale in production, quality control can be simplified through standardization. Franchising is a good illustration of such a process. It is unclear why franchising is absent from poor countries while the need for quality control is so patent. One possible explanation is that franchise contracts are difficult to enforce. Institutional innovations can also be found that facilitate credit checks. Credit reference bureaus exist that disseminate information about opportunistic breaches of contract. Organized exchanges can also be instituted that force market participants to post a bond before engaging in exchange. This bond ensures that contracts are honored, thereby eliminating the need for credit screening and speeding up transactions. The applicability of these innovations to poor countries deserves more research.

References Amselle, J. (1977) Les Négociants de la Savanne, Paris, Editions Anthropos. Aoki, M., Murdoch, K., and Okuno-Fujiwara, M. (1995) Beyond the East Asian Miracle: Introducing the Market Enhancing View, CEPR Publication No. 442, Stanford, Stanford University, October.

210

MARCEL FAFCHAMPS

Axelrod, R. (1984) The Evolution of Cooperation, Basic Books, New York. Bade, J. and Chifamba, R. (1994) ‘Transaction Costs and Institutional Environment,’ The Manufacturing Sector in Zimbabwe: Dynamics and Constraints, Free University of Amsterdam/University of Zimbabwe, RPED Country Study Series, Amsterdam, The World Bank. Barr, A. (1998a) ‘Social Capital and Technical Information Flows in the Ghanaian Manufacturing Sector,’ Oxford Economic Papers, July. —— (1998b) Enterprise Performance and the Functional Diversity of Social Capital, Center for the Study of African Economies, Oxford, Oxford University (mimeograph). Basu, K. (1986) ‘One Kind of Power,’ Oxford Economic Papers, 38: 259–82. Bauer, P. T. (1954) West African Trade: A Study of Competition, Oligopoly and Monopoly in a Changing Economy, Cambridge, Cambridge University Press. Bigsten, A., Collier, P., Dercon, S., Fafchamps, M., Gauthier, B., Gunning, J. W., Isaksson, A., Oduro, A., Oostendorp, R., Patillo, C., Soderbom, M., Teal, F., and Zeufack, A. (1998) Contract Flexibility and Conflict Resolution: Evidence from African Manufacturing, Department of Economics, Stanford, Stanford University, (mimeograph). Braudel, F. (1986) Civilization and Capitalism, New York, Harper and Row. Coate, S. and Ravallion, M. (1993) ‘Reciprocity Without Commitment: Characterization and Performance of Informal Insurance Arrangements,’ Journal of Development Economics, 40: 1–24. Cohen, A. (1969) Custom and Politics in Urban Africa: a Study of Hausa Migrants in Yoruba Towns, Berkeley, University of California Press. Coleman, J. S. (1988) ‘Social Capital in the Creation of Human Capital,’ American Journal of Sociology, 94(Supplement): S95–S120. Collier, P. and Gunning, J. W. (1997) Explaining African Economic Performance, Center for the Study of African Economies, Oxford, Oxford University, (mimeograph). Cuevas, C., Hanson, R., Fafchamps, M., Moll, P., and Srivastava, P. (1993) Case Studies of Enterprise Finance in Ghana, RPED, Washington DC, The World Bank. Daniels, L. (1994) Changes in the Small-Scale Enterprise Sector from 1991 to 1993: Results from a Second Nationwide Survey in Zimbabwe, Gemini Technical Report No. 71, Gemini, Bethesda, Maryland. Dutta, B., Ray, D., and Sengupta, K. (1989) ‘Contracts With Eviction in Infinitely Repeated Principal–Agents Relationships,’ in Bardhan, P. (ed.), The Economic Theory of Agrarian Institutions, Oxford, Clarendon Press. Eddy, E. (1979) Labor and Land Use on Mixed Farms in the Pastoral Zone of Niger, University of Michigan, Livestock Production and Marketing in the Entente States of West Africa, Monograph No. 3. Ellison, G. (1994) ‘Cooperation in the Prisoner's Dilemma with Anonymous Random Matching,’ Review of Economic Studies, 61: 567–88. Ensminger, J. (1992) Making a Market: The Institutional Transformation of an African Society, New York, Cambridge University Press. Fafchamps, M. (1992a) ‘Solidarity Networks in Pre-Industrial Societies: Rational Peasants with a Moral Economy,’ Economic Development and Cultural Change, 41(1): 147–74.

BUSINESS NETWORKS IN SUB-SAHARAN AFRICA

211

Fafchamps, M. (1992b) Non-Convex Transaction Costs, Networks, and Market Power, Stanford, Stanford University, August (mimeograph). —— (1994) ‘Industrial Structure and Microenterprises in Africa,’ Journal of Developing Areas, 29(1): 1–30. —— (1996) ‘The Enforcement of Commercial Contracts in Ghana,’ World Development, 24(3): 427–48. —— (1997) ‘Trade Credit in Zimbabwean Manufacturing,’ World Development, 25(3): 795–815. —— (1998a) Ethnicity and Credit in African Manufacturing, Stanford, Stanford University (mimeograph). —— (1998b) Market Emergence, Trust and Reputation, Stanford, Stanford University (mimeograph). —— (1998c) ‘Risk Sharing and Quasi-Credit,’ Journal of International Trade and Economic Development, August. —— and Lund, S. (1998) Risk Sharing Networks in Rural Philippines, Department of Economics, Stanford, Stanford University, December (mimeograph). —— and Minten, B. (1998a) Relationships and Traders in Madagascar, Department of Economics, Stanford, Stanford University, (mimeograph). —— and Minten, B. (1998b) Returns to Social Network Capital Among Traders, Department of Economics, Stanford, Stanford University (mimeograph). —— and Minten, B. (1999) Property Rights and Trade in a Semi-subsistence Economy, Department of Economics, Stanford, Stanford University, (mimeograph). —— , Biggs, T., Conning, J., and Srivastava, P. (1994) Enterprise Finance in Kenya, Regional Program on Enterprise Development, Africa Region, Washington, DC, The World Bank. —— , Pender, J., and Robinson, E. (1995) Enterprise Finance in Zimbabwe, Regional Program for Enterprise Development, Africa Division, Washington, DC, The World Bank. —— Gunning, J. W., and Oostendorp, R. (1997) Inventory, Liquidity, and Contractual Risk in African Manufacturing, Department of Economics, Stanford, Stanford University (mimeograph). Fisman, R. (1998) Trade Credit and Capacity Utilization in Developing Economies, Washington, DC, The World Bank (mimeograph). Foulds, L. R. (1992) Graph Theory Applications, Springer-Verlag, New York. Gabre-Madhin, E. (1997) Grain Markets in Ethiopia, Food Research Institute, Stanford, Stanford University (mimeograph). Geertz, C. (1963) Peddlers and Princes: Social Change and Economic Modernization in Two Indonesian Towns, Chicago, University of Chicago Press. —— , Geertz, H., and Rosen, L., (1979) Meaning and Order in Moroccan Society, Cambridge, Cambridge University Press. Ghosh, P. and Ray, D. (1996) ‘Cooperation in Community Interaction Without Information Flows,’ Review of Economic Studies, 63: 491–519. Gluckman, M. (1955) Custom and Conflict in Africa, Oxford, Basil Blackwell. Granovetter, M. (1985) ‘Economic Action and Social Structure: The Problem of Embeddedness,’ American Journal of Sociology, 91(3): 481–510. —— (1995a) ‘The Economic Sociology of Firms and Entrepreneurs,’ in Portes, A. (ed.), The Economic Sociology of Immigration: Essays on Networks, Ethnicity, and Entrepreneurship, New York, Russell Sage Foundation, pp. 128–65.

212

MARCEL FAFCHAMPS

—— (1995b) Getting a Job: A Study of Contacts and Careers, 2nd edn., Chicago, University of Chicago Press. Greif, A. (1993) ‘Contract Enforceability and Economic Institutions in Early Trade: The Maghribi Traders' Coalition,’ American Economic Review, 83(3): 525–48. —— (1994) ‘Cultural Beliefs and the Organization of Society: A Historical and Theoretical Reflection on Collectivist and Individualist Societies,’ Journal of Politics and Economics, 102(5): 912–50. Hart, O. (1995) Firms, Contracts, and Financial Structure, Oxford, Clarendon Press. Hayami, Y. (1996) ‘Peasants in Economic Modernization,’ American Journal of Agricultural Economics, 5: 1157–67. —— and Kikuchi, M. (1981) Asian Village Economy at the Crossroads: An Economic Approach to Institutional Change, Tokyo, University of Tokyo Press and Johns Hopkins University Press. —— and Ruttan, V. (1985) Agricultural Development: An International Perspective, rev. edn., Johns Hopkins. Hendley, K., Murrell, P., and Ryterman, R. (1998) Law, Relationships, and Private Enforcement: Transactional Strategies of Russian Enterprises, College Park, IRIS, University of Maryland (mimeograph). Himbara, D. (1994) ‘The Failed Africanization of Commerce and Industry in Kenya,’ World Development, 22(3): 469–82. Hoogeveen, H. and Tekere, M. (1994) Entrepreneurship: Who is a Successful Entrepreneur?, Amsterdam, Free University of Amsterdam/University of Zimbabwe, RPED Country Study Series, The World Bank. Jones, W. O. (1959) Manioc in Africa, Stanford, Stanford University Press Kandori, M. (1992) ‘Social Norms and Community Enforcement,’ Review of Economic Studies, 59: 63–80. Kranton, R. E. (1996a) ‘The Formation of Cooperative Relationships,’ Journal of Law, Economics, and Organizations, 12(1): 214–33. —— (1996b) ‘Reciprocal Exchange: A Self-Sustaining System,’ American Economics Review, 86(4): 830–51. Lang, W. W. and Nakamura, L. I. (1990) ‘The Dynamics of Credit Markets in a Model with Learning,’ Journal of Monetary Economics, 26: 305–318. Ligon, E., Thomas, J. P., and Worrall, T. (1997) Informal Insurance Arrangements in Village Economies, Department of Economics, University of Warwick, Coventry (mimeograph). Lorenz, E. H. (1988) ‘Neither Friends nor Strangers: Informal Networks of Subcontracting in French Industry,’ in Gambetta, D. (ed.), Trust: Making and Breaking Cooperative Relations, New York, Basil Blackwell. Macharia, K. (1988) Social Networks: Ethnicity and the Informal Sector in Nairobi, Institute for Development Studies, University of Nairobi, Nairobi, Working Paper No. 463. Marris, P. (1971) ‘African Businessmen In a Dual Economy,’ Journal of Industrial Economics, 19: 231–45. Meillassoux, C. (1971) The Development of Indigenous Trade and Markets in West Africa, Oxford, Oxford University Press.

BUSINESS NETWORKS IN SUB-SAHARAN AFRICA

213

Milgrom, P. R., North, D. C., and Weingast, B. (1991) ‘The Role of Institutions in the Revival of Trade: The Law Merchant, Private Judges, and the Champagne Fairs,’ Economics and Politics, 219: 1–23. Mitchell, J. C. (1969) Social Networks in Urban Situations: Analyses of Personal Relationships in Central African Towns, Manchester, Manchester University Press. Montgomery, J. D. (1991) ‘Social Networks and Labor-Market Outcomes: Toward an Economic Analysis,’ American Economics Review, 81(5): 1408–18. Mumbengegwi, C. (1994) ‘Indigenous and Small Scale Enterprises,’ The Manufacturing Sector in Zimbabwe: Dynamics and Constraints, Free University of Amsterdam/University of Zimbabwe, RPED Country Study Series, Amsterdam, The World Bank. Narayan, D. and Pritchett, L. (1996) Cents and Sociability: Household Income and Social Capital in Rural Tanzania, Policy Research Department, Washington DC, The World Bank (mimeograph). North, D. C. (1990) Institutions, Institutional Change, and Economic Performance, Cambridge, Cambridge University Press. Platteau, J. (1994a) ‘Behind the Market Stage Where Real Societies Exist: Part I—The Role of Public and Private Order Institutions,’ Journal of Development Studies, 30(3): 533–77. —— (1994b) ‘Behind the Market Stage Where Real Societies Exist: Part II—The Role of Moral Norms,’ Journal of Development Studies, 30(4): 753–815. —— and Abraham, A. (1987) ‘An Inquiry into Quasi-Credit Contracts: The Role of Reciprocal Credit and Interlinked Deals in Small-scale Fishing Communities,’ Journal of Development Studies, 23(4): 461–90. Poewe, K. (1989) Religion, Kinship, and Economy in Luapula, Zambia, Lewinston, The Edwin Mellen Press. Putnam, R. D., Leonardi, R., and Nanetti, R. Y. (1993) Making Democracy Work: Civic Institutions in Modern Italy, Princeton, Princeton University Press. Risseeuw, P. (1994) ‘Firm Growth in Zimbabwe 1981–1993,’ The Manufacturing Sector in Zimbabwe: Dynamics and Constraints, Free University of Amsterdam/ University of Zimbabwe, RPED Country Study Series, Amsterdam, The World Bank. Shapiro, C. and Stiglitz, J. E. (1984) ‘Equilibrium Unemployment as a Worker Discipline Device,’ American Economic Review, 74(3): 433–44. Shillington, K. (1989) History of Africa, New York, St. Martin's Press. Spagnolo, G. (1995) Social Relations in the Workplace: A ‘Linked Games’ Approach, Working Paper No. 76, Working Paper Series in Economics and Finance, Stockholm School of Economics, Stockholm. Staatz, J. M. (1979) The Economics of Cattle and Meat Marketing in the Ivory Coast, University of Michigan,. Livestock Production and Marketing in the Entente States of West Africa. Udry, C. (1990) ‘Rural Credit in Northern Nigeria: Credit as Insurance in a Rural Economy,’ World Bank Economic Review, 4(3): 251–69. —— (1994) ‘Risk and Insurance in a Rural Credit Market: An Empirical Investigation in Northern Nigeria,’ Review of Economic Studies, 61(3): 495–526.

214

MARCEL FAFCHAMPS

Velenchik, A. D. (1995) ‘Apprenticeship Contracts, Small Enterprises, and Credit Market in Ghana,’ World Bank Economic Review, 9(3): 451–75. Williamson, O. E. (1975) Markets and Hierarchies: Analysis and Antitrust Implications, New York, The Free Press, Macmillan. —— (1985) The Economic Institutions of Capitalism, New York, The Free Press, Macmillan. Department of Economics, Stanford University, and DECRG, The World Bank.

8 Risk and Insurance in Transition: Perspectives from Zouping County, 172

Jonathan Morduch and Terry Sicular The process of economic development is the process of the evolution of institutions. Economic institutions typically start as community-based personalized arrangements that facilitate exchange in the absence of strong markets and a legal infrastructure. In many of the world's richer countries, those institutions have evolved in a way that supports the emergence of internationally-integrated anonymous markets guided by a clear rule of law and well-understood accounting and measurement standards. The process has often been slow: in Europe, for example, centuries intervened between the establishment of institutions like medieval trade fairs and the formation of today's European customs and monetary unions. The mission of development policy since World War II has been to radically compress this evolution. Nowhere has this ambition been so great and adopted policies so radical as under Communism. Communist governments have reshaped economies to create distribution and production systems designed to catapult relatively poor nations into the industrialized world. While they pursued this objective by way of economic planning rather than markets, they shared the principle of pursuing anonymity: to each according to their needs (and contributions) irrespective of personal histories and customary hierarchies. The promise of Communism, however, was unfulfilled, and the dismantling of the Soviet system and dramatic reform in China after 1978 have been accompanied by renewed efforts to reshape these economies. Change has been especially rapid in China, where GDP per capita in constant 1987 dollars (adjusted to maintain international purchasing power parity) increased from $700 in 1980 to $1,376 in 1990 and $2,370 by 1995 (World Bank 1998). This growth has been the result of a burst of entrepreneurial activity in an environment where markets have become ascendant even while the

172

We would like to thank Masahiko Aoki, Avner Grief, Keijiro Otsuka, and other participants at the World Bank conference at Stanford University in February, 1999, for their helpful comments. We gratefully acknowledge financial support from the World Bank, the National Science Foundation (under awards SES-9211260 and SES-8908438), the Social Science and Humanities Research Council of Canada, and the Committee on Scholarly Communication with China. Morduch received additional support from the John D. and Catherine T. MacArthur Foundation. All views presented are those of the authors, who are responsible for any errors.

216

JONATHAN MORDUCH AND TERRY SICULAR

government retains partial controls. In rural areas, for example, the reforms have eliminated most planned targets and quotas and permitted the expansion of trade fairs and even wholesale markets. Households have responded by diversifying into a broad range of economic activities, including high-value crop production, new non-agricultural businesses, and off-farm employment. This expansion and diversification has brought new wealth to rural areas. It has concurrently created new economic risks and provided rural households with new opportunities to smooth income fluctuations. The reshaping of state and collective institutions in transitional economies has also, however, created holes in financial and insurance systems. Even in well-established market economies like Japan, Europe, and North America these key markets are often incomplete and imperfect. In transition economies financial and insurance systems are notoriously lacking. For example, while insurance companies in China have begun to provide coverage for life, health, and property losses (especially in urban areas), the rural population has little or no access to formal insurance against major sources of agricultural risk such as damaged crops due to drought, floods, hail, and pest infestation, or livestock losses due to disease. How have households coped with increased economic risk and loss of pre-existing financial and insurance mechanisms? Are community-based informal institutions emerging to help fill gaps? What are the prospects for the development of insurance markets? This chapter reports findings from a four-year study (largely inspired by these questions) conducted in Zouping County, Shandong Province. Zouping is a fairly representative county of northern China, and the study followed a stratified random sample of rural households for four years, from 1990 through 1993. As shown below, information from this longitudinal survey reveals that, in principle, risk sharing among households could substantially reduce the volatility of household consumption. The evidence, however, shows only modest signs of such risk sharing activity. Inter-household transfers are relatively infrequent and insufficient to smooth consumption. While such transfers respond to income declines and disasters, on balance they redistribute towards households that are better off. Moreover, counter to a common assumption that informal risk sharing occurs among households in the same community, we find that when risk sharing occurs it is more often associated with relationships outside the village, not within. The evidence indicates that instead of coping with risk through informal transfers, households rely primarily on their own devices, for example, they rely on saving and borrowing and on diversification of their sources of income. Even with these devices, many households remain considerably exposed to risk.173 The lack of effective insurance here echoes the findings of

173

The lack of strong informal insurance mechanisms has been found in many other studies of risk and exchange in poor economies (Morduch 1999a).

RISK AND INSURANCE, ZOUPING COUNTY, CHINA

217

Jalan and Ravallion (1999), who show that 40 percent of income shocks are translated into consumption shocks for the poorest decile of households in a large longitudinal survey collected in China between 1985 and 1990. The limited role of informal risk sharing and risk smoothing mechanisms raises questions regarding formal programs. The final section of this paper reports on the findings of in-depth interviews with officials at national, provincial, county, and village levels about formal insurance and relief programs. We investigate explanations for weak insurance provision and identify possibilities for innovative insurance arrangements that could draw on both communities and markets.

8.1 The Chinese Context Land tenure systems in China have undergone various transformations since the Communists took power in 1949. In the late 1940s and early 1950s land was redistributed from landlords to small farmers. Collectivization began in the mid-1950s, initially on a voluntary and small-scale basis, but by the start of the Great Leap Forward in 1958 the scope and scale of collectives was pushed to a high level. In the wake of the famine that followed in the late 1950s and early 1960s, collectives were scaled back. The organizational structure of communes, production brigades and production teams was retained, but the primary locus of economic decision making and distribution now shifted from the commune level to the production team. This basic organizational structure remained more or less unchanged until the late 1970s.174 In 1980, there were over 50,000 communes in rural China, each with about 3,000 families (Whyte 1991: 368). The average commune was made up of about 100 production teams organized into 15 production brigades. The brigade was typically made up of a large village or a collection of adjoining small villages, thus building on familiarity and shared histories. The communes and brigades were responsible for meeting production targets determined by planners, and they also built basic infrastructure and factories to supply productive inputs like fertilizer. The production teams managed both farm production and income distribution. Team income (in cash and kind) was distributed on the basis of work points, with more difficult tasks earning more points. Whyte (1991: 371) provides an example where skilled, prime age males might earn 10–12 points per day and less skilled workers might earn 7–9 points daily. At the same time, members who were old, ill, or too young might still earn 4–6 points. A family's allocation would then be determined by the product

174

In the late 1960s and mid-1970s, some attempts were made to move the accounting unit from the team to the production brigade, but these attempts were met with resistance and few collectives made this change (Whyte 1991: 372).

218

JONATHAN MORDUCH AND TERRY SICULAR

of the team's total output multiplied by the family's share of total points. Families could thus share idiosyncratic risks with the 20–50 households in the team, and a mechanism was in place to address losses due to illness and aging. Some studies suggest that risk sharing was especially evident in the collective distribution of basic foodstuffs. Putterman (1993, ch. 5), for example, provides evidence that within collectives grain was often distributed on a per capita basis, not on the basis of work points, while greater differentiation was made in the distribution of cash income.175 This system broke down with the reforms that began in 1978. The reforms expanded the emphasis on private plots and promoted the development of markets. Decollectivization led to the abandonment of collective distribution mechanisms and the work-point system, implicitly dismantling collective mechanisms for risk sharing. At the same time, the government has maintained controls on insurance markets, prohibiting private insurance companies and so preventing private enterprises from filling the void by providing formal insurance. The state-owned People's Insurance Company provides only very limited lines of agricultural insurance (Section 8.4 below). These developments raise the question of how China's rural households protect themselves from risk in the new environment. To what extent do they save and borrow to smooth consumption in the face of transitory income shocks? Have they developed or revived customary informal insurance arrangements? For example, do neighbors engage in reciprocal gift exchange conditional on need, a common traditional form of protection from risk? On the one hand, such developments would not be surprising. The strong role of the village in rural life historically—and the fact that many Chinese villages are made up of just a few extended families (so-called singlesurname villages)—promote possibilities for informal insurance arrangements that are absent in more heterogeneous communities.176 Jalan and Ravallion (1999), for example, call the Chinese village ‘a natural coinsurance group,’ and a recent anthropological study by Andrew Kipnis (1998) finds fairly active gift exchange following the reforms. On the other hand, many villages have historically been marked by social fissures—the same tensions between richer and poorer households

175

Collective risk sharing also took place through the provision of social welfare by production teams and brigades. For example, collectives provided free health care at collective health clinics.

176

Heterogeneity cuts two ways with respect to enhancing informal insurance. Most of the recent economic literature on tensions in informal insurance focuses on difficulties in enforcing intertemporal contracts (e.g. Coate and Ravallion 1993 and the papers cited in Morduch 1999a). Strong interpersonal ties can enhance the use of social sanctions that can keep informal contracts from falling apart. On the other hand, a greater degree of insurance is possible when community incomes are less likely to covary. Heterogeneity may then aid performance of informal insurance.

RISK AND INSURANCE, ZOUPING COUNTY, CHINA

219

that the Communists were able to exploit in their rise to power before 1949. Moreover, Confucian moral views, which are still seen as providing a continuing cultural foundation in rural China, stress strong commitments to family but are far more muted with respect to commitments to the local community. Decollectivization, the development of markets, and the removal of state quotas have allowed inequality and stratification to emerge within communities. The resulting heterogeneity creates potential gains from mutual insurance, but it can also inhibit cooperation and the easy flow of information. The village described by Kipnis, Fengjia Cun in Zouping County, is dominated by several branches of the Feng family (Fengjia Cun translates literally as ‘Feng family village’), strengthening the possible use of social sanctions there. But Kipnis's work, like the influential studies of Bronislaw Malinowski (1922) and Marcel Mauss (1967), emphasizes the social role of gift-giving, especially its role in establishing social hierarchies and bonds between community members. While Kipnis (1997: 71) notes that ‘gifts to the elderly, the ill, and women recuperating from childbirth acknowledged that an already existing obligation could be counted on in time of need,’ he minimizes economic motivations for making transfers and, in particular, strategic motives based on risk sharing. From the vantage of economic theory, though, it is difficult to ignore insurance-based motives for gift exchange. If possibilities exist for large gains from trade, why would such trade not take place?

8.2 The Survey Data and Economic Setting Our study177 takes a close look at these relationships between households and communities through a broader, longitudinal study of 16 villages in Zouping County (including the village in which Kipnis conducted his research). The data used in this study are from a stratified random sample of 259 farm households in 16 villages in this county, which is situated south of the Yellow River in central Shandong Province. The survey followed the households over four years, providing 1,036 observations covering the calendar years 1990 through 1993. Zouping is relatively typical of rural counties in northern China during the reform period. In the 1980s it carried out decollectivization, greatly reduced planning, and promoted the development of markets and of diverse economic activities. The 1990s have seen a growing importance in non-agricultural activities and waged employment, but agriculture has remained the primary source of household income.

177

This section draws heavily on Morduch and Sicular (2000). For a more detailed description of the county and survey sample see Sicular (1998) and Walder (1998).

220

JONATHAN MORDUCH AND TERRY SICULAR

When we initiated the survey, Zouping was one of the few counties to which foreign social scientists were given access for in-depth data collection. Recently several additional longitudinal data sets, mainly collected by official Chinese agencies, have become available. While our sample is comparatively small, the data set has certain advantages relative to other existing data sets. In particular, our survey questionnaire was tailored to deliver consistent definitions of income and to provide rich information on a wide array of economic and social variables, with particular attention paid to inter-household transfers, saving, borrowing and lending, and political variables. Furthermore, close supervision and careful consistency checks yielded a data set of relatively high quality.178 Official statistics show that Zouping is an unexceptional rural county.179 Zouping experienced rapid economic growth during the 1980s, but its per capita GDP in the early 1990s was still 18 percent below the national average (the national average is pulled up by industrial production in major urban centers). In 1990–93 the income of rural households in the county averaged 701 yuan per capita (in constant 1990 prices), slightly below the national average of 717 yuan. Rapid development of non-agricultural activities in the county has caused agriculture to decline in importance as a source of income and employment. In 1990 agriculture contributed 56 percent of county GDP, but by 1993 agriculture's share had fallen to 30 percent. For rural households, however, agriculture continued to contribute the great majority of income: in the early 1990s agriculture contributed 6570 percent of the net income of rural households in the county. Cultivation of wheat, cotton and corn still accounted for the majority of agricultural output value. The survey sample captures these key features of the county.

178

The data were collected through an independent survey organized by the authors. Trained survey workers verbally interviewed the households. Villages were stratified by income and output levels, and households within villages were selected randomly. Each village was assigned a different interviewer, and the quality of the data thus may vary among villages due to differences in skill and effort among the interviewers. The interviews took place in mid-year and households were asked to recall their activities in the prior calendar year. Consequently the data may contain errors in recall. To minimize the above problems, the questionnaires contained overlapping questions that were designed to allow multiple consistency checks. In addition, the questionnaires were checked for inconsistencies during the course of the survey work, and households were re-interviewed when inconsistencies or other problems were detected.

179

County statistics were obtained through interviews with county officials. National-level information is from State Statistical Bureau (various years). Zouping's GDP per capita in 1990 was 1,280 yuan (at the 1990 official exchange rate, 4.77 yuan\$1; GDP adjusted to maintain purchasing power parity is nearly six times the size of nominal figures). The official statistics on rural household income are not strictly comparable to rural household income statistics from the survey data used in this paper, in part because the method of calculation is different. In the official statistics, retained output is valued at planned prices, while we value retained output at market prices. Also, we calculate household size by person-years rather than by the number of residents at the end of the year.

RISK AND INSURANCE, ZOUPING COUNTY, CHINA

221

The ongoing changes in the early 1990s are reflected in our sample statistics. Between 1990 and 1993, real per capita income in our sample grew at an average annual rate of 8.2 percent, rising from 1,175 yuan to 1,628 yuan (in constant 1990 prices).180 The sample does not include urban residents, so agriculture plays a larger role than in the county as a whole. Nevertheless, the share of income from non-agricultural sources rose from 17 percent to 24 percent, pushed up by a 50 percent increase in the share of wages in total income (from 14 percent to 21 percent). This trend was consistent with the county's strategy of promoting the development of collective township and village enterprises.181 The share of income from sideline businesses remained virtually unchanged over the period. The empirical analysis below focuses on households, rather than individuals. Households are increasingly divided along generational lines, but inter-generational households remain common in the sample. The average adult age is 37, and the average education of adults is 5.3 years. All but a few households in the sample cultivate crops. The majority of households, however, have at least partly diversified into non-agricultural activities. Just over half contain a member holding a waged job, and just under 20 percent engage in a non-agricultural sideline activity (although such sidelines are often quite small in scale).

8.3 Evidence on Risk-coping Mechanisms Income Variability Table 8.1 shows that per capita incomes in Zouping were rising yearly at 8.6 percent on average over the 1990–93 period. This rate is high by historical standards, but close to the rate of growth for rural households in the province and for national income in China as a whole in the early 1990s. Between 1990 and 1993 the net income of rural households in Shandong Province grew by 5 percent annually, and per capita GDP in China as a whole grew at a rate of 11 percent (State Statistical Bureau 1998).182 The standard deviation on income growth within the county is also high, however, and the detailed summary statistics reveal uneven performance among households and over time. The bottom three rows of Table 8.1 give averages for the top and bottom quartiles and the middle two

180

Unless noted otherwise, income includes (1) the value of farm outputs minus costs of material inputs (both purchased and self-supplied), hired labor, and hired mechanized services; (2) wage earnings; (3) net revenues from household non-agricultural sideline activities; (4) net transfers from the government and collective (transfers minus taxes and fees, including implicit quota taxes); and (5) transfers received from other households. Products produced by the household and retained for own use are valued at market prices.

181

Interviews, Zouping County, 1991–94.

182

These and other growth rates are all calculated using constant prices.

222

JONATHAN MORDUCH AND TERRY SICULAR

Table 8.1 Average Real Income Growth, 1990–1993

Mean Standard deviation Means by quartile: Bottom quartile Middle quartiles Top quartile

Income per worker

8.6% 27.5

Income per capita, adjusted for time away from home 16.8% 29.9

−18.7 5.1 42.7

−13.3 13.1 54.5

−10.0 16.4 62.0

Total household income

Income per capita

14.8% 28.5 −13.3 5.2 50.4

21.2% 32.1

Notes: Calculated based on income in constant 1990 yuan. The sample contains 248 households with complete data for all four years (992 observations), and growth rates are available for three years (744 observations). The mean growth rate during the four-year sample period is calculated for each household, and the standard deviations and quartile averages shown in this table are calculated over these household means.

quartiles of each income measure. The growth figures are averaged by household and then averaged by category, and the bottom row shows that very high growth rates for some households in some years are pulling up the average. Average income growth per capita in the top quartile exceeds 40 percent per year. Some households, however, are doing extremely poorly in some years. Average income growth per capita for the bottom quarter shows losses approaching 20 percent or more per year. The broad upward trend in the aggregate data thus masks a large number of income declines. Note that these findings are robust to the way that income is calculated: as shown in Table 8.1, the same qualitative results are maintained whether income is viewed on a per capita basis, per worker, or per capita adjusted for time spent away from home by household members. Growth in mean income accompanied by substantial variability is also seen in the data for total household income. Throughout the rest of the paper, we define income on a per capita basis adjusted for time spent away from home by household members. Table 8.2 shows the roles of agricultural income versus non-agricultural income in explaining the patterns of growth and loss. The table shows that although non-agricultural income is overall less important than agricultural income in explaining levels, it accounts for at least as large a share of income changes. For households in the bottom quartile, however, agricultural income is paramount, explaining nearly 12 percentage points of the −13.3 percent average loss. Overall, agricultural and non-agricultural income are negatively correlated, with a correlation coefficient of −0.19. This suggests potential gains from diversification but also gives evidence of specialization.

223

RISK AND INSURANCE, ZOUPING COUNTY, CHINA

Table 8.2 Contributions of Agricultural and Non-Agricultural Income to Income growth, by Quartile, Percentage Points Household Income

Entire sample Bottom quartile Middle two quartiles Top quartile

5.9 −12.6

Non-agricultural 9.2 −1.1

Household income per capita, adjusted for time away from home Total Agricultural Non-agricultural 16.8 7.3 9.8 −13.3 −11.7 −1.8

11.1

5.2

6.7

13.1

6.3

7.4

50.4

25.8

24.6

54.5

28.2

26.4

Total

Agricultural

14.8 −13.3

Notes: Calculated based on income in constant 1990 yuan. The sample contains 248 households with growth rates available for three years (744 observations). Mean income growth during the four-year sample period is calculated for each household, and the numbers shown in this table are calculated over these household means. The numbers for agricultural and non-agricultural income are equal to the absolute yearly change in each component divided by the total so they sum to growth in total income. Agricultural plus non-agricultural components of growth do not add up exactly to growth in total income due to rounding errors and minor differences in samples.

Both the losses and gains in Tables 8.1 and 8.2 are large, reflecting both the new opportunities in post-reform rural China and the attendant risks. An additional explanation for the volatility is that long-term investments are being made in one year (leading to high current losses) and yielding returns in the next (leading to high current income growth in subsequent years). We did, however, investigate the volatility of income with and without subtracting long-term investments, and found little difference between the two. In addition, negative partial correlations of income across years should partly wash out given that the figures presented are average gains/losses per household over the four-year sample. Still, since the panel is short and investments may play out over periods before and after the sample, it is unlikely that all timing issues will wash out. In the analysis below, we thus focus on specific shocks facing households (deaths, illness, reported crop loss) as well as evidence from the income data.

The Potential for Risk Sharing How well can risk sharing within villages reduce volatility? Table 8.3 gives information on the potential for risk sharing to smooth consumption over time. The table shows that if household income perfectly tracked household consumption (so implicitly households have no consumption-smoothing

224

JONATHAN MORDUCH AND TERRY SICULAR

Table 8.3 The Potential for Risk Sharing: The Average Coefficient of Variation of Household Consumption Under Hypothetical Scenarios Under autarky (consumption = income) Under village-level risk sharing Under county-level risk sharing

Coefficient of variation 29.5 14.0 5.4

Notes: Calculated using per capita income (household size is adjusted for time at home) in constant 1990 yuan. The sample contains 248 households with complete data for all four years (992 observations).

possibilities), the average coefficient of variation of consumption would be 29.5 percent, where the c.v. is calculated for each household over the four years 1990–93 and then averaged over the sample. The second row gives the result of a hypothetical risk sharing arrangement. Here, we consider transfer exchange arrangements designed in each of the 16 villages such that household risk would be mitigated as best possible, subject to the constraint that four-year average income is held constant for each household (i.e. there is no systematic redistribution, only risk sharing). These calculations give perspective on the ability to share risks year-by-year and do not allow borrowing or saving across years. These simple arrangements would cut the average coefficient of variation by half to 14 percent. If, instead, similar arrangements were implemented at the level of the county (allowing greater risk pooling), rather than independently in each village, the coefficient of variation would be reduced by half again, to under 5.4 percent. Such calculations show in a simple way that risk sharing can make a large difference to consumption variability, and we turn to evidence of this kind of risk sharing below. One should bear in mind, however, that we have not considered inter-temporal consumption smoothing above. If, instead, households can use savings and borrowing possibilities to mitigate risk on their own, the gains to risk sharing through reciprocal transfer exchange will be greatly diminished. We return to this argument (and present relevant evidence) below.

Inter-Household Transfers In order to shed light on the kinds of issues raised above, data were collected on transfers received from within the village and transfers received from outside the village. Table 8.4 gives key summary statistics for these data. Transfers here are defined to include the value of both financial and in-kind (such as goods, food, and farm products) transfers received by households. The figures do not, however, include the value of labor exchanged, and transfers of labor are described separately below.

225

RISK AND INSURANCE, ZOUPING COUNTY, CHINA

Table 8.4 Inter-Household Transfers, 1990–1993 Households receiving transfers Percent of observations 27.8 Average value of transfers 235.5 received (yuan) Average per capita house- 1,460.8 hold income (yuan) Average per capita house- 24.6 hold income growth (%) By sub-group in the sample: Percentage in richest 50% 54.0 Percentage in poorest 50% 46.0 Percentage with income 56.5 increase Percentage with income 43.5 decline

Households not receiving All households transfers 72.2 100.0 0.0 65.5 1,206.2

1,277.0

15.8

18.5

48.5 51.5 51.2

50.0 50.0 52.7

48.9

47.3

Notes: Transfers and income are in constant 1990 prices. Income is net of transfers received, with household size adjusted for time at home. Transfers include the value of both cash and in-kind transfers received from households within and outside the village, except for transfers of free labor. Statistics on income growth are available only for three years of the sample, and so statistics based on or classified according to income growth are calculated only over these three years. Households are classified as being in the richest or poorest half of the sample based on average household income over the four years of the panel. Households are classified as having an income increase or decrease based on yearly changes in income, and so here households can change classification over time.

The table shows that slightly more than a quarter of the observations record non-zero transfers. Transfers are thus small on average, although relatively large for those receiving them. The average transfer size is 235.5 yuan, which is 16 percent of the average (pre-transfer) income of households receiving transfers. To whom do transfers go? The bottom of Table 8.4 presents descriptive evidence on whether (1) transfers are more likely to be received by poorer households and (2) whether households facing income declines are more likely to report receiving transfers. (Note that all income figures in Table 8.4 are net of transfers received.) We find that richer households are in fact slightly more likely to report receiving transfers: the richest half of the sample comprised over half of all the observations reporting non-zero transfers (54 percent). On the other hand, the poorest half make up just 46 percent of the observations with non-zero transfers. The average per capita income of households receiving transfers is 1,461 yuan, which is 21 percent greater than that of households reporting no transfers. There is little evidence that suggests risk sharing motives in these aggregated data. If transfers were given as a response to income decline, households experiencing income losses should make up a disproportionately

226

JONATHAN MORDUCH AND TERRY SICULAR

large share of the observations with non-zero transfers. But transfers are in fact more likely to be received by households with increasing income than those with income declines: roughly 53 percent of all observations experienced income growth, and 57 percent of the observations with non-zero transfers experienced income growth. Similarly, 47 percent of all observations had income declines, while 44 percent of observations do so in the sample with non-zero transfers. The average income increase of households receiving transfers was 24.6 percent vs. 15.8 percent for households who do not receive transfers. These descriptive statistics reveal little evidence that transfers are motivated by either risk sharing or progressive redistribution. The fact that income variability changes little when income is calculated with and without transfers further reinforces the conclusion that risk sharing is not a major motive for transfer exchange. The coefficient of variation of income per capita including transfers is 29.5 (see Table 8.3). The c.v. of income excluding transfers is 30.3, only marginally higher. If transfers were used to cope with income variability, one would expect that the c.v. of income would increase more noticeably when transfers are subtracted from income. Table 8.5 disaggregates the data for transfers between those received from within the village and those received from outside. The picture with respect to risk sharing now becomes clearer. The bulk of transfers are received from Table 8.5 Transfers Received Inside versus Outside the Village Households receiving transfers from within the village Percentage of observations 23.6 Average value of within- 150.3 village transfers Average value of outside- 66.4 village transfers Average household income 1,499.2 (yuan) Average household income 30.5 growth (%) By sub-group in the sample: Percentage in richest 50% 54.3 Percentage in poorest 50% 45.7 Percentage with income 59.9 increase Percentage with income 40.1 decline Notes: see notes to Table 8.4.

All households Households receiving transfers from outside the village 9.8 100.0 79.0 35.5 307.5

30.1

1,352.8

1,277.0

15.0

18.5

50.5 49.5 48.4

50.0 50.0 52.7

51.7

47.3

RISK AND INSURANCE, ZOUPING COUNTY, CHINA

227

within the village, with 24 percent of households reporting transfers from within but just 10 percent reporting transfers from outside. Transfers from outside, however, are on average twice as large, with an average size of 308 yuan vs. 150 yuan. As above, the incomes of households receiving transfers are higher than the average for the sample, but they are considerably larger for households receiving transfers from within than for those receiving transfers from without (1,499 yuan vs. 1,353 yuan). Again, both groups have positive income growth on average, but the average growth rates are twice as large for households receiving transfers from within (31 percent) vs. those from outside (15 percent). The fraction of households with income declines among those receiving transfers from within the village is below that for the entire sample (40 percent vs. 47 per-cent), suggesting little tendency toward risk sharing. But the fraction of households suffering losses is larger for the sample receiving transfers from outside (51 percent), opening the possibility of some insurance motives. The evidence thus suggests that the two sources of transfers—from within the village and from outside—are at least in part given for different reasons. The tendency for risk sharing is greatest for transfers coming from outside, which is consistent with a situation in which fellow villagers have limited ability to help each other (typically because shocks often hit the entire village at once), whereas outsiders are in a better position to help. The evidence on within-village transfers is consistent with the observations of Mauss (1967), Malinowski (1922), and Kipnis (1998) that downplay the role of risk mitigation as a central motivation for transfer exchange. Table 8.6 takes advantage of the disaggregated data in order to determine if in-kind transfers exhibit different patterns when isolated. The table also gives evidence on transfers of labor (which are not included in the aggregates in the tables above). Differences should show up here if, for example, households in difficulty are given help securing food or help working their plots, transfers more likely for villagers to exchange than cash. In-kind transfers are fairly frequent, and their patterns differ from that of the aggregated data in a way that shows a tendency toward progressive redistribution. About 8 percent of households report receiving transfers of meat, oil, or eggs, and 8 percent again report transfers of crop products. Transfers of labor are reported by 19 percent of households. The sizes of transfers of goods are 2–4 percent of average income, and labor transfers are 11 percent. Strikingly, in-kind transfers go disproportionately to poorer households in the sample. The average income of recipients is 1,168 yuan per year, vs. 1,374 for non-recipients. The fraction of households receiving in-kind transfers that have had income increases is close to that of households with declines (52 percent vs. 48 percent), however, and the fractions closely mirror those for the total sample (53 percent vs. 47 percent). The presence of risk sharing motives is again elusive, once more echoing the anthropological observations.

228

JONATHAN MORDUCH AND TERRY SICULAR

Table 8.6 In-Kind Transfers, 1990–1993 Households receiving inkind transfers Percentage of observations: Meat, oils, and eggs 8.2 Crops 7.9 Livestock 1.2 Labor 19.2 Total 29.3 Average value of transfers received (yuan): Meat, oils, and eggs 49.5 Crops 33.4 Livestock 27.9 Labor 122.7 Total 114.7 Average household income and income growth: Income 1,168.0 Income growth 14.0 By sub-group in the sample: Percentage in richest 50% 35.3 Percentage in poorest 50% 64.7 Percentage with income 51.8 increase Percentage with income 48.2 decline

Households receiving no in-kind transfers

All households

91.8 92.1 98.8 80.8 70.7

100 100 100 100 100

0 0 0 0 0

4.0 2.6 0.3 23.5 33.6

1,373.6 20.3

1,320.2 18.5

56.1 43.9 53.0

50.0 50.0 52.7

47.0

47.3

Notes: See notes to Table 8.3. Data on transfers of animals are available only for 1991–93, so the number of observations used in calculations relating to animal transfers and total in-kind transfers is 744.

The descriptive statistics above give a sense of the scale and function of transfers. They do not, however, control for important household characteristics like education, age, and location. Tables 8.7 and 8.8 thus extend the analysis with random effects estimates of the determinants of the receipt of transfers. The dependent variable is the total amount of transfers received, as used in Table 8.4 above. The explanatory variables include household income per capita (adjusted for days spent away from home), income growth, whether the household self-reports a ‘disaster’ (including illness), household size (again, adjusted for time away from home), the fraction of adults in the household, and indicators of whether the household contains a member of the Communist Party, a village-level cadre, or a cadre working at the sub-village level. Additional variables (not reported in the tables) include the number of workers per capita, average age (and age squared) of

229

RISK AND INSURANCE, ZOUPING COUNTY, CHINA

Table 8.7 Determinants of Transfers Received, 1990–1993, Random Effects Estimates Variable Household income per capita Household income p.c. growth ‘Disaster’ reported? Adjusted household size Adult male share Communist Party member? Village-level cadre? Sub-village-level cadre? Year fixed effects? Village fixed effects? Year* village fixed effects? R2 Number of observations Number of households

(1) 0.08 (.01)**

(2) 0.08 (0.01)**

(3) 0.08 (0.01)**

−22.2 (11.9)*

−20.6 (12.1)*

−17.3 (12.4)

176.1 (40.0)** 22.5 (11.5)** −99.7 (81.4) −24.9 (31.2)

168.3 (40.9)** 32.6 (12.0)** −136.4 (82.3)* −33.1 (31.3)

183.8 (42.7)** 32.0 (12.0)** −146.3 (82.7)* −33.2 (31.3)

55.3* (31.7) −17.9 (47.0) yes no no 0.12 742 248

63.0** (31.8) −23.8 (48.8) yes yes no 0.17 742 248

62.4** (31.8) −18.0 (48.9) yes yes yes 0.21 742 248

Notes: *** Significant with 99 percent confidence. ** Significant with 95 percent confidence. *

Significant with 90 percent confidence. Standard errors are in parentheses. Calculated based on income net of transfers received in constant 1,990 yuan. Income is in per capita terms (family size is adjusted for time at home). The dependent variable includes the value of all financial and in-kind transfers except of labor. The regressions also include number of working-age adults per capita, average age (and age squared) of adults, average education (and education squared) of adults, and three indicator variables for class background.

adults, average education (and education squared) of adults, and three indicator variables for class background.183 Table 8.7 presents estimates based on a linear model, and Table 8.8 presents results from a Tobit specification. The first specification in Table 8.7 includes dummy variables for year, and the second adds dummy variables for village, increasing the R2 from 0.12 to 0.17. The third specification also includes dummy variables for year interacted with village. The presence of the interacted year and village effects mean that the coefficients on the explanatory variables are net of

183

See Morduch and Sicular (2000) for a description of these variables and their role in explaining earnings.

230

JONATHAN MORDUCH AND TERRY SICULAR

Table 8.8 Determinants of Transfers Received, 1990–1993, Random Effects Tobit Estimates Variable Household income per capita Household income p.c. growth ‘Disaster’ reported? Adjusted household size Adult male share Communist Party member? Village-level cadre? Sub-village-level cadre? Year fixed effects? Village fixed effects? Year* village fixed effects? Wald χ2 (degrees of freedom) Number of observations Number of households

(1) 0.18*** (.03)

(2) 0.19*** (0.03)

(3) 0.19*** (0.03)

−35.2 (35.0)

−22.9 (37.0)

−18.3 (34.8)

364.5*** (93.7) 32.6 (26.9) −414.2** (196.4) −34.5 (72.5)

382.3*** (101.6) 90.9*** (29.1) −506.0** (201.3) −36.7 (74.8)

445.2*** (105.3) 90.5*** (29.2) −553.2*** (201.6) −28.9 (73.5)

32.6 (73.0) −255.1* (133.1) yes no no 88.7 (18)

156.9** (74.4) −270.4* (143.2) yes yes no 169.5 (33)

163.6** (74.2) −264.3* (142.2) yes yes yes 710.4 (60)

742 248

742 248

742 248

Notes: See notes to Table 8.7.

trends over time and the effects of generally good or bad years in the villages. The focus is thus on the responsiveness of transfers to idiosyncratic shocks. Income per capita is a positive determinant of transfers in each specification in Table 8.7. A 100 yuan increase in income increases transfers by 8 yuan. The effects are even stronger in Table 8.8, where the increase is by 18–19 yuan. This result thus counters the view that transfers are motivated in part by redistribution from wealthier to poorer households: once a broader range of variables is considered, redistribution appears to be regressive, tending to go from poorer to wealthier households. Transfers, however, are less likely to go to households with more adult males per capita, indicating a tendency to aid households with traditionally lower earning capacity. Adding a broader range of control variables to the specification also yields considerably more evidence consistent with risk sharing. A decline in income by one percentage point increases average transfer receipts by as

RISK AND INSURANCE, ZOUPING COUNTY, CHINA

231

much as 22 percent in the linear specification. The Tobit results in Table 8.8 are slightly larger but no longer significantly different from zero. Suffering a self-reported ‘disaster’ is also a major determinant of transfer receipts, raising the level of transfers by at least 176 yuan, or by more than 12 percent of pre-transfer income, in the linear specification. The Tobit results yield increases that are twice as large. Thus while transfers and risk sharing are not obviously linked in the broad descriptive data nor in the ethnographic data, associations emerge here. Of the political variables, only status as a village-level cadre (defined as having a current or former cadre in the household) consistently takes a large and statistically significant coefficient. In the linear specification such status raises transfer levels by 55–63 yuan and is consistent with Kipnis's (1998) argument that transfers are often given to forge relationships with important members of the social hierarchy. Neither being a party member nor being a cadre at a lower level helps, however. In fact, the Tobit results indicate that households containing sub-village cadres are less likely than others to receive transfers. These findings are consistent with those in Morduch and Sicular (2000), where a similar pattern is found when explaining income levels. The pattern of coefficients is largely unaffected by the choice of controls for village and time. The results suggest that villagers consider economic conditions outside of their immediate, local circumstances in determining the giving of transfers. In summary, the regression results are consistent with the mixed motives for transfer exchange noted above. There is both evidence of risk sharing and evidence that transfers are made to forge social bonds. The two forces drive in opposite directions. The first force directs transfers to households suffering losses, while the second directs transfers to the established, powerful households in the community.

Saving and Borrowing If transfers afford limited insurance against income declines, how do households cope with income shocks? Tables 8.9 and 8.10 give descriptive data on the levels and distribution of financial saving (net of borrowing). The second column of Table 8.9 shows that, on average, households have accumulated net financial savings (savings held in the form of cash, savings deposits, and bonds, net of borrowing) equal to 77 percent of their current per capita income. This level of deposits should afford households a decent buffer against moderate income declines when having to rely on their own means. Figures on financial gross saving (not net of borrowing), not reported here, are very similar to those on net savings. Households may also ‘save’ through building up and drawing down levels of food stocks and accumulating livestock, equipment, and consumer durables. Following

232

JONATHAN MORDUCH AND TERRY SICULAR

Table 8.9 Net Financial Savings, 1990–1993

Mean Standard deviation 25th percentile Median 75th percentile

Total saving

Annual increase in Total saving as a percentage of current saving income

3,860.1 7,300.7 955.0 2,308.2 4,780.1

77 87.8 27 57 102

156.5 3,949.0 −431.6 317.0 1,319.2

Annual increase in saving as a percentage of current income 2.0 71.4 −12.8 8.4 28.8

Notes: Net financial savings are financial savings plus loans to other households less borrowing, calculated in constant 1990 yuan over 992 observations. Households are categorized in quartiles based on their average per capita income (including transfers received, adjusted for time at home) over the four years.

the trends suggested by interviews in Zouping, however, we focus exclusively on financial saving here. By considering saving net of borrowing, we take into account both self-finance and finance through credit.184 As with income, though, medians are substantially lower than means. The median household can cover 57 percent of current per capita income with saving, and at the 25th percentile coverage falls to 27 percent. At the 75th percentile, however, 102 percent of current per capita income is covered. These figures on cumulative net savings do not translate into high yearly savings rates. The savings rate (that is, the annual increase in saving as a percentage of current income) is just 2 percent during the sample period, well below the national savings rates cited below (see Section 8.5). The median savings rate is higher but still a relatively modest 8.4 percent. Savings rates vary substantially among households and are strongly correlated with the level of income. At the 75th percentile, the savings rate is 29 percent, while at the 25th percentile savings rates are negative, at −13 percent. These low average savings rates for our sample reflect in part the boom in house-building and repair, as well as purchases of consumer durables, that occurred at the time of the survey. Excluding households with purchases of durables and construction of buildings pushes the average rate of savings up to about 7 percent, and the median rate to 11 percent, rates that are noticeably higher than the numbers in Table 8.9, but still lower than the national aggregates. Table 8.10 offers a different perspective on the saving data. Here, the sample has been broken into four equal-sized groups based on their per

184

See Park (1997) for an excellent analysis of household grain stock management in northwest China. Park et al. (1997) describe recent innovations in credit institutions in rural China.

233

RISK AND INSURANCE, ZOUPING COUNTY, CHINA

Table 8.10 Net Savings Rates by Average Household Income Percentile Mean Average household income in: Bottom quarter 3.5 2nd quarter −2.8 3rd quarter 3.3 Top quarter 4.2

25th

Median

75th

−17.2 −6.6 −12.2 −10.3

5.0 7.8 11.1 12.8

27.8 19.4 38.4 32.2

Notes: The net saving rate equals the change in net financial savings divided by household income. These numbers are calculated in constant 1990 yuan over 992 observations. Households are categorized in quartiles based on their average per capita income (including transfers received, adjusted for time at home) over the four years.

capita income. Savings rates increase somewhat in moving from lower to higher income groups, with more dramatic action in the upper tails of the savings distributions than in the means. Again, we see dissaving at the 25th percentile and substantial saving at the 75th percentile. As above, we present regression results in order to control for the broad range of household-level characteristics considered in Table 8.7. Random effects estimates of the determinants of savings rates are presented in Table 8.11 (we do not run Tobits since nearly all households report non-zero savings rates). Levels of income are positively associated with savings rates in the first two specifications, but standard errors are large and the coefficients are not significant—and the coefficient turns negative in the third, preferred specification. Not surprisingly, positive income growth is associated with higher savings rates. The converse is that income declines are associated with dissaving. Suffering a disaster is also strongly associated with dissaving. These coefficients are all uniformly statistically significant. None of the other variables take significant coefficients except for the sub-village cadre variable. One reason may be that, unlike the village-level cadres, officials at lower levels rely on their own resources to get along. Since they are relatively successful within villages, their savings rates are apt to be higher than those of others (who are not party members or cadres of any sort). Like the regressions on transfers, about 80 percent of the variation in net savings remains unexplained. As described above, some of this is likely to be associated with house-building and the purchase of consumer durables, and additional work is required to disentangle how these variables determine rates of accumulation. The evidence here shows that both transfers (especially from outside the village) and savings provide partial buffers in times of risk. These traditional forms of informal insurance have limits, however, and we turn next to market-based institutions.

234

JONATHAN MORDUCH AND TERRY SICULAR

Table 8.11 Determinants of Net Saving Rates, 1990–1993, Random Effects Estimates Variable Household income per capita (×10−6) Household income p.c. growth ‘Disaster’ reported? Adjusted household size Adult male share Communist Party member? Village-level cadre? Sub-village-level cadre? Year fixed effects? Village fixed effects? Year* village fixed effects? R2 Number of observations

(1) 4.38 (3.59)

(2) 4.67 (4.14)

(3) −1.88 (4.08)

0.10** (0.05)

0.12** (0.05)

0.08* (0.05)

−0.20* (0.12) 0.002 (0.03) 0.13 (0.22) 0.04 (0.08)

−0.20* (0.12) 0.01 (0.03) −0.05 (0.23) 0.07 (0.08)

−0.25** (0.12) 0.01 (0.03) 0.03 (0.22) 0.06 (0.08)

0.02 (0.08) 0.30** (0.13) yes no no 0.04 742

−0.001 (0.09) 0.22* (0.13) yes yes no 0.08 742

0.01 (0.08) 0.22* (0.13) yes yes yes 0.18 742

Notes: ** Significant with 95 percent confidence. *

Significant with 90 percent confidence. Standard errors are in parentheses. Calculated based on income (including transfers received) in constant 1990 yuan. Income is per capita (adjusted for time at home). The dependent variable is the change in net saving divided by household income. Regressions also include number of workers per capita, average age (and age squared) of adults, average education (and education squared) of adults, and three indicator variables for class background. Tobit regressions are not used here because almost all households in the sample report non-zero net savings rates.

