Clusters and Regional Development (Regions and Cities)

Clusters and Regional Development Over the past decade the ‘cluster model’ has been seized on as a tool for promoting ...

Author: B. Asheim

62 downloads 2262 Views 3MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

Clusters and Regional Development

Over the past decade the ‘cluster model’ has been seized on as a tool for promoting competitiveness, innovation and growth at local, regional and national scales. However, despite its popularity there is much about the cluster model that is problematic, and in some respects the rush to employ ‘cluster ideas’ has run ahead of many fundamental conceptual, theoretical and empirical questions. As such, there is a need for a more thorough theoretical and empirical evaluation of the notion. This book fulfils that need, assessing the cluster notion and drawing out not only its undoubted strengths and attractions but also its weaknesses and limitations. The editors have brought together a number of leading scholars who address several key questions: How do we define and identify clusters? What is meant by ‘cluster theory’? What are the benefits of clusters? What are their disadvantages? How do clusters evolve? Do clusters improve regional innovativeness and competitiveness, and if so, in what ways? How do clusters relate to the global economy? What are the limits of the cluster concept? What policy options are available to promote clusters? This critical examination of the clusters concept will be important reading for economic geographers, economists, planners and policy makers. Bjørn Asheim is Professor of Economic Geography at the Lund University, Sweden, and Professor at the Centre of Technology, Innovation, and Culture, University of Oslo, His research interests include international comparisons of creative cities, clusters and regional innovation systems. Philip Cooke is University Research Professor and founding Director of the Centre for Advanced Studies, University of Wales, Cardiff. His research interests lie in studies of economics of biotechnology, regional innovation systems, knowledge economies, and policy actions for business clusters and networks Ron Martin is Professor of Economic Geography and Fellow of the Cambridge-MIT Institute at the University of Cambridge. His research covers the theory and empirics of regional growth and competitiveness, local labour markets, the geographies of money and finance, and the spatial evolution of the ‘new economy’.

Regions and Cities (formerly known as Regional Development and Public Policy) Series editors: Ron Martin, University of Cambridge, UK, Gernot Grabher, University of Bonn, Germany, Maryann Feldman, University of Toronto, Canada

Regions and Cities is an international series that aims to provide authoritative analyses of the new significance of regions and cities for economic development and public policy. It seeks to combine fresh theoretical and empirical insights with constructive policy evaluation and debates, and to provide a definitive set of conceptual, practical and topical studies in the field of regional and urban public policy analysis. Regions, Spatial Strategies and Sustainable Development Graham Haughton and Dave Counsell (eds) Geographies of Labour Market Inequality Ron Martin, Philip Morrison (eds) Regional Development Agencies in Europe Henrik Halkier, Charlotte Damborg and Mike Danson (eds) Social Exclusion in European Cities Processes, experiences and responses Ali Madanipour, Goran Cars and Judith Allen (eds) Regional Innovation Strategies The challenge for less-favoured regions Kevin Morgan and Claire Nauwelaers (eds) Foreign Direct Investment and the Global Economy Nicholas A. Phelps and Jeremy Alden (eds) Restructuring Industry and Territory The experience of Europe’s regions Anna Giunta, Arnoud Lagendijk and Andy Pike (eds) Community Economic Development Graham Haughton (ed.) Out of the Ashes? The social impact of industrial contraction and regeneration on Britain’s mining communities David Waddington, Chas Critcher, Bella Dicks and David Parry

Clusters and Regional Development Critical reflections and explorations

Edited by Bjørn Asheim, Philip Cooke and Ron Martin

First published 2006 by Routledge 2 Park Square, Milton Park, Abingdon, Oxon OX14 4RN Simultaneously published in the USA and Canada by Routledge 270 Madison Avenue, New York, NY 10016 Routledge is an imprint of the Taylor & Francis Group, an informa business © 2006 Bjørn Asheim, Philip Cooke and Ron Martin for selection and editorial matter; individual chapters the contributors

This edition published in the Taylor & Francis e-Library, 2006. “To purchase your own copy of this or any of Taylor & Francis or Routledge’s collection of thousands of eBooks please go to www.eBookstore.tandf.co.uk.” All rights reserved. No part of this book may be reprinted or reproduced or utilised in any form or by any electronic, mechanical, or other means now known or hereafter invented, including photocopying and recording, or in any information storage or retrieval system, without permission in writing from the publishers. British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Library of Congress Cataloging in Publication Data Asheim, Bjørn Terje. Clusters and regional development : critical reflections and explorations / Bjørn Asheim, Philip Cooke and Ron Martin. p. cm. – (Regional development and public policy series) Includes bibliographical references and index. 1. Regional planning. 2. Economic development. 3. Industrial clusters. 4. Regional economics. 5. Space in economics. I. Cooke, Philip (Philip N.). II. Martin, Ron (Ron L.) III. Title. IV. Series: Regional development and public policy. HT391.A843 2006 307.1′2–dc22 2005026954 ISBN10: 0–415–34914–1 (Print Edition) ISBN10: 0–203–64089–6 (ebk) ISBN13: 978–0–415–34914–7 ISBN13: 978–0–203–64089–0 (ebk)