8.4 Emerging Insurance Mechanisms and Markets Decollectivization and related economic reforms in rural areas have greatly reduced the role of formal institutions in providing insurance for the rural population.185 Our findings above indicate that the gap in insurance created by the decline of rural collectives and expansion of market-oriented activities has not been filled by informal and individual risk sharing mechanisms.

185

This section covers practices during 1990–94. The information is drawn from interviews with officials at the Ministry of Finance in Beijing (June 15, 1993), Shandong Province Insurance Company (May 28, 1993), Shandong Province Civil Administration (May 28, 1993), Zouping Insurance Company (June 21, 1993 and July 4, 1994), Zouping County Civil Administration Bureau (June 22, 1993 and June 28, 1994), Zouping County Bureau of Finance (July 1, 1993), officials in Xidong Township and households in Weijiapo Village (June 23, 1993), officials in Changshan Township and households in Baojia and Houshi Villages (June 23, 1993), officials in Jiao Qiao Township and Diao Song and Xiao Zhao villages (July 5, 1994), and officials in Libatian township and Liuzhuqiao Village (July 6, 1994).

RISK AND INSURANCE, ZOUPING COUNTY, CHINA

235

The Chinese government recognized that the reforms would create a demand for formal insurance and in the early 1980s resurrected the People's Insurance Company (PIC). Since that time the PIC has taken steps to extend the availability of formal insurance in rural areas. Efforts to sell crop insurance, for example, began in the late 1980s, and by the early 1990s certain, admittedly limited, types of crop insurance were widely available in Zouping and elsewhere in China. The government also provides disaster relief and certain other social insurance programs through the Ministry of Civil Administration and its local county bureaus. Private and foreign insurers were effectively prevented from entering the market, so insurance was only available through these channels. Here we evaluate these sources of formal insurance and future prospects. The PIC is organized along lines that follow the existing hierarchy of government administration in China. Offices of the PIC extend from Beijing through the provinces to the prefectures and finally to the county level. Premia are set at the provincial level in consultation with higher and lower levels. The prefecture, province, and central branches of the PIC provide partial substitutes for missing reinsurance markets through the practice of sharing premia and compensation obligations with higher levels (described further below). Financial management and calculation of profits and losses is at the prefectural level, but the PIC is not a purely profitmaking enterprise. For example, some loss-making lines (notably agricultural insurance) are retained due to the social concerns of policy makers. The insurance company views these losses as payment for the opportunity to sell profitable lines like property and life insurance. But, as a result of the losses, agricultural insurance compensations are limited and coverage is narrow.186 It is impossible to buy coverage against infestation of cotton by boll worms, for instance, despite their common appearance and destructiveness. In addition, the maximum coverage of insurable cotton losses was 70 yuan per mu, although the cost of production is roughly 200 yuan per mu from planting to harvest.187

186

There are two main forms of crop insurance in Zouping County. The first is insurance against damage to cotton due to water-logging, hail, or wind storms. The second is insurance against fires on wheat fields at the time of harvesting and threshing. There have also been some (not very successful) experiments with insuring fruit production. In general, insurance companies have lost money on crop insurance, although not excessively. Both the wheat and cotton insurance contracts share the same features: premia are in fixed amounts per unit land, compensation is limited to low amounts, the township is the agent for transacting the contract, and the terms of contracts are (nominally) uniform throughout the province.

187

One mu is 1/15 hectare or 1/6 acre. It is common for households in Zouping to have 3–4 mu per family member, so a family of four will have 12–16 mu (2–3 acres). The cost of cotton insurance during the study period was 1 yuan per mu, and the insurance company argued that the poverty of farmers prevented charging more.

236

JONATHAN MORDUCH AND TERRY SICULAR

The Zouping Bureau of Civil Administration picks up some of the slack, handling disaster relief as well as old age insurance (yanglao baoxian) and social insurance (shehui baoxian).188 Types of major disasters include windstorms, hail, and pest infestations. Still, disaster relief does not cover all forms of losses, and the distribution of relief is often skewed towards lower-income households, rather than towards households with the greatest losses. Thus scope exists for the expansion of agricultural insurance by the PIC. Two aspects of the design of agricultural insurance programs under the PIC are unusual, and both these features build on the hierarchy of administrative levels in order to overcome missing markets and reduce transactions costs.189 The first feature is that the insurance company refuses to sell most lines of agricultural insurance to individual households. Instead, policies are sold to townships only. Figure 8.1 illustrates the relevant links in the administrative chain. In better-off townships, income from township-owned enterprises is frequently used to pay premia. In worse-off townships, however, collections are taken from households to pay the costs. In neither case, though, does the insurance company or the township deal directly with the households. Instead, the township deals with ‘affairs districts’ (banshi chu), an administrative level between the township and village. The banshi chu in turn deal with village leaders, who then deal with the households. When households face crop losses, again they do not deal directly with the insurance company. The insurance company only compensates the township, and the funds then percolate down through the banshi chu to the villages and, ultimately, to households. The advantages of this arrangement are several. First, it lowers transactions costs for the insurance company by shifting a share of the administrative burden on to the local leadership. This is particularly critical since plots are small and households are many. Unlike in many other countries, land within villages in rural China is allocated fairly equally, with land Fig. 8.1 Hierarchy Of Agricultural Insurance Provision By the People's Insurance Company, 1993

188

The Labor Bureau also provides some old age insurance, aimed at cadres and workers in enterprises rather than farm households.

189

Property and life insurance contracts are sold to individuals, with most property insurance purchased by households with private businesses (geti hu ), and the contracts are not notably different from those in other countries.

RISK AND INSURANCE, ZOUPING COUNTY, CHINA

237

allocations roughly proportional to household size. As a result, nearly all households own some land, and the fact that the farm population is comprised almost entirely of smallholders increases transactions costs at all stages of insurance provision. In addition, the sheer size of the rural population quickly turns any attempt at comprehensive coverage into a major effort. At the same time, rural areas are densely populated and villages are formed in clusters. This, and the relative uniformity of holdings, makes it possible to achieve valuable returns to scale through sales to townships. Second, since local leaders have discretion over the allocation of funds, they can take advantage of their information on the effort of households to penalize households for losses due to moral hazard and to reward those in particular need. Officials at township and village levels confirmed their use of this discretion, although the incidence of pay-outs was too infrequent to confirm these patterns in the longitudinal data. In these two ways selling insurance to the community mirrors advantages found in group-lending contracts used to channel credit to households lacking collateral. These group-lending contracts channel resources to individuals but do so through contracts with groups of clients only, and they have been central to the explosion of interest in programs that lend to poor clients, so-called ‘microfinance’ programs (especially those following the model of the Grameen Bank of Bangladesh). Microfinance has been attempted around the world, including in a growing number of sites in China (Morduch 1999b; Morduch et al.1997). The insurance contracts seen in Zouping are far from optimal, but they suggest that there may be applications of lessons from microfinance for insurance provision, and we return to that possibility below. The third advantage is that adverse selection is mitigated, since the company can enforce that a township purchases insurance for the entirety of a village. Thus, isolated residents (who may be riskier than average) cannot purchase insurance independently—although there may remain a tendency that villages with the greatest risks are more likely to be insured than other villages. While this feature of agricultural insurance in China has advantages, it places discretion in the hands of local leaders and so opens the possibility of misfeasance. This possibility is mitigated by joint inspection of damage by township parties, local officials, and county insurance company staff, and by placing lists of recipients of compensation and amounts received prominently in villages. The lists are posted jointly by township and local officials, but while the practice is strongly suggested, it is not mandatory. (The Zouping Bureau of Civil Administration also distributes funds in this way, and distribution lists are supposed to be posted in villages. Staff at each level in the hierarchy then visit the level below to verify disbursements.) An example is given by responses to a drought in Xidong Township. Xidong experienced a severe drought in 1989, and it continued to be dry

238

JONATHAN MORDUCH AND TERRY SICULAR

through 1992, especially in the mountainous sections of the township. Thirty villages had trouble obtaining enough drinking water in 1989, and, overall, grain output (mostly wheat and corn) fell by about 30 percent. The County Bureau of Civil Administration responded with grain relief and reductions in grain quotas to be delivered to the state. The county bureau also set up a reserve fund that allowed disbursal of credit at subsidized interest rates.190 The process of disbursement begins with a household's application for a relief loan from the reserve fund. Approval of the loan by the village leader is required, and, because most households lack adequate collateral, a guarantor must be found. For small amounts, the village leader or some other respected individual typically acts as guarantor, but a work unit must guarantee larger loans. The township-level Civil Administration staff then investigate the application, relying heavily on village-level cadres, whose stamp is required on applications. Funds then are disbursed directly to the households, bypassing the hands of village leaders. Such loans and relief helped in Xidong, but households typically also had to rely on finding alternative employment in the nearby city of Zhou Cun and also reduced their consumption levels. In particular, households delayed house repairs, furniture purchases, and expenses on smaller household items. The second feature of agricultural insurance that draws on existing hierarchies is the practice of ‘dividing insurance’ (fenbao) mentioned earlier. Since China lacks the possibilities for re-insurance available in OECD economies, local insurance companies need a way to diversify their risk. Lacking markets for that risk, the practice of dividing insurance achieves much the same purpose. Under the system, the prefecture turns over 30 percent of its premia and insurance burden to the province-level insurance company. The province-level company then turns over 30 percent of its premia and burden to the national company. The well-understood hierarchy, from local levels up to Beijing, facilitates the smooth functioning of this division. These practices are responses to difficulties that arise in selling insurance in such an environment. Worldwide, efforts to provide crop insurance have yielded far more examples of failure than success (at least as measured by profitability), and the experiences in Zouping are no exception. Not long ago, similar problems were common in rural credit provision: demand for credit by low-income households was great, but public provision of such credit experienced many more failures than successes. The

190

There are two types of funds for disaster relief. The first is a savings fund (jiuzaifupin chujinhui ) and is comprised mainly of deposits from individuals, functioning like a credit society. The second is a circulating fund (jiuzaifupin zhouzhuanjin ), and it is comprised mainly of money deposited by levels of government above the township-level. The use of funds has been encouraged by the central government, and funds are used for a variety of purposes within local governments.

RISK AND INSURANCE, ZOUPING COUNTY, CHINA

239

microfinance movement has, however, provided a relatively successful model for rural credit provision. Below we draw on experiences in Zouping, as well as on lessons from microfinance, to suggest ways to improve insurance performance.

Proposals: Experience Rating and Joint Liability It was noted above that insurance contracts in Zouping have some features that are similar to those of microfinance programs. Those features are not employed with an eye to optimality, however, and we consider steps that could aid the efficiency and profitability of agricultural insurance in regions like Zouping. The first stumbling block is that the provincial office of the insurance company mandates that all households buying insurance for a given crop should pay the same price. This simplifies record keeping, but it creates problems given the agro-climatic heterogeneity of the province. In practice, though, the local insurance company has crafted ways to get around the regulation. The principal tactic is to lend money to townships to cover insurance purchases and to then vary the terms of the loans to effectively vary the price of the insurance. This helps the insurance company stem adverse selection, but they could do much better through more careful pricing. One proposal, then, is to use ‘experience rating’ to help set premia. Unlike some other risks like fire, health or death, the risks in agriculture arise regularly (with each harvest) and in ways that are reasonably well understood. Thus, one way to move toward a more fair and efficient crop insurance program is to take the history of risks into account in setting premia. In the USA, this sort of ‘experience-rating’ is used to determine taxes paid by employers for unemployment insurance and to determine automobile insurance rates. It is also common in health insurance contracts. If such experience rating were implemented on a household by household basis, the transactions costs would likely be prohibitive. If implemented at the township level, however, much of the important heterogeneity would wash out and free rider problems could emerge. The most natural unit to consider is a smaller unit such as the village or, in large villages, a unit akin to the pre-reform production team. The use of villages or teams can provide protection against problems of moral hazard. If the scheme is appropriately structured, local information and enforcement abilities can be harnessed to improve repayment rates and limit liabilities. The fee structure could be fairly simple. There might be, say, ten categories of riskiness. For each insured activity, villages would be initially categorized at a level which would make their premia roughly actuarially

240

JONATHAN MORDUCH AND TERRY SICULAR

fair (relative to expected compensation). Then, at regular intervals, the village would either be rewarded by lower premia if they perform better than expected or penalized if they perform worse. By moving away from uniform rates, adverse selection is mitigated, and by creating adverse consequences in response to adverse outcomes, moral hazard is mitigated as well. An advantage would be that, as with microfinance contracts, those who have information about behavior (neighbors) have incentives to use it to minimize misfeasance (since, as losses mount, so do future premia). This further reduces the moral hazard problem, in much the same way that group lending practices have lowered default rates in other contexts (e.g. the Grameen Bank). The proposed program would not address issues of adverse selection within the village, and these could pose difficulties. What would happen if a farmer with inherently riskier production methods decided to start growing a crop that had previously been grown only by ‘safer’ farmers? By adding variability to the distribution of his group's total expected production, the farmer would clearly impose a negative externality on the others in the same insurance group. One might be tempted to leave this problem to individual households to work out (à la Coase), which would be possible if information is well known and contracts were enforceable. Neither assumption is fully plausible, but the problem has an analogue in microfinance group lending contracts with joint liability, and it has not been a major impediment there. The broad proposal outlined here is of course speculative, but the Chinese government is presently experimenting with insurance reform, and the contractual innovation sketched above could provide a constructive way forward.

8.5 Concluding Remarks While China's reforms have brought great economic opportunities for many households, they have also brought increased economic risks. These risks arise in part due to the dismantling of rural collectives, once the major risk sharing institution in rural China, alongside increased commercialization and exposure to markets. These changes raise the question of whether informal insurance mechanisms similar to those observed in other countries have emerged. Our data from Zouping County show some evidence consistent with informal insurance through inter-household transfers, but the scale remains limited. Household savings also provide an avenue for coping with risk, but savings rates in our sample are modest. Both saving and risk sharing behavior have been affected by substantial changes in the demographic context, pushed by the sharp decline in fertility rates and increasing life expectancy. The result has been a decline in

RISK AND INSURANCE, ZOUPING COUNTY, CHINA

241

family size and an aging of the population. In 1965, 6.4 children were born over a woman's reproductive lifespan and 40 percent of the population was under age 15. By 1995 the fertility rate had fallen to 1.9 births per woman, and just 26 percent of the population was under age 15. In the absence of the collective, the main locus for risk sharing has been the family, and the shrinking of large families, especially extended families living together or in separate households, raises the demand for alternative risk sharing mechanisms. With encouragement from the state, formal insurance markets are developing (particularly markets for property, life, and health insurance), but as discussed above agricultural insurance remains a losing proposition for the People's Insurance Company, and coverage is limited despite the fact that the economy remains deeply rooted in agriculture. The shifting demographic context also increases the need for old age insurance. Investment in children can no longer be the sole form of saving for retirement, and the domestic savings rate (as a percentage of GDP) has increased from 24.8 percent in 1965 to 42.0 percent in 1995 (World Bank 1998). Increased investment opportunities surely explain part of the change in savings rates, but demographic forces have also been found to play a critical role (Kraay 1998). In 1992 rural old age insurance was started in Zouping on an experimental basis, with the Bureau of Civil Administration highlighting pressure created by fertility reductions and de-collectivization as the reason for this experiment. The new policy reflects a growing concern with pension insurance throughout China. These demographic developments highlight the importance of savings and insurance for China's future development, and they will continue raising questions about the roles that communities and markets can play in reducing vulnerability. In this context, we put forward a proposal for innovation in insurance provision. The Chinese government has earned a reputation for its commitment to running localized experiments with new policies and institutions, including microfinance. There are potential lessons for insurance reform to be drawn from microfinance institutions, and we have sketched some ideas for how this might work in principle. The key ideas include the incorporation of experience-rating in contracts, coupled with sales of insurance to villages or other groups, rather than to individuals. These proposals refine current practice, which is to sell crop insurance to townships not to individuals, but revise existing mechanisms so as to achieve greater efficiency and provide broader coverage. The proposal provides a way to meet future challenges by thoughtfully combining community with market.

References Chan, A., Madsen, R., and Unger, J. (1992) Chen Village under Mao and Deng, University of California Press, Berkeley.

242

JONATHAN MORDUCH AND TERRY SICULAR

Coate, S. and Ravallion, R. (1993) ‘Reciprocity without Commitment: Characterization and Performance of Informal Insurance Arrangements,’ Journal of Development Economics, 40(1): 1–24. Friedman, E., Pickowitz, P., and Selden, M., with Johnson, K. A. (1991) Chinese Village, Socialist State, Yale University Press, New Haven. Jalan, J. and Ravallion, M. (1999) ‘Are the Poor Less Well Insured? Evidence on Vulnerability to Income Risk in China,’ Journal of Development Economics, 58(1): 61–81. Kipnis, A. (1998) Creating Guanxi: Sentiment, Self, and Subculture in a North China Village, Duke University Press, Durham, NC. Kraay, A. (1998) ‘Household Saving in China,’ unpublished manuscript, The World Bank, Washington, D.C. Malinowski, B. (1922) Argonauts of the Western Pacific, Routledge, London. Mauss, M. (1967) The Transfer: Forms and Functions of Exchange in Archaic Societies, Norton, New York. Morduch, J. (1999a) ‘Between Market and State: Can Informal Insurance Patch the Safety Net?’ World Bank Research Observer, 14(2): August. —— (1999b) ‘The Microfinance Promise,’ Journal of Economic Literature, 37(4): December. —— and Sicular, T. (1998) ‘Rethinking inequality decomposition, with an application to rural China,’ unpublished manuscript, Harvard University and the University of Western Ontario. —— and Sicular, T. (2000) ‘Politics, Growth, and Inequality in Rural China: Does it Pay to Join the Party?’ Journal of Public Economics, forthcoming. —— Park, A., and Wang, S. (1997) ‘Microfinance in China,’ Poverty and Development (Beijing, in Chinese), October. Nee, V. (1989) ‘A Theory of Market Transition: From Redistribution to Markets in State Socialism,’ American Sociological Review, 54(5): 663–81. —— (1996) ‘The Emergence of a Market Society: Changing Mechanisms of Stratification in China,’ American Journal of Sociology, 101(4), January: 908–49. Oi, J. (1989) State and Peasant in Contemporary China, University of California Press, Berkeley. Park, A. (1997) ‘Risk and Household Grain Management in China,’ unpublished manuscript, Department of Economics, University of Michigan. —— , Brandt, L., and Giles, J. (1997) ‘Giving Credit where Credit is Due: The Changing Role of Rural Financial Institutions in China,’ unpublished manuscript, Department of Economics, University of Michigan. Potter, S. H. and Potter, J. (1990) China's Peasants: The Anthropology of a Revolution, Cambridge University Press, New York. Putterman, L. (1993) Continuity and Change in China's Rural Development: Collective and Reform Eras in Perspective, Oxford University Press, New York. Sicular, T. (1998) ‘Establishing Markets: The Process of Commercialization in Agriculture,’ in Walder, A., (ed.), Zouping in Transition: The Process of Reform in Rural China, Harvard University Press, Cambridge, MA. State Statistical Bureau (various years) China Statistical Yearbook (zhongguo tongji nianjian), China Statistical Publishing House, Beijing.

RISK AND INSURANCE, ZOUPING COUNTY, CHINA

243

Walder, A., (ed.) (1998) Zouping in Transition: The Process of Reform in Rural China, Harvard University Press, Cambridge, MA. Whyte, M. (1991) ‘The Commune as a Social System,’ in Dernberger, R., Dewoskin, K., Goldstein, S., Murphy, R., and Whyte, M., (eds.), The Chinese: Adapting the Past, Facing the Future, Center for Chinese Studies, University of Michigan, Ann Arbor, MI. World Bank (1998) World Development Indicators 1998 (CD-ROM), The World Bank, Washington, D.C.

This page intentionally left blank

Part III Governance of Local Commons

This page intentionally left blank

9 Water Community: An Empirical Analysis of Cooperation on Irrigation in South India Pranab Bardhan

9.1 A controlled supply of water is crucial for much of Asian agriculture.191 This is particularly the case in many of the semiarid and uneven-rainfall areas of south Asia. In 1853, when Marx and Engels first discussed, in correspondence, the problems of ‘Asiatic’ societies, they immediately agreed on the particular importance of public irrigation in these societies, necessitated by what they called ‘climatic and territorial conditions’, and of the particular social organizations this gives rise to.192 Even though their emphasis on the idea of the centralized hydraulic state was somewhat misplaced, it was a valuable insight which gave rise to a large anthropological literature on the relationship between irrigation and social and political organization.193 While there is a considerable amount of work on the management of large canal systems and the structure and practices of the irrigation bureaucracy, economists have only recently started paying attention to issues of local community-level cooperation and other institutional arrangements that are key to substantially improving the existing levels of utilization of irrigation potential. Community institutions can have various functions in different irrigation systems: for example, they aim at pooling efforts and resources in constructing and maintaining field channels at the local outlet level, at regulating water allocation and monitoring violations; in cases of tank irrigation also at desilting, weeding, and stopping

191

The data used in this chapter are from a survey designed with the help of Nirmal Sengupta at the Madras Institute of Development Studies, who has supervised the data collection process and in general been very helpful at the data analysis stage as well. Jeff Dayton-Johnson also helped in the survey design. In the data collection and coding process we have received invaluable assistance from R. Manimohan, A. Raman, and J. Jeyaranjan. The data collection effort was funded by the MacArthur Foundation. Thanks are also due to Samuel Bowles, Jeff Dayton-Johnson, Gershon Feder, Yujiro Hayami, Elisabeth Sadoulet, and A. Vaidyanathan for comments on an earlier draft. The assistance of Laura Giuliano in the data analysis is gratefully acknowledged.

192

In this they were carrying forward one strand of European thinking on Asian history and society, that was particularly fostered by Adam Smith and John Stuart Mill. For an account of the roots of Marx's thinking on this question in earlier European tradition as well as a critique of the ingredients of the concept of ‘Asiatic mode of production’, see Anderson (1974).

193

For an early survey of this literature see Hunt and Hunt (1976).

248

PRANAB BARDHAN

encroachments on tank beds; at repairing, maintaining, and controlling water allocation from public and community tubewells; at controlling groundwater overexploitation with privately owned pumps in areas with fragile aquifers, and so on. Water reform in the sense of building or promoting such community institutions of cooperation is at least as important in south Asia as land reform in rural development.194 But the history of local community-level cooperation in water management in south Asia is rather mixed.195 There are several documented examples of successful local community water management (although usually at a rather low level of organizational form) in different parts of the region, some of them going many hundreds of years back in history and still surviving, but there are also numerous cases of failure of cooperation, leading sometimes to an anarchical scramble for water. It is important to understand the conditions working for and against sustainability of local cooperation in situations of general social and economic interdependence. The usual game-theoretic models of cooperation among self-interested agents in repeated situations of strategic interdependence provide some useful insights. But, in view of their admissibility of multiplicity of equilibria, many of the comparative-static questions cannot be satisfactorily resolved without recourse to contextual data analysis. Besides, these models are often couched in too restrictive a framework, inadequate in capturing some of the salient issues of real-world cooperation (or lack of it); for a long time we will have to depend on arduous empirical work to give us clues about the building blocks to more satisfactory theoretical models and policy action. This chapter concentrates on an analysis of the empirical determinants of cooperation in irrigation communities196 on the basis of a large survey conducted in the South Indian state of Tamil Nadu. In Section 9.2 we describe the variables and the data and the effects that we expect the independent variables may have on cooperation, and in Section 9.3 we report the actual results of the statistical analysis.

9.2 In all data were collected from 48 villages spread over six districts in Tamil Nadu, with a selected unit (called ‘ayacut’) of irrigation system in each

194

I have elaborated on this theme in Bardhan (1984, ch. 16).

195

Apart from the more well known accounts in Ostrom (1990), Wade (1987), and Vaidyanathan (1984), see also Jayaraman (1981), Easter and Palanisami (1986), Sengupta (1991), and Vaidyanathan (1994).

196

For a partially similar analysis of determinants of the adoption of distributive rules and of cooperative maintenance effort with data from 54 farmer-managed surface irrigation systems in central Mexico, see Dayton-Johnson (1998).

COOPERATION ON IRRIGATION IN SOUTH INDIA

249

village (either a tank or a branch of canal of roughly about 50 hectares size). Half of the irrigation units belonged to canal systems, and the other half to more traditional tank systems. Within tank systems, half belonged to what are called ‘isolated or chain tank’ systems, and the other half to ‘system tanks’ (where the tanks were, unlike in the former case, linked to larger irrigation units). Within each system the villages were randomly chosen. Within each village a stratified sample of ten farmers (stratified by land size classes) were chosen. Most of the analysis in the chapter is, however, on the basis of data for the irrigation unit or the ‘ayacut’ as a whole; in some cases we have derived data from the individual farmers' responses. In general on matters of cooperation differences across villages were much more prominent than intra-village differences among the ten farmers. First a few words on the description of the general functioning of the water users' organizations. The majority of the organizations surveyed are traditional and informal community organizations that have been in existence for some time (27 of the 48 surveyed are reported either as ‘traditional’ or at least 20 years old). But only 13 of these units have formal associations (and ten of these have formal associations not at the village or sluice levels but at the more aggregative zonal level). Twelve of the irrigation units surveyed have been set up (relatively recently) and are run by the Public Works Department (PWD). In another 12 canal-based irrigation units, although PWD is the overall official agency, it is the traditional village committee that manages irrigation matters at the village level. There is a great deal of variation among the organizations with respect to the appointment of guards as monitors and enforcers, frequency of meeting, mobilization of collective labor, mobilization of funds, method of cost-sharing, and involvement in non-irrigation activities.197 Most of the users' organizations meet at least once every one to two years; only nine reported never meeting at all. The units run by the PWD reported meeting at most once a year (and often only every two years), while most others meet once or twice a year during the cropping season. Five of the units surveyed have organizations which reported having regular meetings at least ten times a year. One function of the water users' organization in many of the units surveyed is to mobilize community labor for the purpose of maintaining and repairing the field channels. Generally, it appears that community labor is most common in those organizations that are traditional or have existed for over 40 years. While most of these units mobilize community labor both for regular maintenance and for emergency repairs, several units reported mobilizing community labor only in the event of emergency

197

Unfortunately, the data and descriptions regarding some of these organizational variables were not systematic or complete enough to allow reliable coding and inclusion in our subsequent regression analysis.

250

PRANAB BARDHAN

repairs. Community labor does not appear to be used systematically in any of those units where the PWD runs the organizations. About three-quarters (37 in number) of the units surveyed have some formal system of fundraising. Of these 28 units have a system of dues, fines and/or taxes. Such a system is most prevalent in those units which are canal-based. In the tank-based systems, an alternate system of fundraising is possible, the sale of collective resources such as fish and trees. Nineteen of the tank-based units have collective funds which are mobilized this way; ten of these units supplement the collective fund with a system of dues and/or tax collection. About half of the water users' organizations report participating in other, usually village-wide, activities, such as conducting the temple festival. This does not appear to be true of any of the organizations which are run by the PWD. Since issues like cooperation and some of its determinants involve the quality of social and economic relations, much of the data are more qualitative (sometimes with yes–no answers) than quantitative, from which we had to construct variables that therefore take only a limited range of values. The description of the dependent and independent variables, including their mean values with standard deviation and the range of variation, is given in Table 9.1. We have used three alternative variables as indicators of cooperation198 within the community on matters of irrigation: (1) quality of maintenance of distributaries and field channels (we call the variable DISTFDCH); (2) absence of conflict in water allocation in the ayacut in the last five years (we call this variable H20COOP); and (3) extent of violations in water allocation rules (we call the variable VIOLH20). From Table 9.1 we can see that there is a considerable amount of variations among the villages in the values of these variables. Let us now describe and comment upon the explanatory variables that we have used. One obvious variable to consider is that relating to the group size: we have used a variable called NUMUSERS, which is the number of households199 using the particular irrigation source (the number varies between 11 and 279, with a mean of 53). The common presumption in the literature on local commons is that cooperation works better in small groups. The early work of Olson (1965) suggested that collective action is easier to organize in smaller groups. In a game played by rotating irrigators (turntakers and turnwaiters) with self-enforced rules (i.e. without formal guards), Weissing and Ostrom (1990) show in their theoretical model that in equilibrium an increase in the number of irrigators is associated with an increase in the stealing of water, other things remaining the

198

We shall later comment on how these may not always be good indicators of cooperation among the farmers.

199

We have also used an alternative variable, NUM_ACRE, the number of households using the irrigation source per acre of the ayacut area. But usually this variable is less significant than NUMUSERS in our regressions.

251

COOPERATION ON IRRIGATION IN SOUTH INDIA

Table 9.1 Dependent and independent variables Variable name

Mean

DISTFDCH

No. of observations 45

Min

Max

1.31

Standard deviation 0.92

0

2

H2OCOOP

48

0.67

0.48

0

1

VIOLH2O

48

0.52

0.50

0

1

NUMUSERS

48

52.67

54.29

11

279

NUM_ACRE

48

0.36

0.02

0.06

1.07

GINI

48

0.41

0.11

0

1

CASTE75

48

0.69

0.47

0

1

AVGACCES

48

3.65

1.20

2

7

TOPOGRPH

48

0.46

0.50

0

1

LINED

48

1.96

0.82

1

3

CANAL

48

0.50

0.51

0

1

PWDDECID

48

0.25

0.44

0

1

TAILEND

48

0.75

0.44

0

1

Description of variable Index of quality of maintenance of distributaries and field channels No conflict over water within village in the last 5 years† Water allocation rules frequently violated by at least one group† No. of beneficiary households using this irrigation source No. of beneficiary households per acre of ayacut area Gini coefficient of landholding of beneficiary households in ayacut At least 75% of sampled farmers are members of the same caste group† Number of months there is access to irrigation No equal access to water because of topographical nature of the ayacut† If system is partially or fully lined If ayacut is in a canal system† If PWD takes all decisions on water allocation† If village is situated at tailend of the irrigation system†

252

PRANAB BARDHAN

PRIORAPP

48

0.44

0.50

0

1

LINKAGE

48

2.46

0.74

1

3

MARKET

48

2.39

0.70

1

3

OTHIRRIG

48

0.26

0.18

0

0.66

TRADLORG

48

0.56

0.50

0

1

GUARD

48

0.38

0.49

0

1

SHRPROP

48

0.19

0.39

0

1

LEGALRGT

48

0.33

0.48

0

1

WMARKET

48

0.44

0.50

0

1

Note: † in the description of a variable indicates that the variable takes only two values, 0 and 1.

No conflict with other villages over water† Index of connection with urban areas Measure of extent to which farmers are market oriented Estimated fraction of total irrigated land held by sampled farmers outside the ayacut Where irrigation organization has been there for 20 or more years† There exists at least one guard in the ayacut† Cost-sharing proportional to landholding† When formal water rights exist, as opposed to customary rights† If there is a market for well water†

COOPERATION ON IRRIGATION IN SOUTH INDIA

253

same.200 In general, models of repeated games suggest that cooperative strategies are more likely to emerge and be sustained in smaller groups. In small irrigation communities peer monitoring is easier, the ‘common knowledge’ assumption of models of strategic decisions is likely to be more valid, shared norms and patterns of reciprocity are more common, social sanctions may be easier to implement through reputation mechanisms and multiplex relationships, and even hydrologic needs of farmers may be relatively similar. On the other hand, there may be some positive economies of scale in larger groups, particularly in matters of pooling resources, appointing guards, lobbying with officials, and so on. Another possibly important explanatory variable is heterogeneity, both social and economic. The only variable in our dataset that relates to social homogeneity or heterogeneity is CASTE75, which takes a value of 1 in villages where at least 75 percent of the sampled farmers are members of the same generic caste group (in most cases a ‘backward’ caste group) in the village. By this crude measure, 69 percent of the ayacuts in our dataset are socially relatively homogeneous. In terms of economic heterogeneity we have a variable called GINI, which measures the Gini coefficient of inequality of landholding of farmers in the ayacut area. The mean value of GINI in our dataset is 0.41. Theoretically, the relationship between heterogeneity and successful collective action is rather complex.201 On the one hand, there is the well known suggestion of Olson (1965) that in a heterogenous group a dominant member enjoying a large part of the benefits of a collective resource is likely to see to its provision even if he has to pay all of the cost himself (with the small players free riding on the contribution of the large player). On the other hand, there are cases where the net benefits of coordination of each individual may be structured in such a way that in situations of marked heterogeneity or inequality some individuals (particularly those with better exit options) may not participate, and the resulting outcome may be more inefficient. In addition, the transaction and enforcement costs of cooperative arrangements may be high in situations of intra-village heterogeneity, conflicts, and inequality. Internalization of cooperative norms is more difficult under such circumstances; the degree of confidence or trust that individuals have in the likelihood that others will play their part in a cooperative agreement may be low. In anticipation that the relationship between inequality and cooperation may be non-linear, we have introduced a squared Gini coefficient (called GINISQU) in our set of explanatory variables.202

200

Of course, other things do not remain the same. We shall see in our data that the existence of guards is positively (though not very significantly) correlated with NUMUSERS, and guards have a positive influence on cooperation.

201

For a theoretical analysis of some of the complexities, see Baland and Platteau (1997, 1998), and Dayton-Johnson and Bardhan (1998).

202

For a theoretical analysis of a U-shaped relationship between wealth inequality and cooperation in the context of a two-player non-cooperative model of conservation of a common-pool resource, see Dayton-Johnson and Bardhan (1998).

254

PRANAB BARDHAN

Another kind of factor that affects cooperation in water management is the physical condition of water availability. In conditions of extreme scarcity arrangements of cooperation often break down. When there is some significant access to water, it pays the irrigators to cooperate in maintaining field channels, and in obeying allocation rules. In our dataset the villages with acute water scarcity in general have less cooperation, but (probably for historical reasons) the waterscarce villages are also more likely to have canal-based irrigation; so the effect of water-scarcity is mixed up with that of possible bureaucratic inefficiencies in the release of canal water (more on this below). We have defined a variable AVGACCES to indicate the number of months in a year that the farmers in the ayacut have access to water: it varies from two to seven months, with a mean of 3.7 (2.8 for the canal areas).203 We have also used two other variables relating to physical conditions: one is TOPOGRPH to indicate where the topographical nature of the ayacut precludes equal access to water for all the farmers; and the other is LINED to indicate where the irrigation channel is lined. For government involvement we have used two alternate variables, CANAL (to mark half the villages where it is an administered canal system), and PWDDECID (where the PWD of the state government makes all the decisions about water allocation and distribution). In the canal area villages there is not merely less water availability, water is also more inequitably distributed: in 19 out of 24 canal area villages in our dataset (as opposed to 2 out of the 24 villages served by the tank systems) there is evidence of such inequity of water supply or access. In general, the water release cycles may be more unreliable from the farmers' point of view when they are administered by PWD officials. Of course, there is potentially a major problem of endogeneity here: does the PWD get involved when irrigators' cooperation has failed?204 After some background checking we have found that the villages where PWD takes all the decisions are all located in Coimbatore district and are precisely those where, due to primarily physical reasons of long-term water scarcity, a large-scale system of inter-river basin water transfer scheme had been undertaken by the government (called PAP—Parambikulam Aliyar Project). We therefore think the problem of endogeneity may not be that serious here.

203

It is well known in the irrigation literature that no single-dimensional variable can adequately capture the quality of water supply, which depends on many factors (like volume, timing, crop produced, and cultivation practice). As an alternative to the variable AVGACCES we have also tried a variable indicating if water is ‘sufficient’ for the paddy crop in the primary season, but this variable turned out to be less significant in the statistical analysis than AVGACCES.

204

Alternatively, the farmers may get involved in cooperation, when the PWD fails to do its job, but this looks unlikely from the details of the description of the irrigation units in the survey.

COOPERATION ON IRRIGATION IN SOUTH INDIA

255

A few variables relate to the locational context of the ayacut in question. The variable TAILEND refers to the cases where the village is situated at the tailend of the irrigation system, which, other things remaining the same, may unite the farmers of the village in their struggle to get larger amounts of water away from the more favorably located villages. In our dataset 75 percent of the villages are at the tailend of their respective system. In general only 44 percent of villages report that there is no conflict with other villages over water. As an alternative to TAILEND, we have also used a variable PRIORAPP to indicate those villages where no water conflict is reported with other villages. The locational context is also important in the matter of the exit options open to the villagers in conforming to cooperative arrangements. We have used LINKAGE as a variable to be a proxy for the connection of the village to urban areas or transport and communication modes (like bus and telephone). We also have a variable MARKET as a measure of the extent to which the farmers are oriented to selling their produce in the outside market. (Since how much a farmer sells his produce for may be determined by his other characteristics, i.e. it may be endogenous, we have used a predicted value of the MARKET variable in the regressions for determinants of cooperation). A somewhat different kind of exit-option variable may be indicated by how much access an ayacut member has to water sources outside the ayacut.205 The variable OTHIRRIG is the estimated fraction of the total irrigated land (of the sampled farmers in the ayacut) that is outside the ayacut. For all of these exit-option like variables our prior is to expect a negative effect on cooperation.206 The history of cooperation in a village may matter, as cooperation may be self-reinforcing or ‘habit-forming’, as Seabright (1997) explains in a theoretical model. We have a variable TRADLORG to indicate villages where the water users' organization has been there for 20 years or more. These villages are characterized by more use of community labor in maintenance works and emergencies, are more likely to hire guards (called ‘neerani’ in much of this area) for monitoring and enforcement (indicated generally by the variable GUARD), and to use cost-sharing proportional to landholding (indicated generally by variable SHRPROP). Since the use of guards and a cost-sharing regime may be endogenous, we have used in the cooperation regressions predicted values of these variables, PGUARD and PSHRPROP. We have also asked questions about the farmers' perception about the process of rule crafting. For example, in villages where at least four out of

205

We have tried another variable to indicate access of farmers to alternative, private, sources of water within the village (like pumpsets and open wells). But in the dataset this variable is highly correlated with variables like CANAL. One possible reason may be that government canals increase the subsoil water for private wells.

206

Except in the case of MARKET, where the negative effect may be modified by the market opportunities for produce raising the return to cooperation on water.

256

PRANAB BARDHAN

ten sampled farmers believe that the water rules were crafted by the elite, we indicate that by our variable ELITERUL. (We also had a variable, FAIR, for villages where the rules are generally perceived as fair, but we dropped this variable as it almost coincides with cases of cooperative behavior.)

9.3 The regression results are in Tables 9.2 (A, B, and C), 9.3 (A and B), and 9.4 (A and B). (For some of the independent variables the predicted values are estimated on the basis of first-stage regressions in Table 9.5.) We are primarily concerned with the signs of the regression coefficients.207 Let us start with DISTFDCH, the index of quality of maintenance of distributaries and field channels, as the dependent variable. This index is uniformly lower in ayacuts with higher inequality in landholding: the coefficient for GINI is negative in all of the three alternative cases in Table 9.2 and highly significant in two of the three cases. It is similarly lower, and significantly so, with ELITERUL, i.e. for villages where rules are perceived to be crafted by the village elite.208 As our discussion in the preceding section suggests, the effect of the number of users of the irrigation system can be ambiguous. In the ordered logit estimates the coefficient for NUMUSERS is negative but not very significant in one case, and significant at the 10 percent level in another; in another case when we take the number of users per acre of the ayacut area as an alternative, the coefficient is positive but not very significant.209 AVGACCES, the number of months in a year there is access to irrigation, is positively and significantly related to DISTFDCH in all three cases. The interpretation is that when there is access to an irrigation system over a longer period in the year the return to investing effort in maintenance of field channels is higher (and may outweigh the increased maintenance costs that may follow from higher use).210 The negative coefficient of LINED at first seems surprising, but probably means that when the channels are

207

Since this is a logit analysis the signs are to be appropriately interpreted as indicating directions of changes in the probability of the dependent variable taking on a particular value with changes in the values of an independent variable, even though in our statement of the results we sometimes do not spell it out fully.

208

From the data of the 480 sampled farmers it is clear that when an average farmer believes that the water rules have been crafted jointly (as opposed to by the elite or by the government), he is more likely to have positive comments about the water allocation system and about rule compliance by other farmers.

209

In the corresponding OLS regressions neither NUMUSERS nor NUM_ACRE is at all significant.

210

There is reason to believe that the access to irrigation refers more to general accessibility to water sources and structures than to the flow of water through the field channels, so the reverse causation (i.e. better maintained channels implying better access to irrigation water) is an unlikely explanation of the positive coefficient. In fact in general in the canal areas AVGACCES is very low, and yet DISTFDCH is high.

257

COOPERATION ON IRRIGATION IN SOUTH INDIA

Table 9.2 Ordered logit estimates Dependent Variable: DISTFDCH Variables NUMUSERS NUM_ACRE GINI AVGACCES LINED PWDDECID TAILEND PRIORAPP LINKAGE OTHIRRIG PGUARD PSHRPROP ELITERUL No. of observations Log likelihood Pseudo R2

(A) Coefficients −0.0308 −31.1481* 5.0160* −2.0352 8.0913*

−4.2195* −4.8767 9.0919* 11.6709 −8.8760* 45

Level of Significance 20% 5% 5% 15% 5%

5% Insignificant 5% 15% 5%

−15.4788 0.5900

(B) Coefficients

Level of Significance

3.5369 −12.0576 3.3115* −6.2620* 10.8124*

20% 15% 5% 5% 5%

−6.0943* −4.7874* −4.7635 7.1193*

5% 5% Insignificant 5%

−6.6006* 45

5%

−14.6912 0.6109

(C) Coefficients −0.0339*

Level of Significance 10%

−23.2620* 4.5637*

5% 5%

4.8454* 1.5790 −2.2385 −4.2900*

5% Insignificant Insignificant 5%

10.3531* 9.6847* −7.9717* 45

5% 10% 5%

−16.5396 0.5619

Notes: * indicates significance at 10% level or below. PGUARD and PSHRPROP are predicted values of the variables GUARD and SHRPROP respectively.

258

PRANAB BARDHAN

Table 9.3 Logit Estimates Dependent Variable: H2OCOOP Variables

(A) Coefficients NUMUSERS −0.0868 GINI −106.6778* GINISQU 76.2725* CASTE75 2.9169* AVGACCES 1.4893 LINKAGE −3.2398 PGUARD 22.4376 No. of observations 48 Log likelihood −10.3061 Pseudo R2 0.6627

(B) Level of significance Coefficients 15% −0.0533* 5% −77.4133* 10% 59.8977* 10% 2.1678 20% 20% −2.0861* 15% 15.3198* 48 −11.7129 0.6166

Level of significance 5% 5% 10% 15% 10% 10%

Notes: * indicates coefficients that are significant at 10% level or below. GINISQU is the square of the value of the variable GINI.

Table 9.4 Logit Estimates Dependent Variable: VIOLH2O Variables NUMUSERS GINI GINISQU CASTE75 AVGACCES TOPOGRPH TAILEND LINKAGE PMARKET OTHIRRIG PGUARD No. of observations Log likelihood Pseudo R2

(A) Coefficients 0.0350* 25.3600 −34.9862 −0.8536 −0.8960 1.1632 2.1425* 5.8283 −7.5257* 48 −15.7310 0.5266

(B) Level of significance Coefficients 5% 0.0289* Insignificant Insignificant Insignificant 15% −1.0435* Insignificant 1.1202 −0.4413 5% 0.499* 15% 4.5255 5% −3.4537* 48 −20.0658 0.3961

Level of significance 5%

5% Insignificant Insignificant 10% 15% 5%

Notes: * indicates coefficients that are significant at 10% level or below. GINISQU is the square of the value of the variable GINI; PGUARD and PMARKET are the predicted values of the variables GUARD and MARKET respectively.

259

COOPERATION ON IRRIGATION IN SOUTH INDIA

Table 9.5 First-Stage Logit Estimates Independent variables

Dependent variables GUARD Coefficients Level of significance NUMUSERS 0.0305 20% GINI 9.8339 15% AVGACCES −0.8961 15% CANAL −4.2407* 5% LINKAGE 1.5349* 10% TRADLORG 3.0781* 5% LEGALRGT WMARKET No. of obser- 48 vation Log likelihood −13.6321 Pseudo R2 0.5707

SHPROP Coefficients 0.0265* 9.8831*

Level of significance 5% 10%

4.1930*

5%

2.6285* 2.0355*

5% 10%

MARKET Coefficients −4.5713 0.4339 2.8389*

48

2.0298* 48

−13.7212 0.4076

−32.8864 0.2917

Level of significance 15% Insignificant 5%

5%

Note: * indicates significance at 10% level or below.

lined they are more likely to be found occasionally in a ‘broken condition’, compared to channels which are not lined at all. (In fact, of the 15 units that are not lined at all, only two report the field channels being in bad condition.)211 An alternative explanation, suggested to us by Dayton-Johnson on the basis of his observation in Mexico, is that some poor villages are unable to properly take care of the modern water control structures that tend to accompany lined canals, either because they could not get replacement parts, or because they did not know how to repair them, or both. Contrary to the general impression about the negative effect of government involvement, the coefficient for PWDDECID is positive and significant in all three cases. It is possible that when PWD is actively involved in water allocation and in maintenance works above the outlet, returns to farmers' collective maintenance efforts on channels below the outlet are higher. Looking more closely at the detailed data we find that of the 12 units where PWD is actively involved, in half of them the field channels are well maintained,212 and in the other half of the units they are badly maintained. It is interesting to observe that in the latter half the farmers have more private sources of water (pumps and open wells) than in the former half (and the GINI coefficient of land

211

Following a suggestion of Gershon Feder we ran a regression only for the 30 units which are lined (partially or fully). The results are similar to those reported in Table 9.2, except that the significance of some of the variables diminishes (partly because of the reduced number of observations).

212

In four of these six units it seems that the PWD, rather than a traditional village organization, looks after the field channel maintenance as well (though the farmers make some cash contributions). This means good maintenance may not always be simply interpreted as an indicator of cooperation among the farmers.

260

PRANAB BARDHAN

distribution in the ayacut is considerably higher, i.e. as expected, inequality of landholding goes with more access to pumps and open wells). The coefficient for TAILEND is positive but not significant. But the coefficient for PRIORAPP is negative, and in one case highly significant. This suggests that water conflicts with other villages may be helping the farmers of the ayacut unite in cooperative maintenance effort. The variable for urban connections, LINKAGE, has a highly significant negative coefficient in all the three alternate cases in Table 9.2. This is consistent with our prior about the adverse effect of exit options on cooperative behavior discussed in the preceding section. The variable OTHIRRIG, indicating access to alternative sources of irrigation outside the ayacut for the farmers, has an appropriately negative sign, but it is not significant. The quality of maintenance is understandably highly positively correlated with the existence of a guard, whose enforcement of water distribution rules increases the benefits of field channel maintenance.213 Since the appointment of a guard is itself dependent on some of the factors that enter into the determination of cooperative behavior, we use here only the predicted value, PGUARD, in this second-stage regression.214 In the first-stage logit analysis for GUARD, reported in Table 9.5, we use as instruments the variables CANAL and TRADLORG: in canal areas there is much less use of a guard, whereas in ayacuts with a traditional organization appointing a guard is quite common.215 The only other variable that is significant (at the 10 percent level) in the first stage regression for GUARD is LINKAGE: it seems that better urban-linked villages use more guards. But Table 9.2 suggests that, controlling for the positive effect on PGUARD, LINKAGE has a significant negative effect on maintenance of field channels. In Table 9.2 we also see the positive (but not always significant) effect of the proportional cost-sharing rule on maintenance: cost-sharing

213

In some alternative regressions (not reported here) we have found that the interactive effect of PGUARD and NUMUSERS has a significant negative coefficient, indicating that the effectiveness of the guard system in encouraging cooperation decreases as the number of users increases.

214

Of course, a two-stage regression is somewhat problematic when the second stage is a non-linear logit model. One can use OLS as a check to confirm the results. But the problem with using OLS is that with dependent variables that are grouped, the error terms cannot be assumed to be normally distributed. The assumption is even worse, and the results more biased, if OLS is used when the dependent variable takes on only 0–1 values, as in the case of H2OCOOP and VIOLH2O. In the case of DISTFDCH we have used OLS as a check, and the results are largely confirmed (except for the variable NUMUSERS, as we have noted in note 19).

215

We presume this is determined more by historical-conventional factors. In fact in areas where the government takes the decisions the enforcement of rules is usually by government officials or police, not by guards. It is also the case that in canal areas the water available to the ayacut is from a flow (rather than from a storage reservoir as in the case of tanks), and guarding against water theft may therefore be more serious in the case of tanks than in the case of canals. In any case the TRADLORG variable is not statistically significant as a direct determinant in the cooperation regressions, and we can therefore legitimately use it as an instrumental variable.

COOPERATION ON IRRIGATION IN SOUTH INDIA

261

proportional to landholding may be perceived as more fair than, for example, cases where all farmers have to bear the same cost even though the larger farmers get more of the benefit, or where there is no rule.216 Since the SHRPROP rule may itself be dependent on factors that enter into the determination of cooperative behavior, we use in Table 9.2 only the predicted value, PSHRPROP. The first-stage logit analysis for SHRPROP is reported in Table 9.5. SHRPROP is positively and significantly related to the number of ayacut members, inequality of landholding among them, canal areas, ayacuts where the irrigation organization is old, and where formal water rights is more prevalent.217 Since GINI has a positive effect on SHRPROP (possibly indicating social pressure for a redistributive adjustment of the costsharing rule to take account of wealth disparities) and PSHRPROP has a somewhat significant positive effect on DISTFDCH, we can say that the effect of inequality of land distribution on the quality of maintenance on field channels is twofold: on the one hand, the direct effect is negative, but the indirect effect, working through the costsharing rule, is positive.218 Similarly the effect of NUMUSERS is twofold: the direct effect is negative (though not always significant), but the indirect effect, working through the cost-sharing rule, is positive. Now we turn to the dependent variable in Table 9.3, H2OCOOP, i.e. where there has been no intra-village conflict over water over the previous five years. GINI is again significantly negative, but GINISQU is significantly positive, indicating the kind of U-shaped relationship between inequality and cooperative behavior we have discussed in the preceding section. The coefficient for CASTE75 is positive in both cases of Table 9.3, and significant in one, confirming that social homogeneity (in the form of 75 percent or more of the farmers belonging to the same caste group) is conducive to cooperation. The urban connection variable, LINKAGE, is, as before, negative in its effect on cooperation, though significant in one of the two cases. Also as before, PGUARD has a positive effect, but it is significant in one of the two cases. The government involvement variable, PWDDECID, did not turn out to be significant in this set of regressions, but again looking at the detailed data, as in the case of the field channel maintenance regressions, in half of the 12 units where PWD takes all decisions, there was cooperation in water allocation and in the other half there was conflict (and the former associated with more equality in land distribution and fewer private pumpsets and open wells than in the latter).219

216

Only about 1 in 5 villages has the proportional cost-sharing rule in our dataset.

217

In one-third of the villages in the dataset formal water rights, as opposed to customary rights, exist. Again these are likely to be determined by historical-conventional factors, and we can therefore use this variable as a legitimate instrument.

218

In some alternative regressions (not reported here) we have found that the interactive effect of SHRPROP and GINI is positive in the DISTFDCH regression, suggesting that the higher the level of inequality the more positive is the influence of proportional cost-sharing on the quality of maintenance.

219

Exactly the same pattern is observed in the other 12 canal-based units.

262

PRANAB BARDHAN

In Table 9.4 the dependent variable is the opposite of cooperation; it is VIOLH2O, i.e. where water allocation rules are frequently violated by at least one group. (Both for the ayacut as a whole and for the sampled farmers our data definitely suggest that the rule violations are more often by the better-off farmers.)220 Here the signs of the coefficients for GINI, GINI-SQU, CASTE75, TAILEND, and OTHIRRIG are as expected from the earlier regressions, but they are not significant. NUMUSERS is positive and highly significant, indicating the difficulty of preserving cooperative behavior in large groups. AVGACCES, the duration of access to water, is negatively, and in one of the two cases highly significantly, related to violations, indicating that water scarcity encourages violations. As before, the urban connection variable, LINKAGE, has a negative impact on cooperation, and PGUARD has a positive effect. In one of the two regressions, we have used PMARKET as an alternate outside connection variable to LINKAGE. MARKET is a measure of the extent to which farmers are market-oriented (particularly in terms of sales of produce). Since this may be endogenously determined, we have used PMARKET, the predicted value of MARKET, in the second-stage regression. The first-stage regression is reported in Table 9.5. The only two variables that are significant (with a positive coefficient) in this first-stage regression are CANAL and WMARKET (indicating cases where there is a market for well water in the village),221 the two variables chosen as instruments. From the second-stage regression we can see that like LINKAGE, PMARKET increases the probability of violation of water allocation rules. TOPOGRPH (the case where the topographical nature of the ayacut precludes equal access to water), TAILEND, and OTHIRRIG have coefficients of expected signs, but they are generally not significant. From the list of independent variables we had to drop PWDDECID and ELITERUL, because in both cases their values perfectly predicted the value of the dependent variable. This means in all the villages where PWD decides on water allocation and distribution, frequent rule violations are reported:222 this

220

For example, in 29 out of 48 villages rule violations are mainly by the rich farmers/upper castes, as reported in the ayacut-level survey. Also at the level of the sampled farmers, the average farmer perception is that the rich (in terms of landholding, the top four out of the ten sampled farmers in the village) comply with rules significantly less than the poor (the bottom four out of the ten sampled farmers), particularly in water allocation and financial cost-sharing.

221

About 44 percent of the villages in our dataset have a market for well water.

222

We also have evidence in our dataset that the farmers contribute to the village collective fund much less when the government is involved, presumably because the government is then assumed to be the financier of last resort. For evidence of government involvement weakening cooperative effort in watercourses in Pakistan, see Byrnes (1992).