Contents

List of tables and figures Notes on contributors Preface 1 The rise of the cluster concept in regional analysis and policy: a critical assessment

vii ix xvii

1

BJØRN ASHEIM, PHILIP COOKE AND RON MARTIN

2 What qualifies as a cluster theory?

30

PETER MASKELL AND LEÏLA KEBIR

3 True clusters: a severe case of conceptual headache

50

ANDERS MALMBERG AND DOMINIC POWER

4 In search of a useful theory of spatial clustering: agglomeration versus active clustering

69

FIORENZA BELUSSI

5 Cutting through the chaos: towards a new typology of industrial districts and clusters

90

IVANA PANICCIA

6 Entrepreneurs as agents in the formation of industrial clusters

115

MARYANN FELDMAN AND JOHANNA L. FRANCIS

7 Innovation, learning and cluster dynamics BART NOOTEBOOM

137

vi Contents 8 Do clusters or innovation systems drive competitiveness?

164

JAMES SIMMIE

9

The role of clusters in knowledge creation and diffusion: an institutional perspective

188

DAVID B. AUDRETSCH AND ERIK E. LEHMANN

10 Do clusters ‘think’?: an institutional perspective on knowledge creation and diffusion in clusters

199

MICHAEL STEINER

11 Spaces of knowledge flows: clusters in a global context

218

MERIC S. GERTLER AND DAVID A. WOLFE

12 Spatial and organizational patterns of labor markets in industrial clusters: the case of Hollywood

236

ALLEN J. SCOTT

13 Cluster and hinterland: when is a proactive cluster policy appropriate?

255

G. M. PETER SWANN

14 Putting clusters in their place

272

NICK HENRY, JANE POLLARD AND PAUL BENNEWORTH

Name index Subject index

292 296

Tables and figures

Tables 2.1 5.1 7.1 7.2 8.1 8.2 8.3 8.4 8.5 8.6 11.1 12.1 13.1 13.2 13.3 13.4

Cluster publications 1953–2004 Distinctive features of the generic typology Sources of collaboration Networks for exploration and exploitation UK competitive industries Interlinked trading clusters in England and Wales UK competitive sectors not interlinked with other industries Reasons for location of innovative firms in five European cities Local co-operation arrangements by type of collaborator Growth in key indicators of competitiveness of non- and leading innovators Typology of knowledge bases and flows Union locals and guilds with collective-bargaining authority in contemporary Hollywood, 2002 Approximate relationship between network value and network size Effect of new entry on cluster stakeholders Origin of components for typical personal computer Relationships between degree of specialization in clusters and total transportation

32 100 148 156 170 171 173 177 179 182 230 249 259 261 263 265

Figures 1.1 1.2 4.1 4.2 4.3 4.4 12.1

Increasing returns and economic localization Porter’s cluster model Agglomeration versus active clustering Main definitions of industrial districts or clusters Map of industrial district or cluster Evolution of industrial districts or clusters Monthly employment in SIC 78 (Motion Pictures), Los Angeles County, January 1983 to March 2002

5 10 78 80 82 83 241

viii Tables and figures 12.2 Residential locations of members of IATSE Local 80 in Southern California 12.3 Residential locations of members of the Writers’ Guild in Southern California 12.4 Locations of agents, casting directors, and managers in Los Angeles 13.1 Different interpretations of the cluster concept 13.2 The lattice/grid transport network 13.3 The orbital transport network 13.4 Throughput-delay curves

242 243 247 257 264 264 266

Notes on contributors

Bjørn T. Asheim ([email protected]) is Professor of Economic Geography at Lund University, Sweden, and Professor at the Centre of Technology, Innovation, and Culture, University of Oslo, Norway. He is co-founder and deputy director of the new Centre of Excellence in innovation system research (CIRCLE) at Lund University. He was previously Professor of Human Geography, University of Oslo, Norway. He has recently co-edited Regional Innovation Policy for Small-Medium Enterprises (2003) and is the author of numerous articles in economic and industrial geography including papers on industrial districts, regional innovation systems and learning regions. He is an Editor of Economic Geography and on the editorial board of European Planning Studies and the Journal of Economic Geography. His current research is on international comparisons of creative cities, clusters and regional innovation systems. David Audretsch ([email protected]) is the Ameritech Chair of Economic Development, Director of the Institute for Development Strategies at Indiana University. He also directs a research program on Entrepreneurship, Growth and Public Policy at the Max Planck Institute in Jena, Germany, and is a Research Fellow of the Centre for Economic Policy Research, London. His research focuses on the links between entrepreneurship, government policy, innovation, economic development and global competitiveness. He has acted as consultant to numerous organizations and governments, including the World Bank, the U.S. State Department, the United Nations, Commission of the European Union, and the OECD. He has published thirty books and more than one hundred scholarly articles in leading academic journals. He was awarded the 2001 International Award for Entrepreneurship and Small Business Research by the Swedish Foundation for Small Business Research. Fiorenza Belussi ([email protected]) is Professor of Business Economics at Padua University. She has undertaken major international research on the issue of industrial districts, clusters, learning regions and local production systems in Europe and in the eastern countries. She has recently co-edited The Evolutionary Patterns of Local Production Systems (2000) and

x Notes on contributors Technological Evolution of Industrial Districts (2003), and she is the author of numerous articles and papers concerning the economic of innovation in small firms, the evolution of industrial policies at local level, the emergence of business networks, and the organizational restructuring of service sectors and creative industries. Paul Benneworth ([email protected]) is a Research Councils UK Academic Fellow at the Centre for Urban and Regional Development Studies at the University of Newcastle, and a visiting Senior Research Fellow at the Department of Human Geography at the Radboud University, Nijmegen, the Netherlands. Paul is a board member of the Regional Studies Association, and edits Regions, the newsletter of the RSA. Paul’s research concerns knowledge-based regional economic development in old industrial regions, particularly the role of universities in regional economic development and high technology entrepreneurship in less successful places. Philip Cooke ([email protected]) is University Research Professor and founding Director (1993) of the Centre for Advanced Studies, University of Wales, Cardiff. His research interests lie in studies of economics of biotechnology, regional innovation systems, knowledge economies, and policy actions for business clusters and networks. His books include: The Associational Economy (with Kevin Morgan, 1998), The Governance of Innovation in Europe (2000), Knowledge Economies: Clusters, Learning and Cooperative Advantage (2002), and Regional Innovation Systems (2004). He was adviser to the UK Government’s Biotechnology Clusters, has been EU adviser on Regional Foresight, Universities and Regional Development, and in 2004–5 chaired the EU committee on Constructing Regional Advantage. He is OECD adviser on Knowledge Economies, and UNIDO adviser on Innovation Systems. He is an Academician of the UK Academy of Social Sciences, and a Distinguished Research Fellow of the University of Ottawa Management School. Maryann Feldman ([email protected]) is Professor of Business Economics at the University of Toronto where she holds the Jeffrey Skoll Chair in Innovation and Entrepreneurship. Her research interests focus on the areas of innovation, the commercialization of academic research and the factors that promote technological change and economic growth. A large part of Maryann Feldman’s work concerns the geography of innovation – investigating the reasons why innovation clusters spatially and the mechanisms that support and sustain industrial clusters. Her books include The Geography of Innovation (1993), The Handbook of Economic Geography (2000), Innovation Policy in the Knowledge-Based Economy (2001), Institutions and Systems in the Geography of Innovation (2002) and The Economics of Science and Technology (2002). Johanna L. Francis ([email protected]) is a researcher at the World Bank in Washington DC, and a doctoral student in the Department of Economics at

Notes on contributors xi Johns Hopkins University. Her research focuses on the relationship between entrepreneurship and capital formation and the impact of entrepreneurial saving decisions on the distribution of wealth. She has co-authored several articles on cluster formation considering the role of the entrepreneur as the critical motivator of innovation and economic change. Meric Gertler ([email protected]) is Professor of Geography and Goldring Chair in Canadian Studies at the University of Toronto. With David Wolfe, he co-directs both the Program on Globalization and Regional Innovation Systems at University of Toronto’s Munk Centre for International Studies, as well as the Innovation Systems Research Network. Meric’s research interests include regional innovation systems, the globalization of technologies and industrial cultures, and the development of business clusters. His recent publications include Manufacturing Culture (2004), Innovation and Social Learning (with David Wolfe, 2002), The Oxford Handbook of Economic Geography (with Gordon Clark and Maryann Feldman, 2000), and The New Industrial Geography (with Trevor Barnes, 1999). Nick Henry ([email protected]) is a Reader in the Centre for Urban and Regional Development Studies at the University of Newcastleupon-Tyne. He combines this role with a part-time post as Senior Consultant in Regeneration and Economic Development for an employee-owned company public policy consultancy, GHK Consulting Ltd. His primary research interests are in theorizing, implementing and evaluating models of regional development in the ‘advanced’ economies. Conceptually he is interested in models of new economic geographies including clusters, learning regions, networks and value chains, multicultural economic development and alternative economic geographies. Over the past decade, he has investigated these through a variety of academic and consultancy projects. Currently he is Human Geography Editor for the RGS/IBG (Royal Geographical Society/ Institute of British Geographers) Book Series with Blackwell Publishers. His books include: The Economic Geography Reader (co-edited, 1999) and Knowledge, Space, Economy (co-edited, 2000). Leïla Kebir ([email protected]) is an economist specializing in regional economics. She has completed her PhD thesis at Neuchâtel University (Switzerland). Her main research interests are regional development, regional resources, competitiveness and territorial systems of production. She is a member of the European Research Group on Innovative Milieus. She is currently a postdoctoral researcher at the Ecole des Hautes Etudes de Sciences Sociales (EHESS), Paris. Erik E. Lehmann ([email protected]) is Professor of Business and Organization at the University of Augsburg. He took his PhD in economics at the University of Rostock in 1999. He worked as Assistant Professor in the Department of Economics at the University of Konstanz, and Research Fellow in Entrepreneurship, Growth and Public Policy at the