COOPERATION ON IRRIGATION IN SOUTH INDIA

263

may be because the rules are typically rigid and insensitive to local needs, farmers are less normatively committed, officials are bribed to look the other way, and so on. In contrast, in the villages where the village elite, rather than government officials, crafts the rules, there is no violation of rules reported. Since, as we have noted before, the betteroff are the more frequent violators of rules, they tend not to violate rules crafted by themselves. In all of the 12 PWD-run units there are rotational water allocation rules by which the farmers are allotted a certain number of hours of water access per acre or are allowed access to water only in alternate weeks. In all of these 12 units these rotational rules are frequently violated and the rich farmers particularly appropriate more water than is their due. And yet in half of these units (particularly where the inequality in land distribution among the farmers is low), as we have seen before, field channels are well-maintained and there was no incidence of water conflicts within the village in the last five years. This means the inflexible rules of the government (enforced by corruptible agents) are frequently violated without necessarily damaging intra-village cooperation, suggesting again that when rules do not enjoy the backing of community norms, rule obedience is not necessarily an indicator of cooperation among farmers. Summing up over the seven regressions in Tables 9.2, 9.3, and 9.4, we can say that in our dataset cooperative behavior in an irrigation community is by and large significantly related (negatively) to inequality of landholding, to urban or market connections, and (positively) to duration of access to water, monitoring by guards, and in some cases to social homogeneity, small group size, proportional cost-sharing rule, and collective adversarial relation with other villages over water. PWD involvement in water allocation and maintenance can have a positive effect on field channel maintenance (if farmers do not have much access to their own private sources of water like pumpsets and open wells); but such involvement is in more water-scarce areas, and too much water-scarcity may not be conducive to cooperation. Government involvement also encourages violations of specific inflexible rules by the farmers and means fewer cases of ayacut communities appointing their own monitoring guards, which adversely affects cooperation. When the rules are crafted by the village elite, the latter violates the water allocation rules less; otherwise the elite is the more frequent violator of rules. Sometimes when rather inflexible rules are made and enforced by the government, their violations are not necessarily inconsistent with cooperative behavior among farmers themselves. When an average farmer believes that the water rules have been crafted jointly (as opposed to by the elite or by the government), we find that he is more likely to have positive comments about the water allocation system and about rule compliance by other farmers.

264

PRANAB BARDHAN

References Anderson, P. (1974) Lineages of the Absolutist State, NLB, London. Baland, J. M. and Platteau, J. P. (1997) ‘Wealth Inequality and Efficiency in the Commons: The Unregulated Case’, Oxford Economic Papers, 49. —— and—— (1998) ‘Wealth Inequality and Efficiency in the Commons: The Regulated Case’, Oxford Economic Papers, 50. Bardhan, P. (1984) Land, Labor, and Rural Poverty, Columbia University Press, New York. Byrnes, K. J. (1992) ‘Water Users Associations in World Bank-Assisted Irrigation Projects in Pakistan’, Technical Paper no. 173, The World Bank. Dayton-Johnson, J. (1998) ‘Rules and Cooperation on the Local Commons: Theory with Evidence from Mexico,’ Ph. D. dissertation in University of California at Berkeley. —— and Bardhan, P. (1998) ‘Inequality and Conservation on the Local Commons: A Theoretical Exercise’, CIDER Working Paper no. 096-071, University of California at Berkeley. Easter, K. W. and Palanisami, K. (1986) ‘Tank Irrigation in India and Thailand: An Example of Common Property Resource Management’, Department of Agricultural and Applied Economics Staff Paper, University of Minnesota, August. Hunt, R. and Hunt, E. (1976) ‘Canal Irrigation and Local Social Organization’, Current Anthropology, 17, September. Jayaraman, T. K. (1981) ‘Farmers’ Organizations in Surface Irrigation Projects: Two Empirical Studies from Gujarat', Economic and Political Weekly, 16(39). Olson, M. (1965) The Logic of Collective Action: Public Goods and the Theory of Groups, Harvard University Press, Cambridge, MA. Ostrom, E. (1990) Governing the Commons: The Evolution of Institutions for Collective Action, Cambridge University Press, Cambridge. Seabright, P. (1997) ‘Is Cooperation Habit-Forming?’, in Dasgupta, P. and Maler, K. G. (eds.), The Environment and Emerging Development Issues, Clarendon Press, Oxford. Sengupta, N. (1991) Managing Common Property: Irrigation in India and the Philippines, Sage Publications, New Delhi. Vaidyanathan, A. (1984) ‘Water Control Institutions and Agriculture: A Comparative Perspective’, Indian Economic Review, 20(1). —— (1994) ‘Transferring Irrigation Management to Farmers’, Economic and Political Weekly, 29(47). Wade, R. (1987) Village Republics: Economic Conditions for Collective Action in South India, Cambridge University Press, Cambridge. Weissing, F. and Ostrom, E. (1990) ‘Irrigation Institutions and the Games Irrigators Play: Rule Enforcement without Guards’, in Selten, R. (ed.), Game Equilibrium Models, ii, Springer-Verlag, Berlin.

10 State, Community and Market in the Deterioration of a National Irrigation System in the Philippines Masao Kikuchi, Masako Fujita, and Yujiro Hayami This chapter aims to show the possibility of a mismatch between state and local communities resulting in a major deterioration in the conservation of local commons, based on a case study of a national irrigation system in the Philippines. While we share a new paradigm which recognizes the real potential of rural communities to manage irrigation systems, we wish to highlight the danger in hastily handing over systems from state to community management without due consideration of the limits of local communities' organizational capacity. Our analysis will show the need for ‘community-enhancing’ policies for the efficient management of relatively large-scale local commons.

10.1 The Roots of the Problem In 1968, Garrett Hardin advanced his classic thesis of the ‘tragedy of the commons’ highlighting the danger of overexploitation and exhaustion of common-pool or common-property resources, such as forestry, in-shore fisheries, and irrigation, which share with public goods the attribute of non-excludability but not the attribute of non-rivalness. His policy prescription was the conversion of commons under communal ownership either to state management or to private management with the assignments of well-defined property rights. However, in the quarter century since then, case studies have been accumulated to illustrate the ability of rural communities to conserve common-pool resources adequately, while the disadvantages of state management have increasingly been recognized—the inefficiency of distant bureaucracy in their use of local information and initiative, the high cost of privatization, and its inequitable distributional consequences (Wade 1988; Feeny et al.1990; Ostrom 1990; Ostrom and Gardner 1993; Ostrom and Walker 1994; Bardhan 1993; Baland and Platteau 1996). An emerging new paradigm is to advocate the delegation of the management of commons to the initiatives of local communities as much as possible. This paradigm has formed the basis for a pervasive mode of ‘handing over’ the management of surface-flow gravity irrigation systems from state agencies to groups of local beneficiaries—commonly called ‘irrigators’

266

MASAO KIKUCHI, MASAKO FUJITA, AND YUJIRO HAYAMI

associations'—since the 1980s (Small and Carruthers 1991). A more fundamental cause for this move has been the financial crisis of state irrigation administration in developing economies. In the 1970s following the so-called ‘World Food Crisis’ of 1973–74, the construction of gravity irrigation systems boomed on the basis of large inflows of foreign aid. However, as international food markets turned into a low-price regime after the collapse of the commodity boom in 1981, the aid flow to agriculture was reduced and the maintenance of irrigation systems which had expanded in the boom period became a heavy burden on government budgets. The handing-over of state agencies' operation and maintenance as well as irrigation fee collection to irrigators' associations were envisioned as an efficient way to save government resources—for which the new paradigm provided a convenient rationale. Yet, the application of the new paradigm has usually been structured as designing incentive schemes for organizing irrigators' associations, such as payment to community work for maintenance activities and giving a share of collected irrigation fees to the associations, while little support has been provided for institutional developments. Such a design has been based, at least in part, on an overly optimistic interpretation of local communities' response to economic incentives. It has not duly recognized the limit of the communities to organize collective actions by themselves for the maintenance of relatively large irrigation systems encompassing many villages; this fact has underlain the deterioration of the national irrigation system under our present investigation, to be illustrated below, as well as many other systems in the Philippines and other developing economies.

10.2 Outline of the System The irrigation system under our investigation (henceforth referred to as ‘the System’) is designed to serve about 4,000 hectares of paddy fields on the east coast of Laguna de Bay, the largest lake in the Philippines. Its construction was begun in 1953 and completed in 1958. Until then, farmers in this area had mainly practiced a single rain-fed cropping of rice. The commencement of the System's service resulted in nearly complete conversion to irrigated double rice cropping. This provided a basis for a subsequent diffusion of modern high-yielding varieties and heavy application of fertilizers—the so-called ‘Green Revolution.’ The System is based on surface flows of water diverted from the Sta Cruz River which originates in Mount Banahao gathering several tributaries along the northeastern side of the mountain (Fig. 10.1). Irrigation water diverted by a dam is sent to fields in the service area through a network of canals and laterals ending in the Laguna de Bay. The network consists of a main canal, five laterals (A to E) and 16 sub-laterals (Fig. 10.2).

THE DETERIORATION OF A NATIONAL IRRIGATION SYSTEM

267

Fig. 10.1 Location Of the NIA Irrigation System

There are about 2,400 farmers in the System encompassing 37 villages in 5 municipalities. Within the System, most cultivable lands are paddy fields, except for small elevated spots of coconut groves under which houses are clustered. Relatively more coconut groves and less paddy fields are found in the upper rather than the lower streams of the System. In fact, the main canal runs all the way through a coconut belt with scattered niches of paddy fields. Five main laterals flow through the coconut belt for a few kilometers after the head gates on the main canal, which demarcates the western edge of a large, contiguous coconut area covering the foot of Mount Banahao and other small mountains in the region. In the coconut area, tree crops such as banana, lanzones, papaya, and coffee are inter-planted. Some farmers in the upper reaches of the main laterals use the canal water for irrigating these tree crops. As a gravity system, the System is typical among those found in tropical monsoon Asia. The supply of water is based purely on the diversion of a natural river flow with no reservoir and no pumping system being installed. Canal and lateral embankments are formed of mud, except for

268

MASAO KIKUCHI, MASAKO FUJITA, AND YUJIRO HAYAMI

Fig. 10.2 Canal Layout Of the NIA Irrigation System

concrete structures to support water gates and water-gauging facilities at the junctures of laterals and the main canal. Needless to say, there is no device for metering the take-in of water to individual farms or fields. In fact, most fields are irrigated by direct flows of water from upper to lower paddy fields without being channeled through laterals and farm ditches. As the result, the total length of canal serving 4,000 hectares was as little as 80 kilometers, implying an average canal density of only about 20 meters per hectare. The agency that administers this System is the National Irrigation Administration (NIA). NIA is a ‘Government-Owned and Controlled Corporation’ established in 1964 by Republic Act 361. NIA's hierarchy from its headquarters to field level is separated into two divisions: one for the construction of new systems and the major rehabilitation of existing systems and another for the operation and maintenance of existing systems (Fig. 10.3). Each construction or major rehabilitation project is managed by a ‘project office’ established during the project period. The operation and maintenance of national irrigation systems are handled by ‘irrigation systems offices,’ and NIA's assistance to communal systems is administered by ‘provincial offices,’ both under the super-vision of ‘regional offices.’ The systems office managing this System also manages four other national systems and reports to the regional office covering the fourth region of NIA. In terms of its service area, the size of this System is close to the average of the 171 national systems under NIA's auspices.

THE DETERIORATION OF A NATIONAL IRRIGATION SYSTEM

269

Fig. 10.3 Organization Of the (NIA) As Of December 1995

Note: Numbers of offices and systems shown in parentheses. As such, water supplied from the System is endowed with the attributes of ‘local commons’ which are available for use by the people in a local community but are exhaustible if used in excess of their reproductive capacity. Since water resources for the System are limited to the capacity of the diversion from the river flow, it is exhaustible through overexploitation. Thus, an abuse by farmers in upper streams has negative externality for downstream farmers. Yet, no incentive mechanism exists to prevent an individual user from abusing water, because water take-in to his field cannot be metered and, hence, is not chargeable in proportion to his consumption. It is costly to organize action to save water (by such means as rotating water supply among users) or to augment water supply (by such means as removing silts and cutting weeds in canals) because everyone is tempted to be a free rider on others' conservation activities. This problem is exacerbated in the case of gravity irrigation systems by an

270

MASAO KIKUCHI, MASAKO FUJITA, AND YUJIRO HAYAMI

asymmetry in the distribution of the means and benefits of conservation activities; typically, farmers at the head end of the system can reduce abuse in their water use but receive no benefit from it, whereas those at the tailend receive benefits from the head-enders' water saving but are not in a position to reciprocate to them (Ostrom and Walker 1994).223 Indeed, deterioration in the quality of gravity irrigation systems is universal in developing economies (Chambers 1988).

10.3 Deterioration in the System's Performance Our investigation into this System's performance was triggered by the observation of a serious deterioration in the irrigation service to a village in which we have conducted periodic surveys over the past three decade (Hayami 1978; Hayami and Kikuchi 1981, ch. 5; 2000). This village is located at the tailend of Lateral C. Despite its disadvantageous position within the System, this village had experienced no major problem in the service of NIA until the mid-1980s. In the village survey conducted in 1987, we witnessed a slight symptom of deterioration. At the time of the 1995 survey, the deterioration had become very evident; farmers complained of NIA's irrigation reaching only about 20 percent of their fields; many of them had purchased irrigation pumps, but several fields were left unplanted. In the subsequent two years we tried to identify factors underlying the System's deterioration. The first attempt made was to collect data by means of interviews with the headmen of all the 37 villages (barangay captains) located within the System's service area. Results of the survey revealed the fact that the deterioration was not unique to our study village but was common in the lower reaches of major laterals. As shown in Fig. 10.2, a national highway passes through roughly the middle of the System's service area. According to the headmen interviewed, most villages located above the highway did not suffer water shortages even after the System's deterioration. The villagers with assured water supply are classified as being in the upper reach, whereas those located below and who suffered from water shortage are classified as being in the lower reach of the system.224 The aggregation of village headmen's data indicate that the service area in the upper reach continued to be well irrigated in both the wet and dry

223

The tendency for the head-enders to abuse water is enhanced by the nature of rice plants which are relatively tolerant to excess water application but intolerant to drought.

224

This classification does not coincide exactly with the geographical demarcation by the highway. In the service area of Lateral E, for example, three villages are located above the highway, of which only the one closest to the main canal did not experience water shortage and, hence, was classified into the upper reach, while the remaining two villages are classified in the lower reach.

THE DETERIORATION OF A NATIONAL IRRIGATION SYSTEM

271

Fig. 10.4 Irrigation Ratio Per Year For the Entire System, Upper Reaches, and the Lower Reaches Of Lateral C

seasons from the 1970s throughout the 1990s, maintaining the irrigation ratio (defined as the ratio of total area irrigated for wet and dry seasons in a year to the service area) at a level between 180 and 190 percent (Fig. 10.4). In contrast, the ratio in the lower reach was only about 150 percent as early as the 1970–85 period, reflecting a much less sufficient water supply to downstream than to upstream farms. Moreover, the ratio dropped sharply from 153 percent in 1985 to 116 percent in 1996.

Decreased Public Support What factors did underlie this significant decline in water supply to the lower reaches? There was no environmental reason to be suspected. Both the total amount and the seasonal pattern of rainfall remained the same, though subject to large annual fluctuations. Examination of satellite images of the watershed area in Mt Banahao revealed no appreciable depletion of forests between February 27, 1976 and February 11, 1992. It is, therefore, very likely that degradation in the System's performance was brought about mainly by a decline in operation and maintenance (O&M) activities. Indeed, NIA's expenditure for O&M in this System (excluding staff salaries) declined by almost one half from the early 1980s to the late 1980s (Fig. 10.5). The number of NIA personnel engaged in O&M was also reduced sharply. Eight water masters in charge of water

272

MASAO KIKUCHI, MASAKO FUJITA, AND YUJIRO HAYAMI

Fig. 10.5 Operation and Maintenance Expenditures Per Hectare In the NIA System Excluding NIA Staff Salaries, Deflated By GDP Deflator (1995 = 100), Three-Year Moving Averages, 1980–1993

allocation and scheduling and 20 ditch tenders for manual operations were allocated to this System before the mid1980s. In 1995, only 2 water masters and 11 ditch tenders were employed. Correspondingly, the physical structures of the System were seriously dilapidated. As of 1995, the spill gate at the dam was out of order such that it did not shut properly, leaving a gap of about 10 centimeters from which water, which would normally be diverted to the main canal, was running away. The silting at the dam site apparently made the water-holding capacity of the dam less than designed. Of the five head gates diverting water from the main canal to laterals, four had some defects with parts missing or damaged. Many turn-out gates along the laterals were in a similar condition. Many sections of the main canal and the laterals were in obvious need of desilting and weed cutting. These observations leave little room to doubt the hypothesis that insufficient O&M activities since the early 1980s or even earlier had accumulated to produce degradation in the performance of the System. A lowered intensity of O&M activities was not unique to this System but was pervasive in irrigation systems under NIA's auspices. It is only one symptom of reduced public support for agriculture in the Philippines as well as in developed economies throughout the world during the last decade and half of the twentieth century. Public investment in the infrastructure for agricultural production, such as irrigation and research in

THE DETERIORATION OF A NATIONAL IRRIGATION SYSTEM

273

Fig. 10.6 Rice Price and NIA Project Expenditures, 1965–1995

Sources: Rice prices from IRRI (1995), deflators from ADB (1998), and project expenditures from the NIA Central Office. agricultural technology, were greatly intensified especially for rice and wheat in the years of high food prices following the ‘World Food Crisis’ of 1973–74. However, in the 1980s when these investments were transformed into high productivity, as epitomized by the success of the Green Revolution, world food markets entered a low price regime; this is clearly illustrated in the movements of both the export price of Thai rice (as a

274

MASAO KIKUCHI, MASAKO FUJITA, AND YUJIRO HAYAMI

yardstick of international markets) and the domestic price of rice in the Philippines (upper diagram in Fig. 10.6). The situation was exacerbated in the Philippines due to prolonged political instability from the 1980s to the early 1990s following the downfall of the Marcos administration. NIA's O&M budget was doubly affected by the decline in rice prices. It has been stipulated since 1974 that NIA should be autonomous in financing its current expenditures for the O&M of existing systems from its revenue, which consists mainly of irrigation fees collected from farmers.225 On the other hand, capital expenditures for the construction of new systems and major rehabilitation of existing systems were supposed to be financed by government investment, mainly based on foreign aid. The irrigation fees are usually paid in paddy—the rates per hectare being 100 kg in the wet season and 150 kg in the dry season. Thus, the NIA's revenue for financing O&M decreased sharply corresponding to the decline in the price of rice during the 1980s. Another significant revenue to NIA's current account is a 5 percent set-aside from the administrative overheads of the ‘project fund’ originally allocated for the construction of new systems. In addition, part of the project fund, which was specifically designated for the use of O&M projects (mostly foreign-funded), is usually allocated to existing systems, not on a project basis but in proportion to their service areas, to pay for the costs of small repairs. As both the international aid agencies and the national government became reluctant to assist irrigation projects under apparent oversupply of rice, the set-aside for administrative overheads from the new construction funds also shrank. The relationship between the decline in the project fund and the decline in rice prices in the 1980s should be evident from Fig. 10.6. A deterioration in the performance of the System due to insufficient O&M activities was thus a reflection of the financial crisis of the NIA which emerged in the 1980s as a result of the very success of the Green Revolution.

Differential Performances Across Laterals Deterioration in the service of irrigation was, however, not uniform within the System but varied greatly across laterals, as observed in the changes in irrigation ratio during the last 10 years for which NIA's data are available

225

By the revision in 1974 of Republic Act 361, the NIA was requested to be autonomous in financing its current expenditure with the limit of the government's subsidy up to 6 million pesos. By the 1980 revision, the government's subsidy was removed but the NIA was allowed to receive 5 percent of the project fund for new construction and major rehabilitation as a set aside for administrative overhead costs. In addition to irrigation fees, the NIA raise current revenues from such activities as rental of bulldozers and other machines. See Fujita (1998).

THE DETERIORATION OF A NATIONAL IRRIGATION SYSTEM

275

Fig. 10.7 Changes In the Ratio Of Irrigated Area Per Year To Service Area In the NIA System By Lateral, Three-Year Moving Averages, 1985–1994

(1985–94, Fig. 10.7). Major decreases in the irrigation ratio occurred in Laterals C and D, showing that the System's performance deteriorated first in Lateral D during the 1980s, followed by Lateral C during the 1990s. In contrast, a significant improvement was recorded in Lateral E. The irrigation ratio also increased slightly in Lateral B, especially in the 1990s when Lateral C's performance deteriorated rapidly. The ratio in Lateral A was low but remained relatively stable. The differential performances among laterals were, to some extent, based on differences in their physical and geographical conditions. For example, the lower rate of irrigation in Lateral A than in Lateral B is explained by the fact that the service area of the former is extended widely in the lower reach of the system below the highway, whereas that of the latter is limited to the area above the highway. However no appreciable degradation occurred in Laterals A and B. It is understandable that the performances of Laterals C and D, located below the main canal, deteriorated relative to A and B, corresponding to a contraction in the water supply which resulted from insufficient O&M activities. It is a major puzzle, however, that the performance of Lateral E improved despite its location at the tailend of the main canal.

276

MASAO KIKUCHI, MASAKO FUJITA, AND YUJIRO HAYAMI

10.4 Goals and Consequences of Organizing Irrigators' Associations When public support for irrigation began to decrease significantly in the 1980s, NIA tried to mobilize local communities' contributions to O&M activities. For a given amount of water diverted from the river, water supply to crops could be increased if laterals and ditches were cleaned by farmers' collective work. Efficiency in the use of scarce water may be enhanced if a rational water rotation could be agreed upon and observed by farmers. The differential performances of irrigation service among laterals seem to reflect both success and failure in the organization of local initiatives for these activities. In this section the patterns and sources of the differential institutional performances will be examined through a comparison of Laterals C and E. When NIA was pressed strongly to establish its financial autonomy in the 1980s under waning public support, farmers' participation into O&M activities was recognized as a necessary condition to prevent the Systems' performance from deteriorating. The NIA therefore tried to organize farmer-beneficiaries into an irrigators' association (IA). It was also believed that it would be more efficient to delegate the collection of irrigation fees to IAs. NIA's institutional development strategy consisted of two steps: first, to organize farmers into a turn-out service area group (TSA group) and second, to federate TSA groups along each lateral into an IA. Both TSA groups and IAs in national systems are organized according to the guidance of the NIA, each under the management of five elected officers including a president, a vice president, a secretary, a treasurer and an auditor. Contracts between the NIA and IAs are categorized into three types in terms of O&M responsibility turned over from the NIA—Type 1 if an IA agrees to take responsibility for cleaning canals; Type 2 if an IA takes charge of collecting irrigation fees; and Type 3 if an IA agrees to take all the operation and maintenance responsibilities on the condition that the IA establishes property rights on the canal structures and amortizes their construction costs. In order to mobilize local participation and initiatives, the NIA designed various incentive schemes for IAs. First, if an IA agrees to take responsibility for cleaning the canals within its territory, the NIA pays 400 pesos per month (which is roughly equivalent to 3 person-days of farm work) per kilometer of canal length. For repair work which costs more than 500 pesos, the NIA usually grants the contract to an IA. For a smaller repair costing less than 500 pesos, an IA is supposed to mobilize members' labor without pay but materials and machines are supplied by the NIA. When an IA enters a Type 2 contract, the IA is entitled to receive 2 percent of collected fees if it is able to collect more than 50 percent of total invoices (and to receive 15 percent if the collection rate exceeds 90 percent). Under the promotion of NIA, four IAs, one each for Laterals A, B and E, and another for Laterals C and D combined were created, covering

THE DETERIORATION OF A NATIONAL IRRIGATION SYSTEM

277

about 70 percent of the total service area. These IAs were registered at the Security Exchange Commission, one in 1987 and three in 1991, to establish legal status. Among the four IAs, however, those of Laterals B and E continued to be active, the former entered into a Type 1 contract and the latter entered into both Type 1 and Type 2 contracts. Meanwhile, the IAs of Laterals A, C and D stopped functioning only a few years after formation. There is little doubt that the success and the failure in the organization of IAs resulted in the differential performances across laterals as observed in Fig. 10.7. The successful IA in Lateral E was established in 1983, federating six TSA groups in the area above the highway. Later, one TSA group below the highway joined the IA in 1993. While canal cleaning, small repairs, and irrigation fee collection are the important activities of this IA, an even more important function is to establish a water rotation schedule and to monitor its implementation. It holds a regular monthly meeting at which IA officials and TSA leaders discuss prospective water needs and reach agreement on a rotation schedule. The schedule is drawn up in consultation with NIA's field staff so that it is consistent with the system-wide rotation. Leaders take turns in monitoring for 24 hours a day during a crop season to prevent violation of the rotation occurring. Though specific penalties on violations are not stipulated, it is said that the leaders' warning and persuasion are sufficient to suppress illegal water take-out. As the result, conflicts and disputes on the use of water, which were common before the formation of the IA, were largely eliminated in the service area of Lateral E. As farmers recognized the benefit of the IA, their cooperation and participation were enhanced, resulting in as increased rate of fee collection as well as additional contributions to O&M activities proportional to farm sizes. The increased revenue to the IA provided a basis for further improving the quality of irrigation to produce greater satisfaction of members. Thus, a virtuous circle operated between institutional development and irrigation service in Lateral E. The case of Lateral C represents a sharp contrast to that of Lateral E. Similar to the case of Lateral E, the NIA tried to organize farmers in Lateral C into an IA in the early 1980s. This attempt failed, presumably because water supply along Lateral C was relatively abundant so that farmers did not feel the need to organize themselves for water allocation. Later, in the water shortage that emerged in 1990–91, the NIA was able to establish an IA for Laterals C and D combined in consideration of the overlapping service areas between the two laterals, because some of service areas from the two laterals are inter-meshed. In organizing and activating farmers' participation, the NIA appointed eight farmers in the area as ‘Irrigators Organizers’ with an honorarium of 500 pesos per person for a half year. The NIA's water masters and the Irrigators Organizers called farmers together for meetings in each of

278

MASAO KIKUCHI, MASAKO FUJITA, AND YUJIRO HAYAMI

16 turn-out service areas in Lateral C and 14 in Lateral D, explaining the benefits of an IA and urging their active participation. Nevertheless, this IA was not able to advance as far as taking over from the NIA responsibility for cleaning laterals and ditches (Type 1 contract). The rotation of irrigation was also attempted under the guidance of the NIA, but totally failed without being able to reduce rampant illegal water take-outs. The NIA's officials blamed low morale and a lack of cooperative spirit among the people in Laterals C and D, as compared with those of Lateral E. More concretely, the IA's president (who was formerly elected by TSA group leaders but de facto appointed by the NIA) complained about the lack of participation by local leaders, especially those in the lower reaches, at meetings, despite his repeated calls and solicitations. However, the perception of people in the lower reaches was different. The president of the TSA group in the tailend village in Lateral C explained that leaders in his group were eager to participate when the IA was created. Their enthusiasm waned, however, as the IA was incapable of enforcing a water rotation schedule. They complained to both the NIA and the IA about rampant illegal water take-outs by upstream farmers resulting in sheer water shortage in the lower reaches. In fact, they even went as far as padlocking water gates in the upper reaches to prevent upstream farmers taking out water during the turn of downstream users. However, the chains and locks were soon destroyed. Yet, despite knowing who carried out these illegal actions, neither the NIA nor the IA was able to stop them. Dismayed by such problems, leaders in the tailend village (who were relatively large farmers) purchased pumps to cope with the deteriorating water supply from the NIA, rather than wasting their time and energy in activating the IA. Thus, local participation and irrigation service in Lateral C deteriorated in a vicious circle. Why then was Lateral E able to escape such a vicious circle and instead to ride on a virtuous circle? Based on a field study of an irrigation system in India, Robert Wade (1988) advanced the hypothesis that villages at the tailend of the system tend to better organize collective action for economizing on the use of water, because of higher rates of return to such actions under stronger constraints on water supply. This hypothesis seems consistent with the efficient IA organization in Lateral E which is the tailend lateral along the main canal (Fig. 10.2). However, it does not explain why an IA was successfully organized in upper-stream Lateral B, even if it was not as active as the one in Lateral E. Also, it does not explain why farmers in Lateral E successfully organized themselves when they faced a water shortage in the early 1980s and why a similar attempt failed in Lateral C when farmers faced the same problem in the 1990s. In Lateral E, also, conflicts and disputes among farmers on the use of water were common and often fierce, culminating sometimes in bloodshed and even a murder case before the formation of the IA. Why was Lateral E but not

279

THE DETERIORATION OF A NATIONAL IRRIGATION SYSTEM

Lateral C able to make the shift from the negative-sum to the positive-sum cooperative game? We have examined several factors which might explain the difference between the areas with active IAs (Laterals B and E) and the areas with inactive IAs (Laterals A, C, and D), such as the depth of the underground water table as a measure of the cost of using irrigation pumps, and the distribution of farm sizes and the ratio of non-farm households as measures of homogeneity of communities. Differences in those variables are, however, not statistically significant at conventional levels (Fujita 1998). The factor found to be significant was the size of the IAs. As shown in Table 10.1, active IAs for both Laterals B and E are much smaller than IA's for A and C–D—less than a third on average in terms of the length of lateral (including sub-laterals) and service area as well as the numbers of farms and villages in the service area. A is distinctively larger than the others, but since C and D are combined to form one IA, both the territory and the member size of this IA are not quite so different from those of A's. The structures of Laterals B and E are simpler—having only one sublateral—than those of A and C which are attached with nine and three sub-laterals, respectively. Technically, the larger and the more complicated the system, the more difficult it is to monitor whether the system is operated and maintained properly, including identification of where and by whom illegal water take-outs are practiced. More serious should be the greater difficulty of groups which have a larger membership organizing collective actions for the supply of public goods, since it is easier and more profitable for members in larger groups to become free riders (Olson 1965). In this respect, the IAs of A and C–D which include much larger numbers of irrigator-farmers Table 10.1 Length of Laterals, Service Areas, and Number of Sub-Laterals, Villages and Farms in the Five Laterals of the NIA System, 1996 Length of lateral Service area (ha) Number of sub- Number of vil(km) laterals lages Active IA: Lateral B Lateral Ea Average Inactive IA: Lateral A Lateral C Lateral D Average a

6.0 4.9 5.5

240 227 234

1 1 1

29.3 12.8 8.0 16.7

1,602 588 336 842

9 3 1 4

Data at the formation of IA in 1983, excluding the lower reach extended since 1993.

2 3 2.5 17 7 5 10

Number of farms

110 300 205 1,359 309 327 665

280

MASAO KIKUCHI, MASAKO FUJITA, AND YUJIRO HAYAMI

than those of B and E should have found it more difficult to reach a consensus on the rules of water use and public works as well as enforcing the rules for their systems' maintenance by members. It is likely that even a more critical determinant for the effective organization of an IA than the size of its membership is the number of villages in its territory. In the rural sector of developing economies, a village is the basic unit of life and work. It is a community characterized by strong social interactions among people, tied closely by both blood and locational affinity. The community is ruled mainly by traditional norms and conventions, which are enforced by villagers' moral obligations to neighbors as well as their fear of social opprobrium and ostracism which may result from violations of the rules. This community mechanism of rule enforcement is considered to be effective in conserving local commons.226 However, its effectiveness tends to be limited within a small community, typically one village or one tribe. While the solidarity of a village community may be effective in restraining its members from exhibiting opportunistic and free riding behaviors among each other, it often promotes their exercising opportunism against outsiders (Hayami 1997: 266–8). Indeed, it is not uncommon to observe that local commons, such as forests and communal irrigation systems, under the auspices of one village are relatively well maintained, whereas those encompassing several villages are seriously depleted to a condition akin to the ‘tragedy of the commons.’ This difficulty of organizing people across different communities should multiply as the number of villages increases, even if a mutually beneficial scheme of cooperation can be envisaged. There seems to be little doubt that the organization of IAs in Laterals A and C–D was handicapped by having a large number of villages within each of their territories, when compared with Laterals B and E. The differential performances across laterals have been explained so far in terms of success and failure in organizing IAs. It must be recognized, however, that the differential performances were amplified by the optimizing behaviors of NIA's bureaucracy. To understand this process, it is necessary to understand the position of the systems office directly in charge of the System's operation and maintenance within the organization of the NIA (Fig. 10.3) as well as the goals and means of this office. As already explained, the NIA is requested to cover its current expenditure including administrative overheads and O&M costs by its current revenue, consisting mainly of irrigation fees, administrative set-asides from project funds for new construction and special project funds for O&M.

226

Especially, if a rural community, such as a village, is largely self-sufficient with little exposure to urban market activities, the expectation should be strong for its members to indefinitely continue their collective action in the future; this is akin to the situation of infinitely repeating games for sustaining positive-sum cooperative outcomes (Bardhan 1993; Seabright 1993; Ostrom and Walker 1994).

THE DETERIORATION OF A NATIONAL IRRIGATION SYSTEM

281

This financial autonomy in a current-account basis is also applied to each systems office. Collection of irrigation fees above a systems office's current expenditure budget is highly praised since the excess adds to the revenue of the higher echelons of the NIA's hierarchy. On the other hand, the current deficit results in a serious discredit against the managers (‘irrigation superintendents’) of the systems office. The deficit of a systems office can be covered by a loan from a regional office based on surpluses in other systems. However, such a loan is not easily structured when few other systems enjoy surpluses under the depreciation in the value of irrigation fees collected in paddy. Therefore, under the depressed rice market of the 1980s and 1990s, systems offices were forced to cut both their staff and their O&M expenditures. However, it was difficult for many offices to avoid chronic delays in the payment of staff salaries for several months. Under such conditions, it is no wonder that the systems office was hard pressed to increase collection of the irrigation fees. The willingness of farmers to pay the fees depends on the quality of the irrigation service. Yet, the option for the systems office of improving the service to farmers by increasing O&M expenditures was closed under stringent budgetary constraints. Thus, the office opted to allocate more of the available water to areas where irrigation fee collection rates were high at the expense of areas with low collection rates; this was achieved by diverting water to the laterals with active IAs from those with inactive IAs. Indeed, it is an open secret that the water masters open the head gates to Laterals B and E for longer hours at times of higher water demand than those to other laterals. As a result, the water supply to areas with active IAs increased further, whereas water shortages became even more serious in areas with inactive IAs. Such behavior by NIA staff created distrust among farmers in the latter areas and induced them to purchase irrigation pumps, while at the same time withholding the payment of irrigation fees; this further motivated the diversion of more water away from those areas. Thus the vicious circle between inactive local participation and deterioration in irrigation service in Laterals A, C, and D was greatly amplified by the systems office's short-run optimization behaviors. The short-run optimization by the NIA office was not able to prevent its collection of irrigation fees from declining either. Instead, the rate of collection from the System's service area decreased from 73 percent in 1987 to 50 percent in 1992, and further down to 38 percent in 1995.

10.5 Costs and Benets of the System's Management Despite the major deterioration in the System, rice production in its service area was maintained, because, in the areas where the supply of water from the NIA was drained, farmers continued to irrigate their field

282

MASAO KIKUCHI, MASAKO FUJITA, AND YUJIRO HAYAMI

by purchasing private pumps. Because underground water tables in the areas adjacent to Laguna de Bay are shallow and easily replenishable, farmers in the lower reach of the System could switch the source of irrigation at a modest cost with little worry about exhausting underground water. Yet, the switch entailed a cost which could have been saved if the System were properly maintained. We will try to illustrate by simple calculations what level of costs may be involved in inadequate O&M activities in the System or what level of benefits may be forthcoming from appropriate expenditures for O&M.

Private Benets and Costs to Farmers First, the private costs and benefits associated with farmers' switching from NIA's gravity irrigation to pump irrigation are calculated in per-hectare basis in Table 10.2 (middle section). In the situation where the NIA's service has already been terminated, one benefit to farmers from the use of pump irrigation is an increase in farm income resulting from the shift from rain-fed single-cropping to irrigated double cropping (row 1). This benefit is calculated by assuming the same yields for NIA's surface irrigation and pump irrigation because the yield difference is not found to be statistically significant at conventional levels according to the data collected from village headmen within the systems. Farmers' costs in annual flow terms includes the interest and depreciation of capital investments for the purchase of a pump and the digging of a well as well as operating expenses for fuel and labor (row 3). Three alternative interest rates are tried in the calculation: 100 percent per year is a typical rate in informal credits in the rural areas of the Philippines, 25 percent is a modal rate of commercial banks' lending with collateral, and 10 percent is a typical rate of concessional loans from government. Division of row (1) by row (3) produces benefit–cost ratios for introducing pumps where water supply from the NIA is not available. The results show that even for a farmer who has no access to anything other than informal credits, the benefit from introducing a pump is almost twice as high as its cost; if he has good collateral to enable him to receive credit from commercial banks, his benefit rises to more than three times his cost. Thus, it is no surprise that many farmers, especially the larger ones with a fair amount of land and other real assets, were quick to purchase pumps when the NIA's water supply dwindled. The situation is very different in the areas where sufficient water supply from the NIA continues to be available. In this case, the benefit for a farmer's switching from the NIA's surface water to pumped-up underground water is only the saving of the irrigation fee (row 2) which he has to pay in the event of his continued reliance on the services of the NIA. Division of row (2) by row (3) shows that the benefit from the use of a pump is lower

283

THE DETERIORATION OF A NATIONAL IRRIGATION SYSTEM

Table 10.2 Benefits and Costs to Farmers of Switching from the Surface to Pump Irrigation, and Social Benefits and Costs of NIA's Investment in the Systema

Annual flows (P/ha) Income increase with (1) over rainfed farmingc Irrigation fee (=suf- (2) ficient O&M expenditure)d Cost of pump irriga- (3) tion to farmerse Capital cost of sys- (4) tem constructionf Private benefit–cost ratio to farmers NIA service termi- (1)/(3) nated NIA service contin- (2)/(3) ued Social benefit–cost ratio Capital cost excluded (3)/(2) Capital cost included (3)/(2+4) b c d e f

Interest rate assumedb i=100% i=25%

i=10%

40,832

40,832

40,832

2,000

2,000

2,000

21,607

12,607

10,807

79,196

9,783

3,274

1.9

3.2

3.8

0.1

0.2

0.2

10.8 0.3

6.3 1.1

5.4 2.0

Notes: a Costs and benefits are all evaluated at 1995 prices. 100% per year is a typical rate in the rural informal money market; 25% is the typical lending rate (with collateral) of commercial banks; and 10% is the rate of government concessional loans. Assume the same yields per ha with NIA irrigation and with pump irrigation (3.9 mt in wet season and 4.8 mt in dry season) and rain-fed yield (2.9 mt in wet season alone). 2 cavans (1 cavan=50 kg) for wet and 3 cavans for dry season per ha, valued at P8.00/kg of rough rice price in 1995. Investment costs ×i+maintenance costs+fuel costs for wet and dry+operating labor cost. Initial investment costs for constructing the system in 1958 (including the interest for the average gestation period of 1.5 years).

than its cost under the condition of continued water supply by the NIA, even if concessional loans might be provided to farmers from the government. It is not surprising, therefore, that few farmers in the areas with adequate surface water supply purchased pumps. Clearly, farmers in the areas where the gravity irrigation service deteriorated were worse off relative to those in areas of continued adequate service, even though their loss was recovered to a significant extent by the introduction of pump irrigation.

Social Benets and Costs What level, therefore, of costs and benefits would the NIA's poor system maintenance have produced to society? Calculations summarized in Table 10.2 (bottom section) may shed light on this question.

284

MASAO KIKUCHI, MASAKO FUJITA, AND YUJIRO HAYAMI

If the NIA had continued to supply sufficient water, it should have been unnecessary for farmers to shift to pump irrigation. Therefore, the cost of pump irrigation (row 3 in Table 10.2) can be considered the benefit from adequate water supply from the NIA system. The cost of O&M activities per hectare that would enable the NIA system to continue supplying sufficient water to its service area is assumed to be the irrigation service fee set by the NIA intending to cover the System's O&M (row 2). Under the NIA's financial framework, this is the maximum amount that can be spent on O&M. The level of actual O&M expenditures in the System in the 1990s was less than 40 percent of this ‘ideal level.’ Division of row (3) by row (2) produces a benefit–cost ratio for the adequate maintenance of the System when the investment in the System's construction is treated as a sunk cost. The calculated ratios are very high, ranging from 5 to 11 for different interest rates assumed. Since the true social rate of discount is likely to lie somewhere between the rates of government concessional loan and a commercial bank's lending—certainly much lower than the rate of informal credit—the results seem to imply that one peso spent for the existing system's maintenance produces additional income of as much as about 6 pesos to society. If the capital cost of construction (row 4) is included in the denominator, the benefit–cost ratio is reduced to nearly one at private banks' lending rate and to two even at the concessional loan rate. These results suggest that investments in the construction of new gravity systems do not increase economic welfare in society unless the social rate of discount is lower than private banks' lending rate. However, the calculation for the case of including capital cost for construction of the system is not relevant in the present context which is focused on the NIA's strategy with the existing system. Although these calculations are very crude and somewhat over-simplified, the results unambiguously show that the NIA is investing in the O&M activities of existing systems at far less than a socially optimum level.

10.6 Institutional Development versus Increased O&M Expenditure While it is evident that the NIA's O&M expenditure in the late 1980s and the early 1990s were much less than the social optimum, it does not follow that the NIA should return to the organizational structure which prevailed until the early 1980s, whereby the NIA's systems office employed a large field staff directly undertaking the necessary O&M activities, with little effort being made to mobilize local participation and initiatives. The NIA's attempt to organize farmers in service areas into IAs and to hand over O&M activities to them represents one way to achieve greater efficiency in the maintenance of its systems. However, success has been limited to the small laterals with simple structures, involving small numbers of farmers and villages, while the

THE DETERIORATION OF A NATIONAL IRRIGATION SYSTEM

285

NIA's attempt to organize IAs in larger laterals, with more complicated technical designs and social structures, has largely failed. It is easy to understand that leaders in local communities have to face progressively greater difficulty in organizing IAs, as the size of membership and the number of villages involved become larger. The time and energy required to achieve consensus among larger populations in many different villages are often excessive for the leaders, who are usually busy with their own farming and other business. Assistance and advice from NIA staff with technical expertise as well as a certain discretion on the allocation of water and the O&M budget could be highly instrumental in achieving cooperation and consensus among people across different villages. The NIA has long recognized the importance of such institutional extension activities and has employed specialists with the title of ‘institutional development officers’ (IDO) mainly for the assistance of communal systems, where several successful cases have been reported. Since the early 1980s, when the NIA tried to organize IAs in national systems, some IDOs have been allocated for that purpose. A manual prepared by the Institutional Development Department at the NIA Headquarters specifies a standard that the full-time work of one IDO for one year should be allocated to each lateral for establishing an IA. Yet, under the stringent budgetary constraint, no full-time IDO has ever been allocated to any one of the five laterals in the System. Activities for the promotion of IAs were left largely to water masters. It is easy to imagine that the institutional extension activities by water masters, who are busy with dayto-day operations, are much less than sufficient, especially after a contraction of their number due to the systems office's financial crisis. In 1995–96, none of the IDOs was employed in the systems office, whereas only five IDOs were assigned to the Regional Office for Region IV that oversees 35 national systems. They were more than fully occupied by taking care of the already well-functioning IAs with little time to spend organizing new IAs or activating inactive ones. While recognizing the limitation of government employees to promote the organization of local participation and initiatives, is it not unreasonable to expect that the IAs of Laterals A and C–D could have been somewhat more active, even if less active than in B and E, if the institutional extension activities by IDOs had been provided at a standard level as specified in NIA's manual? In this section we will try to assess the efficiency of the NIA's investment in the institutional development of IAs in relation to that of increased O&M activities by NIA itself.

Economic Returns to Institutional Development In order to illustrate the potential economic gains from public investment in institutional development, benefit–cost calculations are undertaken as shown in Table 10.3.

286

MASAO KIKUCHI, MASAKO FUJITA, AND YUJIRO HAYAMI

Table 10.3 Social Benefits and Costs of NIA's Institutional Development vs. Increased O&M Expenditure for Irrigators' Associations (IAs) in the Systema Institutional development Success probability=0.5 Interest rate assumed i=25% i=10% Total present value (million pesos): Benefitb (1) 24.2 67.0 Costc (2) 1.8 3.4 Benefit–cost (1)/(2) 13.8 19.5 ratio 338 Internal rate of returns (%) b c

Success probability=0.2 Interest rate assumed i=25% i=10%

Increased O&M Success probability=1.0 Interest rate assumed i=25% i=10%

9.7 1.8 5.5

48.4 18.5 2.6

173

26.8 3.4 7.8

134.0 46.2 2.9

130

Notes: a Costs and benefits are all evaluated at 1995 prices. Annual benefit = Row (3) of Table 10.2 × 1500 ha × success probability. Benefit is assumed to emerge from the 3rd year. Cost for increased O&M=(Row (2) of Table 10.2–average O&M of 1985–94) × service area. The average O&M expenditure for 1985–94 is P680/ha, and service area is assumed to be 3,500 ha. Cost for institutional development=sum of the costs of 3 IDOs for the first 2 years and 1 IDO thereafter. The IDO costs are the sum of yearly salary and other costs for institution building per IDO per year (254,500 pesos/IDO).

THE DETERIORATION OF A NATIONAL IRRIGATION SYSTEM

287

In this calculation, it is assumed that, in order to establish an effective IA in a lateral, which it is difficult for the initiatives of local people alone to organize, one full-time IDO must be allocated to the lateral for two consecutive years (instead of one year as specified in the NIA manual). In addition, it is assumed that, in order to sustain the effective functioning of an IA after establishment, one IDO must continue to work one-third of his/her time in each lateral. Our calculation assumes that, if effective IAs could be successfully organized by those IDOs' institutional extension activities, the O&M of the System would be adequately performed to assure sufficient water supply to its service area, thereby eliminating farmers' need to use pumps. However, the success should be a matter of probability depending on many ad hoc elements such as the qualities of IDOs and local leaders as well as personal matches (or mismatches) between them. Therefore, in our calculation in Table 10.3, two alternative probabilities are assumed as 50 and 20 percent on the success of the institutional development project. One success out of five such trials is considered a rather conservative assumption according to the judgment of several NIA officials even for the cases of assigning one IDO for only one year. The annual flow of benefit from the NIA's investment in the institutional development project, if successful, will be the saving of pump costs arising from the shift back from the area presently irrigated by pumps to that served by NIA's surface irrigation (row 3 in Table 10.2). Multiplication of the annual benefit flow per hectare to the area presently irrigated by pumps (1,500 hectares of service area and 3,000 hectares of crop areas for two seasons) produces the system-wide total benefit flows, which is converted into the total present values at the beginning of the institutional development project by discounting future annual flows by alternative interest rates assumed (row 1 in Table 10.3). Note that this calculation is likely to produce conservative estimates of total benefit because it assumes the benefit arises only in the area presently under pump irrigation. In fact, it is likely that, so long as the NIA's O&M expenditure remains at its present level, the System will continue to deteriorate, resulting in progressively larger areas diverted to pump irrigation in the future. The cost of the institutional development project to effectively organize IAs in Laterals A, C, and D is assumed to consist of the two-year salaries of three IDOs for the first two years of the project and one IDO for subsequent years, plus other costs in support of their extension activities including transportation, holding extension meetings, an honorarium for local leaders' assistance, and various logistics. More concretely, one IDO's salary per year is assumed to be 170,000 pesos and other costs to be 84,500 pesos per IDO per year, according to information collected from the NIA. The total cost of three IDOs for the first two years and one IDO

288

MASAO KIKUCHI, MASAKO FUJITA, AND YUJIRO HAYAMI

for subsequent years are discounted and summed up to total present values at the beginning of the project (row 2 in Table 10.3). The social profitability of the institutional development project can be assessed by the ratios between the total present values of benefit and cost as well as the internal rates of returns calculated under alternative assumptions. The estimates of the benefit–cost ratio (row 1 divided by row 2) show that even under the most conservative assumptions (only one success out of five projects tried, the social rate of discount equal to the interest rate of a commercial bank loan), one peso spent for the support of IAs is expected to produce 5.5 pesos to society; this corresponds to the conservative estimates of the internal rate of return being as high as 173 percent per year, which is about seven times higher than the commercial banks' lending rate. Comparable calculations are made for the case of increasing NIA expenditure on O&M activities by its own staff to a sufficiently high level so that the area presently under pump irrigation will return to the services of the System. Benefit flows from the increased O&M expenditure are assumed to be the same as for the case of institutional development except that the probability of success is assumed to be one since the increased O&M activities are carried out by the NIA itself. The corresponding annual cost flow is assumed to be the difference between the irrigation fee and average O&M expenditure for 198594, that is multiplied by total service area. These benefit and cost flows are converted into present values under the same assumptions as for the case of institutional development. The calculation shows that the benefit–cost ratios of increasing expenditure for NIA's direct O&M activities are higher than 2.5 and its internal rate of return is as high as 130 percent. Yet, the rate of returns to increased expenditure for O&M by the NIA is significantly lower than that of investment in institutional development even for its success probability of only 20 percent. Such results imply that the present strategy of the NIA to hand over O&M activities to IAs can increase social economic welfare, if it is accompanied by appropriate public investment in the development of IAs. Despite such a high social rate of return being likely to be forthcoming, the NIA has invested very little in supporting the development of IAs. This underinvestment may be explained partly by the much higher discount rate on future incomes held by NIA officials than the social rate of discount. It may also be explained by the low rate of private returns to the NIA. Private benefits to the NIA from the success of institutional development would be increases in the collection of irrigation fees corresponding to increases in the area actually serviced by the NIA System. If this benefit is compared with the NIA's costs as assumed in the calculation in the social rate of return, the private benefit–cost ratio for NIA's institutional development investment is less than one if commercial banks' lending rate is

THE DETERIORATION OF A NATIONAL IRRIGATION SYSTEM

289

applied for discounting and is close to one even if the government's concessional banks' lending rate is applied, even under the assumption of one success out of five trials; this corresponds to the internal rate of return being insignificantly larger than the concessional loan rate (Table 10.4). As repeatedly explained, the financial crisis of the NIA was created by a reduction in public support for irrigation from both the external and the internal sources. Under the crisis, which is such that even salary payments to its regular staff are often delayed, the NIA official future discount rate was likely to be much higher than the interest rate of concessional loans from the government. Finding itself in such a situation, it is no wonder that the NIA neglected investment in the institutional development of IAs. The low rate of private returns to the NIA from increased expenditure on the NIA's direct O&M activities is even more pronounced with its benefit–cost ratio lower than one half and its internal rate of return being negative. In such a situation, it is natural that the NIA has reduced its O&M expenditure under the prescribed financial autonomy.

10.7 Toward a Community-Enhancing Approach Deterioration of the gravity irrigation system that we have investigated represents an organizational maladjustment to reduced public support for agriculture. Although the System was under the auspices of a state agency, the NIA, its efficient management requires the cooperation and participation of local beneficiaries, because the System has attributes of local commons or common-pool resources. So long as the NIA had been able to command abundant resources based on generous support from foreign donors and the national government, the System had been maintained and operated decently, if not efficiently. However as external support was contracted under a depressed rice market, the NIA faced a financial crisis. Pressed by the need to economize the use of public resources, the NIA rightly tried to hand over a part of its resource-allocation functions to local communities by organizing irrigators' associations (IAs). This attempt succeeded in relatively small areas but failed in many others, resulting in the deterioration of the overall system's performance.227 The failure stemmed from the difficulty of organizing local communities sufficiently strongly to prevent farmers from becoming free riders, and participating in activities such as illegal water take-out and non-payment of irrigation fees. The failure implies a dysfunction of the organizational reform mechanism. According to Albert Hirschman (1970), deterioration in the performance of

227

Such a failure is not unique to this System but can be found in many other systems in the Philippines (Oorthuizen and Kloezen 1995).

290

MASAO KIKUCHI, MASAKO FUJITA, AND YUJIRO HAYAMI

Table 10.4 Private Benefits and Costs to NIA of Institutional Development vs Increased O&M Expenditure for Irrigators' Associations (IAs) in the Systema Institutional development Success probability = 0.5 Interest rate assumed i=25% i=10% Total present' value (million pesos): Benefitb (1) 3.1 9.9 c Cost (2) 1.8 3.4 Benefit–cost (1)/(2) 1.8 2.9 ratio 50 Internal rate of returns (%) b c

Success probability = 0.2 Interest rate assumed i=25% i=10%

Increased O&M Success probability = 1.0 Interest rate assumed i=25% i=10%

1.2 1.8 0.7

6.1 18.5 0.3

4.0 3.4 1.2

14

Notes: a Costs and benefits are all evaluated at 1995 prices. Annual benefit=Row (2) of Table 2×1500 ha×correction rate. Correction rate=0.8. Row (2) of Table 10.3.

−28

19.8 46.2 0.4

THE DETERIORATION OF A NATIONAL IRRIGATION SYSTEM

291

an organization may induce reform either through members' collective protests or action demanding appropriate reform (‘voice’) or escape by individuals from the organization (‘exit’). In the case of this System, the voice was raised, for example, by downstream farmers complaining to the NIA as well as protesting to upstream farmers against their abuse of water, or even by padlocking head gates and channeling water illegally. Such a voice, often supported by force and involving bloodshed, was effective in inducing the organization of active IAs in Laterals B and E, which are characterized by small and simple system designs as well as by the low numbers of farmers and villages involved. However, the same voice and force totally failed to achieve the institutional development in Laterals A, C, and D involving much larger numbers of farmers and villages and thereby finding it difficult to reach agreements and consensus among themselves. This contrast is consistent with a hypothesis that the effectiveness of a voice for organizational reform may depend, to a significant extent, on the benefit that the reform is expected to produce, but it depends more critically on the inner structure of the organization to determine the cost of mobilizing a sufficiently strong voice. The failure in the System's reform also resulted from the NIA's inability to appropriately respond to the voice of downstream farmers. The situation could have been very different if the NIA had advanced adequate support for institutional development to the communities in larger laterals that are difficult to organize with the initiatives of local people alone. For NIA officials, the urgent financial difficulty should have been more binding than the pressure of farmers' voices; this adds to the evidence in support of the hypothesis that government officials' time discount rate on future incomes can be much higher than the social rate. One of the major factors underlying weakness in the voice of farmers was the existence of a relatively easy exit from the system via the purchase of irrigation pumps. In the long run, this exit option will be very damaging to the NIA, as it has meant the diminution of the source of irrigation fees. The long-run hazard, however, did not induce the NIA to undertake reform for the provision of institutional development supports, because the short-run saving of current expenditure was over-riding in their financial crisis. Instead, they took advantage of the weak voice of farmers in the areas of deteriorated service to maximize their short-run revenue by diverting water to farmers with better service and, hence, higher rates of irrigation fee payment. This myopic behavior of the NIA further encouraged farmers to exit the system by purchasing pumps on the market. If this market exit option had not been available, the farmers' voice could have been elevated sufficiently strongly in the long run to form effective IAs even in large laterals. During the time when the NIA was trying to form an IA in Lateral C, farmers in its lower reaches protested by voice and by force to farmers in the upper reaches against their abuse of water.

292

MASAO KIKUCHI, MASAKO FUJITA, AND YUJIRO HAYAMI

However, they soon gave up and opted to exit with pumps. If pumps had not been available at a modest cost through the market, they should have had no other option but to escalate the voice to violence, even involving major bloodshed. Such a negative sum situation would have eventually forced upstream farmers to accept compromise and cooperation with downstream farmers. In fact, this was the process by which villages in Japan established detailed rules and their effective enforcement on the use of common-pool resources, such as forests and irrigation water, in the seventeenth century when the scarcity of natural resources was elevated under continued population pressure upon closed land frontiers (Tamaki and Hatate 1974; Tamaki 1983; Hayami 1997: 252–3). The problem is that, until very recently, farmers in villages along this NIA system had had little chance to learn how to organize collective action for water management. Before the NIA irrigation was extended in 1958, farmers had practiced rain-fed rice production for which coordination in the use of water had been unnecessary. After that time, the irrigation system had been operated and maintained by the NIA with little involvement of farmer-beneficiaries, except for irrigation fee payments. It was quite natural that farmers were not able to organize adequate collaboration across many villages along the lateral within a short period when the NIA's maintenance activities were contracted rather abruptly; this is a situation common to many villages in Southeast Asia where the traditional abundance of natural resources has been rapidly eroded by explosive population growth. Such a social structure of village communities, plus the availability of an exit option through the market makes it a very difficult undertaking to organize rural people for the management of local commons in Southeast Asia as well as many other developing economies in the world today. The new policy orientation of delegating the management of local commons from state agencies to rural communities has the potential to promote both efficiency and equity. However, if it is promoted for the motive of saving government resources alone without recognizing the severe constraint in organizing rural people, positive consequences are unlikely to be forthcoming. Major public support is called for to overcome local communities' organizational capacity limit. While the financial autonomy of state agencies, such as the NIA, should continue to be respected, an additional public investment in support of communities' institutional development in a separate account from day-to-day administration may be justified on the ground of seemingly much higher social returns than the private returns to state agencies from such investment. Similar to the advocacy for the market-enhancing role of government (Aoki et al.1997), the community-enhancing activities should have to be borne by government either directly or, perhaps, through NGOs. Needless to say, the concrete policies that can really enhance communities' capacity can only be designed through

THE DETERIORATION OF A NATIONAL IRRIGATION SYSTEM

293

in-depth investigations into highly elusive norms and organizations of the rural communities.

References ADB (Asian Development Bank) (1998) Key Indicators of Developing Asian and Pacific Countries, vol. 29, Asian Development Bank, Manila. Aoki, M., Kim, H., and Okuno-Fujiwara, M. (eds.) (1997) The Role of Government in East Asian Economic Development, Oxford University Press, Oxford. Baland, J. M. and Platteau, J. P. (1996) Halting Degradation of Natural Resources: Is there a Role for Rural Communities? Oxford University Press, Oxford. Bardhan, P. (1993) ‘Symposium on Management of Local Commons,’ Journal of Economic Perspectives, 7(Fall): 87–92. Chambers, R. (1988) Managing Canal Irrigation, Oxford & IBH Publishing, New Delhi. Feeny, D., Berkes, F., Mccay, B. J., and Acheson, J. M. (1990) ‘The Tragedy of the Commons: Twenty-two Years Later,’ Human Ecology, 18(1): 1–19. Fujita, M. (1998) Tojokoku Kangai System ni okeru Seifu to Kyodotai (Government and Community in the Management of Irrigation Systems in Developing Countries; A Case of the Philippines), Ph.D. thesis submitted to Aoyama-Gakuin University, Tokyo. Hardin, G. (1968) ‘The Tragedy of the Commons,’ Science, 162: 1243–48. Hayami, Y. (1978) Anatomy of a Peasant Economy, International Rice Research Institute, Los Baños, Philippines. —— (1997) Development Economics: From the Poverty to the Wealth of Nations, Oxford University Press, Oxford. —— and Kikuchi, M. (1981) Asian Village Economy at the Crossroads, University of Tokyo Press, Tokyo (published in 1982 for North America by Johns Hopkins University Press). —— —— (2000) A Rice Village Saga: Three Decades of Green Revolution in the Philippines, Macmillan, London and St. Martin's Press, New York. Hirschman, A. O. (1970) Exit, Voice, and Loyalty: Responses to Decline in Firms, Organizations, and States, Harvard University Press, Cambridge, MA and London. IRRI (International Rice Research Institute) (1995) World Rice Statistics 1993–94, International Rice Research Institute, Los Baños, Philippines. Olson, M. (1965) The Logic of Collective Action, Harvard University Press, Cambridge, MA. Oorthuizen, J. and Kloezen, W. H. (1995) ‘The Other Side of the Coin: A Case Study on the Impact of Financial Autonomy on Irrigation Management Performance in the Philippines,’ Irrigation and Drainage Systems, 9: 15–37. Ostrom, E. (1990) Governing the Commons, Cambridge University Press, New York. —— (1992) Crafting Institutions for Self-Governing Irrigation Systems, Institute for Contemporary Studies Press, San Francisco.

294

MASAO KIKUCHI, MASAKO FUJITA, AND YUJIRO HAYAMI

Ostrom, E. and Gardner, R. (1993) ‘Going with Asymmetries in the Commons: Self-Governing Systems can Work,’ Journal of Economic Perspectives, 7(Fall): 93–112. —— and Walker, J. (1994) Rules, Games, and Common-Pool Resources, University of Michigan Press, Michigan. Seabright, P. (1993) ‘Managing Local Commons: Theoretical Issues in Incentive Design,’ Journal of Economic Perspectives, 7(Fall): 113–34. Small, S. E. and Carruthers, I. (1991) Farmer-Financed Irrigation: The Economics of Reform, Cambridge University Press, Cambridge. Svendsen, M. (1993) ‘The Impact of Financial Autonomy on Irrigation System Performance in The Philippines,’ World Development, 21(6): 989–1005. Tamaki, T. (1983) Mizushakai no Kozo (Structure of Hydraulic Society), Ronsosha, Tokyo. —— and Hatate, I. (1974) Hudo: Daichi to Ningen no Rekishi (Environment: History of Land and Man), Heibonsha, Tokyo. Wade, R. (1988) Village Republics: Economic Conditions for Collective Action in South India, Cambridge University Press, Cambridge and New York.

11 Evolution and Consequences of Community Forest Management in the Hill Region of 228

Keijiro Otsuka and Towa Tachibana Community management of local natural resources, such as forest, water, and rangeland, has been receiving increasing attention in the development literature as a potentially effective institutional arrangement to regulate the use of these resources (Baland and Platteau 1996; Bardhan 1993; Hayami 1997). Having perceived the failures of central government's ownership and control, international donor agencies and the governments of many developing countries began to adopt the participatory community approach, in which local resource users are entrusted to manage natural resources (FAO 1989; World Bank 1996). The promotion of this new policy has been supported by new developments in economic theory, particularly in the game-theoretic fields, which suggests the possibility of avoiding the tragedy of the commons under a variety of conditions (Pearce 1992). Further, several authors argue that community management is potentially an effective institutional arrangement to attain efficient use of local natural resources, particularly if it is granted formal and assured land rights (Runge 1981; Ostrom 1990; McGranahan 1991; Bromley 1992; Ostrom and Gardner 1993; Baland and Platteau 1996). Under the private ownership every owner must exert effort to protect his/her property. In the case of forest resource management, we expect that the cost of protection will be particularly high if the resources are minor forest products, such as firewood, fodder, and grasses.229 This is because it is highly costly to identify the property from which these minor products are extracted.230 The cost of protecting forests, however, may be economized by

228

We would like to thank Bob Allen, Yujiro Hayami, and other participants of the World Bank Conference in Stanford (1999). We are also grateful to our collaborators in Nepal, Ridish Pokharel, Santosh Rayamajhi, and Hari Upadhyaya. This research was partly supported by a grant from the Japan Economic Research Foundation.

229

Tiny private forests are developed near farmers' residence in some places and fodder trees are often planted on edges of private fields in the hill region of Nepal, because it is easy to oversee such forests and plantations.

230

It is likely that the cost of protection will be lower if major forest products are timber, because felling and hauling timber trees illegally may well be detected by someone in the village community. Moreover, the production of high-quality timber trees requires silvicultural operations such as pruning, singling, thinning, and weeding. Kijima et al. (2000) demonstrate that timber trees were more actively planted and thinned under individualized management than traditional collective management in post-war Japan. Similar findings are made in timber forests and plantations in the inner Tarai region of Nepal by Sakurai et al. (1998).

296

KEIJIRO OTSUKA AND TOWA TACHIBANA

organizing collective action such as patrolling and guarding the community resources under the common property regime.231 It is also argued that community management will be efficient in traditional communities where people know each other and are tied by kinship and other enduring personal relationships (e.g. Hayami 1997). In such circumstances, the transaction costs of formulating management rules of community resources and enforcing them may not be too high and, hence, the use of resources may be effectively regulated. A recent theoretical study based on evolutionary game theory demonstrates that whether forest resources are depleted under a common-property regime depends on the cost of sanction that maintains the cooperative behaviors of its members (Sethi and Somanathan 1996). Further, Aoki (2000), in this volume, argues that to the extent that the threat of social ostracism is credible, cooperation in the commons game can become the norm of behaviors in the village community. In contrast to the development in theoretical studies, there have been relatively few empirical studies exploring the efficiency and effectiveness of local community management of natural resources.232 Given the various theoretical predictions, it is clearly an empirical question whether the community management of natural resources is effective in conserving and protecting common property resources. This chapter attempts to assess the feasibility and the effectiveness of forest resource management under the community management system based on a case study of community forests in the hill region of Nepal. First, we attempt to identify factors facilitating or impeding the formation of community management systems. Second, we explore the consequences of the community forest management on the forest conditions. Specifically we postulate the U-shaped changes in forest resources over time, in which deforestation is followed by reforestation. The deforestation phase can be explained by open access to forest resources coupled with increased population pressure, whereas the reforestation phase can be explained by the emergence of effective community management. This hypothesis maintains that organizing community members, introducing management rules, and enforcing them are costly, and hence community management is introduced only when the benefit of initiating such management exceeds the cost. In other words, we argue that the benefit of initiating community management exceeds the cost of setting up such a system only when forest resources have become sufficiently scarce and valuable.233

231

Community ownership or a common property system is defined as a system whereby resources are subtractable like private goods, and are owned and used jointly by a group of people. It is generally presumed that the exclusion of users is costly to a varying extent. See Ostrom (1990), who proposes to call resources with imperfect excludability common-pool resources.

232

See Wade (1988), Stevenson (1991), Sakurai et al. (1998), and Kijima et al. (2000), Bardhan (Chapter 9 in this volume), and Fujita et al. (1999), among others. See also Arnold (2000) for a recent survey of the literature on community forest management.

233

Our hypothesis is consistent with the induced institutional innovation hypothesis (Hayami 1997) which argues that the increasing scarcity of a resource induces an innovation to save that resource relative to other resources.

COMMUNITY FOREST MANAGEMENT IN NEPAL

297

11.1 Geographical and Historical Background In Nepal there are five topographical regions: to range them from low to high altitudes, they are Tarai, Siwalik, Middle Mountain, High Mountain, and High Himal. The hill region corresponds to Middle Mountain, which lies between the altitudes of 700 and 2000 meters. This region has been the center of the country, although the Tarai plain has emerged as an industrial center since the eradication of malaria in the 1950s. Because of its rugged geographical conditions, the majority of inhabitants in the hill region of Nepal have unfavorable access to urban markets. Even now it is not exceptional to walk over hilly trails for a few days to reach one's village from the nearest market town. As a result, the main economic activity has been subsistence farming with minimum dependence on purchased inputs, such as chemical fertilizer, and the main source of household energy has been almost exclusively firewood. Furthermore, farmers depend on such minor forest products as leaf fodder and grasses to feed livestock for milking, plowing, and making composts. Compost application is indispensable to maintain the soil fertility of terraced fields located on steep slopes. Even leaf litter collected from the forest is used for bedding materials for livestock to collect nutrients contained in urine. Historically under the feudal Rana regime from 1846 to 1950, local government officials controlled the harvesting of timber trees in the forests. Such control was particularly strict in areas near Tarai and main roads, which had favorable access to timber markets in India. Even in remote areas where market access was unfavorable, permission was required to fell timber trees for own use, such as the construction of houses.234 In contrast, the collection of minor forest products was usually unregulated, which would reflect the abundance of such resources at that time. The feudal system, however, collapsed in 1950, and was followed by the political upheaval and the accelerated population growth. It is alleged that massive deforestation began in the 1950s with the gradual collapse of the feudal forest management system, which resulted in rampant felling of timber trees (Gilmour and Fisher 1991). In order to arrest deforestation, forest areas were nationalized in 1957. With the limited number of forestry officers, however, the government could not implement the nationalization policy effectively except in Tarai and the areas near the hill district headquarters (Graner 1997). According to our interviews with village elders, forest users did not usually perceive that their forests were nationalized

234

It is not clear how strictly the regulation on timber harvesting was applied to the remote hill regions. Graner (1997, ch. 2) suggests that the Rana regime attached great importance to the forests even in the hill region. Gilmour and Fisher (1991, ch. 1) argue, however, that the central government of Rana promoted forest clearance in most areas of the hill region and protected limited forest areas as a source of fuel for manufacturing arms. In our field interviews, many older respondents recalled that, with required payment, it was not too difficult for them to obtain permission for timber harvesting during the Rana regime.

298

KEIJIRO OTSUKA AND TOWA TACHIBANA

until cadastral surveys were conducted one or two decades later. Other factors, such as population pressure, would have been also responsible for deforestation (Gilmour and Fisher 1991). While the loss of forest area has been the major form of deforestation in most developing countries (see, e.g., Deacon 1994), degradation of forest conditions has been the major concern in the hill region of Nepal. Indeed, based on aerial photographs taken in 1964–65 and 1978–79, it is reported that the loss of forest area had been negligible (Metz 1991; Soussan et al.1995). This is because most of the cultivable land had already been converted to farmland in the hill region by the first half of the twentieth century. According to our own estimates based on aerial photographs taken in 1978 and 1992–96, the average proportion of forest areas in village development council (VDC),235 a local administrative unit, changed only from 35.2 percent to 35.7 percent. In general, no visible changes in the area of our sample forests were identified by the analysis of aerial photographs. As will be seen, proportions of crown cover declined in many sample forests for the same period, indicating that the quality of forests has generally deteriorated. In the forests near main roads, timber extraction for market sale by villagers as well as outsiders has been a serious problem. In the majority of hill forests which have unfavorable access to markets, however, timber trees were cut primarily for the construction of houses and funerals in the villages. Other direct causes for declining crown cover were grazing, and collection of firewood and leaf fodder for own consumption. Grazing is a major threat to forests, as cattle eat and step on seedlings and coppices. While the collection of dead and dry branches does not deteriorate the forest condition, cutting live branches for firewood and fodder results in major damage. As the population increased, the intensity of grazing and the extraction of firewood and fodder increased, resulting in degradation of forest conditions. The shortage of forest resources led to a reduction in the size of livestock and increased the time allocated to collecting forest resources from distant forests, which had traditionally been used by different communities, such as hamlets (Kumar and Hotchkiss 1988). Such invasion on other forests has often created conflict among people from different communities. Responding to continued deforestation and increasing conflicts, some communities spontaneously began to form informal forest user groups to manage forests, even though the government possessed legal ownership titles. It should be noted that such informal user groups were formed based on customary access rights to forest, but not on residence within the political boundaries. Thus, it is not uncommon to find cases in

235

The local administrative hierarchy in Nepal consists, from larger to smaller units, of development region, district, VDC, and ward. Usually there are several hamlets in each ward.

COMMUNITY FOREST MANAGEMENT IN NEPAL

299

which several hamlets from different VDCs form an informal user group.236 These informal user-groups began to exclude outsiders from the use of their own forests and to regulate the extraction of timber and other forest products by the user group members.237 Messerschmidt (1990) and Gilmour and Fisher (1991), among others, indicate the burgeoning success of such indigenous forest management systems based on the results of several case studies in the early stages of their evolution. Having observed favorable effects of forest-user group management on forest conditions, the government of Nepal officially declared the promotion of community management of forests in 1987 (Nepal 1988). Since then, the government has handed over use rights of forests to forest user groups, if they demonstrate their willingness and ability to manage forests effectively to the district forest officer (DFO). The submission of a ‘constitution’ and a ‘management plan’ are required, which specify the rules of forest management agreed upon by all community members. Once handed-over, formal forest user groups are allowed to sell forest products to the markets, even though we seldom observed significant sale of those products. Furthermore, the Forest Act in 1993 granted the formal user groups the legal right to punish violators of the constitution, which seems to have further strengthened the community management.

11.2 Basic Characteristics of Sample Forests Nepal consists of five development regions: eastern, central, western, mid-western, and far-western. Initially we have randomly sampled 100 forests from the whole hill region in terms of the relative importance of forest area in each region compared with the total hill area.238 The minimum forest size of 10 hectares was imposed in order to apply aerial photo analysis with sufficient accuracy. Fifty forests were chosen from the non-remote areas and another fifty from remote areas; remoteness being defined by a distance of more than 10km from a main road and 15km from district capitals. Roughly speaking, non-remote forests can be reached within a

236

Hamlet (tol in Nepali language) is not the geographical minimum unit for the membership of a forest user group. Though not so common, we encountered cases in which some households are members of a forest user group and others are those of another user group even within the same hamlet.

237

While the organizational structures of the informal user group are diverse, it is common to select committee members in general meeting to entrust the management activities of forest, such as the selection of trees which are allowed to be felled. There are also cases in which the leaders of local administration are chosen to be the leaders of informal community forest management. In contrast, the formal user groups must select committee members by election.

238

By chance, we happened to select a few forests from the lower hills called the Siwalik mountains at an altitude of 300–700 meters.

300

KEIJIRO OTSUKA AND TOWA TACHIBANA

Table 11.1 Basic Characteristics of Sample Forests No. of sample forests Traveling time to (hours): Nearest local market Nearest ranger office Size of forest (ha) No. of user households Average forest area per household (ha) Proportion of user group management (%) Proportion of formal user groups (%)

Non-remote forest 38

Remote forests 36

1.04 1.29 86.1 187.9 0.46

6.24 5.90 117.8 132.8 0.89

92.1

50.0

74.3

33.3

half-day's walk from the main road. Because of the inconsistency and the unreliability of some data, this study focuses on 74 sample forests.239 As is shown in Table 11.1, non-remote forests are located substantially closer to local markets and offices of forest rangers, who are in charge of local forest management. The average size of community forests in non-remote areas is about 70 percent of those in remote areas. In contrast, the number of user households is much larger in non-remote areas. As a result, the average forest area per user household in non-remote areas is nearly one-half of that in remote areas. Therefore, population pressure on forest resources tend to be higher in non-remote areas. Such areas, however, have better access to alternative sources of energy for firewood, such as kerosene and biogas, and to chemical fertilizer, which is a major substitute for compost. While population pressure will stimulate the emergence of spontaneous community forest management, favorable access to markets may hinder it. According to Table 11.1, the proportion of sample forests managed by user groups, either formally or informally, is much higher in non-remote forests.240 This may be explained by the deteriorating forest conditions in non-remote areas due to severe population pressure that led to the formation of forest user groups. As a matter of fact, the average crown cover in 1978 was 39 percent and 49 percent in non-remote and remote sample forests, respectively. According to the result of our group interviews, leaders of about 70 percent of forest user groups stated that the shortage of forest resources was the single most important reason for initiating user group management.

239

In certain areas, our survey teams had difficulty in locating sample forests chosen by the use of aerial photos. In this study, we have omitted those 26 sample forests, which might not have been investigated by our survey teams by mistake with positive probabilities. We hope that this procedure will not result in a serious bias in the estimation results of regression analyses.

240

Actually there are several informal user groups which are inactive as evidenced by the absence of management rules and activities. Based on the results of resurvey, we treated these cases as no user group management.

COMMUNITY FOREST MANAGEMENT IN NEPAL

301

Another important factor affecting the formation of user group management may be the direct intervention by the DFO. An informal forest user group is a voluntary organization, which manages forests officially owned by the government. Since no formal use rights are granted, the informal group may face difficulties in excluding both outsiders and inside violators. In order to register forest user groups or to acquire the official use rights of forests from the government, an informal user group has to submit a management plan to the DFO in accordance with the guidelines prepared by the Ministry of Forestry. Once approved, the forest is handed over to the user group and secure use rights are granted, unless it grossly violates government regulations, such as felling mature timber trees without permission. The proportion of registered forests managed by formal user groups is definitely higher in non-remote areas (Table 11.1). This seems to reflect the fact that because of budget and manpower constraints, DFOs have primarily assisted those user groups which are easily accessible. Local DFO staff also provide technical support for the forest management, including the provision of seedlings for rehabilitation of degraded forests. DFOs occasionally visit the formally managed forests to check forest conditions and user group activities. If they detect violations of management rules, it is expected that they assist user group committees to implement punishment. In fact, users of the registered forests generally perceive that the enforcement of the rules of community forest management has become more effective since the forest was registered. Another major role of DFOs is to examine whether there are any serious conflicts of management rules among forest users proposed by informal forest user groups. The DFOs refuse to hand-over forests to user groups if conflicts are found. As is pointed out by Aoki (2000) and Baland and Platteau (1996), it is clear that the introduction of the third party, i.e. the DFO, has changed the nature of community management games. As shown in Table 11.2, the relatively small number of informal user groups were formed before the mid-1980s. Since the change in government Table 11.2 The Number of User Groups Initiating Informal and Formal Management by Period 1966–70 1971–75 1976–80 1981–85 1986–90 1991–97 Total

Informal 4 1 4 3 20 21 53

Formal

1 31 32

302

KEIJIRO OTSUKA AND TOWA TACHIBANA

policy in favor of community user group management in 1987, not only were a large number of informal user group management schemes initiated, but also a sizable number of formal user groups were set up. The initiation of some user group management may have been inspired by the opportunity to acquire forest use rights. In fact, several user groups were set up as formal user groups from the beginning with the direct assistance of the DFO, even though the majority of formal user groups had been converted from informal groups.

11.3 Forest Conditions and Management Rules In the standard economic model of common property management (e.g. Dasgupta and Heal 1979), the crux of the management issue is how to prevent socially excessive exploitation of resources arising from the failure of coordination among resource users. In the reality of Nepal, over-exploitation has already taken place and forests have been seriously degraded. At present, a more important management issue is how to regenerate forest resources rather than simply protect them. Thus, the extraction of scarce forest resources is not only regulated but also often prohibited by user groups in the short run. One of the critical questions is whether and to what extent the policy of handing-over forest use rights to the community is effective in enhancing the management efficiency of forest user groups. In practice, informal user groups often perceive that the local forests, which they have used for many years, belong to them. Based on such perception, they voluntarily undertook forest management. Thus, some researchers argue that informal user groups have been successful in protecting forests in a sustainable manner, even though they do not possess official use rights (e.g. Fisher 1989; Fox 1993). Others, however, disagree (e.g. Gilmour et al.1989). According to our field interviews, management rules are more effectively enforced under formal user group management than its informal counterparts, primarily because the enforcement of management rules is supported, implicitly or explicitly, by the authority of the DFO. We conducted an analysis of aerial photos taken in 1978 and 1992–96, in order to obtain information on general forest conditions and their changes over time. The most important indicator of the forest conditions generated by the aerial photo analysis is the proportion of crown cover. The results, however, were not wholly satisfactory, primarily because it is difficult to distinguish between the canopies of big trees and short bushy trees generated by coppices of felled trees. As a result, relatively small differences and changes were found in crown cover across forests under no management, informal and formal user group management and over time (see Table 11.3). A closer examination suggests, however, that forest conditions

303

COMMUNITY FOREST MANAGEMENT IN NEPAL

Table 11.3 Forest Conditions in 1978, 1992–96, and 1997 by Management Regime in 1997 No. of forests Crown cover (%): 1978 1992–96 Biomass per ha in 1997 (ton/ha) No. of regenerated trees per 100 m2 in 1997

No user group 21

Informal 21

Formal 32

42.4 38.6 143

31.4 33.3 135

31.9* 28.1* 131

31.0

33.0

47.4*

Note:* indicates that there is significant difference in means between no user group management and informal or formal management at the 5 percent level.

under no user group management deteriorated more rapidly, even though they were more favorable than those under informal and formal user group management in both 1978 and recent years. These observations are consistent with our hypothesis that deforestation is followed by reforestation due to the formation of user group management. It appears that, though inaccurate, crown cover broadly captures the forest conditions. On the other hand, there is no clear indication that forest conditions were improved under formal user group management. Recall, however, that formal user group management was initiated mostly in the 1990s, so that the impact of formal user group management might not have been revealed in terms of changes in crown cover from 1978 to 1992–96. This is another limitation of the use of aerial photos for our analysis. In the hill region of Nepal, the most important forest resources for local users are not timber but minor forest products, such as firewood, leaf fodder, and leaf litter. It is therefore appropriate to use the biomass of all trees, rather than timber volume, for the analysis of management of forest resources in this region. Thus, we conducted a forest inventory assessment in which we directly measured the density and distribution of various tree species of different sizes in sample forests. We also measured the rate of regeneration of young trees, which are shorter than breast height. Given the facts that user group management was initiated relatively recently and that the major means of restoring the forest conditions is to generate coppices by strict protection, the rate of regeneration seems to be a more appropriate indicator of the effects of user group management than biomass. It must be pointed out that the measurement of the density and distribution of various trees in community forests in the hill region of Nepal is not an easy task, so we applied a sampling formula for the selection of sample plots. We anticipate that some measurement errors are inevitable, particularly

304

KEIJIRO OTSUKA AND TOWA TACHIBANA

when forests are large and their conditions are not uniform (see Rayamajhi and Pokharel (1998) for details of forest measurement). It is interesting to find that the average biomass per hectare was greatest under no user group management in 1997 (Table 11.3). In terms of the regeneration of trees, however, formal management seems to perform best. It is likely that the regeneration rate was low under informal management essentially because forests were so degraded that recovery requires both time and strict protection. On the other hand, regeneration was also inactive under no user group management despite the large stock of biomass, which is likely to reflect the absence of management. In contrast, the regeneration rate is high under formal management despite a meager endowment of biomass. Such differences can be explained by the effect of formal user group management on the regeneration of young trees. Overall, the above observations seem consistent with the hypothesis of U-shaped changes in forest conditions over time, in which forest conditions deteriorate when forest resources are abundant and improve after user group management, particularly that of a formal one, is initiated. The restrictions on cutting timber trees by user groups have been effectively enforced, as they are bulky resources and, hence, the cost of their protection is relatively low. Major management rules on the extraction of minor forest products are either a restriction on the number of days on which specific members of user groups are allowed to collect forest resources or to totally prohibit use for the time being to facilitate regeneration. Slightly less than one-half of user groups employ watchers or adopt rotational patrolling, and mutual supervision is attempted everywhere under any user group management. While a restriction on the collection of dead and dry branches is important to prevent the excessive collection of firewood, more important regulations for the regeneration and recovery of forest conditions are restrictions on cutting green branches and grazing livestock. Cutting green branches and young trees except for deformed ones, very low branches, and those trees grown too densely, must be strictly controlled for the successful regeneration of trees. Table 11.4 summarizes the regulations on the extraction of minor forest products implemented by informal and formal user groups. The numbers simply indicate whether the user groups adopt any regulations or not, as it is difficult to assess the severity or intensity of regulations. For example, we cannot always assume that a total prohibition of resource extraction is stricter than regulated extraction of the resources as a management rule, because it is often desirable to remove some trees, which are densely grown, and low-lying branches to improve forest conditions. Several important observations can be made from Table 11.4. First, as is expected, no management rule is adopted in the forests without a user group. This indicates that forest is open access, at least for community members, unless user group management is introduced. Second, it is clear

305

COMMUNITY FOREST MANAGEMENT IN NEPAL

Table 11.4 Frequency of Sample Forests Adopting Regulations of Forest Resource Extraction/Use by Management Regime in 1997 No user group 21

No. of sample forests Regulations on (%): Green firewood collection 0 (0) Dead firewood collection 0 (0) Grazing 0 (0)

Informal 21

Formal 32

19 (91) 3 (14) 8 (38)

32 (100) 14 (44) 12 (38)

that all the formal groups, as well as almost all the informal groups, adopt some regulations on the extraction of green firewood, which include prohibitions and regulated extractions.241 Since such regulations are of critical importance for the protection of forests, not only formal but also informal user groups implement some regulations. Thus, we cannot distinguish between the two management regimes in terms of the frequency of regulations for green firewood extraction. Third, the regulation of the collecting of dead branches is less strict than that of green firewood, as it does not affect the forest conditions of the future. Fourth, the collection of dead branches is seldom regulated in informally managed forests, whereas it is regulated in nearly one-half of the formally managed forests. Thus, it seems reasonable to hypothesize that formal user groups manage forests more strictly than informal user groups by adopting stricter management rules, even though both of them manage forests to some extent, unlike the case of no user group management.242

241

We gathered that young trees are not allowed to be cut for firewood even in those two informal user groups which do not have explicit rules on the extraction of green firewood. In these user groups, the users can enter their forests only with their neighbor households, usually four to five, for mutual supervision. Furthermore, their forests can be easily seen from their village residential areas.

242

In order to corroborate this hypothesis statistically, we attempted to estimate the simultaneous equation systems in which the selected dummy variable for the presence of regulation (i.e. for the collection of green firewood, and dead and dry branches) is regressed on management regimes, among others, in the first-stage probit regression and the regeneration rate of young trees is regressed on the predicted management-rule dummy in the second-stage regression. This formulation, however, is not workable, because the management regime dummies used as regressors in the first-stage probit regression almost fully ‘explain’ the dependent variable. In other words, the difficulty in estimation arises from the facts that no management rule is adopted without user groups and that either almost all formal and informal user groups adopt some regulations, as in the case of green firewood extraction, or only a very few informal user groups adopt some regulations, as in the case of dead firewood extraction. See Greene (1997: 892) on this point.

306

KEIJIRO OTSUKA AND TOWA TACHIBANA

Grazing is sometimes prohibited or allowed only in areas where trees are mature under user group management. Even if grazing is not explicitly prohibited, grazing is not feasible if the forests are ‘closed,’ as will be explained shortly. While regulations of grazing have become stricter over time, regulations of firewood collection have remained largely the same. There are several ways to regulate forest resource extraction. Generally, the prohibition implies that nobody is allowed to enter the forest except for management purposes organized by the user group committee, rather than simply prohibiting cutting of trees and green branches. Similarly, the rules of regulated use specify ‘open days’ during which specific groups of user households are allowed to collect resources under the supervision of forest user group committee members. Potentially these regulations can be effectively enforced, as those who enter the forest on ‘closed days’ are in principle immediately caught (see, e.g., Chhetri and Randy 1992).243 The extraction of forest resources by a group of people on the same day facilitates mutual supervision. Furthermore, the extracted resources are usually piled up in one location at the end of the open day and are shared equally among the participating households in most cases. This practice reduces incentives to extract forest resources excessively.244 The same regulations were widely employed in community forest management in pre-war Japan (see, e.g. McKean 1992). Punishment rules differ from community to community. If the illegal extraction of firewood is found by the forest watchers and if this is a first offence, a fine may be imposed which is equivalent to the value of the extracted firewood. Second and third offences are subject to escalated punishments, which are stated in the management rules in the case of the formal management. In practice, repeated offences seldom take place or even if they do take place, the escalated clauses with severe punishments may not be applied to fellow community members. According to the field interviews with the committee members of several user groups, illegal activities have been reduced after handing-over, because such activities are more likely to be punished strictly due to the involvement of local DFO staff in forest management. In general, the use of watchers is less common and punishment is less strict in the case of informal forest management. While concrete evidence is hard to come by, we believe that community norms play a significant role in the enforcement of community management rules. Since villages are geographically isolated in the hill region of Nepal, social interactions among villagers are intense. Farmers collectively participate not only in the management of forests but also in the management

243

We found an interesting regulation in which one tree is allocated to each household each year without specifying open and closed days. A household can cut its tree any time but only under the supervision of the committee members.

244

However, other sharing rules seem to have been increasingly adopted, such as the sharing based simply on one's collection of firewood on open days.

COMMUNITY FOREST MANAGEMENT IN NEPAL

307

of local schools, irrigation and drinking water systems, and festivals, funerals, and religious ceremonies. As Aoki argues (2000), linking management ‘games’ with social activities generating social surplus may lead to a situation where cooperative behaviors become the community norm. Yet, in practice, cooperation in forest management is seldom overlapped with cooperation in irrigation and drinking water management or other spheres, such as the management of schools. This is because the relevant groups for the management of forest and water resources are generally different; while forest management tends to be carried out by a certain number of wards, other social activities are often carried out by a single ward or even hamlets within a ward. It seems that although such linkages are likely to play a role in reducing non-cooperative behaviors in forest management, it is difficult to identify the specific social activities that are closely linked with forest management.

11.4 Determinants of the Emergence of User Group Management In order to identify the determinants of the emergence of user group management, we employed two methods of empirical estimation. First, we estimated a probit function in which the dependent variable assumes unity when any types of forest user groups were set up. We did not distinguish between informal and formal user group management in this estimation, because we suspect that the initiation of formal user group management was affected critically by the intention of DFOs, which may be considered exogenous for user groups. Second, we estimated an ordered-probit function in which the dependent variable is a rank order dummy: 1 for informal and 2 for formal user group management. This method attempts to identify factors affecting not only the formation of informal user groups but also their conversion to formal user groups. The same set of explanatory variables are used and include the following: the number of user households; the size of forest per user household; the number of local administration units involved in a single user group; the ratio of households participating in other social activities; homogeneity of social groups represented by ethnic composition; traveling time to the nearest ranger office; crown cover in 1978; and the ratio of forest area in VDCs in 1978 in which our sample forests are located.245 The number of user households, for which we took a logarithm in the actual estimation, is supposed to capture the transaction costs of organizing collective forest management, which is related to the cost of reaching agreement. If the fixed cost component of setting up user group management is

245

When we calculated the ratio of forest area in a VDC, we included the shrub area because this area also produces minor forest products.

308

KEIJIRO OTSUKA AND TOWA TACHIBANA

high, the large forest area may have a positive effect on the initiation of user group management, partly because the total benefit of forest management will be larger and partly because scale advantages may exist in protection activities. In order to take into account the effect of forest area in addition to the effect of the size of user households, we added a logarithm of forest area per user household. If the transaction costs of community forest management increase with the size of user group and forest area, we expect to have negative effects of these two variables. We observed, however, that large community forests were sometimes divided into smaller units and were managed separately by different groups of people.246 To the extent that such division of a forest area is effective in reducing the transaction costs, both the number of households and the forest area per household may not exert a negative and significant influence on the formation of user groups. It must be also pointed out that the data on the number of user households would suffer from measurement errors, particularly in forests under no user group management or open access, because anyone can freely use the resources of open access forests.247 Also note that due to data limitation, we had to use current-period variables for the size of user groups and a few other variables to explain changes in forest management institutions in the past. The transaction costs of user group management would be affected not only by the size of the group and forest area but also by its ethnic homogeneity as well as homogeneity of administrative units. We used the ratio of the dominant ethnic group to represent the effect of ethnic homogeneity among the users. It is usually presumed that collective action can be more easily organized among homogeneous members. In fact, leaders of 85 percent of user groups stated that the user groups were formed by the decisions of users at large rather than by the initiative of selected leaders. Group homogeneity, however, may imply the lack of leadership. We also considered the number of wards (the lowest local administrative unit in Nepal), from which the users come to the forest, as an indicator of group heterogeneity. Since the ward is an administrative unit, which roughly corresponds to a village in other societies, the transaction cost within a ward tends to be low. Thus, we expect that the fewer the number of wards, the lower would be the cost of collective management.

246

In our observation, the number of community forests implementing the division of user groups is increasing. It will be an important issue to what extent such division will continue towards the direction of management by a few or even private management.

247

In the survey of the open access forests, we asked villagers residing near the forest the names of hamlets from which residents come to the forest under investigation. Then, we visit-ed each hamlet and asked the number of households. This method may lead to underestimation because of the possible omission of hamlets and to overestimation because some households in hamlets are not users of the forest under investigation. According to Graner (1997), there are also cases in which user group management excluded socially and economically disadvantaged users.

COMMUNITY FOREST MANAGEMENT IN NEPAL

309

We considered the two variables to represent the forest conditions and access to forest resources in the earlier period. One is crown cover in 1978 obtained from aerial photo analysis. Although this indicator is not very sensitive to changes and differences in forest conditions, it seems to capture major changes and differences. We hypothesize that the lower crown cover in 1978 would lead to a greater probability of inducing the initiation of user group management. Another variable is the ratio of forest area in 1978 in the VDC in which the sample forest is located. The larger the ratio of forest area, the lower would have been the population pressure on forest resources in the sample forests. This is because, even if sample forest was degraded, users might have easily gone to other forests in nearby locations, if the ratio of forest area in the VDC was high. Thus, we expect that a higher forest ratio leads to a lower incidence of user group management. We used the ratio of households participating in other social activities (i.e. management of irrigation, drinking water systems, and local schools) to capture the effects of possible linkages of forest management with other social activities. Finally, we included traveling time to a ranger office as a variable representing the effects of the DFO. If the formation of user group management is truly voluntary, this variable will be insignificant. To the extent that user groups were set up with the expectation of receiving official user rights or with the direct assistance from the DFO, this variable would have a negative effect on the formation of user groups. The estimation results of probit and ordered-probit regressions are shown in the first two columns and the last column of Table 11.5, respectively. It should be pointed out that the estimated coefficients in the probit estimation are converted into marginal effects computed at the means of regressors, while those in the ordered-probit estimation are not. Thus, simple comparison of the estimation results by probit and ordered-probit regressions does not make any sense. The marginal effects of regressors on predicted probabilities of ordered-probit estimation are summarized in Table 11.6. It is found that the number of user households has a weakly significant effect, if we apply two-tailed t-test, whereas forest area per user household is insignificant. It appears that the size of user group per se, measured by the number of households and the forest area, does not decisively affect the cost of organizing the collective action and enforcing management rules. It may well be that both scale advantages and disadvantages coexist in forest management.248 The number of wards has negative and significant coefficients, suggesting that the larger the administrative heterogeneity, the larger will be the

248

Suspecting that scale economies dominate in the range of small-scale forest management operations, whereas scale diseconomies dominate in the range of large operations, we estimated regression functions which added square terms of either the number of forest users or forest areas. The estimated coefficients, however, are not significant.

310

KEIJIRO OTSUKA AND TOWA TACHIBANA

Table 11.5 Determinants of the Emergence of Forest User Group Management: Probit and Ordered-Probit Regressionsa Probitb Intercept 0.235 (0.482) Ln no. of households 0.154* (1.708) Ln forest area per house- 0.050 (0.689) hold No. of wards −0.100** (2.181) Ratio of households in −0.096 (0.522) other community activities Proportion of largest eth- 0.041 (0.135) nic group Traveling hours to ranger −0.030** (2.152) office Crown cover in 1978 −0.129** (2.204) Ratio of forest land in −0.209 (0.550) VDC in 1978 Log likelihood −34.95 Percentage of correct pre- 73.0 diction

b c

Probitb 0.209 (0.428) 0.146 (1.636) 0.045 (.643)

Ordered-probitc −0.178 (0.132) 0.453 (1.650) −0.011 (0.043)

−.100** (2.166)

−0.244* (1.748)

0.019 (.063)

0.337 (.391)

−0.029** (2.100)

−0.123*** (3.637)

−0.123** (2.146) −0.231 (0.609)

−.193 (1.166) −0.048 (0.040)

−35.08 73.0

−68.71 54.0

Notes: a The numbers in parentheses are t-statistics. ***indicates significance at the 1 percent level, **at the 5 percent level, and *at the 10 percent level. Dependent variable is unity where either informal or formal management has been introduced. Ranked choices of formal (=2), informal (=1), and no community management (=0) are considered as dependent variables.

Table 11.6 Marginal Effects for Ordered Probit Regression No management Intercept 0.055 Ln no. of households −0.141 Ln forest area per house- 0.003 hold No. of wards 0.076 Proportion of largest eth- −0.105 nic group Traveling hours to ranger 0.038 office Crown cover in 1978 0.060 Ratio of forest land in 0.015 VDC in 1978

Informal user management 0.014 −0.036 0.001

Formal user management −0.069 0.177 −0.004

0.019 −0.027

−0.096 0.132

0.010

−0.048

0.015 0.004

−0.075 −0.019

COMMUNITY FOREST MANAGEMENT IN NEPAL

311

cost of organizing collective action.249 On the other hand, the ratio of households participating in other social activities and the proportion of the largest ethnic group are insignificant, indicating that the presence of other community activities and ethnic homogeneity in terms of dominant caste do not directly affect the initiation of forest management. The former result was expected because, as was noted above, the sphere of community cooperation for such social activities as irrigation and school management is usually different from that for forest management. Traveling hours to a ranger office is significant in the probit regression and highly significant in the ordered-probit regression. The former finding indicates that the nearer the ranger office to the forest, the more likely it was that user group management was initiated. Furthermore, the higher significance in the ordered-probit regression suggests that proximity to a ranger office was particularly conducive to the initiation of formal user group management. In fact, Table 11.6 shows that the marginal effect of traveling hours to a ranger office is small for the formation of an informal group, but negative and comparably larger for the conversion from an informal to a formal user group. It seems clear that the influence of the DFO on the formation of formal user groups has been pervasive. Crown cover in 1978 has negative and significant coefficients in the probit regression, implying that forest user group management was more likely to be initiated in forests which were more degraded in 1978. This supports our hypothesis that deforestation leads to the formation of forest user groups. This is also consistent with a finding of Fujita et al. (1999) that collective irrigation management in the Philippines is more likely to be organized effectively where water supply is scarce. This variable, however, is insignificant in the ordered-probit regression. This insignificance and the negative sign of the marginal effect for formal user group management seem to suggest that crown cover in 1978 did not significantly affect the transformation of informal user groups into formal groups. The ratio of forest area in a VDC in 1978 is not significant. It appears that it was not merely the ratio of the forest area but the availability of forest resources that affected the formation of user group management.

11.5 Consequences of User Group Management There are several difficulties in properly assessing the effects of forest user group management on forest conditions. First of all, the representative

249

It must be pointed out that the number of wards may partly capture the effects of forest size and geographical conditions. If the forest is large or lies on the top of a hill and ridges, usually multiple numbers of wards have access to it. For the effective protection and management of such forest, it is necessary to organize all the wards around it, since it is difficult to protect the forest from one or a few sides.

312

KEIJIRO OTSUKA AND TOWA TACHIBANA

indicators of forest conditions (timber volume, biomass, etc.) are stock variables, which change slowly over time. Considering that user group management was initiated relatively recently, we believe that the rate of regeneration of young trees is a more appropriate indicator of the improvement of forest conditions. Second, management rules are highly complex: there are a variety of management rules from no regulation to complete prohibition of resource extraction and they vary from one resource to another. It is therefore very difficult to construct a single indicator of regulation intensity. Third, management rules have often changed over time. Ideally we have to assess the effect of each management rule with due consideration of the periods during which such rule was implemented. Finally, there are potential estimation problems associated with the endogen-ous nature of management rules and institutions. As we have seen in the previous section, the initiation of user group management was likely to be induced by a scarcity of tree resources associated with deforestation. To the extent that degraded forest conditions deter the regeneration of trees and foster the formation of user groups, the impact of user group management on the regeneration of trees may appear negative in the regression analysis, even if it actually contributed to the rehabilitation of forests. In order to avoid this, we have to control for the effects of the forest conditions properly. The forest conditions, however, are also endogenous. In actual estimation, we used the ratio of forest area in a VDC in 1992–96 as a variable representing the pressure on forest conditions in the neighborhood of our sample community forest. The other estimation problem caused by the problem of endogeneity is that there is a possibility that the more effective informal user groups tend to become formal user groups. In that case, even if formal user group management is found to be more effective than informal user group management, it may be due to a self-selection effect rather than the effect of receiving legal use rights from the government with the support from the DFO for community management. Since DFOs do not accept the applications for handing over if the users have any conflicts, this self-selection problem is likely to be pervasive. In this study, as a first step toward coping with the complexities of management rules, we used dummy variables for informal and formal management in regression analyses. As was indicated in the previous section, the current management system reflects the effects of the intervention of the DFO, forest conditions, and other factors. In the estimation of the effects of management systems on forest conditions, we include the proportion of forest area in the VDC in 1992–96, the logarithm of average altitude, and pine forest dummy, in addition to dummy variables for the management system. The proportion of forest area in the VDC in 1992–96 is supposed to capture the external pressure on the sample forests. If there are only a few forests in the neighborhood, many outsiders are likely to invade the forest under investigation. The latter two variables are included to control for the effects

313

COMMUNITY FOREST MANAGEMENT IN NEPAL

of the ecological conditions; the regeneration of trees becomes less active in the higher altitude, and the presence of pine trees deters regeneration. In the first two columns of Table 11.7, we report the results of ordinary least squares (OLS) and weighted least squares (WLS) estimation, in which the dependent variable is the logarithm of the number of regenerated trees per hectare. The number of regenerated trees is a weighted average with weights in proportion to the height of regenerated trees. In the weighted least squares regression, we used the number of sample plots in the forest inventory as weights. Since R2 is not a proper indicator of the goodness of fit in weighted regressions, we report the correlation coefficients between the actual and estimated dependent variables in the table. The fits of these regressions are not very high, which is likely to reflect the measurement errors of the regeneration rate of young trees. Except for the proportion of forest area, however, all the independent variables have the expected signs. The two ecological variables have generally significant effects on regeneration. The most important finding is that the coefficients of the formal forest-user group (FFU) dummy are positive and highly significant, while those of the informal forest-user group (IFU) dummy are positive but only marginally significant according to a one-tailed ttest. Furthermore, the former coefficients are numerically larger than the latter Table 11.7 Determinants of Regeneration Per Hectare in 1997a Model I OLS Constant 12.809 (7.156) IFU dummy 0.264 (1.191) FFU dummy 0.571*** (2.746) Proportion of forest −0.484 (0.916) area in VDC, 1992/ 96 Ln average altitude −0.640*** (2.534) Pine forest dummy −0.364** (2.131) R2 0.323 Correlation coeffi0.571 cients

b c

WLSc 10.772 (6.249) 0.360* (1.657) 0.651*** (3.289) −0.339 (0.691)

Model IIb OLS 11.909 (5.590) 0.428 (1.357) 0.839** (2.003) −0.471 (0.918)

WLSc 9.675 (4.549) 0.545** (1.784) 0.936*** (2.395) −0.409 (0.845)

−0.374* (1.534) −0.356** (2.166) 0.301 0.559

−0.538** (1.905) −0.357** (2.149) 0.310 0.562

−0.241 (0.844) −0.347** (2.162) 0.280 0.534

Notes: a Dependent variable is Ln (regeneration per hectare). The numbers in parentheses are t-statistics. ***indicates significance at the 1 percent level, ** at the 5 percent level, and * at the 10 percent level. Predicted values are used for formal forest user (FFU) dummy. Weighted least squares regression with weight being the number of sample plots in forest measurement.

314

KEIJIRO OTSUKA AND TOWA TACHIBANA

coefficients; the estimated coefficients of around 0.6 for formal user group management and around 0.3 for informal management indicate that the formal and informal management systems increases the rate of regeneration of young trees by approximately 80 percent and 35 percent, respectively.250 These results support the hypothesis that formal user group management is more efficient than informal management in improving forest conditions. These findings are consistent with the findings of Upadhyaya and Otsuka (1998): using the data of firewood collection from five user households in each of our sample forest, they find that the regulation of cutting green firewood by formal user groups was significantly effective in reducing the firewood collection by user households. As was pointed out above, one may question that the significant effect of a formal forest-user group is due to selfselection of effective user groups. In order to check this self-selection bias of formal user groups, we have tried the two-step estimation known as the treatment effects model (Greene 1997: 981). Based on the assumption that selfselection occurs only for the formation of formal user groups, we used the probit estimation for formal management as a first step estimation. The explanatory variables are the same as those which appeared in Table 11.5. The predicted dummy variable for formal user group management is used as an instrument in the second-stage estimation for the rate of regeneration. The results are reported in the third and fourth columns of Table 11.7. The two-step estimation does not result in a much greater difference in the qualitative results compared with those from the simple OLS and WLS estimation. That is, the effect of the formal forest-user group dummy is positive and significant even after removing the possible self-selection bias.

11.6 Concluding Remarks This study attempted to explore the causes and the consequences of forest user group management based on case studies in the hill region of Nepal. Despite the difficulty in measuring the conditions of forests, their changes over time, and the size of user group members, we obtained some useful empirical evidence. First, we found that the formation of user groups was induced by a shortage of forest resources. This finding suggests that when forest resources are abundant, user group management is less likely to be practiced, which leads to deforestation and shortages of forest resources. The degraded forest conditions, in turn, stimulate the formation of user

250

Since a logarithm is taken for the dependent variable, exponential of estimated dummy coefficients must be taken to assess the quantitative impacts of the formal and informal management institutions.

COMMUNITY FOREST MANAGEMENT IN NEPAL

315

groups, because it is costly to sustain livelihoods and subsistence farming without a sufficient supply of forest resources in the hill region of Nepal. Second, there is no indication that the size of forest and the size of user group members hindered the initiation and the effectiveness of user group management. Unlike the usual presumption, the size of management unit does not seem to increase the transaction cost of user group management significantly. It may well be that not only scale disadvantages but also scale advantages exist in the protection, as a small number of people can enforce protection of large forest areas. Third and most importantly, we found that management of community forests by formal user groups seems to be more effective in regenerating trees than management by informal user groups, not to mention the case of no management. It seems clear that the formation of formal user groups, which are granted official use rights of forests and are supported by the local forestry department, is critically important for the effective management of community forests. In fact, our field surveys strongly indicate that handing-over can be an effective policy instrument, only if it is supplemented by the supporting activities of the local forestry department. In other words, community initiatives and government supporting activities are complements but not substitutes. Thus, community management will not function well without governmental support. Even with such support it will not function well either, if the forest resources are not scarce, unless truly strong support measures are provided. Herein lies the dilemma of forest management policy: in order to prevent deforestation as soon as it begins to appear, the government needs to allocate large efforts and resources to community forest management.

References Aoki, M. (2000) ‘Community Norms and Embeddedness: A Game-Theoretic Approach,’ in this volume, Chapter 4. Arnold, J. E. M. (2000) ‘Devolution of Control of Common Pool Resources to Local Communities: Experiences in Forestry,’ in de Janvry, A., Platteau, J.-P., and Sadoulet, E. (eds.), Access to Land, Rural Poverty, and Public Action, Oxford University Press, Oxford. Baland, J.-M. and Platteau, J.-P. (1996) Halting Degradation of Natural Resources: Is There a Role of Rural Communities? Clarendon Press, Oxford. Bardhan, P. K. (1993) ‘Symposium on Management of Local Commons, Journal of Economic Perspectives, 7: 87–92. —— (1999) ‘Water Community: An Empirical Analysis of Cooperation on Irrigation in South India,’ in this volume, Chapter 9. Bromley, D. W. (1992) ‘The Commons, Property, and Common-Property Regimes,’ in Bromley, D. W. (ed.), Making the Commons Work, ICS Press, San Francisco.

316

KEIJIRO OTSUKA AND TOWA TACHIBANA

Chhetri, R. B. and Randy, T. R. (1992) User Group Forestry in the Far-Western Region of Nepal, International Centre for Integrated Mountain Development, Kathmandu. Dasgupta, P. S. and Heal, G. M. (1979) Economic Theory and Exhaustible Resources, Cambridge University Press, Cambridge. Deacon, R. T. (1994) ‘Deforestation and the Rule of Law in a Cross-Section of Countries,’ Land Economics, 70: 414–30. Fisher, R. J. (1989) ‘Indigenous Systems of Common Property Forest Management in Nepal,’ EAPI Working Paper No. 18, Environment and Policy Institute, East–West Center, Hawaii. FAO (Food and Agriculture Organization of the United Nations) (1989) Community Forestry, Rome. Fox, J. (1993) ‘Forest Resources in a Nepali Village in 1980 and 1990: The Positive Influence of Population Growth,’ Mountain Research and Development, 18: 89–98. Fujita, M., Hayami, Y., and Kikuchi, M. (1999) ‘The Conditions of Farmer Participation in Irrigation Management: A Cross-Section Analysis for the Philippines,’ mimeo, International Rice Research Institute, Los Banos, Philippines. Gilmour, D. A. and Fisher, R. J. (1991) Villagers, Forests, and Foresters, Sahayogi Press, Kathmandu. —— King, G. C., and Hobley, M. (1989) ‘Management of Forests for Local Use in the Hills of Nepal: Part 1, Changing Forest Management Paradigm,’ Journal of World Forest Resource Management, 4: 93–110. Graner, E. (1997) The Political Ecology of Community Forestry in Nepal, Verlag fur Entwicklungspolitik, Saarbrucken. Greene, W. H. (1997) Econometric Analysis, 3rd edn., Prentice Hall, New Jersey. Hayami, Y. (1997) Development Economics: From the Poverty to the Wealth of Nations, Clarendon Press, Oxford. Kijima, Y., Sakurai, T., and Otsuka, K. (2000) ‘Iriaichi: Collective vs. Individualized Management of Community Forests in Post-War Japan,’ Economic Development and Cultural Change. Kumar, S. K., and Hotchkiss, D. (1988) ‘Consequences of Deforestation for Women's Time Allocation, Agricultural Production, and Nutrition in Hill Areas of Nepal,’ IFPRI Research Report No. 69, International Food Policy Research Institute, Washington, DC. McGranahan, G. (1991) ‘Fuelwood, Subsistence Foraging, and the Decline of Common Property,’ World Development, 19: 1275–87. McKean, M. A. (1992) ‘Management of Traditional Common Lands (Iriaichi) in Japan,’ in Bromley, D. W. (ed.), Making the Commons Work: Theory, Practice, and Policy, ICS Press, San Francisco. Messerschmidt, D. A. (1990) ‘Indigenous Environmental Management and Adaptation: An Introduction to Four Case Studies from Nepal,’ Mountain Research and Development, 10: 3–4. Metz, J. J. (1991) ‘A Reassessment of the Causes and Severity of Nepal's Environmental Crisis,’ World Development, 19: 805–20. Nepal, Ministry of Forests and Soil Conservation (1988) Master Plan for the Forestry Sector, Nepal, Kathmandu.