xii Notes on contributors Max Planck Institute of Economics, Jena. His research interests focus on entrepreneurship, technology spillovers, corporate governance, and firm financing. He has published some 36 papers and articles on these and related themes. Anders Malmberg ([email protected]) is Professor of Economic Geography and Director of the Centre for Research on Innovation and Industrial Dynamics (CIND) at Uppsala University, Sweden. Between 1990 and 1995 he was the research co-ordinator of the European Science Foundation Scientific Programme on Regional and Urban Restructuring in Europe. He has held visiting positions at Durham University, UK and the Swedish Collegium for Advanced Study in the Social Sciences (SCASSS). His research focuses on the intersection between industrial transformation and local/regional economic development. His recent work has dealt particularly with the role of the local milieu in promoting knowledge creation and firm competitiveness, and more generally the causes and effects of regional specialization and industry agglomeration. He has published numerous papers in academic journals in economic geography and related fields. He is a member of the Royal Society of Sciences at Uppsala. Ron Martin ([email protected]) is Professor of Economic Geography at the University of Cambridge, UK, and Research Associate of the Centre for Business Studies and a Fellow of the Cambridge-MIT Institute there. His research interests include: regional growth and competitiveness; financial markets and regional development; economic theory and economic geography; local labour markets; the geographies of the ‘new economy’, and the geographies of public policy. He has published some 25 books and more than 135 papers on these and related topics. Recent books include Money and the Space Economy (1999), The Geographies of Labour Market Inequality (with Philip Morrison, 2003), Putting Workfare in Place: Local Labour Markets and the New Deal (with Peter Sunley and Corinne Nativel, 2005), and The Competitive Advantage of Regions (with Michael Kitson and Peter Tyler, 2005). Ron has undertaken research for the European Commission and the UK government on clusters, regional growth, and city competitiveness. He is an Academician of the UK Academy of Social Science and a Fellow of the British Academy. Ron is listed by the American Economic Association (2003) as one of the world’s most cited economists. He has edited several journals, including Transactions of the Institute of British Geographers and Regional Studies, and is currently an associate editor of the Journal of Economic Geography and an editor of the Cambridge Journal of Economics. Peter Maskell ([email protected]) is Professor of Regional Economics at Copenhagen Business School, Denmark, and Director of DRUID – The Danish Research Unit for Industrial Dynamics. He is Chairman of the Governing Board of DIME – the EU Network of Excellence on Dynamics of Institutions and Markets in Europe. Peter is a member of Academia Europea

Notes on contributors xiii and former Chairman of the Danish Social Science Research Council, Research Director at Centre for Business and Economic Research and former member of the European Science Foundations Standing Committee for the Social Sciences, and of the Danish Parliaments Standing Committee on Research. His research interests include: industrial dynamics; economic and spatial organization; economics of knowledge; and globalization. He has published extensively within these areas. He serves on the board of several corporate enterprises and has undertaken major research projects on innovation, industrial specialization and economic performance. Bart Nooteboom ([email protected]) is Professor of Innovation Policy at Tilburg University, the Netherlands. His research interests include: innovation, entrepreneurship, organizational learning, inter-firm collaboration and networks, and trust. He has published six books and 250 papers on these and related topics. He was member of a government committee on technology policy and of a variety of committees for the Ministry of Economic Affairs. He was director of a research-institute/PhD school for Economics, Business and Geography. He is a member of the Royal Netherlands Academy of Arts and Sciences. He was awarded the Kapp prize for his work on organizational learning and the Gunnar Myrdal prize for his work on trust. Ivana Paniccia ([email protected]) is Lecturer in Industrial Economics at the Luiss University of Rome, and Head of the Economic Analysis Department of the public utilities’ ‘watchdog’ of the Municipality of Rome. Her main interests are in the field of regional economics, international business economics and economics of public utilities regulation. She has undertaken various national and international research projects in applied economics, including EC-funded projects, and published various papers in academic journals in economic geography, organization studies and related fields. Her most recently published book is Industrial Districts: Evolution and Competitiveness in Italian Firms (2002). Jane Pollard ([email protected]) is a Senior Lecturer in the Centre for Urban and Regional Development Studies at the University of Newcastleupon-Tyne. Her research interests include geographies of money and finance, the role of different financial intermediaries in regional economic development and the changing nature and practices of economic geography. Her current research focuses on postcolonial economic geographies and the social, cultural and religious elements of financial networks. She is a co-editor of Knowledge, Space, Economy (2000). Dominic Power ([email protected]) is Associate Professor of Economic Geography at Uppsala University, Sweden. His research is concerned with regional and industrial competitiveness, and innovation dynamics. His chief empirical focus has been the workings of the cultural industries: in particular the music, design and fashion industries. He has