COMMUNITY FOREST MANAGEMENT IN NEPAL

317

Ostrom, E. (1990) Governing the Commons: The Evolution of Institutions for Collective Action, Cambridge University Press, Cambridge. —— and Gardner, R. (1993) ‘Coping with Asymmetries in the Commons: Self-Governing Irrigation System Can Work,’ Journal of Economic Perspectives, 7: 93–112. Pearce, D. G. (1992) ‘Repeated Games: Cooperation and Rationality,’ in Laffont, J.-J. (ed.), Advances in Economic Theory, Cambridge University Press, Cambridge. Rayamajhi, S. and Pokharel, R. (1998) ‘From Deforestation to Reforestation: Common-Property Forest Management in the Hill Region of Nepal,’ a report submitted to the Environment and Production Technology Division, International Food Policy Research Institute, Washington, DC. Runge, C. F. (1981) ‘Common Property Externalities: Isolation, Assurance, and Resource Depletion in a Traditional Grazing Context,’ American Journal of Agricultural Economics, 63: 595–606. Sakurai, T., Pokharel, R., Rayamajhi, S., and Otsuka, K. (1998) ‘Private, Collective, and Centralized Community Management: A Comparative Study of Timber Forest and Plantation Management in Inner Tarai of Nepal,’ a paper presented at the workshop on Land Tenure and the Management of Land and Trees, Tokyo Metropolitan University, Tokyo. Sethi, R. and Somanathan, E. (1996) ‘The Evolution of Social Norms in Common Property Resource Use,’ American Economic Review, 86: 766–88. Soussan, J., Shrestha, B. K., and Uprety, L. A. (1995) The Social Dynamics of Deforestation: A Case Study from Nepal, Parthenon Publishing Group, New York. Stevenson, G. G. (1991) Common Property Economics: A General Theory and Land Use Applications, Cambridge University Press, Cambridge. Upadhyaya, H. K. and Otsuka, K. (1998) ‘Community Forest Management in Hill Region of Nepal,’ a paper presented at the workshop on Land Tenure and the Management of Land and Trees, Tokyo Metropolitan University, Tokyo. Wade, R. (1988) Village Republics: Economic Conditions for Collective Action in South India, Cambridge University Press, Cambridge. World Bank (1996) World Bank Participation Source Book, World Bank, Washington, DC.

12 Liberal Reforms and Community Responses in Mexico Alain de Janvry, Céline Dutilly, Carlos Muñoz-Piña, and Elisabeth Sadoulet

12.1 From State-led to Community-Led Governance The Ejido Before the 1992 Reforms Rural communities have, in Mexico, an extraordinarily important, and yet largely unrecognized, role in influencing the way in which a huge endowment of natural resources is used and local public goods are delivered to a majority of the rural population. In contrast to Africa and northeast Asia (Hayami and Platteau 1997), these communities generally do not have their roots in tribal or ancestral village origins. They are instead mostly creations of the state, products of the political settlement that followed 10 years of peasant-led armed revolution. This settlement led to formation of the ejido communities as defined in the constitution of 1917. In some cases, land that had been expropriated from indigenous communities by expansion of the haciendas was restituted, giving a tribal basis to the ejido communities. But in the vast majority of cases, communities were created by pulling together in ejidos the former workers of expropriated haciendas or ad hoc groups of rural households demanding access to land. A dual form of property rights was thus created, each covering today approximately half of the Mexican territory. On the one hand, a sector of private farms with a ceiling on land ownership equivalent to 100 hectares of irrigated land (the ‘pequeñas propriedades’). On the other hand, a social sector, generically referred to as the ejido sector, that includes indigenous communities with restituted lands and new ejido communities created by the reform, with ejido communities outnumbering indigenous communities 15 to 1. In this ejido sector, all land is owned by the community. Land in ejidos is also accessed under a dual system of land rights: individual plots to which ejidatario households have access in usufruct; and common property resources (CPR) to which all ejidatarios have access as regulated by the community. To understand how the ejido community was organized to manage its affairs, it is important to relate this to the purposes of the 1917 reforms (Gordillo et al.1998). Because the land reform occurred as the outcome of a peasant-led revolution that left one million dead, creation of the ejido

REFORMS AND RESPONSES IN MEXICO

319

sector had both political and economic objectives that would eventually run at cross purposes. The political objective was to contain peasant demands within the institutionalized political process and to incorporate them as clients of the ruling political party. This was done by using corporatist organizations, such as the National Confederation of Peasants (CNC), itself a part of the ruling political party, to mediate the relation between peasants and the state. The state thus had a strong pro-active role in managing the political life of the community. The economic objective was to organize production more efficiently than in the hacienda, particularly for the delivery of staple foods to the urban sector while the private sector was specializing in high value crops. Land reform beneficiaries were, however, considered unable to manage their own economic affairs, and they were thus placed, like infants, under the tutelage of the state. This applied not only to the delivery of public goods in the community and the management of common property resources, but also to production in individual plots. On those, access to the market and to essential services such as credit and insurance was mediated by community leadership, itself under strong control of the state. While tutelage could have been a rapid transition toward independent management of community affairs, submission to state control was maintained over the years and it took until the liberal reforms of 1992 for this to happen. To achieve both political and economic objectives, the communities were thus placed under strong control of the state. Governance in ejidos was organized through three internal bodies that reflected this main mise of the state (Baitennmann 1998). 1.

2.

3.

The general assembly. This was the highest authority in the ejido, constituted of all official members (ejidatarios) in the ejido. Meetings were to be held not less than once a month, with decisions taken on a majority vote. Decisions concerned internal rules and regulations of the ejido, election of the members of the other two internal bodies, authorization and regulation of production and marketing decisions, resolution of disputes, and changes in ejido membership and land allocation. Whenever issues of access to land or production were to be discussed, a representative of the Secretariat of Agrarian Reform (SRA) had to be present. The executive board, composed of six elected members under the leadership of a president. While elected by the assembly, this president could exercise considerable influence over ejido affairs. Because of the importance of political clientelism in accessing public goods and services, the president of the executive board also had close links with the ruling political party. The oversight council, also with six elected members, in charge of providing checks and balances on the performance of the assembly and the executive board.

320

A. DE JANVRY ET AL.

Not only was the state thus importantly present in the internal management of community affairs, SRA regulations additionally severely constrained what communities and ejidatarios were allowed to do. For instance, ejidatarios had to work their lands directly as they were not allowed to hire wage labor. They could not rent the land or sell it. Absences from the ejido for more than two years led to loss of the right to land. Ejidatarios could not transfer their land plots to more than one heir to prevent atomization of the land.251 The state was also involved in ejido arbitration matters since family controversies and boundary conflicts with other ejidos or private farms had to be settled in administrative tribunals under the authority of the SRA. Access to credit was obtained from the state agricultural bank through the ejido, not individually, with all ejidatarios co-liable for the total amount of credit received. Credit was granted for specific crops with specific agronomic practices, leaving little leeway for individual initiatives. Marketing of staple crops was through the parastatal CONASUPO that made payments of ejidatario debts directly to the state agricultural bank. Parastatals were also mediating institutions for access to inputs and sale of specialized crops such as coffee, cocoa, sugar cane, and tropical fruits. All borrowers from the official bank were required to purchase crop insurance from a single government insurance agency. Mechanisms of evasion were of course plentiful, from lack of participation in assemblies, to corruption, and to the development of active secondary markets for basically all factors and services controlled by state intervention. For instance, land was commonly rented and sometimes sold through informal arrangements to other community members and sometimes to outsiders. Wage labor was hired to replace family members who had migrated. Assemblies were often held without a representative of the state and minutes falsified to claim that the official had been present. The names of ejidatarios who had left would continue to appear on the list of participants to assemblies so they would not lose their rights. Secondary markets and illegal practices were thus important in adapting the highly regulated management of community affairs and individual income strategies to the desires of community members. Efficiency and equity costs were of course enormous, but they allowed gains compared to strict implementation of the law. Thus, while land legally belonged to the community and its management was under the jurisdiction of the community, the state had strong control over community affairs. Community cooperation was a state regulated more than a voluntaristic act. In spite of secondary markets and corruption, the ejido sector had become a truncated peasant economy and peasant

251

This was not the case in indigenous communities where community membership by birth gave right of access to land to all members, including the many descendants of any original member.

REFORMS AND RESPONSES IN MEXICO

321

community in the grips of a model of political and social control (de Janvry et al.1997).

The Ejido Since the 1992 Reforms While the ejido sector showed strong production performance through the 1960s, principally based on extensive programs of public investment in large-scale irrigation projects, multiple state-imposed constraints on community and individual initiatives gradually brought production to stagnation and welfare to poverty. In addition, democratic opening eroded the ruling party's monopolistic control over the ejido and undermined the effectiveness of the ejido as an instrument of political control. The costs of economic stagnation and extensive public subsidies could no longer be justified by political gains for the ruling party. There were also strong pressures from urban and business interests to gain access to ejido lands through an open land market. In a general context of structural adjustment reforms pursuing economic liberalization, the constitution of 1917 was amended in 1992 to allow privatization of individual plots (with associated greater security of tenure and the possibility of using land as collateral for credit), freedom of managing individual and community land as seen fit, and the right for individuals or the whole community to associate with private interests outside the ejido in pursuit of business advantages. The reforms were expected to change drastically the structure of governance in the ejido, promoting community-led as opposed to state-led management. The role of the president of the executive board was to decline while democratic representation was to increase through greater power being attributed to the assembly. The president could only be elected for a three-year term without renewal and his/her duties were to become largely administrative. The assembly, however, also lost importance with individualization of decision making over land transactions, land use, marketing decisions, and access to supporting services such as technical assistance and credit. For instance, after certification of land rights to individual ejidatarios under PROCEDE (the Program to Certify Ejidatario Rights and Land Plots), land transactions no longer need authorization from the assembly, but only signatures from two witnesses and notification to the National Land Registry (RAN). Disputes among ejidatarios can be settled directly with the Office of the Agrarian Attorney General (Procuradoria Agraria) without referral to the assembly. And assemblies are only required to meet once every six months instead of monthly. By severely diminishing restrictions on individual behavior and the role of the state in ejido affairs, the reforms can thus have a contradictory impact on the role of the community and the determinants of cooperation. Where cooperation is deemed worthwhile by ejidatarios and they have the ability to cooperate, the role of the community could be enhanced,

322

A. DE JANVRY ET AL.

evolving from mandatory cooperation to voluntary (endogenous) cooperation. For the delivery of public goods, collective arrangements in support of production (e.g. for accessing credit and technical assistance or reducing transactions costs in participating to markets), and the management of common property resources, cooperation may thus become more effective. By contrast, where cooperation is deemed a cost more than a benefit, or where structural conditions are unfavorable to cooperation, the role of the community may instead wane. The reforms in themselves thus do not guarantee a greater cooperative outcome. Laws codifying the relative roles of the assembly and the executive board are unenforceable by the SRA and unable to curtail the power of local bosses per se. Laws codifying decision making processes within the ejido are also inadequate to guarantee democratic behavior. For instance, secret ballots are not required and a show of hands is still the norm, facilitating coercion by ejido bosses. Under the reforms, ejidatarios have the options of: (1) obtaining certificates of land ownership that allow them to rent land and sell it internally to the community. These certificates are granted by the PROCEDE program and participation of an ejido to this program requires majority approval in the assembly; (2) obtaining full titles over their certified land plots, allowing them to additionally sell land to outsiders, full titling is costly and requires a two-thirds majority approval in the assembly; and (3) dissolving the ejido as a community organization with economic, political, and social purposes. In the process of certification, ejidatarios also have the power to appropriate common property resources in individual ownership if this is ratified by the assembly, unless these lands are designated as protected for environmental purposes. Appropriating common property resources in individual ownership allows members to expand their individual land plots and/or to incorporate new members into the ejido. If the ejido were dissolved, common property resources would return to the state. Observations show that certification has progressed rapidly, with some 75 percent of the ejidos either completed or in progress. Titling has, by contrast, been practically nonexistent, indicating that ejidatarios are concerned in preserving the ejido community, at the cost of restricting land transactions with outsiders. Based on case studies, Goldring (1998) observes that ejidatarios value the ejido as an organization for its support in production, access to common property resources, as a source of social and political capital, and to give them a sense of community identity that they feel is worth preserving. Hence, the issue of the role of the community and of the ability of the community to cooperate in managing its resources and its affairs is not made mute by the reforms. There is, however, no longer any guarantee that the level of cooperation achieved will be, at a minimum, as mandated by the state. Cooperation is now endogenous to community determination. In some cases, cooperation may thus wane and in others it may be consolidated by the need and opportunity of assuming

REFORMS AND RESPONSES IN MEXICO

323

leadership over community affairs without state tutelage. It is this issue of community response to economic and political liberalization, and in particular of adjusting cooperative practices and of preserving or dividing common property resources, that we address in this chapter.

12.2 The Ejido Sector Today Today, the ejido sector (including both new ejidos created by the land reform and lands restituted to indigenous communities) is composed of 28,058 communities with 3.5 million ejidatario households, 18 million individuals, and some 70 percent of the rural population. It is consequently a very important sector in terms of population, control over natural resources, share of agricultural production, and also incidence of poverty. The data we use here to characterize the ejido are based on two nation wide surveys of the ejido sector. The 1994 survey was done by the SRA and the University of California at Berkeley (see de Janvry et al.1997); the 1997 survey was done by the SRA and the World Bank (see Cord et al.1998). The surveys are panels of ejidos and ejidatario households that include 286 ejidos and 1,634 ejidatario households. The 1994 survey occurred before the PROCEDE program was initiated and largely before the communities had adjusted management practices, while the 1997 survey was done after much adjustment had already happened. We describe the ejido in terms of importance of individual plots and CPR, quality of governance, provision of public goods, use of CPR, and quality of cooperation using the 1997 data. Table 12.1 shows that CPR are important since they account for 69.9 percent of the total land in ejidos, while individual plots in usufruct account for 28.4 percent. Not all ejidos have CPR: as can be seen in Table 12.2, 21 percent do not have any. Among the 79 percent that do, individual plots per member cover 11.3 hectares and CPR per member 35.2 hectares (Table 12.1). For some ejidos with extensive grazing and forest areas, CPR cover the majority of the territory: thus 47.2 percent of the ejidos have more than half their total territory in CPR (Table 12.2). Table 12.3 shows that the two most important uses of CPR are pastures and forests: 77.6 percent of all ejidos have CPR pastures that account for 51.6 percent of all ejido land; 50.2 percent of all ejidos have forested land and it covers 36.5 percent of all ejido land. Land in CPR is also used for individual cultivation of crops since these lands can be allocated by the Assembly to individual households, including non-ejidatario residents of the ejido. Table 12.3 shows that 25.6 percent of the ejidos have some land used in this fashion, but this only occupies 5.6 percent of total ejido land. Finally, 5.8 percent of the ejidos have collective farming activities in CPR,

324

A. DE JANVRY ET AL.

Table 12.1 Access to Land in the Ejido, 1997 Forms of tenure

Land distribution in the sample % of the land

286

Total area per ejido (ha) 965

Land in ejidos that have the corresponding form of tenure Number of % of the Land per member (ha) ejidos land

28.4

252

58.0

286

2,380

69.9

226

76.9

286 286 286

46 12 3,403

1.4 0.4 100.0

274

3.3

Number of ejidos Individual plots in usufruct Common property resources (CPR) Urban zones Other Total

Individual CPR

11.3 35.2

325

REFORMS AND RESPONSES IN MEXICO

Table 12.2 Importance of Common Property Resources (CPR) Across Ejidos, 1997 Share of land in CPR None 0–10% 10–50% 50–90% > 90% Total

Number of ejidos 60 34 57 71 64 286

% in the sample 21.0 11.9 19.9 24.8 22.4 100.0

Table 12.3 Use of CPR Land in Ejidos, 1997 Types of use CPR pastures CPR forests Individual cultivation of CPR land Collective cultivation of CPR land Other uses Total CPR land

Number of ejidos 173 112 57

Share of ejidos (%) 77.6 50.2 25.6

Share of land (%) 51.6 36.5 5.6

Area (ha) 1993 2201 658

13

5.8

0.3

146

42 223

18.8 —

6.6 100

1060 3916

usually special women and youth projects, or collective projects to generate resources in support of social activities. They occupy a minute 0.3 percent of the land. Use of CPR lands for individual cultivation, and eventual appropriation of CPR land in individual ownership under PROCEDE, are importantly affected by population pressure over the land. This pressure originates in the fact that ejidatarios were not legally allowed to divide their plots among heirs. The child or spouse who inherited the usufruct right to the plot of land thus became the replacement ejidatario. Children other than the one who officially inherited the land either had to migrate or stayed in the ejido as posesionarios (individuals with access to a plot of land in the ejido, but no right to vote in the Assembly or to use the CPR) or as avencidados (individuals who live in the urban area of the ejido where they may have usufruct or ownership of an urban lot but no access to ejido land and no voting rights). Since more than half of the ejidos were created before

326

A. DE JANVRY ET AL.

1940, non-ejidatario households that reside in the ejido have become numerous in spite of occasional permission granted by SRA to incorporate new members. These two categories of non-members eventually represent large groups, usually kinship related to ejidatarios, with a great desire to gain access to land in the ejido and to official rights to participate in ejido affairs. Overall, landless avecindados are very important, representing 29.7 percent of the ejido population, and casually landed posesionarios represent 10.5 percent of the ejido population, while ejidatarios as a consequence only represent 59.9 percent (Table 12.4). Hence, population pressure to access land and become official ejido members is high. In ejidos with avecindados, these households match ejidatarios one to one. In ejidos with posesionarios, there are four posesionario households for every ten ejidatarios. The quality of governance and cooperation in managing ejido affairs are characterized in Table 12.5. One can see that quality of governance and cooperation are, overall, weak: 46 percent of the ejidos do not have internal rules, 27 percent of the ejidos fail to achieve a quorum at the first call of an assembly, and 35 percent of the ejidatarios do not participate in assemblies. Cooperation applies to the provision of services to maintain or improve CPR, to the extraction of resources in grazing and forestry, and to the provision of local public goods. Here again, we see that rules to increase provision and limit extraction are uneven and weak overall. It is only in the provision of local public goods that significant percentages of participation Table 12.4 Population Pressure for Access to Land in the Ejido Categories of households Number of households All ejidos (286 ejidos) Ejidatarios 103 Posesionarios 18 Avecindados 51 Posesionarios/ejidatario Avecindados/ejidatario Ejidos with posesionarios (130 ejidos) Posesionarios 38 Posesionarios/ejidatario Area cultivated per posesionario (ha) Ejidos with avecindados (189 ejidos) Avecindados 76 Avecindados/ejidatario

Distribution of households Other statistics 59.9 10.5 29.7 0.2 0.5

0.4 5.0

1.0

REFORMS AND RESPONSES IN MEXICO

Table 12.5 Indicators of Cooperation at the Ejido Level, 1997 Percentage 1 Symptoms of quality of governance and cooperation in ejidos (286 ejido-level observations) Ejido has internal rules (yes/no) Rules have been actualized as required by law (yes/no) Rules are registered with government (yes/no) Quorum is achieved at first call of assembly (yes/no) Percentage of ejidatarios who come to assembly Ejido keeps registry of agreements (yes/no) 2 Cooperation in pasture management (159 ejidos with common property pastures) Existence of rules on provision Ejidatarios who use pastures must pay fees (yes/no) Ejidatarios must participate to repair of fences (yes/no) Ejidatarios must participate in weed control (yes/no) Ejidatarios must participate in surveillance of ejido borders (yes/no) Existence of rules on extraction Animals are rotated among fields (yes/no) Ejido has rules limiting the number of animals per member (yes/no) Ejido has problems of intrusion of outsiders into pastures (yes/no) Symptoms of quality of cooperation Pastures have had an increase in the presence of weeds since 1994 (yes/ no) Pastures have had an increase in erosion gullies since 1994 (yes/no) 3 Cooperation in forest management (106 ejidos with common property forests) Existence of rules on provision Ejido had reforestation projects in the last three years (yes/no) Existence of rules on extraction Ejido imposes individual limits on timber extraction by ejidatarios (yes/ no) Ejido imposes fines or punishments if limits are not respected (yes/no) Ejido allows avecindados and posesionarios to extract timber (yes/no) Symptoms of quality cooperation Percentage of ejidatarios who respect limits on timber extraction (%) Illegal extraction is observed (yes/no) 4 Cooperation in the provision of public goods: maintenance or construction of infrastructure (1,634 individual ejidatario household observations) Symptoms of cooperation: participation in maintenance activity when activity exists (yes/no) Roads Irrigation infrastructure Public lighting School Health center Social center Animal corrals Other activities Total

54 76 68 73 65 85

15 42 14 23 12 13 23 26 27

36 31 70 58 85 14

63 24 7 57 32 25 8 92 79

327

328

A. DE JANVRY ET AL.

are observed. For instance, 63 percent of the ejidatarios participate in road repair activities and 57 percent in school maintenance. However, even vital public services, such as the maintenance of irrigation infrastructure, are weakly attended.

12.3 Liberal Reforms and Changes in Community Cooperation With the advent of the reforms, the determinants of cooperation should become increasingly endogenous to the community, i.e. explained by the characteristics of the community and its environment. However, there was plenty of opportunity to bend the rules of state-mandated cooperation before the reforms. Peasant communities are indeed notable for foot dragging (Scott 1976) and we have seen that there was extensive bending of the rules and failure to comply with state-defined cooperative practices. Hence, a reasonable amount of adjustment of cooperation to community and contextual characteristics may have already occurred before the reforms. In this section, we assess what are the determinants of cooperation across communities, and how much of a change in cooperation can be attributed to the reforms, with some ejidos increasing cooperation in response to endogenous incentives while others decreased cooperation as state impositions were removed. There are several symptoms of cooperation that we observe in the two periods: the frequency of assembly meetings, whether a quorum is achieved when an assembly is first called, the percentage of members assisting assemblies, and whether the ejido has formal internal rules in compliance with

REFORMS AND RESPONSES IN MEXICO

329

SRA requirements. From the standpoint of endogenous cooperative behavior, the most meaningful indicator is the percentage of ejidatarios that participate in assemblies since this is a clearly voluntaristic act as opposed to an administrative decision (number of meetings, internal rules). Data on assistance to assemblies suggest considerable heterogeneity in levels of cooperation across ejidos. In 1997, 23 percent of the ejidos had less than 50 percent of their members attending assemblies, 45 percent between 50 percent and 80 percent attendance, and 32 percent above 80 percent attendance. Assistance to assemblies, remained essentially constant over the period, with an average of 66 percent in 1994 and 69 percent in 1997. However, there were many changes across ejidos: • • •

37 percent of the ejidos kept roughly the same attendance (a change of ±10 percent in participation); 29 percent of the ejidos increased attendance (more than 10 percent increase in participation); 34 percent of the ejidos decreased attendance (more than 10 percent decrease in participation).

These data thus indicate that there are different degrees of cooperation across communities. How can this be explained? In the literature (Ostrom 1992; Baland and Platteau 1996), cooperation has been looked at as a binary matter: either the expected gains from cooperating exceed the fixed costs of cooperation, and cooperation occurs; or expected gains do not cover costs and there is cooperation failure. If, instead of binary levels of cooperation, we observe a continuum in degrees of cooperation, this suggests that other explanations are necessary. A possible explanation is as follows. For an ejido community, the expected gains from cooperation are a function of the quantity and quality of the resource to be collectively used. The community thus makes an effort to induce members to cooperate, including attendance at assembly meetings, that is proportional to the quantity and quality of the resources it controls. The level of this effort depends on the expected gains from cooperation and on the costs of enforcing cooperation which are a function of a set of characteristics of the ejido. For an individual community member, the cost of cheating on attendance is the disutility created by failing to attend meetings that results in ostracization of the member and loss of social capital in the community, with the attendant negative consequences for access to mutual insurance, information, employment, etc. through other community members. This disutility varies across individual members. Good citizens and poorer members take it very seriously, while opportunists and richer members do not attach as much importance to it. This schedule is represented by UU in Fig. 12.1, with community members ranked by decreasing order of the importance they attach to community pressure to conform. For a community

330

A. DE JANVRY ET AL.

Fig. 12.1 Equilibrium Attendance At Assemblies

member, the cost of participating in the assemblies is the opportunity cost of time measured in utility. Across community members, this disutility schedule is represented by the downward sloping curve cc in Fig. 12.1. Even if the cost of attending meetings is the same for all members, the disutility of this cost should decline with the ranking of individuals according to the decreasing level of importance they attach to community pressure to conform if both are associated with wealth. Intersection of the UU and cc schedules, if it exists, implies that there is an equilibrium percentage of participants n to assembly meetings. This equilibrium percentage increases with the expected benefits from cooperation and declines with the costs of enforcing cooperation. This gives us the following econometric specification of an equation explaining degree of participation at assembly meetings across ejidos: Endogenous variable: percentage of members participating in assembly meetings. Exogenous variables: •

Expected benefits from cooperation Quantity of the resource: hectares of CPR per ejidatario. Effective price of resources extracted=transactions costs to market=distance to the nearest population center. • Variable costs (difficulties) of enforcing cooperation – Group size: number of ejidatarios, number of posesionarios and avecindados, group heterogeneity (disparity in plot sizes). – Availability of exit options: size of migration network. – –

REFORMS AND RESPONSES IN MEXICO

– – – –

331

Social cohesion: ejido with an indigenous majority, age of the ejido. Degree of control over the resource: existence of external border conflicts. Facilitators of enforcement: ejido has collective sources of income. History of cooperative behavior: past symptoms of cooperation such as frequency of assembly meetings, frequency of achieving a quorum on first call of an assembly, assistance to assemblies, and existence of internal rules.

Results in Table 12.6 show that cooperation in 1994 was already quite well explained by the expected endogenous determinants of cooperation. Expected benefits from cooperation matter: both market integration (the negative of distance to market) and the quantity of CPR per ejidatario increase participation to assemblies. The variable costs of enforcing cooperation reduce participation: group size (the number of ejidatarios) creates a disincentive to participate. The number of non-members in the community (avecindados) also plays negatively as it undermines the quality of cooperation by increasing the number of potential cheaters. Group heterogeneity, measured by disparity in plot size, creates higher costs of monitoring and reduces participation. Social cohesion under the form of an homogenous ethnic population helps cooperation. And lack of control over external borders, that is whether the ejido is subjected to open access abuses, deters the effectiveness of cooperation and reduces incentives to participate at meetings. Three years is only a short time span for behavior to adjust to the new mode of cooperation implied by devolution. Hence, the levels of cooperation observed in ejidos cannot be expected to have reached a new equilibrium, making the levels of cooperation observed in 1997 difficult to explain. There are several variables that no longer affect cooperation. The role of expected benefits from cooperation was undermined in 1997 by a severe crisis of price incentives for the major staples produced by the ejido (Cord et al.1998). Consequently, the differential levels of cooperation across ejidos should be found more on the cost side than on the side of expected benefits. The number of non-members in the ejido (avecindados) and existence of external border conflicts have been resolved by the reforms, making them less important as determinants of cooperation. By contrast, the total number of ejidatarios, now including the new members incorporated under the reforms (see next section), still plays a negative role. The size of the communities for effective cooperation thus remains a major issue. The exit option, represented by the size of the migration network to which members of an ejido have access (the percentage of adults with past migration experience in the ejido), is increasingly available to ejidatarios since the reforms offered greater freedoms to rent land and to allocate labor as pleased. As a consequence, we see that greater exit options

332

A. DE JANVRY ET AL.

Table 12.6 Determinants of Cooperation in 1994 and 1997: Assistance to Assemblies (Robust Ordinary Least Squares) 1994 Coefficient Expected benefits from cooperation Market integration −0.002 (distance to nearest population center) Quantity of resource: 0.0004 ha CPR/ejidatario Variable costs of enforcing cooperation Group characteristics Group size: Number −0.001 of ejidatarios Group size: Number −0.00008 of avecindados Group heterogeneity : −0.008 Plot size disparitya Exit option (size of −0.42 migration networkb) Social cohesion Ejido has indigenous 0.097 majorityc+ Age of the ejido (years −0.001 since foundation) Ejido has external −0.057 border conflicts+ Facilitators of cooperation Ejido has sources of 0.013 incomed+ Goodness-of-fit: R2 0.32 Number of observa- 179 tionse

a b c d e

t-stat

1997 Coefficient

t-stat

−2.71

0.00009

0.12

1.90

0.0003

0.98

−5.28

−0.001

−6.65

−2.25

0.0002

1.10

−1.63

0.002

0.57

−1.09

−1.50

−3.62

2.17

0.067

1.44

−1.26

−0.001

−1.28

−1.88

0.0009

0.02

0.41

0.025

0.76

0.26 205

Notes:+ Dummy variable. Ratio of largest to smallest individual plot in the ejido. Migration network=percentage of adults with migration experience in the ejido. Indigenous=individual who speaks an indigenous language. Ejido earns income from common property resources. The number of observations differs in 1994 and 1997 due to missing data in 1994.

now undermine cooperation, a regularity rationalized by Hirschman (1970). Other variables such as ethnicity remain positively related to cooperation, but have dropped in significance. We must carefully refrain from making statements on the future role of these variables until a new cooperative equilibrium has been reached. We thus conclude that two powerful regularities

333

REFORMS AND RESPONSES IN MEXICO

Table 12.7 Effect of Group Size on the Percentage of Members Who Assist in Assemblies (Partial Results from Robust Ordinary Least Squares Regression) Community size: number of ejidatarios 1–50 50–75 more than 75

Cumulative effect Coefficient −0.06 −0.74 0.73

t-statistic −0.28 −1.81 3.34

Slope −0.06 −0.80 −0.07

t-statistic −0.28 −3.78 −3.45

Fig. 12.2 Cooperation and Community Size: Percentage Of Members Who Assist At Assemblies and Number Of Ejidatarios

dominate the effectiveness of cooperation under devolution: group size and exit options, both of which are detrimental to cooperation. More complete identification of the determinants of cooperation under the new rules will have to wait until a new equilibrium is reached. We can explore further the role of the size of the community in affecting the quality of cooperation in 1997 by introducing a spline in the relation between assistance to assemblies and the number of ejidatarios.252 Seeking by tatonnement the optimum breaking points in the number of ejidatarios on a best fit basis, we locate those at community sizes of 50 and 75 (Table 12.7). The results are represented in Fig. 12.2. We see that there is no impact of

252

A spline regression fits a line with discontinuities in the slope at specific breaking points. The slope between breaking points is the cumulative value of the regression coefficients in that and all previous intervals.

334

A. DE JANVRY ET AL.

group size on the quality of cooperation when the ejido has less than 50 members since the slope is not significantly different from zero. There is then a significant precipitous drop in the quality of cooperation as membership exceeds 50 until it reaches 75. In this range, the marginal effect of community size on assistance to assemblies is that for every 10 additional members, the percentage of members who assist in assemblies decreases by 8 percent. In communities with more than 75 members, the quality of cooperation continues to deteriorate as group size increases. The marginal impact of group size declines to a drop in assistance in assemblies of only 0.7 percent for every additional 10 members, but this effect is still very significantly different from zero. Seen from the angle of quality of cooperation, ejidos of 50 or less would thus be optimum. If there are economies of scale in the provision of public goods, 50 would be preferred to a smaller number. Beyond this, there exists a tradeoff between moral hazards (quality of cooperation) and the cost of delivery of public goods. Visits to ejidos in Michoacan (e.g. Paso del Muerto) led us to observe that 50 or less is indeed an effective size for organizing cooperation, with the optimum size affected by a number of idiosyncratic factors including the existence of scale economies for the delivery of particular public goods in specific ejidos. Since we have panel data, we can use the observed changes in cooperation between 1994 and 1997 to explain why cooperation improved in some ejidos while it declined in others. The results are presented in Table 12.8. Since nonzero changes in the explanatory variables over the period are endogenous, changes in cooperation are explained by the benefits and costs of cooperation in 1994, the specific purposes of cooperation during the transition period, and the ejidos' history of cooperative behavior as observed in 1994. Again, a full adjustment cannot be observed, but differences across ejidos are already notable. We see again that ejidos with a larger membership and with greater exit options are the ones where the intensity of cooperation declined. Cooperative behavior in the transition was used to resolve community problems. We thus see that the existence of external border conflicts induces a cooperative response now that ejidatarios have more control over community affairs. As we saw in Table 12.6, by 1997 the problem of external border conflicts had been solved. The past history of cooperation also helps improve cooperative behavior, a regularity rationalized by Seabright (1993). The negative effect on the percentage of members who assisted assemblies in 1994 represents a saturation effect: those ejidos that were already at high levels of participation necessarily increased participation less. However, those that had internal rules in 1994 under state-led cooperation show a greater increase in cooperation. Hence, there was learning to cooperate that helped sustain higher levels of cooperation when the community acquired more self-determination rights. We conclude that state-led cooperation was already importantly determined by endogenous behavior and that a new cooperative equilibrium

335

REFORMS AND RESPONSES IN MEXICO

Table 12.8 Determinants of Change in Cooperation Between 1994 and 1997: Change in Assistance at Assemblies (Robust Ordinary Least Squares)

Expected benefits from cooperation Market integration (distance to nearest population center) Quantity of resource: ha CPR/ejidatario Variable costs of enforcing cooperation Group characteristics Group size: Number of ejidatarios Group size: Number of avecindados Group heterogeneity : Plot size disparitya Exit option (size of migration networkb) Social cohesion Ejido has indigenous majorityc+ Age of the ejido (years since foundation) Ejido has external border conflicts + Facilitators of cooperation Ejido has sources of incomed+ Cooperation in 1994 % of members who assist at assemblies + Quorum achieved on first call + Frequency of assembly meetings + Ejido has internal rules + Goodness-of-fit: R2 Number of observations

a b c d

Change 1994–97 Coefficient

t-statistic

0.0009

1.50

0.000

0.01

−0.001 0.000 0.007 −1.22 0.03 −0.0007 0.052

−0.90 1.74

0.003 −0.53 −0.025 0.021 0.051 0.37 176

0.67 2.27 1.72

Notes: + Dummy variable. Ratio of largest to smallest individual plot in the ejido. Migration network=percentage of adults with migration experience in the ejido. Indigenous=individual who speaks an indigenous language. Ejido earns income from common property resources.

under the new rules has not yet been reached, making community-led cooperation difficult to explain. Nevertheless, where conditions are favorable to cooperation (small size, fewer exit options, and a history of cooperative behavior), the level of participation further increased with the liberalization of community behavior. Structural features of the communities, such as size of membership, play negatively, suggesting that many ejidos were created too large for successful endogenous cooperation. In our sample, 35 percent of the ejidos have more than 100 families and 14 percent more than 200. If, as we observed, 50 families is an optimum group size for effective cooperation, then only 34 percent of the ejidos qualify. If there are economies of scale in the provision of public goods and the management of CPR, the optimum size may be larger. Yet, if moral hazards

336

A. DE JANVRY ET AL.

dominate in determining the outcomes of cooperation, reducing the size of the larger ejidos would seem to be a necessity in order to adjust to the new rules of community behavior.

12.4 Preservation or Division of the Commons and the Role of Cooperation The 1992 reforms opened the possibility for ejidos to expand the area worked in individual plots by enclosing CPR lands, and to incorporate additional members, presumably from among the avecindados and posesionarios resident in the ejido, as full-rights ejidatarios. These possibilities have been exercised highly unequally across ejidos. We analyze in this section the determinants of these two decisions.

Determinants of the Decision to Appropriate CPR Resources The incentive for individual community members to press—or not—the assembly to enclose CPR land for individual use is based on a comparison of the profitability of using land individually relative to accessing land as a CPR. Land in plots is used for crops while land in CPR is used for pasture (say). Profitability of land in individual plots thus depends on the total factor productivity of resources in agriculture, on the relative price of agricultural vs. livestock products, and on the size of the initial land allocations in private plots. The lower the initial allocations, the higher the marginal productivity of land in individual plots, and the greater the incentive to enclose more land. Profitability of land in CPR depends importantly on the quality of cooperation in managing these lands (McCarthy et al.1998). If the quality of cooperation in the ejido is low, lands are overgrazed, and the profitability of animal husbandry is low. Hence, the lower the ability to cooperate, the greater the incentive to appropriate more land privately since the opportunity cost of land in CPR is low. Incentives to appropriate also depend on the quantity of CPR resources available per ejidatario, with greater availability lowering the marginal return from CPR land. Finally, appropriation can also be done for the purpose of settling new members as ejidatarios, usually family members who could not inherit land under the previous legislation. Hence, the greater the population pressure measured by the total ejido population per ejidatario, the greater the incentive to appropriate. Econometric analysis of the determinants of appropriating CPR land in individual plots is reported in Table 12.9. Appropriation, cooperation, individual plot size,

337

REFORMS AND RESPONSES IN MEXICO

Table 12.9 Appropriation of Common Property Land, 1994–1997 Number of observations

Probit appropriation of the commons=1

OLS share of CPR appropriated Median regression share of (only those who did) CPR appropriated (only those who did) 59 59 Coefficient P-value Coefficient P-value

161 Coefficient P-value Role of cooperation (1994) Ejido has for- −0.630 0.080 −0.059 mal rules for CPR use + −0.048 0.040 0.006 Number of meetings held per year Role of initial allocations in individual plots (1994) Hectares of −0.001 0.866 −0.000 plot land per ejidatario Role of CPR endowments (1994) Hectares of 0.007 0.020 −0.001 common land per ejidatario Ejido is com- 003.391 0.144 000.540 munal +1 0.157 −0.066 CPR land per −0.266 ejidatario (ejido communal) Role of population pressure (1994) Ejido popula- 0.006 0.316 0.004 tion per ejidatario Agroecological region 0.621 0.048 0.349 Humid or semi-humid tropical region + Arid or semi- 0.499 0.137 −0.080 arid region + Mills ratio 0.067 Prob > χ2 0.000 R2 or pseudo 0.19 0.45 R2 Correctly pre- 69 dicted observations (%)

Notes: 1 Communal ejidos hold all land in CPR. +

Dummy variable.

Tobit share of CPR appropriated (all sample) 161 Coefficient

P-value

0.704

−0.308

0.057

−0.385

0.066

0.511

0.005

0.633

−0.021

0.127

0.963

−0.003

0.393

−0.002

0.560

0.154

−0.001

0.000

0.001

0.628

0.191

000.220

0.567

001.627

0.041

0.134

−0.067

0.320

−0.144

0.042

0.049

0.003

0.450

0.006

0.069

0.014

0.375

0.086

0.611

0.002

0.528

−0.199

0.362

0.191

0.341

0.932 0.41

0.000 0.16

338

A. DE JANVRY ET AL.

and CPR endowments per ejidatario in 1997 are all simultaneous. Hence, we use the levels of cooperation, individual plot size, and CPR endowments in 1994 as explanatory variables of the decision to appropriate land during the 1994–97 period. The data are truncated since a number of ejidos did not appropriate CPR land. To analyze these data, we use several specifications of the decision to appropriate: a probit of the binary choice to appropriate or not, an OLS and a LAD (median regression) on the share of CPR land appropriated among the ejidos that did appropriate, and a Tobit on the share of CPR appropriated across the whole sample. Each estimator has specific advantages and disadvantages. The Heckman probit-OLS approach allows correction for selectivity bias, but has strong distributional assumptions and weak identification when the variables explaining the decisions to appropriate and how much to appropriate are the same. The LAD is distribution fee, consistent in the face of heteroskedasticity, and less sensitive to outliers than OLS because it minimizes the deviations around the median rather than the square of the deviations around the mean. It requires bootstrapping to calculate the distribution of estimated parameters and give less precise estimates. The Tobit makes strong distributional assumptions and also assumes that the determinants of the decision to divide and the quantity divided are the same. With no perfect estimator, results should consequently be read in terms of consistency across alternative specifications. Results show that ejidos that are able to cooperate in managing CPR have less incentive to divide: those with formal rules for CPR use in 1994 and with more meetings per year in 1994 divided less frequently between 1994 and 1997 (probit) and divided a smaller share of the CPR (LAD and Tobit). Cooperation failure was thus an incentive to appropriate the commons. Smaller initial allocations of private plots create incentives to compensate by appropriating CPR. PROCEDE thus offered the communities a historical opportunity to enclose CPR in compensation for the perceived insufficient size of individual plots. Enclosures were also more frequent (probit) when the ejido has greater endowments of CPR per ejidatario. A larger endowment, however, naturally implies that the percentage of the CPR that is appropriated declines very quickly with size (OLS and LAD), so that only in the very small ejidos did divisions occur. Both scarcity of land in individual plots and abundance of CPR land thus enter into the decision to appropriate land. Communal ejidos had no individual plots in 1994, in general because agriculture is practiced using slash-and-burn on plots that are itinerant on communal land. These ejidos thus maintain land in common. However, soil fertility in individually cultivated plots is a common property resource (see Lopez 1998), creating incentives to abuse extraction and hence pressures to divide if land is scarce. Pressure to divide is logically less among communal ejidos with larger CPR endowments per ejidatario since the commons problem on soil fertility is less severe. Communal ejidos apparently chose to divide (Tobit), but the share that is divided declines with

REFORMS AND RESPONSES IN MEXICO

339

size even more than for the other ejidos (in Table 12.9, the effect ‘CPR land per ejidatario if the ejido is communal’ is additive to the effect ‘hectares of common land per ejidatario’). Population pressure is another important incentive to appropriate CPR. In all cases, it increases the percentage of the land that is being appropriated. Finally, the agroecological context matters. In tropical areas, agriculture is generally practiced in slash-and-burn, creating a common property problem for individual plots. This increases the incentive to appropriate the land in individual tenures. Data in Table 12.10 indicate some of the consequences of appropriating the commons. Incorporation of new members and appropriation of the commons led to increased inequality. While inequality of individual plots remained the same in ejidos which did not divide CPR land (the ratio of the largest to the smallest plot changed from 7.2 to 8.5 ha), it increased sharply in the ejidos that did appropriate land, where it increased from 7.6 to 16.7 ha. Hence, this suggests that division of CPR lands was not used to compensate for inequalities. On the contrary, access to additional land apparently led to increasing disparities among members. Data in Table 12.8 also show that the incorporation of additional members had sharply contrasted implications for individual members whether there was division of CPR land or not. Division of CPR land allowed the incorporation of new members while maintaining nearly unchanged the size of individual plots, as the area in individual plots declined by only 2.9 percent. This, of course, was achieved at the cost of a sharp decline in the availability of CPR land per Table 12.10 Consequences of Individual Appropriation of CPR Land Ejidos which had CPR and Ejidos which had CPR and Test of difference did not divide divided partially or totally Impact on inequality of land in plots across ejidatarios: ratio of largest to smallest individual plot 1994 7.2 7.6 1997 8.5 16.7 ** Impact of incorporation of new members on access to land 1994a 14.9 17.3 a 1997 10.9 16.8 % change 94–97 −26.8 −2.9 b 1994 32.7 57.6 b 1997 21.6 31.6 % change 94–97 −33.9 −45.1 b

Notes: a individual plot land per member (ha). CPR land per member (ha).

340

A. DE JANVRY ET AL.

member that fell by 45 percent. By contrast, ejidos that incorporated new members without dividing the commons incurred a loss of 26.8 percent in land in plots per member. Common land per member also fell as the number of members increased, but by only 33.9 percent. Protecting the commons was thus done at the cost of less land under individual control per member.

Determinants of the Decision to Incorporate New Members The decisions to appropriate common land and to incorporate new members are jointly determined and consequently respond to the same determinants at the level of the reduced form. The determinants of incorporation are analyzed econometrically in Table 12.11. Incorporating new members implies giving them access to CPR, with negative externalities on current members. The higher the quality of cooperation, the better managed the CPR are, and hence the higher the opportunity cost of incorporating new members. For this reason, ejidos with higher quality cooperation are observed to incorporate fewer new members. Incorporation of new members is also more likely and represents a higher share of the existing membership if current members have higher individual land endowments and if the ejido has more common land. Larger CPR endowments make incorporation much easier since the opportunity cost on current members is lower. Communal ejidos did incorporate new members at an even higher pace. This is to be expected because community spirit is much higher in these ejidos. Population pressure plays, as expected, a role in the growth in membership (Tobit). Finally, older ejidos are more prone to incorporating additional members. The age of the ejido is positively correlated to population pressure since older ejidos have accumulated more descendants with demands for access to land on their own.

12.5 Conclusions The ejido communities have extraordinary importance in Mexico for their control over extensive natural resources, their high share of irrigated agricultural land, the large fraction of rural population which they contain, and the presence of high levels of poverty. These communities did not evolve endogenously over a long historical period as witnessed in Africa, Asia, and other parts of Latin America such as Guatemala and Peru, but were the product of the state as part of the settlement of a bloody peasant-led revolution at the beginning of the century. They were consequently crafted for political control and expected to perform cooperatively under strong control of the state. Following the debt crisis and structural adjustment, the reforms of 1992 broke this equilibrium, devolving to the communities, control over management of CPR and the delivery of local public goods, and allowing them to decide on individual appropriation or preservation of

341

REFORMS AND RESPONSES IN MEXICO

Table 12.11 Incorporation of New Members in the Ejido, 1994–1997 Probit Inc members = 1

OLS % inc. in members (only those who did) 67* Coefficient P > |z|

Median regression % inc. in members (only those who did) 67* Coefficient P > |z|

161* Coefficient

P > |z|

−0.675

0.066

−0.219

0.268

−0.50

0.080

−0.039

0.683

−0.013

0.281

−0.02

0.170

0.006

0.341

0.002

0.863

0.015

0.024

−0.091

0.903

0.449

0.531

1.287

0.072

0.002

0.196

0.002

0.432

0.005

0.004

−0.103

0.765

−0.122

0.777

0.622

0.005

0.000

0.856

0.001

0.898

−0.004

0.071

0.015

0.288

0.016

0.398

0.029

0.049

0.258

0.006

0.325

−0.000

0.977

0.100

0.066

0.210

−0.448

0.128

−0.159

0.574

0.045

0.858

0.679

−0.477

0.124

−0.179

0.428

−0.236

0.406

−0.131 0.17

0.803

161* Coefficient P > |z| Role of cooperation (1994) Ejido has for- −0.299 0.315 mal rules for CPR use + Number of −0.039 0.094 meetings held per year Role of initial allocations in individual plots (1994) Hectares of plot 0.019 0.061 land per ejidatario Average number 3.990 0.030 of intra-ejido land transactions 1992–97 Role of CPR endowments (1994) Hectares of 0.005 0.050 common property land per ejidatario Ejido is com0.939 0.010 munal + CPR land per −0.006 0.062 ejidatario if ejido is communal Role of population pressure (1994) Avecindado 0.017 0.333

households per ejidatario

Years since ejido 0.007 foundation Agroecological region Humid or semi- 0.349 humid tropical region + Arid or semi0.129 arid region + Mills ratio 2 R or pseu- 0.12 do R2 Correctly predicted observations (%)

67

Notes:+ Dummy variable. *

Number of observations.

0.12

Tobit % inc. members (all sample)

0.08

342

A. DE JANVRY ET AL.

the commons. We analyzed in this chapter how communities adjusted their behavior from state-led to community-led cooperation, and how they took advantage of the opportunity offered by the reforms to appropriate CPR as individual tenures and to incorporate new members with decision making rights. Using data from surveys of ejido communities before participation to the land reform program (1994) and after (1997), we observed highly heterogeneous responses to the reforms across communities, with some improving levels of cooperation while it declined in others, and with some appropriating CPR and incorporating new members while others preferred to preserve the status quo in one or both of these options. We observed, in particular, that failure of ejidos to achieve high levels of cooperative behavior in the management of CPR created incentives toward greater private appropriation of common land. Hence, in the context of the reforms that gave communities opportunities to appropriate CPR in individual tenures, the low prevailing levels of cooperation in a majority of ejidos have contributed to rapid appropriation. This was undoubtedly rational given the prevailing levels of cooperation. However, maintaining extensive grazing and forestry resources in common property can have advantages if there are economies of scale and opportunities for geographical risk spreading, as is generally the case in extensive grazing and forestry operations as found in the ejido sector (see, e.g. Wilson and Thompson 1993; and Nugent and Sanchez 1995). Preservation of these advantages of common property when they exist could be secured if communities were given proper assistance to enhance cooperation, particularly to adjust to the new policy context whereby cooperation must become communityled as opposed to state-led. We have seen that, given the historical context in which the land reform was designed, the ejido had not been organized to maximize the potential for successful cooperation in the management of community affairs, but principally as an instrument of political control and state intervention in resource use. Hence, careful thinking needs be given to recrafting the ejido as an institution to optimize cooperative behavior. We found, for instance, that size beyond a threshold of 50 members weakens endogenous cooperation, suggesting that most ejidos were created too large to sustain effective cooperation under the new rules. Given the importance of the natural and human resources located in the ejido sector, careful redesigning of the ejido communities for effective cooperation following devolution could thus have major productive, environmental, and welfare effects.

References Baitennmann, Helga (1998) ‘The Article 27 Reforms and the Promise of Local Democratization in Central Veracruz’, in Cornelius, Wayne and Myhre, David

REFORMS AND RESPONSES IN MEXICO

343

(eds.), The Transformation of Rural Mexico, Center for U.S.–Mexican Studies, University of California, San Diego. Baland, Jean-Marie and Platteau, Jean-Philippe (1996) Halting Degradation of Natural Resources: Is There a Role for Rural Communities? Oxford University Press, Oxford. Cord, Louise et al. (1998) Economic Adjustment and Institutional Reform: Mexico's Ejido Sector Responds, The World Bank, Latin America and Caribbean Region, Mexico Country Management Unit, Washington DC. de Janvry, Alain, Gordillo de Anda, Gustavo, and Sadoulet, Elisabeth (1997) Mexico's Second Agrarian Reform: Household and Community Responses, Center for U.S.–Mexico Studies, University of California, San Diego, La Jolla. Goldring, Luin (1998) ‘Having Your Cake and Eating It Too: Selective Appropriation of Ejido Reforms in Michoacán’, in Cornelius, Wayne and Myhre, David (eds.), The Transformation of Rural Mexico, Center for U.S.–Mexican Studies, University of California, San Diego. Gordillo, Gustavo, de Janvry, Alain, and Sadoulet, Elisabeth (1998) ‘Entre el control político y la eficiencia: evolución de los derechos de propriedad agraria en México’. CEPAL Review, 66: 149–66. Hayami, Yujiro and Platteau, Jean-Philippe (1997) ‘Resource Endowments and Agricultural Development: Africa vs. Asia’. CRED, University of Namur, Belgium. Hirschman, Albert (1970) Exit, Voice, and Loyalty, Harvard University Press, Cambridge, MA. Lopez, Ramón (1998) ‘The Tragedy of the Commons in Côte d'Ivoire Agriculture: Empirical Evidence and Implications for Evaluating Trade Policies’, The World Bank Economic Review, 12(1): 105–31. McCarthy, Nancy, de Janvry, Alain, and Sadoulet, Elisabeth (1998) ‘Land Allocation Under Dual Individual–Collective Use in Mexico’, Journal of Development Economics, 56(2): 239–64. Nugent, Jeffrey, and Sanchez, Nicolas (1995) ‘The Local Variability of Rainfall and Tribal Institutions: The Case of Sudan’, Economics Department, University of Southern California, Los Angeles, California. Ostrom, Elinor (1992) Governing the Commons: The Evolution of Institutions for Collective Action, Cambridge University Press, New York. Scott, James (1976) The Moral Economy of the Peasant: Rebellion and Subsistence in Southeast Asia, Yale University Press, New Haven, CT. Seabright, Paul (1993) ‘Managing the Local Commons: Theoretical Issues in Incentive Design’, Journal of Economic Perspectives, 7(Fall): 113–34. Wilson, Paul, and Thompson, Gary (1993) ‘Common Property and Uncertainty: Compensating Coalitions by Mexico's Pastoral Ejidatarios’, Economic Development and Cultural Change 41: 299–318.

13 Community Arrangements to Overcome Market Failures: Pooling Groups in Japanese 253

Jean-Philippe Platteau and Erika Seki

13.1 Introduction The agrarian society of Japan has shown a remarkable ability to evolve collective mechanisms to solve various problems arising from the absence of important markets or from pervasive market failures. These problems range from the production of local public goods to the overcoming of ecological and other externalities through the smoothing of income and consumption flows. Thus, there is impressive evidence about collective efforts aimed at constructing and maintaining irrigation systems as well as settling conflicts over the use of water at community level; about successful attempts to coordinate the use of common property resources such as forests and meadows; and about various informal insurance schemes often operated through hierarchized networks of village relationships (Smith 1959; Hayami and Kikuchi 1981; McKean 1986; Aoki 1998). It is not only in agricultural but also in fishing communities that the Japanese people have evinced considerable skill in devising and running collective mechanisms. In feudal times, for example, in order to limit the number of entrants into the local fishery and to ensure proper management of the resource, fishermen's associations or guilds came into existence and adopted strict rules of access and use. Membership in each guild was rigidly limited to persons born in the village and all members had to follow strictly the regulations of the guild, which established fishing zones, set seasonal limits and imposed restrictions on fishing gears and techniques (Ruddle 1987: 25). An institutionalized system of village or local sea

253

Thanks are due to Jan Willem Gunning (University of Amsterdam), Avner Greif (Stanford University), Jean Tirole and Georges Molins-Ysal (University of Toulouse), Tim Besley and Robin Burgess (London School of Economics), and François Maniquet (University of Namur) for their useful comments on a previous version of the paper or in the course of seminar presentations. Our greatest debt is towards Frederic Gaspart and Jean-Marie Baland (University of Namur), Masahiko Aoki (Stanford University) and Jonathan Morduch (Princeton University) who spent generous time discussing with us various points addressed in the paper. Also, we are greatly indebted to the Fisheries Cooperative Associations of Iwase and Shinminato for their continuous moral and logistical support. The field work would not have been possible without the financial assistance of CRED (University of Namur) and the Institute of Advanced Studies, United Nations University (Tokyo).

POOLING GROUPS IN JAPANESE FISHERIES

345

tenure was thus born in which the generally acknowledged sea territory of each coastal village was a seaward extension of its terrestrial territory. This system became codified at the beginning of the eighteenth century, at least in Okinawa and other main islands of the country (Ruddle 1987: 13–14). Interestingly, the erstwhile fishermen's guilds eventually transformed themselves into the modern Fishery Cooperative Associations (FCAs) that came to form the backbone of the coastal fishery management system sanctioned by Japanese law from the mid-twentieth century. Under this comanagement system, a unique achievement in the present-day world, what the FCAs do is to implement and enforce national and prefectural legislations and regulations which they are free to complement by locally made rules and mechanisms as they deem appropriate in the light of evolving local conditions (Asada et al.1983; Ruddle 1987; Yamamoto and Short 1991). Japanese fishing communities are remarkable not only because they operate as decentralized agencies of coastal water management but also because, to an unusual extent, they tend to rely on cooperative practices to solve various problems. Within each FCA, fishermen are typically grouped according to the type of technology used. These groups are called ‘self-management institutions’ in official terminology and they numbered 1,339 in the late 1980s. As shown by the 8th Fisheries Census conducted in 1988, 147 groups, i.e. 11.0 percent of the total, are engaged in some form of pooling of income and input costs. The proportions are higher for the demersal fishery (21.8 percent) and the multiple-species fishery (19.6 percent) while they are lower for sedentary and pelagic species (8.4 and 7.4 percent, respectively) (Hasegawa 1993: 29). The purpose of this chapter is to understand the rationale and to assess the viability of pooling mechanisms in the specific context of a common property resource, a coastal fishery in Japan. The first question refers to the possibility that income- and input cost-pooling may help fishermen overcome various kinds of market failures, particularly in the insurance and labor markets. To the extent that these advantages of pooling exist, the second question arises as to why some groups which have actually attempted pooling succeeded while others eventually failed, thus presumably causing substantial welfare losses to their members owing to some sort of ‘community failure.’ The question is all the more interesting to raise as the conditions for collective action are a priori favorable since (1) groups of Japanese fishermen are rather small; (2) they are homogeneous in the sense that members are natives of the same village and use the same technology; and (3) they are entrusted by law to design and enforce rules for the proper management of their fishery, including the task of allocating water space (see above). As a matter of fact, these conditions relating to size, homogeneity, and legal support are often cited as critical success factors of collective actions in the specialized literature (Olson 1965; Esman

346

JEAN-PHILIPPE PLATTEAU AND ERIKA SEKI

and Uphoff 1984; Ostrom 1990; Sandler 1992; Tang 1992; Baland and Platteau 1996, 1997, 1998). To meet the above objectives, we chose to study the functioning characteristics and the particular histories of three groups of fishing units operating in Toyama Prefecture. We have selected groups with contrasted experiences and records regarding coordination of efforts in general, and pooling in particular. Since these groups are engaged in the same fishery—they aim at the same species, the shiroebi, or Japanese glass shrimp254—and use the same harvesting technology (with one exception to be spelt out in due time), they face roughly similar technical and environmental conditions, thus enabling us to control for a large number of background variables liable to affect collective action outcomes. Two of the three groups of fishermen selected for our study operate in the same area called Shinminato, use the same harbor facilities, and belong to the same cooperative known as the Shinminato Fishery Cooperative where they are the only two groups entitled to fish shiroebi. One of these groups, which we will henceforth call Group A, used to be the only shiroebi-fishing group in Shinminato. Since 1992, the other group, called Group B, was also granted the right to fish shiroebi in Shinminato after six long years of negotiation with Group A under the auspices of the local FCA.255 Note, however, that the latter group is allowed to fish that species during only part of the year (between June and September) during which period it operates alternately, from day to day, with Group A. The third group in our sample, called Group C, operates in another area in Toyama Bay, known as Iwase, and is the only shiroebi-fishing group belonging to the local cooperative association. While Group A started pooling in 1962 and has done so continuously since that time (thus representing a pooling experience of more than 35 years), Group C started at about the same time but drastically reduced the extent of pooling in the 1970s. As for Group B, it adopted pooling as soon as it entered the shiroebi fishery in 1992 but abandoned the system after one year only (in fact, two boats continue to pool to the present date). The number of fishermen's groups selected is admittedly very small. Yet, this is a deliberate choice given that we wanted to record systematically, over a period of a few months, the daily landings, current expenditures, and asset maintenance and repair costs for each boat comprising the three groups concerned (aggregate daily landings and fish prices are obtainable from the two local FCAs concerned). Furthermore, extensive discussions

254

Shiroebi is a small shrimp that belongs to the family of pelagic species and the stock of which is migratory and highly mobile.

255

Previously, Group B had fished other species between June and September. Since the late 1980s, the economic importance of these fisheries has gone down due to either depressed unit prices or to species extinction. On the contrary, the shiroebi benefited from growing demand for fresh products from urban consumption centers.

POOLING GROUPS IN JAPANESE FISHERIES

347

with boat skipper-owners, crew members, and their families were necessary to elicit their opinions about the performances and problems of their group as well as to gain a good understanding of their mode of operation (structure of decision making, conflict settlement, etc.). Participant observation and in-depth interviews appeared to be the only reliable technique to achieve that objective since confidence must be gradually built up before fishermen feel able to speak out openly about matters that form an important part of their life. One of the authors (Erika Seki) therefore spent ten months, from October 1996 to August 1997, on the different sites selected for the study and she actually succeeded in being accepted within the specific milieu of Toyama's marine fishermen, thereby ensuring the reliability of both the quantitative data and qualitative information collected. It bears emphasis that no enumerator was employed for the study as all the data were either extracted by Seki or self-recorded under her supervision by the fishermen themselves, their wives, or even their children. The remainder of the chapter is comprised of four sections. In Section 13.2, different, potentially complementary hypotheses are put forward in order to highlight the economic functions underlying the attributes of the most sophisticated sample group (Group A). The possibility that these attributes are somehow interrelated and linked, forming a cluster to be considered as a whole, is central to the theoretical framework developed in this section. In Section 13.3, we then proceed to test these hypotheses against available empirical evidence, focusing on the respective roles of insurance and profitability motives. The evidence used takes on the form of perceptions and subjective statements by the actors themselves, or the form of objective facts, whether they consist of historical events or of quantifiable aspects of present-day reality that we have systematically observed on the ground. In Section 13.4, the question is addressed as to why Group A has succeeded in implementing sophisticated mechanisms of coordination and pooling while the other two sample groups have (largely) failed. The critical role of local status or rank considerations occupies centre stage in the argument elaborated to answer that question. Section 13.5 is the conclusive section where the major findings of the study are summarized and final considerations are offered.

13.2 Economic Functions of Group Attributes: Theory A Summary Description of Observed Group Attributes Japanese fishermen communities tend to organize themselves collectively with respect to a number of tasks, which may actually vary considerably from one group to another. This is amply illustrated in the case of our

348

JEAN-PHILIPPE PLATTEAU AND ERIKA SEKI

restricted sample of three fishing groups in Shinminato and Iwase which we have selected for intensive study. In Table 13.1, the main attributes of the three sample groups are summarized with regard to various aspects of their functioning, namely distribution of incomes and expenditures, work relations, information processing, access to the resource, and determination of output and effort levels. Group A has the most complete pooling system: both incomes and operating expenses are shared equally among fishing units while investment costs are borne by the whole team but only insofar as nets are concerned (boats and other equipments are individually financed). Group B also adopted a pooling arrangement but gave up after one single year of experimentation. Two units, however, have decided to continue pooling according to the pattern initially adopted by the entire group. This pattern differs from that followed by Group A mainly in the way costs are shared. As a matter of fact, the purchase and the maintenance of nets is not financed collectively although costs in kind for moderate repairs are jointly borne (repair work is then carried out collectively). Moreover, Group B does not give any collective allowance for fuel and food expenses. In other words, both investment costs and operational costs are individually borne by Group B's owners. The pooling arrangement used by Group C in Iwase is more complex insofar as the catch proceeds are considered to consist of two distinct parts, one that is pooled and the other that is not. The first portion is Table 13.1 Functioning Characteristics or Attributes of the Three Sample Groups Attributes Pooling Income pooling

Group A

Group B

Group C

Full pooling

Input cost pooling Work coordination Synchronization of fishing hours and number of hauls Collective net repair Exchange of current information Regulation Control of access to fishing space Control of fishing effort Sharing of knowledge and expertise

Full pooling

No pooling except among Partial pooling two units No pooling Partial pooling

Yes

Yes

Yes

Yes Systematic

Limited Minimal

Limited Minimal

Well-developed

Imperfect

Imperfect

Effective control Actively encouraged

No control Minimal

No control Minimal

POOLING GROUPS IN JAPANESE FISHERIES

349

obtained by computing the minimal value of the entire landings of the group, which is done by applying to total landings the lowest price that has been obtained during the period considered. This imputed ‘minimal’ value is then divided among member units in strict proportion to individual landings. As for the surplus value of the group's catches, it is shared equally. In actual fact, the incidence of pooling does not appear to be large since, during the period between April and August 1997, the fraction pooled accounted for as little as 10 percent of the total value of landings. On the other hand, all expenditures except fuel costs and expenses incurred for a carrier boat are borne individually. Pooling is not the only attribute for which we observe wide variations across groups. In fact, functions of work coordination, regulation as well as sharing of knowledge and expertise—which will be commented upon as we go along in the following discussion—are much more developed in the full-pooling group (Group A) than in the other groups. What is important to emphasize is that all these attributes and the corresponding mechanisms may be interconnected in the sense that some of them enhance the effectiveness of others through the overcoming of incentive problems. In attempting to uncover the economic functions underlying all the aforementioned attributes, it is therefore important to analyze a well-functioning group, such as Group A, as a cluster of possibly interrelated attributes, or as a coherent structure the elements of which are mutually reinforcing (for a similar approach, see Aoki 1994). On the face of it, two kinds of considerations, insurance and profitability, may motivate fishermen to adopt a pooling arrangement. If the two motives are simultaneously present, we may expect a series of self-reinforcing effects to be at play, as depicted in Fig. 13.1. In the remainder of this section, such effects are discussed in some detail. In order to help the reader better articulate the discussion with the figure, numbers between square brackets are given in the text that correspond to various self-reinforcing effects numbered from [1] to [6] in Fig. 13.1. The direct instrumental links between various group attributes and insurance or profitability objectives are not given any number, however. It is the purpose of Section 13.3 which follows the theoretical exposition in Section 13.2 to provide detailed empirical evidence aimed at identifying the motives and effects that are relevant in the case under study.

Insurance Motives (i) The most obvious rationale of an income-pooling arrangement, as economic theory has taught us, is insurance against risk of income fluctuations. Since this risk is especially high in fisheries, particularly when small-scale harvesting technology is being used, it does not seem surprising that Japanese fishermen are tempted to at least try to pool their incomes.

350

JEAN-PHILIPPE PLATTEAU AND ERIKA SEKI

Fig. 13.1 Sophisticated Fishermen's Groups As Clusters Of Interlinked Attributes

Notes: Solid arrows correspond to direct effects while dotted arrows correspond to indirect effects.

POOLING GROUPS IN JAPANESE FISHERIES

351

Information problems are generally assumed away in models of voluntary reciprocal transfers constructed by economists to show the possibility of informal risk-pooling: agents, it is assumed, can perfectly observe each other's income to spot defections and, as a result, they are motivated to consider the effects of their current actions on their reputation which itself determines the range of future income opportunities open to them (see, e.g., Coate and Ravaillon 1993; Besley 1995: 2165–72). When information is imperfect, however, they may be tempted to free ride on the assistance efforts of others. This they can do by shunning effort when the potential catch is large in their fishing spot, or by understating their true income. Insofar as, in both Shinminato and Iwase, misreporting of output is difficult because all fishermen unload their catches in a centralized place (the jetty nearby the local cooperative's building), the main method of free riding is labor shirking. True, this problem is mitigated by the limited size of self-management groups, yet moral hazard persists and needs to be surmounted if a risk-sharing scheme is to be viable. It can be argued that the arrangements adopted by Group A to synchronize fishing operations aim precisely at solving this problem. As a matter of fact, synchronization and coordination of fishing operations help create conditions—skippers collectively decide whether or not to operate on a particular fishing day and, when they do, they start and stop fishing at the same time while setting a maximum number of individual hauls—that produce a lot of useful information at low cost. Boats operate in one or two rather concentrated areas with the consequence that they remain in sight of each other, thereby allowing easy mutual monitoring. True, this is not sufficient for effective labor input to be accurately assessed, yet, as we shall argue in Section 13.4, competition for rank in pooling groups ensures that imperfect effort observation is not really a problem if group heterogeneity remains tolerable. In addition, pooling requires that incomes of all member fishing units are made common knowledge within the group [1]. (ii) In the pooling arrangement adopted by Group A, the costs of current inputs such as fuel and food expenditures are subtracted from aggregate gross output of the group before sharing takes place. The classical argument used to explain such an arrangement is that it eliminates distorsions in the use of current inputs if they are observable (Adams and Rask 1969; Newbery 1975; Bliss and Stern 1982). The idea is indeed that the disincentive to apply inputs under an output-sharing arrangement is exactly offset by the subsidy to the input cost under the equal output- and cost-sharing rule (Hayami and Otsuka 1993: 79), a rule that is equivalent to the above mechanism applied by Group A. Since fuel and food expenditures are rather difficult to measure, the assumption of input observability is not satisfied. Yet, the authorized deductible expenditures on these two items

352

JEAN-PHILIPPE PLATTEAU AND ERIKA SEKI

correspond to a lump sum that is identical for all member fishing units. In this way, incentives to misreport this kind of expenditure or to misuse fuel are controlled, and the benefits of cost sharing can be reaped (Platteau and Nugent 1992: 404). This shows that insurance-based considerations are not sufficient to account for the pooling arrangement observed (in Group A): incentive considerations must be brought into the picture to explain its cost-sharing aspect. Turning now to the treatment of maintenance and repair expenditures on fishing assets, particularly nets, cost-sharing is also practiced whether in the form of collective labor efforts or pooling of the cash expenditures involved. Here, two explanations can be put forward which can be possibly combined, one based on insurance and the other on profitability considerations. The insurance argument is that owner-skippers are keen to spread the risk of net damage in order to stabilize their net incomes. Indeed, it is not only catches and, therefore, gross incomes that fluctuate randomly but also input costs, particularly when they are associated with the use of equipments which themselves constitute a risk. Since insurance markets do not exist for such volatile and fragile assets as fishing nets, owner-skippers look for informal, community-based insurance mechanisms. As for the profitability argument, it is the same as that used above with respect to current inputs: unless expenditures on net repair and maintenance are borne collectively, each skipper has an incentive to underuse his gear given that only a portion of the marginal productivity of net use will accrue to him and his crew. This said, insurance and profitability considerations may be interlocked in the following sense: to the extent that insurance against the risk of net damages induces skippers to take more risks—they are more willing to cast their net in relatively productive but dangerous fishing spots—it will have the effect of increasing the expected level of catches, thereby enhancing profitability. Bearing in mind that pooling of costs is combined with pooling of income, it nevertheless appears to have an ambiguous impact on risk taking if risk taking is positively correlated with labor effort. There are indeed two effects running in opposite directions: the pooling of incomes acts as a disincentive to apply effort and, therefore, to take risks while the pooling of net repair expenditures induces skippers to take more risks and, therefore, to apply more effort. Note that, if information conditions created by work synchronization have the effect of limiting the dampening of incentives caused by the pooling of incomes, the second effect is likely to outweigh the first so that effort and risk taking will increase. In the case where effort and risk taking are not correlated, pooling unambiguously induces skippers to take more risks, yet as less labor is likely to be applied, the ultimate effect on expected output cannot be predicted.

POOLING GROUPS IN JAPANESE FISHERIES

353

Whichever the effect, like in the case of income fluctuations, incentive problems arising from pooling net repair and maintenance expenditures must be somewhat controlled to make the insurance scheme viable. Hence the crucial role of synchronization of fishing operations as a way to monitor the risk behavior of group members. (iii) There is still another attribute (of Group A) that may be analysed as the result of an insurance motive, viz. the diffusion between members of knowledge and expertise about fishing techniques and the appropriate design of nets.256 As a matter of fact, skills and expertise can become rapidly outdated if a major technological breakthrough occurs. In order to insure against such risks of sudden depreciation of human capital, skippers may be willing to enter groups with a balanced age distribution. As evident from our interviews with members of Group A (not those of Group B), the interest of older skippers lies in their ability to count on young people's readiness to learn new methods (such as the use of the GPS) in the event of a technological shock. As for the latter, they are motivated to cooperate with the former because, in the absence of radical technical change, they can improve their mastery of existing techniques (such as how to adjust the balance of a net) by observing from close quarters the elders' behavior and by taking direct advice from them. It is revealing that, in the case of Group A, pooling has been adopted at a time when fishermen were confronted with a major technological breakthrough involving the introduction of electronic devices for navigation and stock-detecting purposes and the shift from cotton to nylon nets. These changes brought about new uncertainties against which skippers were apparently eager to collectively insure.257 There is another expertise-related risk which is intergenerational in nature and which at least one skipper (from Group A) has expressed the need to insure against. The fear is that death may befall a skipper before he could transmit his skills to the son who is expected to succeed him in the management of the family boat (bear in mind that most boat owners are simultaneously skippers). This fear is justified by the fact that sea accidents are frequent and often fatal. Since deaths by accident are random, teaming up with other skippers so that an apprentice does not exclusively

256

Thus, the mounting of technically efficient nets requires advanced skills that can be developed and acquired only by continuous invention and experimentation of innovative designs. Moreover, although fishing boats are equipped with sophisticated electronic devices—such as echo-sounder, satellite-driven navigators, and radar—their manouvering demands considerable expertise on the part of skippers and, to a lesser extent, of the crew members also.

257

One skipper explicitly admitted that, with mechanization of fishing, he lost his leadership in terms of catch performances.

354

JEAN-PHILIPPE PLATTEAU AND ERIKA SEKI

depend on his father or on any single master to learn the requisite skills is an obvious solution to the above problem. In the above, it is essentially insurance considerations that have been assumed to lead fishermen to enter an incomepooling scheme that is made sustainable by various tasks of work coordination and synchronization. We now consider the theoretical possibility of a reverse hierarchical relationship between income pooling and work coordination. In this scenario, motivated by the search for increased profitability, fishermen undertake to coordinate their fishing operations, yet to be effective such coordination needs to be accompanied by an income-pooling scheme that thus becomes instrumental to a profitability-increasing collective mechanism.

Protability Motives (i) When information is imperfect, losses of potential output can arise from (at least) two different sources. First, good fishing locations may be bypassed by all the boats and, second, the resource available in plenty in some fishing sites may be underexploited due to the unwillingness of skippers present on those sites to share information with fellow skippers. Systematic searching procedures achieved through coordination of individual efforts may be expected to ensure a better coverage of potential fishing areas and thereby reduce the first type of inefficiencies. Moreover, group coordination entails savings on capital costs because fixed equipment (e.g. radar) need not be installed on each and every boat to make searching operations effective.258 As for the second source of output losses, it can be removed by information exchanges aiming at directing boats to the most promising spots. The latter solution, however, is fraught with the well known free rider problem. A straightforward way to overcome that problem consists of a reciprocal arrangement whereby each skipper agrees to communicate to others the coordinates of large congregations of fish which he has come across. Given the repeated nature of interactions among skippers, each of them would have an incentive to abide by such a commitment even though conveyance of strategic information entails a positive cost. This is because a skipper who falls on a good spot today and shares this information with his colleagues may expect them to reciprocate in the future when the situation is reversed. Nonetheless, such a decentralized mechanism of information exchange can work only if violations of the prescribed commitment can be perfectly detected, a condition which is hardly verified in reality.

258

In actual fact, one boat in Group A is not equipped with radar, which would actually prevent it from going to sea in bad weather if it could not rely on the cooperation of the other fishing units (benefit emphasized by the skipper concerned).

POOLING GROUPS IN JAPANESE FISHERIES

355

True, centralized landing and marketing of catches enable fishermen to acquire some knowledge about the actual performance of their colleagues and to confront it with the latter's observed behavior regarding disclosure of strategic information while being out at sea. Yet, the problem is that from actual landings it is hard to infer the resourcefulness of the fishing location used by a lucky skipper. As a matter of fact, a skipper enjoying a good catch may contend that he has indeed fallen on a good spot, yet the resource potential of this spot could be more or less fully tapped by a single boat so that there is no point in calling in fellow skippers. Since this statement is impossible to verify by others, free riders may go undetected and unpunished. As a consequence, any arrangement based on reciprocal voluntary exchanges of information about resource availability is likely to be unsustainable. Yet, information sharing can succeed if it is combined with an income-pooling scheme [2]. This is because each skipper then has an incentive to disclose strategic information to other group members: combined with income-pooling, information sharing becomes a selfenforcing mechanism. The same argument applies with respect to the attribute of knowledge- and expertise-sharing (see above): pooling helps to enhance the incentives of skippers to share their fishing expertise and skills [3]. (ii) When income-pooling is interlinked with information sharing, all fishing units belonging to the group will not necessarily flock to the most promising fishing spot. In point of fact, a pooling group is induced to optimally allocate its member units between the available spots. This result obtains because the pooling of incomes induces skippers to coordinate their decisions with a view to achieving the maximum aggregate level of catches or, in game-theoretical terms, because it has the effect of transforming a PD (Prisoner's Dilemma)-structured game into a Chicken game (see Appendix I for a formal proof), thereby enabling members to select one of the Pareto-optimal allocations. If the most promising spots are also the most risky, say, because the risk of damage or loss of nets is comparatively high, in the absence of a pooling mechanism (and assuming risk-averse fishermen) we may expect a congregation of boats to occur in the least, rather than in the most, productive locations. When available, such a mechanism indeed has the double effect of creating an incentive to allocate boats efficiently among existing spots and of insuring fishermen against the above risk. Pooling, however, is not the only way to avoid congestion costs when information is sufficiently spread (through a deliberate process of information exchange or not) to cause a congregation of boats in certain locations. Another, much more predominant method consists of laying down a rotating scheme, providing clear rules of sequential access to the attractive location(s). Under such a scheme, since they know their order of access to such locations (which usually varies from day to day), all skippers are better

356

JEAN-PHILIPPE PLATTEAU AND ERIKA SEKI

able to assess the catches they can achieve in all the available locations than if order of access was purely random. As a consequence, boats will be more scattered over the fishing space. For example, a fisherman who only has the second turn on the most productive spot will decide to fish in a less productive spot if the first turn on the latter brings him a higher income. Yet, if access to the best spot is random due to lack of coordination, he will stay there as long as the expected income exceeds the income he can earn by having the first haul on the second spot (see Appendix I). Another advantage of rotating schemes is that they help minimize the risk of boat collisions, net entanglements, and costs of conflicts if they include spacing rules designed to maintain an optimal distance between boats. Such rules are needed because, as observed in Toyama Bay, the order of access to the most attractive location is fixed only for the first trip on any given day. As a result, all skippers want to save as much time as possible in an attempt to improve their position in the subsequent net-casting sequences (and to fetch high prices by landing their catches early). One obvious way of achieving this is by getting close to the preceding boat in the sequence so as to start fishing operations as quickly as possible after completion of the previous haul. Spacing rules are also useful when the allocation of spots follows another model in which, instead of waiting in a queue for getting access to the most favored fishing spot, skippers who do not have the first turn(s) prefer to cast their net in a proximate second-best (or third-, or fourth-best) location. In this case, indeed, they are easily tempted to trespass the frontier separating this second-best location from the most favored one, thereby increasing the risk of accidents. On the face of it, pooling makes rotating arrangements including spacing rules, largely redundant. Why then are such arrangements and rules in existence in the case of Group A which is pooling? Clearly, aspects other than profitability considerations must be brought into the picture to account for such a situation. As will be seen later (Section 13.4), interpersonal comparisons of performances and local status considerations are the missing link that is needed to answer that question. To the extent that these considerations contribute to create a certain amount of competition among members (even though total catches are equally divided among them), equalization of opportunities to catch fish remains important. (iii) Profitability can also be increased via scale economies.259 Indeed, cooperation in physical operations related to the searching of nets lost and the mending of nets damaged entails substantial economies of scale which have the effect of increasing the effectiveness of mutual insurance against net damages and losses. There are actually four different kinds of scale

259

In fact, the gains from division of labor can themselves be considered as scale economies.

POOLING GROUPS IN JAPANESE FISHERIES

357

economies that can be exploited through such cooperation. First, by working together in searching operations for nets lost at sea, a group of fishing units is able to systematically probe the area where the net searched has been reportedly lost. As a result, the time needed to find a lost net or the probability of retrieving it is much higher when several boats participate in the searching process than when only the boat concerned is involved. Operations aimed at spotting and targetting fish schools are obviously subject to the same scale economies. Second, there are scale economies in repair and maintenance operations. Given their large sizes (average length is around 200 meter and maximum depth is approximately 100 meter), nets are more easily handled (unrolled, spread, packed, etc.) by a group of cooperating people than by the single crew of the unit owning the damaged net. Third, precisely because repair and maintenance tasks are more effectively carried out in group, the work can be more expeditiously completed with the consequence that the gear does not remain immobilized for a long time. This is an important advantage insofar as it enables equipment owners to be content with a limited number of spare nets and thereby save on investment costs. Fourth, by working collectively, fishermen can more easily communicate among themselves their ideas about ways to improve the design and structure of nets. In other words, the interaction framework created by the requirement of collective net repairs tends to encourage cross-fertilization of innovative ideas and to stimulate their experimentation. Scale economies in R&D efforts actually extend to the whole realm of fishing skills and techniques. There are indeed many components to fishing skills and they are not necessarily concentrated in one person (for example, one fisherman may be adept at detecting the presence of fish, another one at anticipating changes in weather conditions, and a third one at handling nets quickly). Participation in a group therefore allows fishermen to reap the benefits of technical complementarities between their respective skills and experiences. Note incidentally that this last advantage could not be obtained if net repairs were entrusted to a specialized agent centralizing net repair operations and selling his services in the market. The fact that the technology embodied in a net is very specific to his operator is another source of market failure in net repair activities across most artisanal fisheries of the world. To be effective, a net must indeed be exactly tailored to the user's specific needs and skill characteristics. As in the case of information exchanges, coordination of net repair and maintenance activities, including searching operations out at sea, is vulnerable to the free rider problem. At first sight, a simple decentralized mechanism of mutual assistance whereby fishermen would commit themselves to help fellow group members in the event of a loss of, or damage to, a net, is liable to reap the underlying scale economies. Their commitment would be credible to the extent that they have an incentive to abide by

358

JEAN-PHILIPPE PLATTEAU AND ERIKA SEKI

their promise in the expectation of future benefits when it will be their turn to suffer damages and losses. Unfortunately, things may not be so simple if information is imperfect so that detection (and punishment) of violations is costly. Physical presence in searching and mending operations is no doubt easy to verify, yet the quality of efforts applied in the course of such operations does matter a lot and is much more difficult to measure. For example, lack of concentration and conscientiousness in searching for a lost net in the portion of the sea ascribed, or in repairing the portion of the net allotted, may be difficult to observe or, at least, to assess with sufficient precision. Hence the fragility of any arrangement based exclusively on voluntary promises of reciprocal assistance. Again, the interlinking of such a mutual assistance scheme with income-pooling may overcome such incentive problems [4]. This is because the adding of an income-pooling scheme to other group attributes causes coordination of net repair and maintenance works to take on a public good aspect. Everyone having an interest that all nets in the group are as productive as possible—since individual incomes depend on the performances of the whole group—the incentive for mutual assistance is enhanced. Moreover, since the groups observed in our sample have a relatively small size, the ‘incentive dilution’ argument—when a group is small, the cost of free riding tends to be high for an individual member—can be used to the effect that the members' incentive to violate their promise by shirking on labor quality is likely to be insignificant (see Olson 1965). (iv) Profitability gains may accrue to fishermen if they succeed in influencing fish prices through deliberate control of landings. The auctioning mechanism used to dispose of fish in Shinminato harbor actually places Group A's members in the position of a perfectly discriminating monopolist rather than that of a classical monopolist. The sale procedure is an original two-stage auctioning mechanism that consists of a ‘morning auction’ and an ‘aota auction’. The ‘morning’ auction starts immediately after the landing of the first hauls of the day which are brought back to shore by each boat between 5.30 and 6.00 a.m. All subsequent landings are disposed of through the ‘aota auction’. Each auction is of the Dutch type, implying that it is descending.260 The critical difference between the ‘morning’ and the ‘aota’ auctioning procedures is

260

The precise steps are as follows. (1) An auctioneer declares a total quantity to be traded; (2) he then points to a given portion of that quantity for immediate bidding and proposes a unit price for that portion; (3) the buyers raise their hands to indicate the quantity they are willing to buy at the proposed price; (4) if demand (i.e. the sum of all the quantities indicated by the traders) does not happen to exceed the size of that portion, the requests of all fishmerchants are met and the auctioneer proposes a lower price to dispose of the remaining fish in the same portion. If there is an excess demand, the auctioneer revises the initial unit price upwards. The process continues until the excess demand is eliminated and the portion is completely sold. Finally, once the whole portion is disposed of, the auctioneer indicates a new one to be offered for sale. Another bidding process takes place until the total landings are sold out.

POOLING GROUPS IN JAPANESE FISHERIES

359

that, while the former takes place only after the fish obtained from the first hauls have been unloaded on the jetty, the latter is held even before actual landings from subsequent hauls are realized.261 When landing their first hauls, fishermen are indeed asked to provide an estimate of the total expected landings of the day, and it is on the basis of such estimates that the ‘aota’ auction is held. As a result of this two-stage auctioning mechanism, fish buyers are incited to reveal their willingness to pay for the fish on sale. This is because fish is sold at multiple prices to different buyers and, given the nature of the mechanism used, understating one's willingness to pay involves a risk that one will not be supplied if total production falls short of expectations. In point of fact, the disposal of shiroebi takes place immediately after the landing and, as a rule, the fishmerchant who bid the highest price has priority access to the fish while the other merchants are served as subsequent catches reach the harbor and pay the lower price to which they have committed themselves. If total landings are less than the estimated supply, some low bidders will therefore be left unattended. On the other hand, if total production exceeds the auctioned quantity, a new auction is called for. In the process, a downward-sloping demand curve is gradually revealed, that fishermen become aware of even while continuing their fishing operations. As a group, thanks to coordination of their harvesting decisions, they are thus able to capture the entire producer surplus and, unlike what happens under a situation of classical monopoly, they achieve a Pareto-optimal output. Absent this coordination and fishermen have an incentive to overfish in the sense of harvesting more fish than the optimal quantity. This is because, when skippers compete with each other, each of them is induced to catch as much fish as possible and to bring it back to shore early in order to fetch the highest price (bear in mind that the first fishmerchant to obtain the fish landed is the one who bid the highest price in the ‘aota’ auction). Given that everybody behaves in the same way, however, no skipper can know for sure the cumulative quantity of fish that would have been disposed of before his own catches are landed. As a consequence, the price that he will get for his produce remains highly uncertain and, in deciding how much fish he will catch, he will equate marginal cost with the expected price and not with the actual equilibrium price as would happen with a discriminating monopolist. Owing to this price uncertainty caused by open competition, it can be shown that skippers are likely to overfish in the hope of higher profits.

261

‘Aota auction’ literally means the purchasing of a green or standing rice crop, that is, a forward purchase.

360

JEAN-PHILIPPE PLATTEAU AND ERIKA SEKI

The question then arises as to how fishermen can be induced to act in a coordinated manner. This is a vexed issue to the extent that any agreement aimed at ensuring the advantages of a perfectly discriminating monopolist must not only specify a total output quota and its distribution among individual fishing units, but also the time sequence according to which they will land their catches (since the price obtained depends on the order of arrival to the point of sale). It is hard to see how fishermen could effectively cope with such a problem in a context where price configurations vary from day to day. Due to this volatility of fish prices, any rotating scheme determining a constantly changing order of landing so as to equalize chances to obtain good prices is bound to arouse tensions and invite frustrated fishermen to violate the agreement. The reputation effect occurring through repeated interactions is not likely to be of much help in these circumstances since violators might rightly complain that the agreement has turned out to be harmful for them. On the other hand, and for the same reason that it is difficult to find a scheme that proves equitable ex post to everybody, endless demands for its renegotiation could undermine its enforceability. Once again, income-pooling appears as a mechanism liable to reduce the above moral hazard problem, all the more so if it is interlinked with intra-group synchronization of fishing activities and information exchange [5]. By forcing fishermen to internalize the impact of their actions on the incomes of their fellow members, income-pooling induces them to abide by an output-restraining scheme that does not need to be coupled with a rotating arrangement fixing the order of landings. Work coordination tends to make the control of fishing effort easier because it facilitates the ending of fishing operations as soon as the aggregate production quota is fulfilled [6] and the way in which this quota is de facto distributed among member units at the time of completion is irrelevant. For such an interlinked arrangement—output-restriction cum-income-pooling-cum work coordination—to be effective, however, it is obviously important that all fishing units landing their catches in a given selling point participate in it. This is not always true in Shinminato where coordination and pooling are practiced in the framework of groups of restricted size that operate together during part of the year. A solution which springs to mind and which has actually been adopted in this area (see Section 13.1) is to rotate the right of going out fishing (in the Bay) so that only one group catches fish at a time. In order to minimize waiting time and also to equalize income-earning opportunities in an environment characterized by significant seasonal variations in resource availability, rotation takes place in rather short intervals, namely on a daily basis. The problem nevertheless remains that, even though fishing operations are separate, the benefits of output-controlling effort are not well internalized due to a strong interdependence between daily demand schedules. Because of this interdependence,

POOLING GROUPS IN JAPANESE FISHERIES

361

which is the result of the possibility to store the shiroebi prawns for future disposal,262 one group of fishing units is able to steal the other group's (future) demand by catching and landing fish in excess of current demand. As a consequence, when it is its turn to fish, each group faces a demand that is depressed by the amount preempted on the previous day. Put in another way, when two groups go out to sea on alternating days, there is a duopolistic competition between them that has the effect of eroding the monopolistic power that they could have wielded had they been operating alone (see Seki 2000, for a formal exposition). (v) Control of fishing effort resulting in restricted landings can aim not only at increasing fish prices but also at overcoming the classical problem of the ‘tragedy of the commons’. In the present case, however, this problem is limited since, as explained in Section 13.1, access to fishing space is placed under the effective control of local cooperative fishermen's associations. It is not absent, though. Fishermen with access rights can indeed compete among themselves by trying to increase their catches since the negative externalities they thus create for others are not properly internalized. Belonging to a well-defined group which can exclude non-members, they interact strategically with the consequence of applying an excessive amount of fishing effort, wasting scarce capital resources, and dissipating the rent (albeit on a lesser scale than would be the case under open access). Moreover, excessive effort today may possibly deplete the resource and thereby jeopardize future catches (Baland and Platteau 1996: 26–32). An income-pooling group, by forcing the internalization of externalities, naturally induces fishermen to restrict their efforts so as to maximize the rent from the fixed resource. It is the purpose of the empirical analysis conducted in the next section to clarify whether the two motives, insurance and profitability, are relevant and, if in the affirmative, to determine which of them is possibly more important. More specifically, among all plausible particular rationales for pooling—a desire to smooth income fluctuations and to dampen the risk of net damages; to reap the advantages of scale economies, including division of labor, and coordinated allocation of fishing spaces; to avoid congestion costs; and to appropriate as large a part as possible of the producer surplus through controlling fishing effort and output—we want to know which ones are playing an important role in the design of collective organizations such as those tried by Groups A, B, and C in Toyama Bay.

262

Shiroebi can be stored in deep freezers for future consumption. It is a common practice that some buyers make purchase today for future (say, tomorrow) consumption at sufficiently discounted prices. We witnessed that the first landings always fetch the highest unit prices and they are disposed for immediate consumption, whereas the latter landings are often brought to the deep freezing facilities for future consumption.

362

JEAN-PHILIPPE PLATTEAU AND ERIKA SEKI

13.3 Economic Functions of Group Attributes: Empirical Testing What are the motives behind the attempts at collective organization by Japanese fishermen, particularly the attempts at pooling (net) incomes? In order to shed light on that question, two distinct sets of information are being used. The first type of information consists of perceptions by the fishermen themselves as they could be gathered from intensive interviews organized both individually and collectively, and from day to day interactions with them. As for the second type, it is based on mostly quantitative data obtained from secondary sources (the price series and aggregate production figures available with the local FCAs) or from direct recording (catches per haul, number of hauls per day, fishing locations, damages to nets, net repair and maintenance expenditures, etc.).

The Role of Insurance Motives All the skippers in Groups A and B have been asked about the reasons for pooling. The question was essentially left open even though some clues were sometimes provided if, after some time, the respondent found it difficult to precisely articulate his opinion. Several answers were allowed but ranking them by order of importance appeared rather hard for most of the skippers. In Table 13.2 the frequencies corresponding to the various possible rationales of pooling are given for the two Shinminato groups. The most prominent result emerging from this exercise is certainly the fact that stabilization of incomes was not mentioned a single time by the 12 skippers interviewed. Contrary to expectations, demand for insurance does not therefore appear as a relevant motive for pooling incomes in the surveyed communities. This can be due to two different reasons: fishermen are not risk-averse, or their incomes are highly correlated so that pooling would not allow them to adequately insure against income fluctuations. A simple test aimed at uncovering the degree of risk aversion of sample Table 13.2 The Rationales of Pooling as Perceived by the Sample Skippers (Frequencies of Answers) (a) To avoid crowding on limited spots (b) To reduce the stress of racing (c) To share information (d) To avoid overfishing (e) To stabilize unit prices (f) To cooperate in net search and repair (g) To stabilize income

Shinminato A (7 skippers) 7

Shinminato B (5 skippers) 4

2 5 1 2 6

3 2 0 2 4

0

0

363

POOLING GROUPS IN JAPANESE FISHERIES

skippers leads to an ambiguous conclusion: a significant minority of them appear to be risk takers while the others can be considered as risk-averse. It may therefore be difficult for a group of skippers to agree on a scheme that is exclusively focused on an income-stabilization objective. Turning now to the second reason, the following must be noted: the coefficient of variation of gross incomes (considering incomes earned at the end of each haul over a ten-month period) earned by the different fishing units belonging to Group A varies between 0.5738 and 0.7631, while the coefficient of variation of the pooled income for the same group is 0.5999. For Group B, the range of variation is 0.5298–0.9249 for the (observed) incomes of individual member units, to be compared with a value of 0.6874 if income would have been pooled among them. As for Group C, the range is 0.7945–1.1846 for (observed) individual incomes, as against 0.6777 in the event of pooling. Clearly, as these figures demonstrate, there is not much insurance benefit to be obtained from pooling incomes. A simple look at the covariance matrices for individual incomes in the three sample groups (not shown here) bears out the above conclusion. In Table 13.3, we have presented the coefficients of correlation between the gross incomes of each individual fishing unit and the aggregate incomes of the rest of the group to which it belongs. Their significance levels are indicated in brackets. It is noteworthy that all correlation coefficients are significantly positive and often quite high, especially in the case of Group A which is the only one to actually practice income-pooling! It can be argued that income covariances are likely to be endogenous since, through time, cooperation in Group A may have caused income processes to stabilize at a higher mean and lower variance, partly driving the high covariances across boats. Yet, it is hard to believe that covariances may have been much lower than they are now, or low enough to have made income risk reduction an original motivation for cooperation. Statements by the fishermen (see above) are also to be taken seriously and, when asked about the circumstances that surrounded the formation of their collective, Group's A Table 13.3 Covariances Between the Gross Incomes (Per Haul) of Each Individual Fishing Unit and the Aggregate Incomes of the Rest of the Group to Which It Belongs (For the Three Sample Groups) Group A Group B Group C

Boat 1 0.6440 (0.000) 0.4528 (0.012) 0.5373 (0.000)

Boat 2 0.7250 (0.000) 0.2753 (0.141) 0.4771 (0.000)

Boat 3 0.8442 (0.000) 0.6146 (0.000) 0.6340 (0.000)

Boat 4 0.7216 (0.000) 0.4229 (0.020) 0.4673 (0.000)

Boat 5 0.6933 (0.000) 0.4705 (0.009) 0.4389 (0.000)

Boat 6 0.8513 (0.000) 0.5996 (0.000)

Boat 7 0.5466 (0.001)

364

JEAN-PHILIPPE PLATTEAU AND ERIKA SEKI

fishermen never mentioned the need to smooth out income fluctuations while they explicitly referred to other motives (see below). If income-pooling is not motivated by insurance considerations, it is probably instrumental to profitability-enhancing objectives. That fishermen in Group A are aware of this role of pooling comes out of several remarks or comments made in the course of the interviews. In particular, on several occasions they have pointed out that the fact of pooling incomes induces them to better share strategic information and to be more conscientious in net repair works. Are fishermen sensitive about the insurance advantages of pooling net repair expenditures? This does not seem to be the case if we simply look at Table 13.2. As a matter of fact, what they (often) mention as a prime reason for pooling is cooperation in net search and repair works rather than insurance against risks of net damage. Their statements in this respect must not be narrowly interpreted, though: as more detailed discussions with them testify, they have numerous aspects in mind when they refer to the importance of cooperating in net search and repair operations, and some of these aspects may have to do with insurance concerns. Thus, there have been recurrent references to the fact that, if a net is not quickly repaired or found in the (probable) event of an accident at sea (which implies that search and repair operations are undertaken collectively), they face the risk of losing their income for several days or even more. This entails the additional risk of losing good crew laborers who rapidly become impatient in cases of prolonged boat idleness, all the more so because they are then required to make a lot of unpaid repair work. Fishermen appear to be aware of the fact that pooling of net repair expenditures bears upon their risk behavior, especially when they have to decide whether to operate in dangerous but promising fishing spots. However, different fishermen assess this impact in diverse ways: some say that they are ready to incur more risks as a result of pooling while others mention the opposite effect. This is not a surprising finding since, as explained in Section 13.2, the pooling of net incomes may yield two contrary effects that make its influence on risk taking a priori unpredictable. Yet, there is a strong presumption that pooling should encourage risk taking rather than the opposite. Whether hard data bear out this hypothesis is the question which we now address. There are two ways in which a positive impact of pooling on risk taking can manifest itself. On the one hand, skippers may be encouraged to operate more frequently in risky fishing locations and, on the other hand, they may be induced to take more risks in given locations. Our data actually allow us to test the impact of pooling on risk behavior because two of our sample groups operate in the same geographical space (Shinminato) and are therefore subject to the same environmental conditions while they have different insurance arrangements: there is only

POOLING GROUPS IN JAPANESE FISHERIES

365

limited cooperation in mending damaged nets in Group B and no cooperation at all in searching for nets lost (Section 13.2) whereas Group A provides perfect insurance to its members in these respects.263 Let us begin by examining the evidence available regarding the impact of pooling on choice of fishing spots. An effective test is possible since skippers from Shinminato must choose between two locations, called Heta and Oki, which have different return and risk characteristics. Heta is located along the slope of the cliff close to the shore and is characterized by a narrow and complicated valley-like structure of sea bottoms. As a result, it is subject to strong and unpredictable undercurrents that make the risk of tearing or losing nets all the more serious as the area is restricted. Oki, by contrast, is relatively far from the shore and has a deeper and wider sea bottom so that the danger of hitting nets against the surface of the sea bottom is smaller. Our data confirm that the probability of suffering a loss or damage of a net is higher in Heta than in Oki: hauls during which damage occurred represented 8.1 percent of the total number of sampled hauls in Heta (32 out of 394 cases) while the corresponding proportion for Oki was only 4.7 percent (31 out of 658 cases). Unfortunately, Heta is reputed to be a more promising spot not because modal catches are larger there than in Oki but because the chances of obtaining big hauls are better. From our data, it can thus be verified that the upper tail of the distribution of observed catches is much less flat for Heta than for Oki (not shown here). As a consequence, we would expect skippers from the best insured group, Group A, to fish more intensively in Heta than their colleagues from Group B. That such a prediction is amply borne out by the data is evident from Table 13.4 which shows the relative preferences of the fishermen Groups A and B for operating in Heta and Oki as revealed by the number of hauls per fishing day in each location (see the left-hand part of the table). As a matter of fact, hauls in Heta appear to make up 40 percent of the total for Group A, while the corresponding proportion is only 24 percent for Group B. The frequency of fishing in Heta relative to Oki is considerably higher for the leading boat in Group A (1.32 for boat A6) than for the leading boat in Group B (0.75 for boat B3). This maximum relative intensity of fishing in Heta in Group B is actually not much different from the average intensity observed for Group A (0.75 as against 0.67). Since Group A's boats operate more frequently in Heta than those of Group B, presumably because they are better insured against net damages, we would expect them to achieve higher allocative efficiency. To test that prediction, we need to compare (expected) marginal catches in Heta and Oki, for both groups. This will be done at a later stage when we turn to the issue of coordinated allocation of effort between fishing spots.

263

Group B is not an ideal counterfactual, though, because two out of five fishing units comprising it are actually pooling incomes.

366

JEAN-PHILIPPE PLATTEAU AND ERIKA SEKI

Table 13.4 Revealed Preferences for Fishing in Heta and Oki and Observed Performances Therein—A Comparison Between Groups A and B (Based on Data Collected During June–August 1997)*

Group A Boat 1 Boat 2 Boat 3 Boat 4 Boat 5 Boat 6 Boat 7 Average Group B Boat 1 Boat 2 Boat 3 Boat 4 Boat 5 Average

Average of hauls per fishing day Heta Oki Ratio (1) (2) (1)/(2)

Average catch per haul (in kg) Heta Oki Ratio (3) (4) (3)/(4)

1.76 1.44 1.02 1.44 1.45 2.15 1.25 1.52

2.07 2.31 2.76 2.34 2.37 1.62 2.32 2.24

0.85 0.62 0.37 0.61 0.61 1.32 0.54 0.67

77.2 60.6 79.2 88.7 44.2 73.2 90.6 73.2

70.3 77.5 71.4 77.7 64.5 56.4 43.9 68.2

1.10 0.78 1.11 1.14 0.68 1.30 2.06 1.07

1.07 0.47 1.44 0.29 n.a. 0.81

2.28 2.91 1.92 2.94 n.a. 2.56

0.47 0.23 0.75 0.10 n.a. 0.32

110.1 104.0 84.7 50.0 75.0 89.5

97.4 57.4 73.5 50.7 40.0 66.2

1.13 1.81 1.15 0.99 1.88 1.35

Note: * The number of days of observation during which the data presented in the table have been collected is much larger regarding net damage and losses (see the left-hand part) than regarding catches and hauls (see the right-hand part). Indeed, while net damage and losses have been recorded for the entire period from beginning of June to end of August 1997 (with the exception of boat 5 in Group B for which such data are incomplete owing to a lack of cooperation of the skipper), production data could be collected only during discrete subperiods as Erika Seki had to record the landings herself and could not be at Shinminato and Iwase at the same time. Catch and haul data were collected in Shinminato over an effective period of 16 days for Group A (except regarding boat 1 for which we have observations for only 15 days) and only 10 days for Group B. Note that such data are available for boat 5 of Group B since the collection, in this case, did not depend on the goodwill of the skipper. Since the number of fishing days of observation differs so much between the two groups owing to random factors, it is misleading to compare catches per boat over the period.

Let us now look at the second impact of pooling on risk taking: are fisher-men belonging to Group A taking more risks in given fishing locations than their colleagues from Group B? To answer that question, we have constructed Table 13.5 in which information regarding the incidence and costs of net damages and losses is displayed for the two groups operating in Shinminato. The period considered extends over 29 fishing days between June and August 1997 when boats of both groups operate alternately. As can be seen from this table, the probability of net damage, as measured by the ratio of frequency of net damage to the number of hauls achieved, turns out to be higher in Group B (0.067) than in Group A

367

POOLING GROUPS IN JAPANESE FISHERIES

Table 13.5 Frequencies of Net Damage and Labor Time Spent for the Search and Repair of Nets Damaged Over the Period June–End of August 1997, for Groups A and B in Shinminato

1

2 3 4

5

6

7

Total frequencies of damages over 29 days Frequencies of damages per haul Frequencies of damages per boat* Total number of manhours spent for mending or searching nets** Average number of manhours spent, per haul, for mending or searching for nets Average number of man-hours spent, per boat, for mending or searching for nets** Average number of man-hours spent, per net damaged, for mending or searching for nets**

Group A Damage 30

Loss 3

Total 33

Group B Damage 29

Loss 1

Total 30

0.038

0.004

0.042

0.067

0.002

0.070

4.29

0.43

4.71

7.25

0.25

7.50

968.5

367 (42)

1,335.5

602

1,216 (16)

1818

1.23

0.46

1.69

1.40

2.83

4.23

138.3

52.4 (2)

190.8

150.5

304.0 (4)

454.5

32.3

108.3 (14)

40.5

20.8

1,200 (16)

60.6

Notes: * Unfortunately, owing to lack of cooperation of one boat belonging to Group B, data relating to net damage (but not those relating to catches) are incomplete for that group (slightly more than half of the period considered is not covered for the problematic boat). Only the four remaining boats have therefore been taken into consideration while computing the probability of net damage for Group B. ** The figures indicated between brackets refer to the number (total or average) of man-hours spent in net searching operations only while those indicated above the bracketed figures correspond to the total amount of labor-time devoted to either searching or mending nets loss (and possibly retrieved).

(0.038) (see row 2). Yet, this is not definite evidence that skippers of Group A take fewer risks than those of Group B. There are two main reasons for this. First, if skippers of Group A are better at handling nets than skippers of Group B, they experience less damage in spite of the fact that they actually take more risks. That Group A's members have better skills is suggested by the fact that the average number of years of experience in fishing is 27 for skippers of Group A but only 22 for those of Group B (the same measure for crew laborers is 22 for the former but only 15 for the

368

JEAN-PHILIPPE PLATTEAU AND ERIKA SEKI

latter). More importantly, fishermen from Group A have a much longer experience in harvesting shiroebi than Group B's fishermen who only joined that fishery in 1992. This experience is especially advantageous for fishing operations in Heta which is more skill demanding. Here is probably an important part of the explanation as to why the probability of net damage in Heta is considerably higher for Group B (18 hauls with damages out of a total of 88 hauls, that is a probability of damage of 0.20) than for Group A (10 hauls with damages out of a total of 308 hauls, that is a probability of only 0.03), while the incidence of net damages in Oki is about the same for both groups. Second, it may well be the case that damage to nets is more frequent but less serious in Group B while the opposite would be true of Group A. To measure the extent of damages to nets and, thereby, the amount of risk incurred, we must compare the average amounts, per net damaged, of labor time spent in Groups A and B for mending operations. Table 13.5 (row 7) shows that, according to this yardstick, severity of net damage is significantly larger for Group A (32.3 man-hours per net) than for Group B (20.8 man-hours). In fact, the difference between the amounts of risk incurred by skippers of the two groups is likely to be larger than the above figures suggest. Again, two possible reasons may account for such a bias. The first reason is actually the same as that mentioned above in connection with frequency of damages: because of differential skills in handling nets, a given scale of net damage does not reflect identical amounts of risk taken by skippers from Groups A and B. Since Group A's members have presumably better skills, the scale of damages to their nets underestimates the extent of the risks they are incurring, compared to Group B. The second reason lies in the presence of scale economies in net repair activities. Since the incidence of these economies can be presumed to be larger in Group A where the collective nature of mending operations is more pronounced, the actual amount of labor time spent in repairing nets in Group A is likely to be an underestimate of the scale of damages and, thereby, of risks taken by Group A's members when compared with their colleagues from Group B. Note also that, if Group A's members are more skilled in net repair activities, the same bias again occurs. The data reporting the relative frequencies of losses of nets—which are among the most adverse contingencies that can befall a skipper and his crew—bear out the hypothesis that skippers in Group A take more risk than their counterparts in Group B. As a matter of fact, the probability of losing a net is twice as high in Group A as in Group B. This evidence suggests that skippers of the former group are more daring than those of the latter, a fact actually confirmed by qualitative statements made in the course of the interviews.

POOLING GROUPS IN JAPANESE FISHERIES

369

When searching operations are undertaken collectively, scale economies are very important and, as tirelessly emphasized by the fishermen themselves, the probability of retrieving the net lost is much higher than in the case of isolated rescue operations. The (limited) evidence available bears out the cogent hypothesis: the three nets lost by Group A over the 29-day period considered were located and brought back after an average searching time of only 14 hours (that is, two hours per boat). Mending those nets on shore required an average amount of labor of 108 manhours per net damaged. By contrast, the only net lost by Group B could not be found and retrieved after 16 manhours of search (in several attempts) by the boat concerned (‘It is like searching for a contact lens in a swimming pool’, as one retired fisherman puts it). As a result, a new net had to be constructed by the skipper-owner's family with the help of the crew, a task which consumed as much as 1,200 man-hours of labor stretching over several months (see row 7)! In order to insure against the dreadful risk of losing nets and in the absence of an insurance market, there are two possibilities, viz. mutual insurance and self-insurance. Mutual insurance is the option chosen by skippers of Group A while self-insurance under the form of investment in spare nets has been preferred by those of Group B: boats belonging to the latter group have an average number of two spare nets as compared to only one spare net in the former group. Clearly, the solution of mutual insurance is less costly than that of self-insurance, an advantage that arises from the fact that, by collectively organizing for the searching and mending operations, Group A's members can considerably enhance the effectiveness of these operations as well as minimize the adverse consequences of the loss of a net. It is therefore not surprising that they are ready to incur a greater risk of such a loss with a view to increasing their long-term expected catches. Two last comments regarding Table 13.5 deserve to be made. First, since Group B practices pooling of net-mending labor expenditures up to a certain extent (2–3 hours of collective work per net damaged, representing approximately 50 man-hours in total), only minor damages are mutually insured. This characteristic may well account for the twin results that frequency of net damages is comparatively high in Group B while at the same time their scale is comparatively low, thus suggesting the possibility that skippers belonging to this group incur many risks of small damages but few risks of serious ones. Second, if we look at rows 5 and 6 of the table, we can see that there is no significant difference between Groups A and B regarding the amount of labor time spent, either per boat or per haul, for the mending of damaged nets. This is because costs of net repair per haul or per boat are the net outcomes of two contrary trends which cancel out: on one hand, the frequency of net damages is lower in Group A than in Group B, yet, on the other hand, the average amount of repair labor time

370

JEAN-PHILIPPE PLATTEAU AND ERIKA SEKI

per damaged net is higher in the former than in the latter group. Here is an interesting result pointing at the fact that competition for crew labor does not penalize Group A's skippers through the effect of overburdening the crew with extra work for repairing purposes. Thanks to the pooling of repair costs, those boats which, in Group A, suffer from serious damage to their nets are not put at a disadvantage vis-à-vis competing units from Group B. Finally, we have emphasized that a balanced age distribution is a necessary condition if a group is to provide effective insurance against the risk of skill obsolescence among its members. Revealingly, total fishing experience of the skipper and crew (measured by the total number of years of fishing) varies significantly more among the boats comprising Group A than among those of Group B: the coefficient of variation is 0.14 for the former but only 0.05 for the latter. This fact may help explain why skippers of Group A have stressed the importance of skill- and expertise-sharing whereas those of Group B have remained silent on this aspect of collective functioning.

The Role of Protability Motives (i) As is evident from Table 13.2, the desire to avoid the various costs of crowding while operating in attractive fishing spots appears as the main reason (immediately followed by the previously discussed cooperation in net-mending and searching activities) stated by Japanese fishermen for adopting pooling arrangements. Among congestion costs, they often mentioned the risk of mutual entanglement of nets and boat collisions due to excessive proximity, the considerable waiting costs entailed by competition for the first turns in the most advantageous locations (‘previously, we used to spend nights in the boat in order to be the first to occupy the best spots’),264 the psychological and physical stress caused by the need to rush to the coveted place, and the scope for tension in human relations and interpersonal conflicts that racing to the best spots necessarily entails (racing creates a ‘conflicting atmosphere’ that arouses ‘a lot of anxiety’). Reduction of stress is mentioned by several skippers as a major advantage of effort coordination accompanying pooling (see Table 13.2, row 2). In particular, fishing is said to be a more pleasant job when a collective sets a maximum for the allowed fishing time and number of hauls: thanks to collective discipline, ‘we become less stressed and enjoy a more relaxed

264

Shiroebi stock tends to be confined to a rather limited area once it becomes exploitable. The density of fish varies according to precise locations within that limited area and, where it is high, it can actually give rise to acute competition among fishermen.