xiv Notes on contributors published extensively within these areas and recent publications include the co-edited book Cultural Industries and the Production of Culture (2004). Dominic has worked as a policy adviser and consultant to various Nordic government ministries and innovation authorities. Allen J. Scott ([email protected]) holds the rank of distinguished professor with joint appointments in the Department of Geography and the Department of Public Policy at the University of California – Los Angeles. He was awarded a Guggenheim Fellowship in 1986–7, and was granted honours by the Association of American Geographers in 1987. He was elected Fellow of the British Academy in 1999, and was the recipient of the Vautrin Lud Prize for 2003. He has occupied the André Siegfried Chair at the Institut d’Etudes Politiques, Paris, (1999), the First Trust Bank Chair of Innovation at Queen’s University, Belfast (2004), and the Chaire d’Excellence Pierre de Fermat at the University of Toulouse-Le Mirail (2005). His most recently published books are The Cultural Economy of Cities (Sage, 2000) and On Hollywood (Princeton University Press, 2005). James Simmie ([email protected]) is Professor of Innovation and Urban Competitiveness at Oxford Brookes University. He is an economist, sociologist and a qualified town planner. His main research interests are focused on the relationships between innovation and urban regions. He has worked and published extensively in this area particularly with respect to the UK and international comparisons in Europe and North America. Recently he participated in the ESRC ‘Cities: Competitiveness and Cohesion’ programme analysing the reasons for the innovative performance of some of Europe’s most successful cities. This has been followed by major projects analysing the reasons for the relatively poor competitive performance of the English core cities when compared with many of their European counterparts and the economic potential of England’s largest cities. He is currently working on an ODPM-funded study of the competitiveness of the 56 largest English cities. This forms one of the major strands of research in a comprehensive study of the State of English Cities Report. Michael Steiner ([email protected]) is Professor at the Department of Economics at the Karl-Franzens-University of Graz, Austria, and Director of the Institute of Technology and Regional Policy at Joanneum Research. He was also visiting Fellow and Professor at the Universities of Glasgow, Kent at Canterbury, Bocconi, Klagenfurt and Pecs. His main fields of research deal with regional and industrial economics, technological change and economic policy. His recent publications focus on questions of European integration from a transregional perspective and include Clusters and Regional Specialisation (1998), From Old Industries to New Regions: Policies for Structural Transformation in Accession Countries (2003) and Slovenia and Austria: Bilateral Economic Effects of Slovenian EU Accession (2004). He is an expert of the European Commission for innovation and cohesion policy and adviser to governmental institutions at the regional and national level.

Notes on contributors xv G. M. Peter Swann ([email protected]) is Professor of Industrial Economics at the University of Nottingham Business School. Most of his research is about innovation, including work on competitiveness, demand, wealth creation, industrial policy and industrial clusters. He has held several advisory positions with government: as Specialist Adviser to the House of Lords Committee on Science and Technology, as Adviser to the Department of Trade and Industry for the Review of the National Measurement System, and as a Member of the Academic Panel for the Innovation Review in 2003. He was awarded an OBE in the Queen’s Birthday Honours, 2005. David A. Wolfe ([email protected]) is Professor of Political Science at the University of Toronto at Mississauga and Co-Director of the Program on Globalization and Regional Innovation Systems (PROGRIS) at the Centre for International Studies there. He is National Co-ordinator of the Innovation Systems Research Network and the Principal Investigator on its Major Collaborative Research Initiative grant on Innovation Systems and Economic Development: The Role of Local and Regional Clusters in Canada, a comparative study of twenty-six industrial clusters across Canada. He is editor or co-editor of seven books and numerous scholarly articles and public policy reports, including Innovation and Social Learning: Institutional Adaptation in an Era of Technological Change co-edited with Meric Gertler, and Global Networks and Local Linkages, co-edited with Matthew Lucas.