371

POOLING GROUPS IN JAPANESE FISHERIES

working atmosphere, which improves our quality of life as we are relieved from incessant anxiety and rivalry’. In fact, open and completely free competition is viewed by Toyama fishermen as utterly destructive precisely because it is the cause of stress and interpersonal antagonisms. In the next section, we will see that they are not totally opposed to competition, though: competition can be good and stimulating provided that it is maintained within reasonable limits. From Section 13.2, we know that pooling favors a better allocation of effort among different fishing locations through two distinct mechanisms. First, by inducing coordination aimed at achieving maximum aggregate catches, pooling has the effect of reducing congregation of boats either in spots that promise big catches (usually Heta) or in those that are less risky (usually Oki).265 If fishermen are risk-averse and do not practice pooling, they will tend to gather in the safer place (Oki) so as to maximize their expected private utility whereas, through coordination geared towards maximizing aggregate expected utility for the whole group, they would spread their fishing time more evenly between Heta and Oki. However, rules of rotating access are an alternative manner of allocating fishing effort (more or less) efficiently. Second, by providing insurance against the risk of net damage, pooling induces fishermen to operate in Heta more frequently than they would do otherwise. To test whether Group A, the pooling group, actually achieves a better allocation of fishing effort than Group B, the non-pooling group, we estimate and compare the expected marginal products of effort in Heta and Oki for these two groups. We expect marginal productivities for Group A to be more or less equal between the two locations. On the contrary, since Group B does not pool, its marginal productivities should diverge. Yet, as rules of rotating access prevail within this group, the absence of pooling should not cause any substantial efficiency cost owing to lack of coordination (assuming that these rules are more or less effectively enforced). As a result, a difference between expected marginal productivities for this group can be essentially interpreted as a risk premium. The estimating equation used for calculating the marginal products of fishing effort in both locations and for both groups has the following functional form:

where y′ is the marginal product measured in terms of catch per haul per boat; x stands for the order of net-casting considered in the total sequence of hauls performed on a given day; and D , D and D are dummy variables for s

265

d

g

Concentration patterns may, of course, differ from one day to another according to changing expectations of catches and risk.

372

JEAN-PHILIPPE PLATTEAU AND ERIKA SEKI

fishing spots, incidence of net damage, and groups, respectively. Assignments of the dummies are as follows: D = 1 for Heta, D = 0 for Oki; s s D = 1 for hauls with net damage, D = 0 for hauls without net damage; d d D = 1 for Group A, D = 0 for Group B. g

g

The unit of observation is the haul and the marginal product corresponds to the catch per haul which obviously varies, on any given fishing day, according to the turn which the boat concerned had in the total fishing sequence.266 The total number of observations is 142 hauls for Oki and 47 hauls for Heta. The results are shown in Table 13.6. The coefficients of all explanatory variables in the above function are significant at the 5 percent level of confidence. The coefficient of the effort variable measured by 1/x is significantly positive, thereby confirming the expectation of decreasing marginal productivity curves (the fish stock on any given day being a constant, the productivity of additional hauls is expected to decline). The coefficient of the location dummy is also positive, indicating that, for a given amount of fishing effort, marginal catch is higher in Heta than in Oki. The coefficient of the damage dummy has the expected negative sign, implying that marginal catches are significantly lower when the net is damaged. Finally, the coefficient of the interaction term is negative: for a given amount of effort, the marginal productivity of Group A is significantly lower than that of Group B when it operates in Heta. From the estimated equation given in Table 13.6, we have calculated the marginal products of effort for both groups and fishing locations, using the average number of hauls per day as the value chosen for x. We consider the expected values of marginal products defined as weighted averages of the estimated marginal catches in the presence and in the absence of net damages:

where p is the probability of a net damage. Table 13.6 Estimation Results of Marginal Product Functions (Based on Production Data Collected During June–End of August 1997) Fitted equation

t stat

266

16.337 (0.000)

3.305 (0.001)

1.903 (0.059)

−4.038 (0.000)

−2.469 (0.014)

d.f.

R2

189

0.19

Thus, if a boat from Group B made two different hauls in Oki on July 11th, one in the second position and another one in the eighth position (after other boats from the group had hauled in the same spot, once or twice), the corresponding catches will be regressed on x values of 2 and 8, respectively.

373

POOLING GROUPS IN JAPANESE FISHERIES

This formula is computed four times, one for each location and each group. The different probabilities of net damages and losses calculated as ratios of actual occurrences of such hazards to total numbers of hauls, are given in Table 13.7. The estimates of expected marginal productivities according to fishing locations and groups are shown in Table 13.8, together with the average numbers of hauls per fishing day. It is striking that, for Group A, marginal product of fishing effort is almost identical between Heta and Oki, thus suggesting an efficient allocation of fishing effort. By contrast, the marginal product of effort for Group B is significantly larger in Heta than in Oki (the former exceeds the latter by more than one-fifth), indicating under-allocation of fishing effort to Heta. Since Heta is the most risky location, the fact that Group B is only imperfectly insured against the risk of net damages and losses seems to explain much of this inefficient allocation of fishing time. In fact, a moderate amount of risk aversion (corresponding to a coefficient of relative risk aversion of about 1/3) is sufficient to account for the observed discrepancy between expected marginal productivities in Heta and Oki (see Appendix II for details). There are hints in the data that Group B's skippers go out fishing in Heta only when there are clear expectations of bumper catches (based either on high achievements on the previous day by Group A, or on promising signals through electronic devices), and they then take substantial risks as reflected in a high occurrence of net damages (see Table 13.7). Otherwise, they prefer to concentrate their fishing operations in the safer Table 13.7 Probabilities of Net Damage and Losses According to Fishing Locations and Groups* Group A Group B Groups A and B

Heta 0.039 0.238 0.081

Oki 0.054 0.037 0.047

Heta and Oki 0.047 0.084 0.060

Note: * Based on the data of net damages and locations self-recorded by skippers for 29 fishing days between June and August 1997.

Table 13.8 Fishing Intensities and Estimated Marginal Products of Fishing Effort in Heta and Oki, for Groups A and B

Group A Group B

Average number of hauls per day Heta Oki 10.64 15.73 4.15 12.82

Estimated marginal product Heta Oki 50.25 kg 51.34 kg 72.77 kg 57.47 kg

Relative gap 2.17% 21.0%

374

JEAN-PHILIPPE PLATTEAU AND ERIKA SEKI

location, Oki. On the contrary, skippers from Group A are willing to tempt their chance in Heta even when catch expectations are not much higher than in Oki. (ii) Synchronization of fishing activities and exchange of strategic information (which figures prominently in Table 13.2) are often jointly referred to by fishermen as mechanisms capable of generating considerable benefits and somehow related to income pooling. Regarding the latter aspect, a skipper thus remarked that ‘in pooling, it is not enough to increase one's own catch only; we have to increase the aggregate catch and this requires that we exchange information’; and another one to state that ‘pooling implies that we increase the total group's catch by encouraging each other’. Fishermen therefore seem to be aware of the fact that pooling helps overcome incentive problems arising in information exchange (see Section 13.2). Regarding the former aspect, it is worth noting that synchronization and information exchange operations will be effective only if they can be carried out under the leadership of one group member.267 In Group A, revealingly, a clear leadership role is granted to a senior skipper whose expertise is wellestablished (he is much admired for his skills not only at detecting fish but also at mounting and repairing nets). This leadership function is more problematic in Group B as we will substantiate in the next section. (iii) That cooperation in the repair of nets damaged and the search of nets lost is of crucial importance to fishermen has already been underscored while discussing insurance motives for pooling. The point was then made that collective repair and maintenance work allows the job to be done much more expeditiously, thereby reducing both the waste of fishing time and the risk of losing one's crew.268 What needs to be added now is that most fishermen in Groups A and B have strongly emphasized the existence of important scale economies in both searching and mending operations.269

267

According to several skippers, leadership has the additional advantage of eliminating the negative externality of mutual production of noises in individual echo-sounders, which unavoidably arises when the search is undertaken by all the skippers acting in a completely decentralized manner.

268

‘Today, most damages to nets can be fixed within thirty minutes to two hours which may have taken several days of overtime work for the crew before’, said one skipper from Group A.

269

In their own words: ‘nets have become too big to mount and maintain by oneself ’; ‘in old days, I could carry one net on my shoulders whereas we need several strong men to carry one today’; ‘we used to forego two days of fishing because of net repair whereas, nowadays, it has become rare to spend more than one day towards such a purpose’; ‘it is more efficient to repair nets collectively’; ‘overall, it is less time-consuming to repair nets collectively’; ‘the damage that costs 20 man-hours of labor if repair is done individually will cost only 15 man-hours if it is done collectively’; ‘lost nets can be found more easily by undertaking a collective search’; etc.

POOLING GROUPS IN JAPANESE FISHERIES

375

As is evident from Table 13.2, these two considerations appear to justify ample cooperation in net search and repair. (iv) Finally, Group A's skippers have pointed to the advantage of regulating fishing effort, yet not with the purpose of conserving the resource. As a matter of fact, and with the possible exception of one skipper, they are not convinced that the stock of shiroebi shrimps is in any way influenced by the total amount of fishing effort (the same belief is observed among skippers of Group B). For them, fish comes in Toyama Bay from elsewhere (from the open ocean where it spawns and breeds) and the quantity available for the current season is fixed and determined by ecological factors out of human control. If there is less resource available in Shinminato area this year as compared to last year, it is because, for some natural reason, the fish have moved in greater quantities to Iwase waters. This said, fishermen believe that by limiting effort they can cause fish prices to rise. Since they believe the resource stock available to be a fixed quantity for the whole season, they are persuaded that by rationing present fishing effort they do not run the risk of foregoing potential fish output. Their major concern is to obtain the highest possible prices for the available fish over the whole season and, therefore, to avoid overflooding the market on any given day. In other words, staggering harvest is viewed less as a way to smooth incomes intertemporally than as a way to maximize incomes. As evident from the interviews, Group A's members have well understood that, by inducing them to catch as much fish as possible, competition with Group B undermines their regulatory efforts during the period of alternate fishing. Historical antecedents must also be borne in mind at this point. It is indeed in a context of protest against lack of competition among fishmerchants and middlemen, due to either collusive practices (in Shinminato) or the insignificant number of buyers (in Iwase) that pioneer attempts at instituting pooling arrangements have been undertaken in Toyama Bay in the early 1960s.270 Clearly, therefore, enhancing market power vis-à-vis strongly organized merchants is a powerful motive that prompted fishermen to adopt such arrangements. In order to assess if Group A has succeeded in meeting this objective of limiting fishing effort and increasing incomes, we have constructed Tables 13.9 and 13.10. In Table 13.9, we compare catch and price observations

270

Ample opportunities for cheating or exploitation of market power existed as there were no organized auction systems. In order to protest against this situation of unequal bargaining strength, fishermen in Shinminato and Iwase decided to boycott the traders' cartels in 1961 and they exercised strong pressures on the FCAs to organize fish auctions. In Shinminato, this demand was met and an auctioning system was indeed introduced. The effects of this measure were soon felt and the merchants' cartel was dissolved under the pressure of open competition.

376

JEAN-PHILIPPE PLATTEAU AND ERIKA SEKI

Table 13.9 Comparison of Daily Landings of Group A Between the Two Fishing Periods (April–May and June–August 1997)

Average daily landing of fish after the first haul (for all boats) Average daily landing after the subsequent hauls (for all boats) Average daily landing of first haul compared to subsequent hauls Average lowest price of the day

April–May (period I)

June–August (period II)*

470 kg

442.6 kg

Significance of mean differences between the two periods F stat: 0.17 (0.6788)

922.9 kg

1,330.1 kg

F stat: 6.03 (0.0173)

0.51

0.33

1,288.8 yen

826.4 yen

F stat: 99.65 (0.0000)

Note: * During some days of this period, unlike observations in period I, Group A decided to go out fishing even though no auction took place. The corresponding data have been removed from the sample in order not to distort the relationship between catches and (minimum) shrimp prices.

for Group A in Shinminato between the two successive fishing seasons. Catches are differentiated according to whether they result from the first or from subsequent hauls whereas only the minimum prices obtained on a day are reported. As for Table 13.10, it compares the situations in the three sample groups over the two fishing seasons (when the comparison is relevant), yet this time without distinguishing between output from the first and the subsequent hauls. All data have been obtained from the local FCAs. What can be concluded from Table 13.9? To begin with, the average daily catches achieved in the course of the first haul do not significantly differ between the first and the second fishing seasons. The opposite conclusion obtains when the catches resulting from subsequent hauls of the day are compared across the two seasons. The interpretation we draw from these combined findings is the following: once they have assessed the market conditions of the day after the disposal of the initial catches, fishermen of Group A better succeed in limiting the size of their landings when they operate alone (that is, during the first season, from April to May) than when they face the competition of Group B (that is, during the second period, from June to August). The fact that the catches from the initial haul represent more than 50 percent of those from the subsequent hauls during the first period but hardly 40 percent during the second one is another way to look at the same matter (see the third row of the table).

377

POOLING GROUPS IN JAPANESE FISHERIES

Table 13.10 Comparative Catch Performances of the Three Sample Groups Over the Two Fishing Seasons and Minimum Prices Obtained on a Day (1997) (1) Average monthly aggregate catch (kg) (2) Average number of fishing days/month (3) Average daily catch per boat (kg)

Groups A + B Group C Groups A + B Group C Group A Group B Group C (4) Average lowest price of Group A the day (yen) Group B Group C (5) Average daily gross Group A income per group (yen) Group B Group C (6) Average daily gross Group A income per boat (yen) Group B Group C (7) Average monthly gross Group A income per group (yen) Group B Group C (8) Average monthly gross Group A income per boat (yen) Group B Group A + B Group C

April–May (period I) 25,139 30,389 19.5 11.0 188.7 — 450.6 1,271.0 — 700.0 2,006,019 — 2,045,805 286,574 — 340,949 39,117,370 — 22,503,855 5,471,193 — 5,471,193 3,750,439

June–August (period II) 30,430 42,524 25.0 13.7 221.6 170.9 520.1 784.3 809.8 650.0 1,347,199 828,486 2,154,011 192,457 165,697 359,002 17,109,427 10,190,378 29,509,951 2,444,204 2,038,073 4,482,277 4,918,327

It is equally striking that the minimum price of the day—which can be considered as the reservation price below which fishermen refuse to continue fishing—is, on average, significantly higher during the first than during the second season (see the fourth row), thereby suggesting that limitation of fishing effort offers a reward in the form of higher prices. Note incidentally that the insignificant difference between the average catches obtained from the first haul in the course of the two seasons suggests that there is no clear interseasonal variation in supply conditions. If these conditions had varied significantly across the two seasons, more sophisticated tests would have been required to assess the regulatory effect of pooling. From Table 13.10, it is immediately apparent that the resource is much more abundant in Iwase than in Shinminato since landings are considerably larger in the former than in the latter fishing zone (see row 3) and Group C in Iwase has not succeeded in limiting fishing effort. (Bear in mind that this large differential in landings between the two zones may not be mitigated or erased through migrations of fishermen from Shinminato

378

JEAN-PHILIPPE PLATTEAU AND ERIKA SEKI

to Iwase since property rights over local resource stocks are vested in the local FCA, as explained in Section 13.1). The reverse side of the coin is that prices are lower in Iwase (row 4), yet not enough to prevent gross incomes of Iwase fishermen from significantly exceeding those of their Shinminato colleagues. Note that the same inverse relationship between landings and unit prices is also observed while comparing performances of Groups A and B in Shinminato between June and August. Indeed, if the average daily quantity of fish caught per day by a boat belonging to Group A is 30 percent larger than that caught by a boat belonging to Group B,271 when prices are taken into account and gross proceeds rather than physical catches are considered, the advantage of the former group over the latter is only 16 percent. In other words, prices have a stabilizing effect on incomes. Another, quite expected finding is that during the second period members of Group A work more intensely than during the first period in the sense that they go out fishing during a larger proportion of the potential number of days open for fishing: they work during 85 percent of the total number of days allocated to them, as compared with a ratio of actual to potential fishing time of only 65 percent during the first period when they have exclusive access to the water. This, of course, is a natural response to reduction of potential fishing time as a result of shared rights of access to the fishing ground. Notice that Group B's members use a lower proportion of their available time than their colleagues of Group A during the second period, yet use a larger proportion compared to what the latter do during the first period.272 A last comment is in order. In assessing the aggregate catch performance of Groups A and B clubbed together, it may be observed that the increased performance of Group A during the second period is quite considerable since it largely outweighs the negative effect on landings of the comparatively small size of Group B's fleet that operates one day out of two during that period (see rows 1 and 3). Yet, owing to the fall of price between the two periods and despite the fact that, in relative terms, Group A works more intensely during that period (see row 8), the average daily gross income per boat during June–August is only 68 percent of the figure realized

271

This superior average catch performance of Group A must be entirely ascribed to their high comparative achievement in Oki since in Heta Group B is actually more productive (bear in mind that members of the latter group tend to go out fishing in Heta only when there are good prospects of bumper catches). As intensity of fishing in Heta relative to Oki is comparatively low, especially for Group B, the comparative advantage of Group A in the latter location determines the net result.

272

During the second period, Group B's members do not only use a lower proportion of their potential fishing time than their colleagues of Group A, but they also operate less intensely on a given fishing day: thus, our own data, that concern only a part of the whole June–August period (16 out of 38 days for Group A and 10 out of 37 days for Group B), show a figure of 3.77 hauls per boat per fishing day for Group A, as against 3.40 hauls for Group B.

POOLING GROUPS IN JAPANESE FISHERIES

379

by that group during April–May. The monthly gross income per boat aggregated over Groups A and B during the second period (that is, 4,482,277 yen) is smaller than the same statistic obtained for the first period by Group A alone (see, again, row 8).

13.4 Explaining Success and Failure to Pool: Group Heterogeneity and Preference for Local Status Theory That pooling and the whole set of interlinked attributes of which it is a part is an attractive arrangement for the coastal fishermen of Japan is attested by the fact that all the three sample groups at least tried it out. Yet, only Group A has succeeded in making income- and input cost-pooling a sustainable venture lasting for more than 35 years. Group B's attempt, if we abstract from the fact that two of its fishing units have continued to somehow pool together, has been short-lived (it lasted only one year) while, after about ten years of existence, the pooling mechanism of Group C has been drastically amended with the result that only 10 percent of the total value of landings remains subject to equal sharing among members (see Section 13.2). It is therefore interesting to know why Group A has been so successful and what is the nature of the difficulties encountered by the two other groups which eventually led to the breakdown of their pooling scheme. Group size cannot be a valid explanation of Group A's relative success since it actually has the largest size. The present section consists of two steps. First, we build up a simple theoretical framework capable of explaining success and failure to pool simultaneously. And, second, we present empirical evidence, both qualitative (fishermen's perceptions) and quantitative, in support of the theory. Pooling arrangements (or egalitarian partnership schemes) are comparable to voluntary contribution mechanisms in the production of public goods.273 Free riding is therefore a dominant strategy yielding Pareto-inferior outcomes for all. In the pooling groups considered in this study, such a problem is nevertheless limited in so far as relative individual catch performances or skill levels can be well observed (mutual monitoring is easy)

273

Group output is indeed a kind of public good which is nonexcludable and shared equally by the workers. Experimental evidence confirms that equal revenue-sharing elicits low effort levels from workers while group incentive schemes such as target based schemes forcing contracts (in which a penalty wage is paid to workers in case their collective output falls short of a target, otherwise they equally share in total income) yield much more satisfactory results (Nalbantian and Schotter 1997: 320–1). As for empirical studies, the bulk of them show strong positive effects of group incentives on various productivity and financial performance indicators (see Nalbantian and Schotter 1997: 316, for references).

380

JEAN-PHILIPPE PLATTEAU AND ERIKA SEKI

and skippers have the means to exert peer pressure on their colleagues.274 Such was the argument underlying Section 13.2. To this argument we now want to oppose the view that, even in conditions where free riding is hard to detect, it might not represent a serious problem threatening to undermine pooling arrangements. Indeed, fishermen entering into a pooling group become motivated by popularity or esteem considerations and, if relative individual catch performances or relative skill levels can be well measured, they make interpersonal productivity comparisons which they use as the basis for mutual ranking.275 Since they are keen to obtain the best possible ranking, they are motivated to exert much effort, which benefits the entire group (see also Frank 1985: 34–107, 1988: 179–83; Bernheim 1994; Weiss and Fershtman 1998; Postlewaite 1998).276 Competition for rank is therefore a form of peer pressure which elicits high levels of effort under conditions of imperfect effort observability. Cooperation, however, will not be sustainable if the group is too heterogeneous. The above contention that, paradoxically, heterogeneity may help to sur-mount incentive problems because heterogeneous people accommodate themselves into the cooperative norm by simultaneously instituting a culture of ranking is clearly at odds with standard economic theory. This theory

274

As argued by Kandel and Lazear (1992), peer pressure and some form of revenue sharing tend to go hand in hand: ‘peer pressure is expected to be a more effective force in firms in which profits are shared among those in similar circumstances’, with the effect that individuals empathize with those whose income they affect (p. 816). As for mutual monitoring, it appears as a specific application of peer pressure and is effective only when incomes or profits are shared by a very small group. Bear in mind that in the experimental study referred to in the previous note, no peer pressure is allowed for, nor can any work norm be established by workers and enforced upon each other (decisions are completely decentralized and no communication takes place among workers).

275

Yoram Weiss and Chaim Fershtman (1998) consider that ‘Individuals seem to care about their ranking and the esteem of others, even if they derive no clear economic benefits, and are willing to pay respect to others and to modify their behavior accordingly, without receiving any direct benefit’ (p. 802). Contrariwise, in the line suggested by Frank (1985: 48–55), we contend that people who pay respect to others ask for a compensation in terms of direct economic benefits because being in a low rank position generates disutility. Note, on the other hand, that sensitiveness to esteem or popularity is different from altruism. As aptly remarked by Douglas Bernheim: ‘An altruistic person cares about how someone else feels, whereas a person motivated by popularity or esteem cares about how someone else feels about him (or her )’ (Bernheim 1994: 843).

276

There is an interesting analogy between our story of fishermen's pooling groups and the analysis of Japanese industrial firms by Aoki (1988: chs. 2–3, 1990). The gist of Aoki's argument, indeed, is that such firms (called J-firms) are characterized by a horizontal mode of coordination among operating units (with extensive sharing of ex post on-site information) compensated by incentive systems based on rank hierarchies. In other words, a relatively egalitarian mode of operation that makes individual pay independent of specific job performances is coupled with employee competition for promotion in rank throughout their careers. It is noteworthy that each rank is associated with a certain level of pay rather than with a certain level of social esteem and prestige as observed in our fishermen's pooling groups. These J-firms are contrasted with Hfirms from the Western world that combine a relatively more hierarchical approach to coordination with the relatively decentralized market approach to incentives in which specific jobs are related to competitive wages.

POOLING GROUPS IN JAPANESE FISHERIES

381

indeed predicts that when informational asymmetries exist even a perfectly homogeneous group will find it difficult to persist owing to the pervasive presence of incentive problems (in the absence of such asymmetries, group heterogeneity always has adverse efficiency consequences).277 Such problems are expected to be especially acute if income is shared equally among members irrespective of their individual contributions to output. For example, failure of worker cooperatives is typically blamed on dulled incentives as a result of imperfect information. In a recent paper, Michael Kremer (1997) has argued that distortion of incentives is caused by egalitarian wage policies that are usually followed in these cooperatives and that he attempts to endogenize. He thus supports the argument by Henry Hansmann (1988, 1996) or by Lee Benham and Philip Keefer (1991) that heterogeneity among cooperative owners creates inefficiency. To better highlight the critical importance of competition for rank in pooling groups, we present a graphical argument that leads to two different equilibrium outcomes, one corresponding to success in pooling and the other one to failure. Each skipper is confronted with a discrete-choice problem: he decides to enter a pooling arrangement if his utility derived from pooling is higher than his reservation utility (the utility obtained when he operates singly), otherwise he prefers to stay out of pooling. We assume that there are three components in the skipper's utility function: his income, his effort, and his rank. His rank can be simply measured as the relative magnitude of his own level of the yardstick used (e.g. skill levels, or catch performances) compared to the average level obtaining in his pooling group. Regarding effort, we assume that there are no labor-shirking effects arising from pooling so that effort is identical whether a skipper pools or operates singly. This assumption is warranted by the fact, explained above, that relative status considerations prompted by pooling induce skippers to exert a lot of effort.278 Finally, we assume that, under pooling, output is equally shared among all skippers (fishing units) irrespective of their individual performances. In other words, the equal-sharing rule

277

If group heterogeneity is not too great, so our argument runs, local status considerations can substitute for missing information. A more realistic stance is probably to look at rank-based behaviors and good information as mutually reinforcing factors that have the effect of enhancing trust: rank-conscious skippers are induced to make good catches and others, by observing their efforts, are strengthened in their belief that everybody tries his best. Heterogeneity of performances is deemed all the more acceptable by the best-skilled skippers as they hold such a belief. This said, the assumption that effort can be perfectly inferred from boats' landings is probably too strong. And, if that is the case, relative status considerations must be considered as a useful complement to prevailing information conditions.

278

In fact, if rank-related motives are strong enough, we cannot exclude the possibility that skippers would produce more effort by pooling than by operating singly. In the words of Weiss and Fershtman (1998) : ‘Since social status is inherently relative and one person's gain is another's loss, status seeking may lead to excessive effort, as all try to climb up without any change in relative positions’ (pp. 803–4). Here, however, we rule out this possibility because it is too favorable to pooling.

382

JEAN-PHILIPPE PLATTEAU AND ERIKA SEKI

is exogenously given. The idea here is that differentiating shares would involve recognizing publicly that some people are more valuable than others, which is bound to create tensions among skippers and thereby lower their morale and undermine their team spirit. In the words of Douglas Bernheim: ‘Unequal pay invites the inference that one worker is more valuable than another. This inference may offend the less valued workers, causing him or her to become even less productive’ (Bernheim 1994: 866).279 Clearly, the decision to pool or to operate singly is the net outcome of two effects running in opposite directions. For skippers with above-average skill levels, the negative effect operates through the pooling mechanism since the comparatively low level of skill of some skippers results in lower incomes (and higher expenditures) for those with large skill endowments. As for the positive effect, it results from the fact that the latter derive a satisfaction from the privilege of occupying top-ranked positions in the group: having better than ordinary skills brings you respect and prestige within the pooling group. For skippers with below-average skill levels, the same two effects are also at work, yet they operate in the converse way: the pooling mechanism favors this category of skippers while their low position in skill ranking causes them disutility. A direct implication of this utility representation is that skippers, whether top- or lesser-ranked in the skill hierarchy, may have a certain tolerance for skill heterogeneity in a pooling group, the former because they are willing to forego some income in order to gain prestige by working with less productive colleagues, and the latter because they are ready to endure low local rank in order to earn a higher income. However, since tolerance for heterogeneity may be too small to accommodate actual skill differentials, nothing ensures a priori that the pooling solution is feasible. This becomes plain in Fig. 13.2. For the sake of simplicity, let us also assume that there are only two skippers, i and j, and that we look at the choice problem of skipper i in the utility-skill space. Along the horizontal axis, we measure the skill level of skipper j while the level of skipper i himself is constant and marked out with a given point denoted s (relative catch performances could i be measured along this axis without altering the essential argument). In other words, relative position in the skill hierarchy is what is effectively measured along this axis. The horizontal line describes the utility obtained by skipper i when he operates singly (reservation utility). The utility function of the same skipper when he pools with skipper j is described by the curve U pool when there are profitability and insurance gains associated with pooling. When i such gains are not present, the utility from pooling is depicted by the curve U share. i

279

Interestingly, Holmström (1982) has shown that in certain differentiable, monotonic partnerships, no sharing rule can elicit an efficient set of actions.

POOLING GROUPS IN JAPANESE FISHERIES

383

Fig. 13.2 Choosing To Pool Or Not To Pool When There Are Profitability and Insurance Advantages Associated With Pooling, and When Group Heterogeneity and Local Status Effect Are Present

These two curves, which have the form of a butterfly, represent the combined effects of income transfer and local status. The local status curve that underlies them is asymmetrical because the negative rank-related utility from pooling with a better skilled partner continuously increases, in absolute terms, as the difference in skill levels rises, whereas the positive rank-related utility from pooling with a lower skilled partner increases as the differential skill ratio grows, but only up to a point beyond which it starts to decrease, hence the hump on the left-hand side of both curves. The intuition is that social esteem can be gained from interacting with people who admire your better skills or performances, but only provided that they are not too mediocre. If they are of too low a level compared with you, you may not attach any weight to their admiration feelings and even incur displeasure by interacting with them.280 The curve Uishare is constructed in such a way that the head of the butterfly is precisely located on the line at the point corresponding to the skill level of skipper i, s . Indeed, it must be the case that, in the absence of any profitability i and insurance gains from pooling, two skippers who have identical skill levels are indifferent between pooling and not pooling. If such gains exist, the U share curve is lifted up vertically by the amount which they represent. i

280

Frank (1985: 59) adopts a similar position when he writes that ‘it is the comparisons between the occupants of closely ranked positions that are most important’.

384

JEAN-PHILIPPE PLATTEAU AND ERIKA SEKI

It is evident from Fig. 13.2 that, in the absence of intrinsic advantages of pooling (insurance and profitability advantages), skipper i would agree to enter a pooling arrangement with skipper j provided that the latter's skill level is comprised in the [s *, s **] interval. This is because, on the right-hand side of point s , income transfer advantages j j i outweigh prestige losses while, on the left-hand side of point s , prestige gains outweigh income transfer losses. If, for i the sake of convenience, we make the simplifying assumptions that the utility curves (including the reservation utility curves) are identical for both skippers, and that they are symmetrical, skipper j will accept to pool with skipper i provided that their relative skill level is comprised in the same [s *, s **] interval (see Seki and Platteau 1999, for more j j details). In other words, this interval is the mutually acceptable range of skill ratios that makes pooling possible. When intrinsic pooling advantages exist, this interval is enlarged from [s *, s **] to [s +, s ++]. j j j j So far, we have assumed that skill levels are exogenously given. In actual fact, however, the skills of the lower-ranked skipper can be improved, up to a point, through learning from his more experienced colleague.281 There are three points to add to the previous discussion that are suggested by this new effect. First, taking account of the skillenhancing effect of learning, skipper i, the expert skipper, is likely to tolerate more skill heterogeneity in the initial phase of the team life than is indicated in Fig. 13.2. Second, a direct implication of the dynamics of mutual learning with pooling is that, in the event where initial skill heterogeneity is not too large, thus making pooling feasible and actually adopted, this heterogeneity will be endogen-ously reduced over time. On the contrary, if initial skill heterogeneity is too large to be acceptable for high-ranked skippers, it will remain more or less constant since lowranked skippers will not have the opportunity to learn from more skilled colleagues. However, and this is our third point, there may be an endogenous limit to the extent that the pooling mechanism can improve the skills of the lowest-ranked skippers, and this limit arises from the operation of the local status effect. (We assume that the expert skipper does not gain prestige for having trained a partner.) To see this, assume that skipper j starts with an initial skill level equal to s + and that his maximum attainable level exceeds , implying that it is located j on the right of point T, the left-hand side local maximum of the curve Upool . In this case, skipper i will not be induced i to help skipper j realize his full potential in so far as his own utility would decline as a result. As a matter of fact, at the right of point T, any upgrading of skipper's j skills causes a loss of prestige and status to skipper i that exceeds his income gain through the mechanism of pooling. To sum up, because effort to learn may enhance the skill level, a strong preference for high ranking in the skill hierarchy will induce lesser-endowed skippers

281

Frank has also taken this ‘learning effect’ into consideration (1985: 57–8).

POOLING GROUPS IN JAPANESE FISHERIES

385

to put in a lot of learning effort. Yet, their efforts may be frustrated if the better-endowed skippers refuse to themselves apply effort to teach them new skills and are effectively able to prevent further learning effects from occurring (a condition that may not be easy to satisfy in reality). As our graphical argument illustrates, this possibility exists. Incidentally, it may be noted that, given the choice, skipper i would prefer pooling with partners possessing lower or higher amounts of skills than himself—provided that the skill differential is not too big—to pooling with partners having identical skills. From Fig. 13.2, it can thus be seen that any partner-skipper with a skill level lower than s by not i more than ME′ units, or higher than s by not more than E′N units, will cause skipper i to have greater utility than he i could achieve by pairing with an identical partner (skill-wise). In other words, tolerable heterogeneity is better than homogeneity. An important lesson from the above theoretical analysis is that there are two group-specific reasons capable of explaining why one group may succeed in pooling and another one may simultaneously fail in the same general circumstances (ecological, technical, or economic). On the one hand, the skill differential between willing participants in a group may be too wide to be acceptable to the best-skilled skippers and, perhaps, to the lesser-ranked too. On the other hand, it may be the case that leading skippers in one (potential) group assign relatively low value to occupying a top position in the skill hierarchy. As a consequence, their tolerance for heterogeneity will be much smaller than in another group whose expert skippers view the occupation of a top local status position as a source of much symbolic power. Therefore, even though skill heterogeneity is identical in two sets of skippers, one may succeed in forming a pooling group while the other may fail due to different preference profiles of their leading skippers. Note that the same reasoning also applies to the lower tail of the skill distribution: a pooling group may fail to work because its lesserranked skippers are more prone to a ‘shame’ effect than their counterparts in another group. Of course, different surrounding circumstances are also liable to influence the success or failure of pooling groups. This is because the environment (ecological, technical, or economic) may affect the size of the potential benefits to be derived from coordination of efforts or joint operations (whether under the form of insurance or profitability gains). Furthermore, the environment may not only differ from one place to another, but also vary over time. Thus, it may change in such a way as to enhance the incentive to form pooling groups. For example, a technological change that increases the scope for scale economies or the risk of existing technical skills becoming obsolete has the effect of shifting upwards the skippers' utility curve under pooling, as depicted in Fig. 13.2. As a result, pooling becomes more appealing. Other changes, such as the emergence of

386

JEAN-PHILIPPE PLATTEAU AND ERIKA SEKI

market power in the context of changing demand, may as well modify the relative attractiveness of pooling arrangements.

Empirical Evidence One of the most striking findings emerging from in-depth interviews with fishermen is the glaring difference in attitudes, perceptions, and judgments between the members of Groups A and B in Shinminato (see Seki and Platteau 1999, for detailed excerpts from these interviews). If all skippers from both groups have repeatedly pointed to the pervasive incidence of inter-individual comparisons of performances, skippers from Group A tended to stress positive aspects of intra-group competition for rank while those from Group B underlined its negative effects. For the former, inter-individual comparisons act as an incentive to work hard and manage assets cautiously while for the latter, on the contrary, such comparisons arouse anxiety, stress, and tensions that ultimately undermine effort, lower morale, and destroy trust. Just for the sake of illustration, a representative statement of the perceptions of Group A's skippers is: ‘Pooling will not be possible if one thinks of taking it easy while others work hard; though we cooperate, we have a competitive spirit which is healthy for fishermen to maintain; personally, since everyone wants to be the best, I work hard to keep abreast of the others’. Contrast it with the following statements made by Group B's skippers: ‘We always compare each other's landings and discreetly criticize those who bring lower catches. This is sickening’; ‘When I have no catch, or when I damage my net repeatedly, I feel strained because I anticipate the others’ criticism. At least, when I operate on my own, I am free from this kind of pressure'; ‘My crew is rather slow in working. By pooling, we are expected to work as quickly as the others, and that proved to be impossible. It was a terrible distress for us’; ‘Pooling was frustrating not only for those with higher catches but also for those with lower catches. The former found it unjust to share their catches with the latter and the latter did not like feeling ashamed’. It is interesting to notice that Ruth Benedict's contrasting picture of Japanese and American cultures is disproven in this case: indeed, while what she says seems to apply well to Group B—unlike the American people, the Japanese fear competition and the individual comparisons that it entails because of a deep-rooted sense of honor and desire to avoid anxiety and shame feelings—it does not fit at all with the situation observed in Group A (Benedict 1946, ch. 8). The theory outlined in the previous subsection suggests three distinct lines of inquiry to account for the different trajectories of Groups A, B, and C. These three lines appear to be pertinent in the present case study. The first explanation is based on considerations of initial heterogeneity of

POOLING GROUPS IN JAPANESE FISHERIES

387

the sample groups while the second one refers to differences in the pattern of the members' preference profiles and the third one to differences in the environment in which the various groups operate. Let us look at the three of them in succession. The first hypothesis is that, unlike in Group A, heterogeneity of performances is too big to make a pooling scheme viable in Groups B and C. This hypothesis is strongly suggested by statements of skippers from the latter two groups, especially Group B, to the effect that excessive catch variations among participating units are the critical reason accounting for their failure to pool (see above). In actual fact, one skipper from Group B explicitly said that ‘differences in catches were too large for an agreement to be possible on the continuation of pooling’, and in Group C, reference was made to wide differences in individual performances as the main cause for the radical changes they made in order to dilute their pooling arrangement. One way of testing the hypothesis of excessive heterogeneity is to compute catch variations between members belonging to a given group and check whether they are significantly larger in Groups B and C than in Group A. The test can be further refined by comparing such intra-group variations with some norm considered as acceptable by the fishermen themselves, skippers in particular. This test is possible because in the course of interviews explicit ranges of tolerable variations of performances have been mentioned by some skippers. Thus, from a retired skipper who previously worked in Group A, we learn that a catch representing 70 percent or more of the highest catch of the day is an acceptable performance while a catch lower than 60 percent of the highest catch is definitely not; in between these two landmarks, doubt is permitted and disagreement will only occur if the bad performance is repeated. Another skipper from Group A considered it acceptable ‘that the lowest catch in the group works out to about 70 percent of the highest catch provided that the least performing skipper and crew do their best’ (free translation). On the other hand, a skipper from Group B pointed out that a disparity of performances of 50 percent between the highest and the lowest average catches is unacceptable. Disparities in individual catches per haul in the three groups are shown in Table 13.11, where they are measured by the differences between individual performances and the group mean expressed as proportions of the mean itself (see the figures between brackets). The average of the absolute values of these individual gaps has also been computed for each group. Another synthetic group-specific index of average disparity is provided in the form of a coefficient of variation that abstracts from time fluctuations in the catches of individual boats. Finally, the distance between the lowest and the highest catch performances is measured in proportional terms. As is evident from Table 13.11, all indicators converge to show that while disparities are moderate in both Groups A and C, they are quite

388

JEAN-PHILIPPE PLATTEAU AND ERIKA SEKI

Table 13.11 Comparative Degrees of Heterogeneity in the Three Sample Groups, as Measured by Output Performances Group A Average catch per haul * (in kg) Boat 1 78.16 (0.1011) Boat 2 70.78 (−0.0029) Boat 3 72.45 (0.0208) Boat 4 79.07 (0.1139) Boat 5 63.00 (−0.1124) Boat 6 70.10 (−0.0124) Boat 7 62.81 (−0.1151) Group-average 70.98 Average relative gap** 0.0684 Coefficient of variation 0.0843 Proportional distance*** 79.4%

Group B

Group C

92.16 (0.4026) 64.16 (−0.0235) 77.91 (0.1858) 47.23 (−0.2812) 48.73 (−0.2584) — — 65.71 0.2303 0.2619 51.2%

246.77 266.89 266.51 255.39 253.96 223.29 — 252.25 0.0451 0.0583 83.9%

(−0.0218) (0.0580) (0.0565) (0.0124) (0.0068) (−0.1148)

Notes: * Figures between brackets are the distances between individual average catch performances and the group mean expressed as a proportion of that mean. ** The average relative gap is simply the mean of the figures given in brackets and taken as absolute values. *** The proportional distance is the lowest average catch per haul in the group expressed as a proportion of the highest catch.

large in Group B. Contrary to expectations, heterogeneity in Group C turns out to be even smaller than that obtaining in Group A. In addition, it bears emphasis that the proportional distance between lowest and highest average catches is well within the tolerable range in Groups A and C but clearly outside of it in Group B.282 The above tests, however, are not exact because they suffer from an endogeneity bias. Indeed, we know that differentials in skills and performances are not constant but are capable of being narrowed down thanks to

282

Note also that the average catch per haul of the best performing boat in Group B (boat 1) exceeds by 16 percent that of the corresponding best performing boat in Group A (boat 4) while the catch per haul of the second best performing boat in Group B (boat 3) is of the same order as that of the first and second best performers in Group A (boats 1 and 4). At the other end of the spectrum, the opposite situation obtains: the average catch per haul of the two worst performers in Group B (boats 4 and 5) is 21–25 percent lower than that of the worst performer in Group A (boat 5). Catch performances of all members of Group C are distinctly larger because they operate in a different area which is ecologically more favorable.

POOLING GROUPS IN JAPANESE FISHERIES

389

sharing of skills and experiences within a pooling group. For Group A, therefore, the reported disparities do not correspond to initial, ex ante differences, but to ex post differences as they have evolved under the influence of the pooling mechanism itself. In other words, it is likely that initial intra-group catch (and skill) variations were wider when the group started pooling in 1962. By contrast, disparities among members of Group B, because they reflect the current situation in which they operate singly, do not allow for the kind of learning effect which sustained pooling permits. In other words, we overestimate the gap in the extent of intra-group variations between Groups A and B. There is nothing we can do about this because we do not know the catch performances (or skill levels) of Group A's members when they started pooling together (i.e. we ignore initial heterogeneity), nor what the performances of Group B's members could have been had they pooled together for at least a few years. This being said, the gap in the heterogeneity of skills and performances between Groups A and B is so large that it is highly unlikely that it would vanish if we could control for the aforementioned endogeneity effect. Or, to put it in another way, given the extent of this difference, we can safely presume that initial heterogeneity was significantly lower in Group A than in Group B. Moreover, it is revealing that one skipper from Group B, who, incidentally, manages only the third best performing boat in that group (see boat 2 in Table 13.11), explicitly ascribed its failure in pooling to the fact that the least performing units caught only about half of the fish landed by the best performing boat (boat 1)—almost exactly the same proportion as that reported in Table 13.11. Clearly, he did not believe that exchange of skills and expertise in the context of pooling could reduce such a wide initial heterogeneity so as to make it fall within the acceptable range in a not too distant future. Another way to approach the problem of group heterogeneity is by measuring, for each individual unit, the gain or loss it would make by pooling or by withdrawing from pooling (depending on whether the group considered is presently pooling or not); and then assessing the disparities in these gains or losses between the group members. This exercise leads again to the conclusion that disparities in individual gains and losses from pooling are significantly larger in Group B than in Groups A and C (see Seki and Platteau 1999). It is interesting to notice that while Group A (and Group C) formed in challenging circumstances, in the wake of a collective protest against the exploitative practices of fishmerchants (see above), Group B adopted the pooling arrangement as a sort of visceral response to the initiative of Group A, without having a good grasp of the advantages of pooling. In the words of one skipper of Group B: ‘Without much questioning, we decided to pool in order to be as good as Group A; we thought that by emulating its organization, we would soon become able to compete with its fishermen’. As a

390

JEAN-PHILIPPE PLATTEAU AND ERIKA SEKI

result, Group B's members may have lent too little attention to the heterogeneity of performances among them. Also to be pointed out is that differences in catch performances and skill levels can be properly assessed only after people have operated together for some time. Through observations in the course of fishing operations, members gradually learn about these differentials. It may thus happen that heterogeneity of membership is underestimated when pooling is initiated, and this is all the more likely to be the case if the decision to form a pooling group is taken hastily or in a reactive manner as indicated above for Group B. The conclusion is that group heterogeneity, as measured by interindividual differences in catch performances as well as in gains and losses from pooling, is capable of explaining the failure of Group B. Regarding Group C, however, the evidence does not suggest that the same explanation could hold true since, in terms of heterogeneity, this group is much closer to Group A (it is even more homogeneous, in spite of the fact that exchange of skills and expertise is much less intensive owing to the absence of full pooling) than to Group B. The second hypothesis derived from our theoretical framework—groups are more likely to succeed in pooling if their members' preferences accord more importance to popularity and social esteem considerations—may only be tested for a comparison between Groups A and B, unfortunately. Indeed, the information gathered from the qualitative interviews with skippers from Group C in Iwase does not allow any solid inference about preference profiles, an outcome which might be due to the fact that less time was spent in interacting with them. The noticeable finding regarding preference profiles of group members is that there is a clearer leadership structure in Group A than in Group B and that the leader of Group A, boat 1, seems to derive much utility from his position at the top of the group's skill hierarchy. This comes out of both the statements made by this leading fisherman himself and the statements made by the other skippers in the group. Almost everybody in Group A mentioned the fact that the skipper of boat 1 is very keen to be the best among them, yet at the same time expressed a lot of admiration for his skills and his cooperativeness. For instance, there is much reckoning of the great ability of boat 1's skipper to detect fish, or design and handle nets properly as well as of his willingness to share his privileged information and knowledge with his less experienced colleagues. Precisely because he gains much appreciation for his talent and much recognition for the fact that he makes others benefit from it, he is ready to accept disparities in his group in terms of skill levels and catch performances. He is actually rewarded for this income sacrifice in terms of symbolic income under the form of esteem, prestige, and local status. His authoritarian proclivities are more or less accepted because of his intrinsic skills, charisma, and consideration for others. This seems to be

POOLING GROUPS IN JAPANESE FISHERIES

391

an important feature of Group A in so far as boat 1's skipper is also the member who would gain most (income) by giving up pooling (again abstracting from pooling's intrinsic advantages). Note that the skipper of boat 4 is also an outstanding fisherman as witnessed by his catch record, yet he keeps a low profile, partly because he is younger and has learned a lot from the leading, and older, skipper who apparently intends to make him his successor at the head of the group. In Group B, by contrast, even though there is also a leading expert skipper who manages boat 1 (we have seen above that he significantly performs better, on an average, than the leading two skippers in Group A), the material from the interviews does not suggest that he is recognized and appreciated as a leader in the same way as boat 1's skipper is in Group A. He is actually considered as a boastful person who continuously attempts to impress others in a rather forced manner which irritates his colleagues. Moreover, his behavior is rather individualistic and self-imposing. One skipper thus remarked that ‘he often tries to intimidate others while positioning his net’, taking advantage of his selfproclaimed technical superiority. In fact, the disappointing performance of Group B's best-skilled skipper in terms of leadership ought not to be necessarily ascribed to his preference profile, that is, to the fact that he does not attach importance to social esteem considerations. It may well be the case that, even though he is keen in terms of popularity, he does not possess the complementary qualities required for him to be esteemed by his fellow skippers. As a matter of fact, possessing relatively high skills is a necessary but not a sufficient condition for gaining popularity. It should be associated with a demonstrated willingness to act, if only in appearance, for the benefit of the group (e.g. via sharing of skills) and with an attitude of humility and modesty (you should not say yourself that you are good, but let others say it). The leader of Group B, either because he is not greatly motivated by popularity or esteem considerations or because inherited character traits preclude him from attracting sympathy from others, does not appear to be ready to forego any of his income in order to gain social prestige among fellow skippers. On the other hand, we cannot exclude the possibility that preference profiles are somewhat endogenous: once fishermen get together in a pooling group, their preferences evolve in such a way that they pay more attention to social esteem considerations and therefore become more tolerant regarding differences in catch performances. In other words, peer pressure and some form of revenue sharing tend to go together. If that is the case, initial preference profiles may not have a decisive influence on the feasibility of pooling. Now, we have two converging reasons that go towards explaining why Group A has been more successful than Group B in pooling experiments. But the semi-failure of Group C is still left unexplained. The third

392

JEAN-PHILIPPE PLATTEAU AND ERIKA SEKI

hypothesis derived from the theoretical discussion in the previous subsection is that circumstances surrounding Group C are specific and such that potential benefits from coordination associated with pooling are significantly smaller than those achievable by both Groups A and B (which operate in an identical environment). This, in fact, appears to be a valid explanation. Indeed, the economic and technical conditions prevailing in Iwase require less collective action than those obtaining in Shinminato. For one thing, owing to the existence of a price-fixing mechanism working under the aegis of the local FCA, gains from regulating fishing effort in the former area are limited compared with the latter area where fish is disposed of through public auctions. For another thing, the technology used by fishermen in Iwase is less skill-demanding because of the configuration of the sea bottom (no drifting of nets is involved). Moreover, the limited number of fishing spots, which are marked by buoys, makes searching operations much easier to accomplish than in Shinminato. As a result, the gains achievable by coordinating fishing operations and sharing technical knowledge are not impressive in Iwase. The first, market power factor is particularly noteworthy because it may help explain why Group C's members actually started pooling but largely gave it up in the 1970s. As a matter of fact, the exploitative marketing arrangement through which their catches were disposed of and which has triggered their coming together has given way to the above pricefixing mechanism, thus removing a basic motive for pooling. The historical coincidence between the two events, the improvement in the marketing system and the strong mitigation of the pooling mechanism, points to the importance of coordination benefits as a condition for pooling. Lack of attractiveness of pooling for Group C's members apparently results from the combination of the above absence of significant intrinsic advantages of pooling with the large extent of homogeneity of their abilities (which may itself have partly resulted from their experience of pooling together during several years). As evident from Fig. 13.2, fishermen are less induced to enter a pooling group if their partners are too similar to them than if they are somewhat different. Three final remarks are in order. First, in terms of our conceptual framework, the peculiar historical experience of a common struggle against fishmerchants among Group A's (and Group C's) members has favored pooling and other collective actions by (1) increasing tolerance for unequal skill levels and output performances within the group; (2) inducing the most skilled members to help less-performing skippers to improve their skills so that differentials are gradually narrowed down; and (3) enhancing awareness of the potential benefits of cooperation. Clearly, a major lesson is that history matters and, by influencing the above parameters, may determine whether a pooling or a non-pooling equilibrium will be reached.

POOLING GROUPS IN JAPANESE FISHERIES

393

Second, some statements made by skippers from Group B suggest that free riding was a pervasive problem which had the effect of undermining trust and collective discipline in the group. What needs to be stressed at this juncture is that, rather than being a structural, exogenously given trait of Group B's functioning, free riding behavior may actually have been part of the unraveling process that led to the group's collapse. The following scenario is hinted at by a number of remarks and comments issued by the actors themselves. When the lesser-skilled skippers (and crew) realized that, in spite of their earnest efforts, they did not succeed in achieving catch performances deemed acceptable by the highestranked skippers, they got discouraged and tended to gradually relax their efforts, and this, despite the monitoring of these efforts by other skippers (free riding may thus take place under conditions of perfect information). Observing with a relatively good accuracy both these insufficient achievements and this relaxation of efforts (information is quasiperfect), skippers of the highest type were inclined to follow suit by also diminishing the intensity of their efforts. In other words, attitudes towards effort are endogenous to the relative performances of lesser-talented skippers and crew: if the group is too heterogeneous to start with, effort levels may be cumulatively lowered, presumably after a short time, thereby dooming the group to ultimate failure. This said, we cannot exclude the possibility that the lack of cooperativeness of Group B's leader has made the unraveling process more unavoidable by precluding positive dynamic effects of pooling from occurring. Indeed, to the extent that he is reluctant to share his knowledge with his colleagues, the narrowing down of performance differentials that could be expected to follow from learning processes could not take place in this group. Whichever type of agent withdraws first is not important. Group B ended, we were told, when the least skilled skipper left out of a feeling of uneasiness and shame caused by his below-standard performances. Yet, it is obviously peer pressure that is the real mechanism sanctioning disappointing achievements. It may just be a specific feature of the Japanese culture that a low performing member is never openly asked to leave a group. Instead, his better performing colleagues make him feel so bad about his position—he is made to feel ashamed—that he has no other choice than to pull out. And the pressure is all the more effective as it is exercised discreetly and almost silently. This adverse scenario is to be contrasted with the virtuous dynamics of those groups (such as Group A) that start their collective operations with a moderate amount of heterogeneity: in this instance, indeed, since performances of the well-meaning weaker members are considered acceptable by the other members, not only is inclination to work in earnest mutually selfreinforcing but also the gap between their respective achievements should progressively narrow down thanks to transmission of knowledge and expertise from the most to the least talented or experienced fishermen.

394

JEAN-PHILIPPE PLATTEAU AND ERIKA SEKI

Our third and last remark concerns the issue of intergenerational stability of pooling groups. It can indeed be contended that a successful pooling group, such as Group A, may face a risk of disruption when changes of skippers occur due to retirement of founding group members. This will be the case if the skill levels of their sons, unlike their own skill levels, are so diverse that they come to exceed the threshold of heterogeneity tolerance. Intergenerational stability of a successful pooling group cannot therefore be taken for granted. Such a problem ought not to be exaggerated, however. This is because, as we have actually observed in Group A (where one young skipper has taken over from his father), intra-group learning also produces its heterogeneity-reducing effect across generations.

13.5 Conclusion and Final Considerations There are three major lessons to draw from the present study. First, in order to be fully understood, pooling arrangements in fisheries must be analyzed as clusters of interlinked attributes. If this is not done, the crucial fact that pooling, rather than being adopted as a way to reach a given objective, may actually be instrumental to other attributes serving different objectives may escape attention. Second, when pooling, Japanese coastal fishermen appear to pursue both insurance and profitability objectives. Insurance considerations dominate in matters of net repair and maintenance expenditures (no insurance market exists for nets), while profitability advantages include benefits from scale economies (especially in searching activities), regulation of fishing effort, exchange of strategic information, and coordination in access to fishing locations. Our detailed empirical analysis reveals that many advantages of coordination and joint operation associated with pooling are actually reflected in better performances on the part of the pooling group. In particular, allocation of effort between fishing locations characterized by different risks (of net damage) and productivities is much closer to the optimum in the (fully) pooling group than in the group where the attempt to pool has largely failed. Third, initial group heterogeneity is a crucial variable in deciding whether a pooling group will succeed or fail. If the group is not too heterogeneous to begin with, pooling is self-reinforcing inasmuch as group heterogeneity is likely to be gradually reduced through diffusion of knowledge from the more to the less talented or experienced members. If, on the contrary, group heterogeneity is rather large initially, frustrations are bound to develop: higher-ranked individuals will resent the low performances of their fellow members and the latter will feel humiliated by their inability to achieve an acceptable standard of performance compared with the former. As a result, the group will progressively unravel.