Preface

Over the past decade Michael Porter’s notion of ‘clusters’ (as developed for example in his book On Competition, 1998, and in numerous articles) has exerted considerable influence on the study of regional and local agglomerations of industrial specialization, innovation and enterprise. His own studies cover a multitude of clusters in several countries, and this work has stimulated a tidal wave of similar research by geographers, regional economists, regional scientists and planners. At the same time, policy-makers the world over, in the World Bank, the OECD, national governments, and regional and local development agencies, have seized upon Porter’s cluster model as a tool for promoting national, regional and local competitiveness, innovation and growth. Few other models of regional economic success have exerted such an impact on the policymaking arena. But the mere popularity of a construct is by no means a guarantee of its profundity. Seductive and politically popular though the cluster concept is, there is much about it that is problematic, and in some respects the rush to employ ‘cluster ideas’ has run ahead of many fundamental conceptual, theoretical and empirical questions. Despite the popularity of the concept amongst policymakers, there has been inadequate theoretical and empirical evaluation of the notion. Our motivation for this book, then, is the belief that the time is ripe for an incisive and theoretically informed assessment of the cluster notion, drawing out both its undoubted strengths and attractions, but also its weaknesses and limitations. By so doing we should be able better to judge the significance of the cluster concept for understanding regional development, regional innovation, and regional competitiveness. There are several issues here. How do we define and identify clusters? What is meant by ‘cluster theory’? What are the benefits of clusters? What are their disadvantages? How do clusters evolve? Do clusters improve regional innovativeness and competitiveness, and if so, in what ways? How do clusters relate to the global economy? What are the limits of the cluster concept? What policy options are available to promote clusters? These are just some of the questions we wish to explore, by means of papers written by some of the world’s leading regional analysts in this area. As a collection, the papers will provide a definitive account of the cluster notion, as a conceptual framework, an empirical construct, and as a policy tool.

xviii Preface To this end, the book brings together a number of leading international scholars whose work bears directly on the cluster debate. The remit was for thoughtful, provocative, and original pieces that address the sort of key questions listed above. The aim was not to provide a catalogue of case studies, although some authors do invoke empirical evidence and examples to illustrate their arguments. We are grateful to all of the contributors for their enthusiasm for this project. Without them, of course, it would not have materialized. Bjørn Asheim, Philip Cooke and Ron Martin

1

The rise of the cluster concept in regional analysis and policy A critical assessment Bjørn Asheim, Philip Cooke and Ron Martin

Introduction: the cluster craze Over the past two decades or so there has been a veritable flood of interest, within economic geography, economics, and business studies, in industrial localization: the observed tendency for many industries to form specialized concentrations in particular locations. Industrial localization of course is nothing new. It was a key characteristic feature of nineteenth-century industrialization in Europe, the United States, and elsewhere. But for much of the twentieth century, in the face of major shifts in industrial structure, the rise of mass production methods, and the ascendancy of the large, integrated firm, many of these former localized concentrations of specialized activities went into decline and a rather different, more geographically dispersed, pattern of production became the main basis of economic growth: what has commonly come to be known as the transition from Fordism to post-Fordism (Piore and Sabel, 1984). In addition, for some commentators, since the beginning of the 1980s, the accelerating movement towards a globalized, information-technology driven economy has further eroded the significance of location and spatial proximity for business performance and success (O’Brien, 1992; Cairncross, 1997; Gray, 1998; Reich, 2001). However, reality seems to point in the opposite direction: globalization and technological change appear to be promoting rather than reducing the importance of location in the organization of economic life. According to many observers the past two decades or so have witnessed the emergence of new localized production systems of specialized industrial agglomerations, as part of a more general ‘resurgence’ of regions and cities as the loci of contemporary economic development and governance (see, for example, Sabel, 1989; Storper, 1995, 1997; Krugman, 1997; Porter, 1998; Scott, 1998, 2001; Morgan, 2004). Thus as the highly influential business economist Michael Porter puts it (1998, p. 90): In a global economy – which boasts rapid transportation, high-speed communications, accessible markets – one would expect location to diminish in importance. But the opposite is true. The enduring competitive

2 Bjørn Asheim, Philip Cooke and Ron Martin advantages in a global economy are often heavily localized, arising from concentrations of highly specialized skills, knowledge, institutions, rivalry, related businesses, and sophisticated customers. At the same time, these observers argue, increasing global integration itself leads to heightened regional and local specialization, as falling transport costs and trade barriers allow firms to agglomerate with other similar firms in order to benefit from local external economies of scale (Krugman, 1991; Fujita, Krugman, Venables, 1999; Brackman, Garretsen and Marrewijk, 2001; Baldwin et al., 2003), which in their turn are thought to raise local endogenous innovation and productivity growth (see Martin and Sunley, 1998). For these and other related reasons, it has become fashionable within certain academic circles to talk of the ‘localization of the world economy’ (Krugman, 1997) and the rise of a ‘global mosaic of regional economies’ (Scott, 1998). Further, the definition of postFordist economies as ‘learning economies’, in which innovation is a socially and territorially embedded, interactive learning process (Lundvall and Johnson, 1994), also emphasizes the importance of localized industrial agglomerations, in this instance as providing the best context for the promotion of knowledge-intensive innovative firms (Asheim, 1999; Cooke, 2001; Maskell, 2001). These localized concentrations of specialized activity take many different forms, and numerous neologisms have been coined in attempt to capture their salient features. The industrial districts of the so-called Third Italy were one of the earliest prominent types to attract discussion (Brusco, 1989, 1990; Becattini, 1989, 1990; Asheim, 2000; Paniccia, 2002). Meanwhile in California, Allen Scott began to highlight the rise of new industrial spaces (Scott, 1988). Others have preferred to talk of local production systems (Crouch et al., 2001). Still others, focusing on localized agglomerations of high-technology activity, have variously used such terms as local high-tech milieux (Keeble and Wilkinson, 2000), local and regional innovation systems (Asheim and Gertler, 2005; Cooke, 1998, 2001), or even learning regions (Asheim, 1996, 2001; Florida, 1995; Morgan, 1997). However, probably the most influential neologism to have swept through the academic and policy discussions of economic localization is Michael Porter’s notion of industrial or business clusters. According to Porter (1998, p. 8), Clusters are a prominent feature of the landscape of every advanced economy, cluster formation is an essential ingredient of economic development. Clusters offer a new way to think about economies and economic development. Porter (1998, p. 197) defines clusters as Geographical concentrations of interconnected companies, specialized suppliers, service providers, firms in related industries, associated institutions (for example universities, standards agencies, and trade associations) in particular fields that compete but also co-operate.