POOLING GROUPS IN JAPANESE FISHERIES

395

The empirical analysis confirms the theoretical prediction that groups are more likely to fail not only when initial heterogeneity is too large, but also when the members' preference profiles are not compatible with large differentials in individual performances. This may happen because the higher-ranked members do not attach great value to being the best (or little social esteem and respect is granted to them as a result of their above-average achievements) and/or the lesser-ranked ones have a profound dislike for being in an inferior position (or a lot of social stigma is attached to their low performances). Furthermore, groups may not form or succeed simply because the benefits from acting as a collective are not large enough under the specific circumstances (ecological, technical, and economic) which confront them. An immediate implication of the study is that profitability losses (such as those arising from the non-exploitation of pervasive scale economies) are unavoidable and insurance benefits are impossible to reap when the characteristics of potential group members in terms of their relative abilities or performances and their preferences regarding rank are not conducive to sustainable income- and cost-pooling among them. To put it in another way, welfare gains are forsaken if groups are too heterogeneous and/or if tolerance with respect to heterogeneity is too low.283 The fact that in Japan hardly more than 10 percent of the so-called ‘self-management institutions’ have successfully organized into pooling groups—still a remarkable achievement when compared with all other countries where fishing pooling groups are almost never observed—attests that welfare losses due to the inability to set up optimal institutions are substantial. Yet, the extent of such losses cannot be ascertained since it is quite possible that, in the case of some fisheries (particularly, for sedentary and pelagic species), the benefits from coordinating efforts are too low to justify pooling. Because willing pooling groups are pre-formed in the particular context of Japanese fisheries—a feature that directly follows from the peculiar sea tenure system obtaining in Japan—selection mechanisms cannot be easily used to surmount the problem of group heterogeneity.284 If coordination is not sustainable within a given resource-owning group, it is difficult for

283

Of course, homogeneity is not a sufficient condition for successful collective action. What must be stressed is that many other conditions, such as small group size and legal recognition or support (as opposed to harassment) by the government are well fulfilled in the present case study.

284

See Frank (1985: 48–55) for a partial equilibrium analysis of the way workers of unequal productive ability and different tastes for local status sort themselves out in different firms conceived as local hierarchies. As a result of this people's ability to insulate themselves from unfavorable income comparisons by forming local hierarchies of their own choosing, everyone can achieve a better outcome. Indeed, those who value status highly can give up an amount of income that is worth less to them than the status they get in return and those who do not value status highly can accept an income premium that is more valuable to them than the status they give up to get it (Frank 1985: 55).

396

JEAN-PHILIPPE PLATTEAU AND ERIKA SEKI

members to join other different groups with characteristics closer to theirs. Migration is of course possible but there is no guarantee that another group will readily share its right of access to water with a newcomer. Furthermore, if skippers migrate outside their native village, the benefits resulting from network externalities and other advantages of kin-based groups would be lost, thereby making collective action more difficult. Here there is clearly a tradeoff between solving problems of resource management through assigning stable rights of ownership over fishing space, on the one hand, and reaping insurance and profitability gains from joint operation, on the other hand. Only if the introduction of more market and, therefore, more economic mobility, could do a better job—say, by establishing transferable fishing quotas or licences, or by overcoming incentive problems in providing insurance against net damage—would the above community failure be overcome. Another solution is theoretically available to overcome heterogeneity, namely the fixing of differentiated shares that vary according to individual boat performances.285 Apparently, this solution is easy to apply since the output of each boat is individually and costlessly measurable. It is interesting to notice that one of our sample groups that failed to pool according to a rigid egalitarian formula actually stepped back to a more complex arrangement that embodies the principle of differentiated sharing. But the other group that experienced difficulties in pooling abandoned the system altogether (except for two members who decided to continue it). This reflects the fact that unequal sharing is not an easily acceptable rule when participants to the scheme are supposed to work in close collaboration with one another and to interact continuously not only at sea but also on shore: unequal sharing may indeed offend the less valued participants. Finally, the conclusion that substantial welfare losses result from inimical group characteristics rests on the assumption of a family-based ownership structure where fishing units are individually owned by household heads acting as skippers. A conventional business firm owning several boats could apparently reap many scale economies and insurance benefits, yet, leaving aside the equity aspect involved in a redistribution of fishing assets, the problem of how it could deal with the serious incentive problems plaguing fishing activities is not easily solved (Platteau 1989; Platteau and Nugent 1992). The fact of the matter is that individual skippers need to be made residual claimants to be provided with maximum incentives to perform effectively in terms of labor quantity and quality, risk taking, and

285

When an effort is made endogenous, the equal sharing rule may be preserved without necessarily causing the group to collapse. As a matter of fact, given rank competition, individuals of unequal skill abilities may choose to put in varying amounts of effort so that output performances are more or less equalized in spite of skill differences (see Gaspart and Seki, 1999, for a formal treatment). Their model, however, assumes that skill levels are private information.

POOLING GROUPS IN JAPANESE FISHERIES

397

avoidance of various kinds of moral hazard or enforcement problems (e.g. if not properly remunerated, boat skippers could collude amongst themselves to underreport output, overreport input and damages to equipment, etc). In the presence of various market imperfections (including the shortcomings of specialized net repair activities), cooperation among different fishing units therefore seems the only way available to improve productive efficiency (for a given technology) and to insure against risks that the market cannot shoulder. If that solution does not work owing to some sort of ‘community failure’—as appears unfortunately to be the case in almost all small-scale fisheries over the world—the conclusion cannot be escaped that significant welfare losses are being incurred by rural producers.

Appendix I: The Effect of Pooling on Allocation of Effort Between Fishing Locations The advantage of pooling in terms of allocation of fishing effort among different fishing locations can be easily demonstrated using a game-strategic framework in which two boats, called boats A and B, have to choose whether to operate in location 1 or 2. There are four possible outcomes: the two boats can simultaneously operate in either spot 1 or spot 2; or, they can work separately, one in spot 1 and the other in spot 2. Assume that the expected catch for the boat which has the first turn in spot 2, which is the less productive spot, is smaller than the expected catch for the boat which has the first turn in spot 1, but larger than the expected catch for the boat which has the second turn in spot 1. That is, we assume that

bearing in mind that yij is the expected catch for a boat which takes jth turn in spot i. It is also assumed that, when two boats concur to the same fishing location, they have a fifty–fifty probability of having either the first or the second turn. The payoff matrix is therefore as given in Fig. 13.3a. On the condition that the expected catch for an individual boat taking the first turn in spot 2 (while the other boat has the first turn in spot 1) is smaller than its expected catch were it to compete with the other boat for access to spot 1, that is, if , the payoff matrix exhibits a typical Prisoner's Dilemma structure in which each player has a dominant strategy. As is evident from Fig. 13.3a, the Nash equilibrium of the game is an outcome in which both boats operate in spot 1. This is clearly inefficient because the aggregate output at such equilibrium is smaller than what could be achieved if they decided instead to operate in the two spots, with either boat A taking the first turn in spot 1 and boat B the first turn in spot 2, or the converse. These allocations are Pareto optimal, yet will never be chosen by the agents as long as they operate in a decentralized manner.

398

JEAN-PHILIPPE PLATTEAU AND ERIKA SEKI

Income-pooling can nevertheless achieve one of the Pareto-optimal outcomes because it induces boat skippers to coordinate their decisions with a view to reaching the maximum aggregate level of catches. In point of fact, once income is equally shared between the two boats, the payoff structure is altered in such a way as to transform the game from a Prisoner's Dilemma into a Chicken game. Two Nash equilibria then exist that correspond to the above Paretooptimal outcomes: one boat operates in spot 1 and the other in spot 2 (see Fig. 13.3b). So far, it has been implicitly assumed that all skippers in a group have equal abilities and characteristics (e.g. preference profiles over different fishing spots), which is not necessarily true. Now, a group which has chosen to pool incomes has an interest in taking these differential skill and preference profiles into account when allocating participants between the available fishing spots. In other words, to achieve the efficient allocation outcome, it will allocate skippers among different locations in such a way that their specific abilities and preferences best match the technical and risk characteristics of each fishing zone. This is the classical ‘division of labor’ argument. In the above coordinated game, we have pointed out that there exist two equilibria which correspond to mismatch solutions. Yet, if skill or preference profiles differ between the skippers running the two boats, the game may have a unique solution: the skipper of each boat is sent fishing in the place for which he has a comparative advantage. Fig. 13.3a The Payoff Matrix Of a 2 × 2 Game Where Two Boats Have To Choose Between Two Fishing Locations In a Decentralized Manner

Fig. 13.3b The Payoff Matrix Of a 2 × 2 Game Where Two Boats Have To Choose Between Two Fishing Locations In a Coordinate Manner

POOLING GROUPS IN JAPANESE FISHERIES

399

Finally, it should be obvious that a scheme specifying clear rules of access to spot 1 would achieve efficient allocation of fishing effort in the same way as income-pooling. This is because the random factor reflected in the probabilities of getting the first or second turn on this spot has now vanished to give way to a certain order of access. Under the above assumption that the first haul on spot 2 is smaller than the first haul on spot 1 yet larger than the second haul on the latter, the skipper with only the second turn on spot 1 will choose to operate in spot 2 rather than wait for his (second) turn in the best spot. Access rules thus provide for a unique allocation which is efficient.

Appendix II: Estimate of Risk Aversion We estimate the Arrow–Pratt measure of Constant Relative Risk Aversion (CRRA), using the following utility function defined over income (y ): i

We consider that fishermen choose between two fishing spots, Oki and Heta, characterized by the production functions f and g, respectively. Oki is known as a relatively safe fishing spot while catches in Heta are comparatively volatile. The catch in Oki is therefore assumed to be constant at the level f, whereas fishing in Heta may result in either a high catch, g, or a low catch, g, the former being achieved with probability p and the latter with probability 1−p. This framing of the problem allows us to measure the extent of risk aversion by the Arrow–Pratt measure of relative risk aversion r(y):

The von Neumann–Morgenstern expected utility function is:

(1) where x and x are the levels of fishing effort allocated to Oki and Heta, respectively. o h Normalizing the total amount of effort available to one, each fisherman chooses an allocation, x*, by solving the following expected utility maximization problem:

The first-order condition with respect to x is: h

400

JEAN-PHILIPPE PLATTEAU AND ERIKA SEKI

Since x ′(x ) = −1, we can write o

h

By taking logarithms, we obtain the following functional form:

The time unit is the fishing day and our data cover ten days. Note that f is the total catch obtained on a given day from the various hauls made in Oki by boats of both Groups A and B. The low catch level in Heta is assumed to be zero, i.e. g = g = 0. As for the high catch level in Heta, g it is obtained by considering all the positive output performances recorded (per day) for that location. We are interested in deriving a point estimate of β by applying the OLS technique to the above equation. The result is:

which suggests a relatively low degree of risk aversion. The overall fit provided by the OLS estimate (not shown) is poor, not a surprising fact given the small size of the sample. This is a minor issue since we are not interested in making statistical inferences.

References Adams, D. W. and Rask, N. (1968) ‘Economics of Cost-Share Leases in Less Developed Countries’, American Journal of Agricultural Economics, 50: 935–42. Aoki, M. (1988) Information, Incentives, and Bargaining in the Japanese Economy, Cambridge University Press, New York and Cambridge. —— (1990) ‘Toward an Economic Model of the Japanese Firm’, Journal of Economic Literature, 28(1): 1–27. —— (1994) ‘The Japanese Firm as a System of Attributes: a Survey and Research Agenda,’ in Aoki, M. and Dore, R. (eds.), The Japanese Firm: Source of Competitive Strength, Clarendon Press, Oxford, pp. 11–40. —— (1998) ‘General Comments’, in Aoki, M., and Hayami, Y. (eds.), The Institutional Foundation of Economic Development in East Asia, Macmillan, London. Asada, Y., Hirasawa, Y., and Nagasaki, F. (1983) ‘Fishery Management in Japan’, FAO Fisheries Technical Paper, No. 238, FAO, Rome. Baland, J.-M. and Platteau, J.-P. (1996) Halting Degradation of Natural Resources—Is there a Role for Rural Communities?, Clarendon Press, Oxford.

POOLING GROUPS IN JAPANESE FISHERIES

401

Baland, J.-M. and —— (1997) ‘Wealth Inequality and Efficiency in the Commons, Part I: the Unregulated Case’, Oxford Economic Papers, 49(3): 51–82. —— and —— (1998) ‘Wealth Inequality and Efficiency in the Commons, Part II: the Regulated Case’, Oxford Economic Papers, 50(1): 1–22. Benedict, R. (1946) The Crysanthemum and the Sword: Patterns of Japanese Culture, Houghton Mifflin, Boston. Benham, L. and Keefer, P. (1991) ‘Voting in Firms: The Role of Agenda Control, Size and Voter Homogeneity’, Economic Inquiry, 29: 706–19. Bernheim, B. D. (1994) ‘A Theory of Conformity’, Journal of Political Economy, 102(5): 841–77. Besley, T. (1995) ‘Savings, Credit and Insurance,’ in Behrman, J., and Srinivasan, T. N. (eds.), Handbook of Development Economics, 3A: 2123–207, Elsevier, Amsterdam. Bliss, C. J. and Stern, N. H. (1982) Palanpur: The Economy of an Indian Village, Clarendon Press, Oxford. Coate, S., and Ravaillon, M. (1993) ‘Reciprocity without Commitment: Characterization and Performance of Informal Insurance Arrangements’, Journal of Development Economics, 40(1): 1–24. Esman, M. and Uphoff, N. T. (1984) Local Organisations: Intermediaries in Rural Development, Cornell University Press, Ithaca. Frank, R. (1985) Choosing the Right Pond: Human Behavior and the Quest for Status, Oxford University Press, Oxford. —— (1988) Passion Within Reason: The Strategic Role of the Emotions, W.W. Norton & Co., New York. Gaspart, F. and Seki, E. (1999) ‘Sharing, Heterogeneity and Status Considerations: Incentive Theory and Evidence in Japanese Fisheries’, CRED (Centre de Recherche en Economie du Développement), Faculty of Economics, University of Namur, Belgium (mimeo). Hansmann, H. (1988) ‘Ownership of the Firm’, Journal of Law, Economics, and Organization, 4(2): 267–304. —— (1996) The Ownership of Enterprise, The Belknap Press of Harvard University Press, Cambridge, MA. Hasegawa, A. (1993) Gyogyoushigen Kanri no Tebiki (Guideline for Management of Fisheries Resources), National Federation of Fisheries Cooperative Associations (Zengyoren), Tokyo. Hayami, Y. and Kikuchi, M. (1981) Asian Economy at the Crossroads— An Economic Approach to Institutional Change, University of Tokyo Press, Tokyo. Hayami, Y. and Otsuka, K. (1993) The Economics of Contract Choice: An Agrarian Perspective, Clarendon Press, Oxford. Holmström, B. (1982) ‘Moral Hazard in Teams’, Bell Journal of Economics, 13: 324–40. Kandel, E., and Lazear, E. P. (1992) ‘Peer Pressure and Partnerships’, Journal of Political Economy, 100(4): 801–17. Kremer, M. (1997) ‘Why are Worker Cooperatives so Rare?’, MIT, Cambridge MA (mimeo). McKean, M. A. (1986) ‘Management of Traditional Common Lands (Iriaichi) in Japan’, in National Research Council, Proceedings of the Conference on

402

JEAN-PHILIPPE PLATTEAU AND ERIKA SEKI

Common Property Resource Management, National Academy Press, Washington, DC, pp. 533–89. Nalbantian, H. R. and Schotter, A. (1997) ‘Productivity Under Group Incentives: An Experimental Study’, American Economic Review, 87(3): 314–41. Newbery, D. M. G. (1975) ‘The Choice of Rental Contract in Peasant Agriculture’, in Reynolds, Lloyd G. (ed.), Agriculture in Development Theory, Yale University Press, New Haven, CT, pp. 109–37. Olson, M. (1965) The Logic of Collective Action, Cambridge University Press, Cambridge. Ostrom, E. (1990) Governing the Commons—The Evolution for Collective Action, Cambridge University Press, Cambridge. Platteau, J. P. (1989) ‘Penetration of Capitalism and Persistence of Small-scale Organizational Forms in Third World Fisheries’, Development and Change, 20(4): 621–51. —— and Nugent, J. B. (1992) ‘Share Contracts and Their Rationale: Lessons from Marine Fishing’, Journal of Development Studies, 28(3): 386–422. —— (2000) Institutions, Values, and Economic Development, Harwood Publishers. Postlewaite, A. (1998) ‘The Social Basis of Interdependent Preferences’, European Economic Review, 42: 779–800. Ruddle, K. (1987) ‘Administration and Conflict Management in Japanese Coastal Fisheries’, FAO Fisheries Technical Paper, No. 273, FAO, Rome. Sandler, T. (1992) Collective Action: Theory and Applications, Harvester Wheatsheaf, New York. Seki, E. (2000) ‘Rotation Scheme to Share Common Pool Resources: Theory and evidence about an alternating duopoly’, CRED (Centre de Recherche en Economie du Développement), Faculty of Economics, University of Namur, Belgium (mimeo). —— and Platteau, J. P. (1999) ‘Heterogeneity and Social Esteem in Income-Pooling Groups: Some Evidence from Japan’, CRED (Centre de Recherche en Economie du Développement), Faculty of Economics, University of Namur, Belgium (mimeo). Smith, T. C. (1959) The Agrarian Origins of Modern Japan, Stanford University Press, California. Tang, S. Y. (1992) Institutions and Collective Action—Self-Governance in Irrigation Systems, ICS Press, San Francisco. Yamamoto, T. and Short, K. (1991) International Perspectives on Fisheries Management—With Special Emphasis on CommunityBased Management Systems Developed in Japan, Proceedings of the JIFS/IIFET/Zengyoren Symposium on Fisheries Management, National Federation of Fisheries Cooperative Associations (Zengyoren), Tokyo. Weiss, Y. and Fershtman, C. (1998) ‘Social Status and Economic Performance: A Survey’, European Economic Review, 42: 801–20.

Comments

286

Douglass C. North I want to pull together what we have been talking about in these two days. I take it that our task has been to try to arrive at some limited generalizations about the relationship between communities and markets and at some broader generalizations about ways human beings interact over time to create structures of human interaction. Further, we wished to evaluate the effectiveness of those structures. Let me begin by talking briefly about a conference sponsored by the Santa Fe Institute in 1996. This was one in a series of conferences at the Institute on the limits to knowledge in various disciplines. The first in the series had concerned physics. That was a good place to start: physicists have the hubristic view that they can explain anything in the world from the origins of life to the limits of the cosmos. The conference in 1996 concerned economics. At this, I talked about the limits to understanding long-run change. The conference got hung up on a fundamental set of issues about the nature of theorizing which confront us all the time when we are concerned with not just economics but all social science theorizing. A physicist attending the conference kept saying that what is wrong with economists is that we do not have ‘power laws,’ or we have them but do not know what they are. And he said that if we would develop power laws then we would be able to explain not only all of economics but also all the social sciences as well. Some of us tried to explain to him that probably there are not any power laws—or if there are, they are insignificant; most of us do not think that there are any power laws in economics. Paul Samuelson has stated the issue clearly: what permits economics to be a science is whether the world it models is ergodic. Samuelson said that if it is ergodic then we can have a scientific discipline; and if non-ergodic, then we cannot. Robert Solow has pointed out that such a perspective reflects the view that economics is like physics! But the issue is important for our point here because we are concerned with the degree to which we can arrive at generalizations about an enormously complex process. If the world we model in economics is ergodic we should be able to arrive at timeless fundamental principles that have a scientific foundation and we should be able to determine precisely what they are. If it is non-ergodic, and I think it is, then we can arrive at limited generalizations

286

These comments were made at the EDI-sponsored workshop on Community and Market Development, Stanford University, February 5–6, 1999.

404

DOUGLASS C. NORTH

which, in effect, are constrained by the fact that we have lived and continue to live in a continuously evolving world—one that does not go round and round in circular flow but rather has limited parallels between present and past. It is obvious that while we can develop generalizations about what is going on in this world, we have to temper them all the time by the recognition that the world is non-ergodic. What I intend to do, therefore, is to look, as an economic historian, at the evolving nature of this world, to see whether I can also put that in the context of some very limited generalizations that we can make about human interaction. That will allow us to intermingle some theory with history to get some results that I hope will be useful not only in the present, but for us to be able to understand better the human condition over time. It should not surprise anybody who has spent a lot of time trying to understand history, that there is enormous diversity in the human experience. This diversity reflects the phenomenon that we evolved from primates in various physical settings of climates, geography, and diseases. This diversity shows up, first of all, in the language and then in the kind of tools and artefacts we have created which together evolved contrasting trajectories of human beings. Economic history has to be informed by an enormous complex of things—you are concerned not only about tools, techniques, knowledge and beliefs, and so on, but also about how they all interact with each other. In economic history, we try to understand the process of evolution. What it comes down to, at least in terms of our being able to get a hold on it, is that in very different environments, with new experiences, we keep on modifying the ways in which we perceive the world. Our beliefs about that world change, and we construct institutions that try to control the world in which we live in an ongoing, continuous process. But this is neither a process that goes round and round nor one that remains constant. It is a process in which we continually face new and novel situations. Indeed, the systematic development of science and its application to the problems of human scarcity have led to the ‘conquest’ of the uncertainties involved in the physical environment but at the price of increasing complexity and uncertainty of the human environment. The tools, techniques, and institutions we have developed to solve problems in the human landscape may work for us at times to solve the new and novel dilemmas we face; but sometimes we are going to get it terribly wrong. We keep fumbling along, trying to find ways by which we can do better. There is nothing new about this process, but it does require that as economic historians we abandon models based on physics and build models based on evolutionary biology. But even that approach has real limits to it. In the Darwinian process the players do not attempt to produce results that they perceive as having downstream consequences; but human beings are always trying to do just that. They keep on trying to enact policies and

COMMENTS

405

institutions and rules and so on which they believe will have consequences downstream that they find favorable. This accumulation of beliefs and institutions and norms and rules that evolve, together with artefacts, reflects the experience of the past. We retain those that seem to have worked in the past and discard those that do not seem to work. This is not a new idea. Hayak said this a long time ago and it still makes a lot of sense when we think about it today. How is this process evolving? One important point, which has showed up over and over again, is that path dependence is a crucial part of the process. We evolve in the context of the beliefs and institutions that have evolved before, and they play a key role in this evolutionary process that is an incremental process of change (even though we sometimes have punctuated equilibrium conditions where it looks like we have discontinuous change). The evolutionary process is continually being modified by changes in demography and the quantity and quality of human beings, as well as by changes in the stock of knowledge—of which we know much less. We know much more about changes in the technology than we know about the overall changes in the stock of knowledge. Our beliefs are based on what we perceive to be knowledge today; even though often in human history accepted beliefs turn out not to be correct knowledge. What was knowledge at one time may not be knowledge at another time. Demographic changes, changes in the stock of knowledge, and changes in institutions interact in very complex ways that we only vaguely and dimly perceive. Because we have not done our homework, economic and social historians are only now beginning to think systematically and seriously about how these interact over time to produce the kind of complex varieties of situations and cultures that we have. Bob Allen's and Avner Greif's stories fit into this history part very nicely. They both try to make sense out of an enormously complex process by intermingling a lot of different parts of the action together to draw some limited generalizations—Allen by sorting out the issues involved in the increase in agricultural productivity in early modern England and Greif by developing a more nuanced understanding of the movement from personal to impersonal exchange. Let me turn now to examine analytically some fundamentals of human interaction. In Masa Aoki's opening section he looks at a variety of structures—whether political, organizational, or market—that have served as common denominators of this human interaction. He and Avner Greif turn a lot of this into game theory, which I am sometimes enthusiastic about and sometimes not so enthusiastic about. I believe a useful common denominator of human interaction is a transaction cost approach. Transaction costs are fundamentally two things: they are the cost of measuring the characteristics and attributes of whatever is being owned or exchanged, and they are the cost of enforcing an exchange. As Lancaster told us a long time ago, what we exchange is not a

406

DOUGLASS C. NORTH

uni-dimensional good or service but, usually, one with many dimensions that have value to them. Because they do, to the degree that we have been able to isolate and measure those dimensions accurately, we lower the cost of exchange. The basic determinants of measurement costs are technology and the specification of property rights. Technology determines the precision with which we can measure the physical dimensions of what is being exchanged. Property rights determine the precision with which we can measure ownership rights to the asset or resource. The more precisely both are specified the lower the cost of transacting. Enforcement can be by the first party, second party, or third party. First-party enforcement is self-enforcement; second-party enforcement is via retaliation; third-party enforcement is either the kind that Avner Greif talks about when he refers to the community responsibility system or enforcement by the state, a political entity with coercion. Transaction costs are going to differ fundamentally in situations of personal exchange and impersonal exchange; but so are production costs. Smithian specialization and division of labor and the economies associated with large markets entail impersonal exchange, which requires low measurement costs. In personal exchange, you tend to have a low cost of enforcement to the degree that you have repeated exchange among the parties. Enforcement costs in impersonal exchange entail this party enforcement mechanisms and in consequence higher transaction costs. Most of the papers we have been discussing for the last two days have admixtures of these types of transaction costs. The papers were continually concerned with the ease of measuring the attributes being exchanged or with whether the high cost of measurement will lead to all kinds of problems with respect to the way in which the exchange can exist. So obviously, a general issue running through history and in the contemporary world is how human beings have devised instruments, institutions, and norms of behavior that have low cost measurement of either principal–agent relationships or of goods and services being exchanged. With respect to enforcement, to the degree we have devised institutions to be able to undertake low cost enforcement, either through a third party or by the coercive enforcement of the state, we can get low costs of transacting. The shift from personal to impersonal exchange is a monumental shift in the human condition. It is a monumental shift because almost all the institutions and rules by which we devise ways to make transaction costs low must be modified in turn. Some of the papers have mixed personal and impersonal exchange; I will put the two together. What I wanted to do with the history story is to say that human beings have an enormous diversity of experiences by which they attempt to solve problems. But human beings are conditioned by the fact that down deep, in one way or the other, they are concerned with costs of measurement, enforcement, and different kinds of exchange conditions and therefore, as a result, have

COMMENTS

407

followed very different patterns. I think these have been elucidated very nicely in a number of the papers. The empirical papers in this conference have highlighted new ideas, new concepts, and new thoughts for me about this complex process of exchange. I want to talk about a few other papers to illustrate some of the points that I have been making and to re-emphasize some of the points you have been making too. First, Yujiro Hayami. He has played a key role in making this and preceding conferences successful with consistently thoughtful, interesting papers that are a delight to read. The only one I am going to mention here is the one with respect to vegetable marketing in Indonesia. I really wanted to raise that one because it does bring out something that by now, I think, we do tend to be sophisticated about. That is, contrary to Clifford Geertz, the successful emergence of commercial and industrial entrepreneurship has developed from upper peasant ancestry. With sufficient changes in conditions and incentives, such peasants became effective entrepreneurs in a variety of different environments. Yujiro quotes Theodore Schultz in the conclusion of the paper. Long ago, Peter Bauer noted the successful adaptation of peasants when he looked at such changes in Africa. The next paper I am going to mention is the one by Akihiko Ohno on hand-weaving in Laos. What I like about this paper is that it brings out very clearly the different kinds of measurements and enforcement costs and consequent transaction costs that play such a critical role in the way in which hand-weaving evolved in Laos. The paper highlights the variety of strategies involved in relational contracting arrangements to deter post-contractual problems. Then there is Jonathan Morduch's paper on China. China is really such a fascinating place; the issue in this paper is the adjustments occurring with radical alterations in the economic structure. The focus is on one particular county and on the dramatic changes, changes that have led to great income disparities. I think the most important part of the paper concerns the shift away from the collective farm as the basic organizational unit and the void that it leaves in the insurance structure that was an integral part of that structure. These changes in good part are imposed from above, although now with decentralization as a result of Beijing loosening up on these conditions we are witnessing fundamental alterations in the economic/political structure. It is going to be interesting to see how insurance can evolve that replaces the implicit insurance in the collectives. Finally Jean-Philippe Platteau's paper, which Jonathan Morduch commented on. I like this even though Platteau has, of necessity, used a very small sample. It is certainly rich in detail and specifics about some interesting aspects that I never thought of when I was looking at ways by which cooperation either does or does not tend to work. The contrasting patterns of cooperative arrangements on the one hand and the disintegration

408

DOUGLASS C. NORTH

of cooperation on the other are suggestive of lines of inquiry for future research. Let me conclude by talking about something that Bob Allen raised as a question and that Pranab Bardhan talked about. That is centralization vs. decentralization. The issue is not just centralization or decentralization. It is the larger issue about how we can get desirable policies: about how to structure the game—factor and product markets—so as to get socially desirable results. It is a large problem because one of the things economists have implicitly assumed is that if you only get the government out of anything, the markets will work themselves. That is a lot of nonsense. Markets (factor markets and product markets) have to be structured so you get the players to compete in a socially desirable way. And that turns out to be very complicated because the structure is going to be different for every kind of market. Not only is the essential structure for good performance going to differ with every factor and product market, but it is going to change over time. It is going to change over time because the kind of measurement and enforcement conditions that arise from changes in technology, information costs, and polities is going to continue to revise the way the game must be played to maintain efficient markets. So the general issue that was raised about centralization vs. decentralization seems to me to be a larger issue of the kinds of institutions, formal and informal, that will underlie political as well as economic markets to make for effective performance not only at a moment of time but, much more difficult, over time.

Index access to irrigation 254, 256, 262, 263 advance-order system 160, 161–2, 165–6, 168, 179–80, 183 African business networks xxi–xxii, 186–214 agriculture ; China 217, 219, 220, 222; English open field to enclosure transition xx, 42–69; Indonesian peasant communities xxi, 129–52; insurance 235–7, 238, 239–40, 241, 320; irrigation India xxii, 247–64, 278Indonesia 133Japan 107–8, 344Mexican ejido communities 321Philippines xxii–xxiii, 265–94, 311; Japan 107–8, 344; Mexican ejido communities 319, 320, 323, 336, 338–9; Nepal 297; Philippines xx, xxii–xxiii, 70–96, 265–94, 311 assembly meetings 326, 328–31 Austria, agricultural labor productivity 54 bankruptcy 29 Belgium, agricultural labor productivity 54, 55 borrowing 224, 232 brand recognition 209 business networks, Sub-Saharan Africa xxi–xxii, 186–214 capitalism ; early Meiji Japan 121; English enclosure xx, 67 caste system 86, 253, 261 Champagne Fairs 7, 8, 10, 20, 23, 24, 190 China, risk and insurance xxii, 215–43, 407 Chinese traders in Indonesia 129, 134, 136, 147–8 clientalization 103, 105, 163, 178 clientship 103–6, 124 clover production 59–63 clubs 106, 194, 207–8 collective action ; heterogeneity relationship 253; Indian irrigation system 250; Nepalese community forest management 296, 311; Philippines irrigation system 266, 278, 292see alsocooperation; pooling collective problem-solving 122–3 collective punishment 194 collectivization, China 217 colonial policies 196 Commercial Revolution 7–8, 9, 16, 19, 24, 25 commercialization ; English open field agriculture 44; Indonesian vegetable production 132–3; Laos handweaving industry 182 commodity risk 145 common pasture 48 common property resources (CPR) ; community role xvi; Japan 344; Mexican ejido communities 318, 319, 322, 323–6, 331, 336–42; Nepalese forest management 296see also local commons; public goods common rights 60 communal ownership, local commons 265

410

INDEX

communalism, open field agriculture 42–3, 44, 58, 66 communes 217 Communism 215, 219 community ; definitions xv, 97, 207–8; market relationship 3, 4–6, 42, 66 community affiliation xx, 10, 15–16, 24–6, 35–6, 186 community management ; Mexican ejido communities 321; Nepalese forests xxiii, 295–317; Philippines irrigation systems xxii–xxiii, 265–94; rural communities xvi community norms ; contract enforcement xvii–xviii; gametheoretical approach xx, 97–125; Indonesian peasant communities xxi, 129, 142; market relationship xv, xvi–xvii; Nepalese forest management 306–7; outsider hypothesis 154; Philippine agrarian systems 70, 87–9, 92, 93, 94; redistribution of income 175; religion 194–5; rule enforcement 263, 280; social heterogeneity 253 community responsibility system (CRS) xix–xx, 6, 16–36 community-based trust 162–3, 167, 170, 172, 175–7, 179, 182–4 community-enhancing policies 265, 292–3 competition, Japanese fishing communities 371, 375, 386 confiscation of goods 14, 15, 19–20, 21, 27–8, 34 conservation xvi, xix contract enforcement ; community institutions 5, 14, 16, 37; community norms xvii–xviii, 117, 124; community responsibility system (CRS) 16–36; individual responsibility 25, 34–5; Indonesian peasant communities 129, 141–4, 147, 148; insider hypothesis 153; Laos handweaving industry 154, 160–8, 178, 180, 182–3; legal system 7, 187; market societies 3–4; networks 189; Philippines agrarian systems 91–3; pre-modern Europe 8–9, 10, 11, 16–37; Sub-Saharan African business networks 186, 191, 193–5see also enforcement costs contracts ; agricultural labor 82–3, 85, 86, 91, 93, 94; grouplending 237; irrigators' associations 276, 277; Japanese merchant trade 119–20; Laos hand-weaving industry 168–9, 170–2, 173–5, 176–7, 179, 180 contractual flexibility 192 cooperation 296, 407–8; game theory 104, 105, 109, 110, 111; irrigation in India 247–8, 249–50, 253, 254, 262, 263; Japanese agriculture 108; Japanese fishing communities 114–15, 345, 356–7, 361, 364, 392, 397; Mexican ejido communities xxiii, 320–3, 326–36, 338, 340, 342; Nepalese forest management 307, 311; state-regulated 320, 334, 340see alsocollective action; pooling copyholders 58, 59, 61, 64 corruption 320 costs ; community forest management 307–8, 315; costsharing in India 255, 260–1; forest protection 295–6; personal relationships 190; Philippines irrigation projects 282–4,

INDEX

411

287–8; pooling of 345, 348, 351, 352, 370see also transac- ethics 195 tion costs Ethiopia, business networks 190, 204, 205 credit ; China 238–9; collective punishment 194; community ethnic groups ; Nepalese community forest management 308, responsibility system 28; Indonesian peasant commun311; reputation 195; Sub-Saharan African business netities 134, 138–41, 144; institutional innovations 209; works 196–7, 198–9, 201–2, 203, 204–6 Laos hand-weaving industry 178, 179, 180; Mexican ejido communities 320; Philippines 282; pre-modern Europe 7, 8; reputation 190–1; Sub-Saharan African business networks 192, 199, 205–6 crop yields 45, 46, 47, 49, 88, 89 CRS seecommunity responsibility system cultural belief xvii–xviii culture broker hypothesis 154, 174, 175, 183 currency crisis, Laos xxi, 166, 171, 180 debt 8, 11–16, 17–37 decentralized markets 186, 193, 194, 407, 408 decollectivization xxii, 218, 219, 234, 241 deforestation xxiii, 296–8, 303, 311, 312, 314–15 democracy 321 demographics ; China 240–1see alsopopulation growth developing countries ; community definition xv; culture broker hypothesis 154; local commons xvi; market failures xvii; rural industries 153; state irrigation systems 266 DFOs seedistrict forest officers differentiated sharing 396 discrimination 202, 209 district forest officers (DFOs) 299, 301–2, 306, 307, 309, 311 diversification ; China 216, 222; English agriculture 43, 64 division of labour 133–4, 398 double-cropping 80, 81, 132, 266, 282 drought 237–8 economic reforms, China 234–5 economic transaction domain 98, 99, 116, 120, 124 economics ; community norms 111; community relations 183–4; evolutionary 70, 404–5 economies of scale seescale economies ejido communities xxiii, 318–43 embeddedness xx, 98, 105, 113–24 employment seelabor enclosure ; English agriculture xx, 42–69; Mexican ejido communities 338 endogenous cooperation, Mexican ejido communities 322, 328, 329, 331, 334–5 endogenous rules xvi, 97–8, 99–100, 106, 114, 124see alsocommunity norms enforcement costs 406 England ; community responsibility system 17–18, 23, 26–7, 29–31, 32–5; enclosures vs open field agriculture xx, 42–69; pre-modern 6, 8, 17–18, 23, 26–7, 29–35 entrepreneurship xxi, 129, 138, 148–9, 151, 215, 407 equal-sharing rule 379, 381–2

412

INDEX

Europe ; agricultural labor productivity 52–6; economic institutions 215; pre-modern xix–xx, 3–41, 190 evolutionary economic theory 70, 404–5 exchange ; clientship 103–6; community vs market-based 4; contract enforcement 187; family relations 203; quid/quo separation 7–11; reputation mechanism 100–2; SubSaharan African business networks 193; transaction costs 405–6see also economic transaction domain; impersonal exchange exit options ; community participation 330, 331–2, 334; Indian irrigation schemes 255, 260; Indonesian peasant communities 144; Philippines irrigation schemes xxiii, 291, 292 exogenous rules 98, 99, 106, 111, 114, 124 expatriate groups 201–2 expenditure, Philippines irrigation projects 271, 272–3, 274, 280–1, 287, 288–9 experience rating 239, 241 factory system 168, 172 fair courts 23–4 family relationships 203–4 Fishery Cooperative Associations (FCAs) 345, 378 fishing communities xxiii–xxiv, 114–15, 344–402 Flanders 19, 22, 28, 31–2, 54 Folk theorem 93, 98 food consumption 51–2, 135, 136, 137 food markets 266, 273–4 forest management xxiii, 295–317 formal user group management 302, 303–5, 307, 312, 313–14, 315 France ; agricultural labor productivity 53; pre-modern 6 free rider problem ; experience rating 239; game theory 108, 111, 112, 113; Indonesian peasant communities 142; irrigation in the Philippines 269, 279, 289; Japanese fishing communities 351, 354, 355, 357–8, 379–80, 393; screening 200 gama system 87, 89–91, 92, 94 game theory xx, 7, 70, 97–125, 405; cooperation 248; impersonal exchange 11–16; social sanctions 296; tragedy of the commons 295 Ghana, business networks 190, 192, 194, 198, 200, 206, 208 gift-giving 104–5, 218, 219 gossip 164 governance ; community to market-based transition 3; game theory 117, 118; Laos hand-weaving industry 164, 167, 176, 177, 178, 182; Mexican ejido communities 319–20, 321, 326 government ; involvement in irrigation schemes 254, 259, 261, 263; public goods 112; support for community management xxiii, 315see alsostate Green Revolution 89, 94, 266, 273, 274

group size seesize of group/community group-lending contracts 237 guilds xviii–xix, 18, 344–5 harvesting arrangements 73–5, 76–81, 85, 89–91, 92 heterogeneity ; community responsibility system 30; Indian caste system

INDEX

253; Japanese fishing communities 380–1, 382, 383–5, 386–90, 393, 394–6; Mexican ejido communities 330, 331 hired labor 82, 85, 89, 92 human capital 353 Hungary, agricultural labor productivity 54 hunusan system 73, 74, 75–81, 87–91, 92 IAs seeirrigators' associations identity, impersonal exchange 12–13, 26 IDOs seeinstitutional development officers impersonal exchange 5, 7–10, 11–16, 17–36, 103, 406 in-kind transfers 224, 227, 228 incentives ; community irrigation management 266, 276; group heterogeneity 380–1; Japanese fishing communities 351–2, 353, 358, 385, 396–7 income ; China 217–18, 220, 221–3, 225–7, 230–1, 232–3; enclosure effect on 63; land reform in the Philippines 81–2; pooling arrangements in Japanese fishing communities 114–15, 348–55, 358, 360–4, 374, 378–9, 398; redistribution of China 230community norms 175enclosure effect on 63peasant community norms 129see also savings income shocks 181, 217, 218, 231 India ; irrigation xxii, 247–64, 278; Philippines comparison 82, 86 indigenous communities, Mexico 318 individual responsibility 6, 25, 34–5, 36 Indonesia xxi, 129–52, 407 Industrial Patriotic Society (Japan) 122, 123 industrialization 122, 149 inequality ; China 219; cooperation relationship 253; distribution of resources 265, 270; income 81–2; Indian landholding 256, 260, 261, 263; Mexican land appropriation 339 inflation, English agricultural development 48, 49, 52 informal contract enforcement 4, 9 informal forest user groups 298–9, 300–2, 304–7, 312, 313–14, 315 informal risk sharing ; China 216–18, 223–4, 225–7, 230–1, 233, 240–1; Japanese fishing communities 351, 352 informal sanctions, Laos hand-weaving industry 153, 154, 162, 163–4, 182 information-sharing ; community-based trust 164; Japanese fishing communities 114–15, 351, 354–5, 374; SubSaharan African business networks 186, 190, 191, 193, 196, 198–9 innovation, open field agriculture 45, 56, 57, 66 insider hypothesis 153 institutional development officers (IDOs) 285, 287 institutions xv, 215, 408; China 234; complementarity 81; contract enforcement 9, 14, 16, 36; irrigation in India 247; Philippines irrigation projects 285–9; pre-modern Europe 36–7

413

insurance ; China xxii, 215–43, 407; pooling arrangements in Japanese fishing communities 349–54, 361, 362–70, 382–3,

414

INDEX

394–6; Sub-Saharan Africa 192; village networks 344; work and income sharing 88see also agriculture, insurance; risk inter-community impersonal exchange xx, 5, 11, 17, 20–1; community affiliation 16, 24, 36; end to CRS 30, 32 interest, agricultural credit in Indonesia 134, 139, 140–1 intra-community contract enforcement, pre-modern Europe xix–xx, 4–5, 11, 14, 16, 24, 35–6 investment, Philippines irrigation projects 272–3, 284, 288, 289, 292 irrigation ; India xxii, 247–64, 278; Indonesia 133; Japan 107–8, 292, 344; Mexican ejido communities 321; Philippines xxii–xxiii, 265–94, 311 irrigators' associations (IAs) 265–6, 276–81, 284–5, 287–9 Italy ; agricultural labor productivity 54; political fragmentation 6; pre-modern states 8, 10, 17, 19–23, 25–6, 27–9, 32–3 itinerant merchants 181, 182 Japan ; agriculture 117; community norms 121; economic growth 149–50; employment hierarchy 115–16; entrepreneurship 148–9, 151; fishing communities xxiii–xxiv, 114–15, 344–402; forest management 306; industry 121–4; irrigation 107–8, 292; trade 118–20 Kanban (just-in-time) system 123, 150 Kenya ; business networks 194, 195, 196, 197, 198, 200, 208; ethnic groups 202, 205–6 labor ; agricultural productivity 52–6; community irrigation units 249–50, 255; employee referral 190; Japanese industry 115–16, 121–3; Mexican ejido communities 320; Philippine agrarian systems 73–5, 82–94; work and income sharing 87–9, 91, 92, 93, 94see alsowages land issues ; China 217, 236–7; Mexican ejido communities xxiii, 318–19, 320–2, 323–6, 336–40, 342; Philippines 72–3, 74–6, 80–2, 84–5, 91–2, 93–4 landlordism ; eighteenth-century England xx, 63, 64–6; Philippine agrarian systems 72–3, 81, 86, 89, 93 Laos, hand-weaving industry xxi, 153–85, 407 legal system ; contract enforcement 7, 187; impersonal exchange 7, 11; individual responsibility 34–5, 36; market economy 3; pre-modern Europe xx, 9 legislation ; Japanese fisheries 345; Mexican land reform 322; Nepalese Forest Act (1993) 299; Philippines Agrarian Reform Code (1963) 76; Statute of Acton Burnell (1283) 33–4 local commons xvi, xix, 97; community rule-enforcement

INDEX

280; irrigation in the Philippines 269, 289, 292; small ostracism xvi, 108–11, 194, 280, 296, 329 group cooperation 250; social exchange game 106–13; outsider hypothesis 153–4 state to community management 265see also common property resources; public goods Madagascar, business networks 189, 192, 204 Malagas, business networks 191, 192, 206 management seecommunity management manufacturing ; African relational contracting 187–8; contractual flexibility 192; Sub-Saharan African business networks 198, 199, 201 market economy xv, xvi–xvii, 3, 117–18, 149 market failures xvii, 344 market integration 331 marketing ; Indonesian agricultural peasant communities 129, 130, 133–7, 151, 407; Japanese fishing communities 392; Laos hand-weaving industry 158–60 markets ; business networks 208, 209; China 215–16, 218; closed 200; community norms xv, xvi–xvii, xviii; community relationship 3, 4–6, 42, 66; decentralized 186, 193, 194; irrigation in India 255, 262; Nepalese forest management 297, 300; peasant communities 129; rural–urban integration 136, 153, 157, 159, 180, 181–2; secondary 320; state control xix ; structure 408 mechanized agriculture 73–4 Merchant Law System 6, 7 Mexico, ejido communities xxiii, 318–43 microfinance programmes 237, 239, 240, 241 middlemen, Indonesian peasant communities 129, 143 migrant workers ; Laos hand-weaving industry 169, 171–2, 174–5, 176–8; Philippines agriculture 79–80 monopsonistic power 129, 134, 142–3, 146–7 Morocco, business networks 195, 204 multiple equilibria 98, 168, 248 mura hachibu 108 mutual insurance 189, 203–4, 369 Nash equilibrium 100, 101, 103–4, 105, 163, 397–8 National Irrigation Administration (NIA), Philippines 268, 269, 270, 271–2, 274–5, 276–92 nationalization of forests 297–8 Nepal, forest management xxiii, 295–317 nepotism 200–1 net damage 352, 355, 356–8, 364–70, 371, 372–3 Netherlands, agricultural labor productivity 54–5 network capital 207, 208 networks, Sub-Saharan Africa xxi–xxii, 186–214 NIA seeNational Irrigation Administration Nigeria, ethnic groups 205 O&M seeoperations, maintenance activities old age insurance 241 open field agriculture xx, 42–69 operations and maintenance (O&M) activities, irrigation systems 271–5, 276, 280–2, 284, 287–9

415

416

INDEX

ownership ; English agricultural land 42, 64; local commons 265; Mexican land reform 322, 325; Nepalese forest management 295; Philippines land reform 81see also landlordism ‘pastoral factory’ 150 peasant communities ; Indonesia xxi, 129–52, 407; outsider hypothesis 153–4see alsoejido communities; rural areas peer pressure 119, 380, 391, 393 People's Insurance Company (PIC), China 218, 235, 236, 241 personal trust 102, 106 Philippines ; irrigation xxii–xxiii, 265–94, 311; mutual insurance 203–4; Rice Bowl agrarian systems xx, 70–96 PIC seePeople's Insurance Company politics ; community responsibility system 6, 31, 32; ethnic discrimination 202–3; Mexican reforms 318–19, 321–3; Philippines 274 pooling, Japanese fishing communities xxiii–xxiv, 344–402 population growth ; England 51, 55; Mexican ejido communities 336, 339, 340; Nepal 297, 298, 300; Philippines 89; pre-modern Europe 26; Southeast Asia 292 power laws 403 pre-modern economic systems xvii, xix–xx, 3–41, 117 prices ; English agricultural development 52; Indonesian agricultural trade 134–5, 138, 139–40, 142–3, 145–7; Japanese fishing communities 358–60, 375–8; Laos handweaving industry 165, 166, 171 prisoner's dilemma 100, 163, 166, 183, 397 private pump irrigation 281–3, 284, 287, 288, 291, 292 privatization 265, 321 process-based trust 162–3, 167–8, 170, 176, 178, 182–4 productivity ; common property resources 336; labor 52–6; open field agriculture 43, 45, 46–50, 55, 65–6; personal relationships 189–90 profitability ; common property resources 336; pooling arrangements 349, 352, 354–61, 364, 370–9, 382–3, 394–6 property ; private ownership 295; rights xvi, xvii, 3, 187, 265, 318, 406see alsocommon property resources; land issues; landlordism public goods ; free rider problem 112, 113; Mexican ejido communities 319, 322, 326–8, 334; non-excludability 265; state role xvi, xxii; Sub-Saharan Africa 193; voluntary contribution mechanisms 379see alsocommon property resources; local commons public support, irrigation in the Philippines 272–3, 276, 289, 292 Public Works Department (PWD), India 249, 250, 254, 259, 262–3 punishment ; game theory 14, 15–16, 101, 102; Japanese wartime industry 122; Nepalese forest management 306;

pre-modern Europe 29,

INDEX

34; Sub-Saharan Africa 194, 195see also retaliation; sanctions putting-out system 121, 160–2, 164–72, 176–80, 182–3 PWD seePublic Works Department quid/quo separation 5, 6, 7–11, 12 rank 351, 380, 381, 382, 386, 395see alsostatus redistribution of income ; China 230; community norms 175; enclosure effect on 63; peasant community norms 129 referral xxi–xxii, 186, 190–1, 200, 205–6 reforestation 296, 303 relational contracting ; Japan 119; Laos 160–8, 171, 182, 183, 407; Sub-Saharan Africa 186, 187–8 relationships ; community definition xv, 207; Sub-Saharan African business networks 188, 189–92, 193, 198, 208 religion 194–5, 204–5 rents ; English agriculture 46–7, 48–9, 63, 64–5; Philippines agrarian systems 74–5, 89 reputation xviii, 98–102, 142, 190–1, 195, 253 resources ; community participation 329, 330; forest community management 295–6, 300, 303, 304, 306, 314; irrigation in the Philippines 289; Japanese rule enforcement 292; over-exploitation of 302; population growth 292; tragedy of the commons 265; unequal distribution 265, 270see alsocommon property resources; irrigation retailers, Laos hand-weaving industry 175–80, 182, 183 retaliation 11, 20–3, 27, 28–9, 32, 102 revoked embeddedness 120–1 Ricardian surplus 47, 48, 49, 63 rights ; informal forest user groups 299, 301, 302, 312; Philippine land reform 82; property xvii, 3, 187, 265, 318, 406 risk ; China 216–18, 223–4, 225–7, 230–1, 233, 238–41; Japanese fishing communities 349, 352–4, 355, 362–70, 373, 399–400; price 145, 146–7; Sub-Saharan markets 189 rotation schemes, Japanese fishing communities 355–6, 360–1, 371 rules ; community norms 280; enforcement xix, 292; Indian water allocation xxii, 250, 255–6, 262–3; Japanese fishing communities 345, 356; Mexican ejido communities 328–9; Nepalese community forest management 299, 301, 302, 304–6, 312 rural areas ; agricultural labor productivity 53; China 216, 217–41; community management xvi, xxii–xxiii; community norms 98; industrialization 122, 149; irrigation systems 265, 280; Laos hand-weaving industry 153–85; local commons xvi ; Mexican ejido communities 318–43; pre-modern to market economy transition 117–18see alsoagriculture; peasant communities Russia, business communities 208 sainfoin production 57–9, 61–2

417

418

INDEX

sanctions ; community size 201; Laos hand-weaving industry 153, 154, 162, 163–4, 182; Nepalese forest management 296; ostracism xvi, 108, 109–10, 111; small communities 253; work and income sharing violations 89see also punishment savings 224, 231–4, 240, 241 scale economies ; Indian irrigation system 253; Indonesian agricultural trade 134, 135, 137, 138; Japanese fishing communities xxiii, 356–8, 361, 368–9, 374–5, 396; Mexican ejido communities 334, 335, 342 search costs, personal relationships 190 self-insurance 369 share cropping 65, 72, 74–5, 76–7, 91–2, 93 share tenancy 74–5, 76, 83, 86 sharing systems of agriculture 87–9, 92, 93, 94 size of group/community ; community responsibility system 28, 30; contract enforcement xviii; Mexican ejido communities 330–1, 333–4, 335–6, 342; Nepalese community forest management 315; social capital 201 skills ; division of labour 398; Japanese fishing communities 353–4, 357, 367–8, 370, 382–5, 388–92, 394 slash-and-burn 338, 339 small businesses 149–50, 191 social capital 118, 121, 201, 329; business networks 195, 207; English agriculture xx, 63–4, 66, 67; game theory 98, 110, 111, 113, 114, 115; Japanese merchant trade 119 social cohesion 331 social embeddedness xx, 98, 105, 113–24 social exchange domain 98, 106–14, 116, 120–1, 123, 124 social heterogeneity ; community responsibility system 30; Indian caste system 253; Japanese fishing communities 380–1, 382, 383–5, 386–90, 393, 394–6; Mexican ejido communities 330, 331 social homogeneity ; Indian irrigation communities 253, 261, 263; Indonesian peasant communities 143; Japanese fishing communities 345, 392; Nepalese forest user groups 307, 308 social mobility 196, 197–8 social obligation 112–13 social similarity 171–2, 174–5, 176, 180, 181, 183 social status 115, 122, 347, 383, 384, 390–1see alsorank social structure ; contract enforcement 16, 36, 37; Indonesian peasant communities 143–4; local commons 292 socialization 206 Somalia, ethnic groups 205 Spain, agricultural labor productivity 54 Spelsbury case study 56, 57–63, 64 spot market transactions 160, 165–6, 171, 172, 179, 180, 182–3 standardization of products 209 state ; community responsibility system 32–3, 35; contract enforcement 25; control over ejido communities 319–20,

334,

INDEX

340; local commons management 265, 266; public investment 292; role xvi, xixsee also government status 115, 122, 347, 383, 384, 390–1see alsorank sub-game prefect equilibrium (SGPE) 12, 15 Sub-Saharan Africa, business networks xxi–xxii, 186–214 subcontracting 150, 151, 191 subtenancy 85 taxes 65, 250 technical change, open field agriculture 43 technology 116, 353, 385, 392, 406 tenant farmers ; England xx, 61, 64–6; Philippines 72–5, 76, 80, 81–2, 89 tilyadora system 73–4, 75–81 Toyota 150 trade ; closed markets 200; ethnic group reputation 195; European expansion 5–6; Indonesian peasant communities 134–5; Japan 118–20; medieval 190; pre-modern Europe 9–10, 20, 27; Sub-Saharan African business networks 186 ‘traders’ dilemma' 129, 142, 148, 154, 175, 183 tragedy of the commons xxii, xxiii, 265, 280, 295, 361 transaction costs ; casual farm labor 85, 92; Chinese agricultural insurance 236–7; community forest management 296, 307–8, 315; contract enforcement 4, 27, 164; exchange 405–6; guilds xix; Indonesian vegetable trading 134–5, 138, 141; participation 330; relational contracting 162, 407; share rent 74, 75; Sub-Saharan business networks 186, 208; wholesalers 207 transfers, inter-household 224–31 transition economies 216 treatment effects model 314 trust ; community relations xvii, 154; Laos hand-weaving industry xxi, 162–3, 167–8, 170–2, 175–8, 182–4; social heterogeneity 253; Sub-Saharan African business networks 186–8, 208 turnip production 59–63 urban areas, irrigation in India 255, 260, 261, 262 urbanization 55–6 user group management 298–9, 300–15 voluntarism, open field agriculture 56, 60, 63 wages ; Japanese industry 115, 116, 121; Laos hand-weaving industry 170; Philippines agrarian systems 89, 90–1 water management seeirrigation water scarcity 254 weaving techniques 156–7 work and income sharing 87–9, 91, 92, 93, 94 World Food Crisis (1973–4) 266, 273 Zambia, religion 195 Zimbabwe 192, 196, 199–200, 202, 205, 208 m>work and income sharing 87–9, 91, 92, 93, 94 World Food Crisis (1973–4) 266, 273 Zambia, religion 195

Zimbabwe 192, 196, 199–200, 202, 205, 208

419