The rise of the cluster concept 3 There are two core elements in his definition. First, the firms in a cluster are linked in some way. Clusters are composed of interconnected firms and associated institutions linked by commonalities and complementarities. The links are both vertical (buying and selling chains) and horizontal (complementary products and services, the use of similar inputs, technologies, labour, etc.). Moreover, most of these linkages, he argues, involve social relationships or networks that produce benefits for the firms involved. The second key feature is that of geographical proximity: clusters are spatially localized concentrations of interlinked firms. Colocation encourages the formation of, and enhances the value-creating benefits arising from, networks of direct and indirect interaction between firms. Over the past decade the cluster concept has become the standard term in the field. Moreover, Porter has promoted the idea of clusters not only as an analytical concept but also as a key policy tool. As the celebrated architect and promoter of the idea, Porter himself has been consulted by policy-makers the world over to help them identify their nation’s, region’s, or city’s key business clusters or to receive his advice on how to promote them. From the OECD, and the World Bank, to national governments (such as the UK, France, Germany, the Netherlands, Portugal, New Zealand), to regional development agencies (such as the new Regional Development Agencies in the UK), to local city governments (including various US states), policy-makers at all levels have become eager to promote local business clusters. Nor has this policy interest been confined to the advanced economies: cluster policies are being adopted enthusiastically also in an expanding array of developing countries (see Doeringer and Terkla, 1996; World Bank, 2000). Clusters, it seems, have become a worldwide craze, a sort of academic policy fashion item (see Martin and Sunley, 2003). The more so because the concept has become increasingly associated with the so-called ‘knowledge economy’, or what some have labelled the ‘New Economy’. Norton (2001), for example, argues that the global leadership of the US in the New Economy derives precisely from the growth there of a number of large, dynamic clusters or localized concentrations of innovative entrepreneurialism. Porter himself has headed a major policy-driven research programme to ‘develop a definitive framework to evaluate cluster development and innovative performance at the regional level’ in the US in order to identify the ‘best practices’ that can then be used ‘to foster clusters of innovation in regions across the country’ (Porter and Ackerman, 2001; Porter and van Opstal, 2001). In the UK, the promotion of clusters, especially of biotechnology, ICT, and so-called ‘creative industries’ (such as media, design, fashion, film and other cultural products sectors), has figured prominently in central government policy and the economic strategies of the regional development agencies. Likewise, the OECD (1999, 2001) sees innovative clusters as the drivers of national economic growth, as a key policy tool for boosting national competitiveness. Clusters, it seems, have become the new policy panacea. Our purpose in the remainder of this introductory chapter is to provide a synoptic reflective overview of the cluster concept, its relationship to other

4 Bjørn Asheim, Philip Cooke and Ron Martin related ideas in economic geography and economics, the strengths and weaknesses of the concept, and its grip on regional development policy. This then sets the scene for the more focused and detailed essays that follow. The aim of the book is to take stock, in a constructively critical way, of the scope and limits of the concept, as both an analytical and a prescriptive tool.

Conceptualizing clusters As mentioned above, Porter’s cluster concept is one of a number of different streams of literature concerned with the phenomenon of economic localization. It is useful, therefore, to situate his cluster idea within this wider intellectual context. In fact the localization of economic activity has become a central topic of interest in several different fields of economics and economic geography. At the time of its somewhat opportunistic discovery in the 1990s by the business school fraternity, and by Michael Porter at the Harvard Business School in particular, the study of ‘industrial district formation’ (Becattini, 2001) or ‘districtualization’ processes as they are perhaps less elegantly called (Varaldo and Ferrucci, 1996; Lazzeretti, 2003), was already well under way amongst several Italian economists, with regional scientists and economic geographers of a heterodox bent as unofficial observers (Cooke and Da Rosa Pires, 1985; Scott, 1988). The publication of a significant synthesis of the Neo-Marshallian Italian research by Piore and Sabel (1984), also from Cambridge, Massachusetts, not Harvard but MIT, with its prognosis of a ‘second industrial divide’ marking a shift from mass production into an era of neo-craft ‘flexible specialization’ characteristic of industrial districts, gave further legitimacy to research in this field. Of particular importance and value in bridging the features of Italian neocraft production in industrial districts and high technology production in modern industrial spaces was Saxenian’s (1994) anatomization and comparison of Silicon Valley’s success, associated with networking social capital, and Cambridge-Boston’s apparent demise due to the absence of such sociability. However, Saxenian did not deploy the language of ‘clusters’, that term being first used in a regional science context by Czamanski and Ablas (1979), based in turn on Czamanski’s (1974) study of industry clustering using linkage analysis. Indeed, the term evolved out of the study of industrial complexes, consisting mainly of large corporate entities and their linkage structures, inspired by Perroux (1955/1971), whose own growth pole concept derived from his application of the Schumpeterian idea of innovative ‘swarming’ on to a spatial canvas, initially the Ruhr Valley. Thus, from the Ruhr Valley to Silicon Valley, the cluster concept has been part of an ongoing effort to decipher the lineaments of an evolving economic landscape. Approached more systematically, at least five such perspectives, including cluster theory itself, can be distinguished, each having different theoretical foundations and employing different terminology. But all share one thing in common: an emphasis on the role of localization as a source of increasing returns or externalities for firms (Figure 1.1). Interestingly, this focus on the externalities

The rise of the cluster concept 5

Italian Neo-Marshallian Industrial Economics

External economies, inter-firm division of labour, and social capital

Local industrial districts of export-orientated, small-firm, flexible specialization

New Trade Theory and Marshallian Localization Economics

External economies and increasing returns as bases of trade

Geographical agglomeration and localized specialization of industry

New Endogenous Growth Theory

Educated labour and R&D as sources of increasing returns

Localized technological progress and divergent regional growth

Economics of Firm Strategy and Marshallian Localization Economics

External Economies and Competitive rivalry

Local clustering as driver of productivity and competitiveness

Neo-Schumpeterian and Evolutionary Economics

Institutions, innovation and learning

Local entrepreneurial milieux, learning regions and regional path dependence

Figure 1.1 Increasing returns and economic localization: five theoretical perspectives compared

of localization is arguably itself part of more general ‘return to increasing returns’ in economics (Buchanan and Yoon, 1994): geography, it seems, has provided a mechanism for (re)introducing increasing returns into economic theory. The argument is that many of the increasing returns in contemporary industrial economic development are regional or local in origin, in the form of local externalities associated with industrial localization, and that, to understand such issues as trade, competitiveness, innovation, and productivity, we need to examine these local externalities. A second common thread linking these different perspectives is that, in stressing the local nature of many external economies of increasing returns, these authors also – to varying degrees – emphasize the importance of local economic specialization. In fact, what underpins much of this new focus on the external economies of industrial localization is, in effect, a resurrection of Alfred Marshall’s notion of industrial districts that he talked about a hundred years ago (Martin, 2005). External economies of localized specialization were central to Marshall’s discussion of ‘industrial districts’ (Marshall, 1930, 1919). His characterization of these specialized industrial districts was cast in terms of a simple triad of external economies: the ready availability of skilled labour, the growth of supporting ancillary trades, and the development of a local inter-firm division of labour in different stages and branches of production, all underpinned and held together by what he referred to as the ‘local industrial atmosphere’, by which

6 Bjørn Asheim, Philip Cooke and Ron Martin he meant shared knowledge about ‘how to do things’, common business practices, tacit knowledge, and a supportive social and institutional environment. For Marshall, these industrial districts were the logical outcome of the process of economic evolution, whereby an economy progresses by inventing yet further subdivisions of function (‘differentiation’) and more intimate connections between them (‘integration’) (Martin, 2005). He saw industrial districts as an alternative mode of industrial organization to the large integrated firm with its internal economies of scale (Marshall, 1930, p. 266): [External economies of scale are] those dependent on the general development of the industry . . . which can often be secured by the concentration of many small businesses of a similar character in particular localities: or, as is commonly said, by the localization of industry. The links back to Marshall’s work are most evident in the Italian school of industrial economics that has focused on identifying and explaining the success of the specialized, export-orientated, small-firm-based industrial districts of the so-called Third Italy (that area of the country embracing the Veneto, EmiliaRomagna, Toscana, and the Marche). Led by such pioneers as Becattini, Brusco, and Bagnasco, this school has used Marshall’s original work as a basis for a revitalized conceptualization and theorization of industrial districts (see Asheim, 2000; Paniccia, 2002). Thus we find Becattini (1990, p. 40) referring to the Italian industrial districts in very neo-Marshallian terms: Each firm tends to specialise in just one phase, or a few phases, of the production process typical of the district. The firms of the district belong mainly to the same industrial branch . . . defined in . . . a broad sense as it includes upstream, downstream and ancillary industries. But where Becattini and his colleagues depart most notably from Marshall’s conception is in stressing that – at least in the Italian case – the industrial district is a socio-cultural as well as economic entity: ‘a socio-economic territory which is characterised by the active presence of both a community of people and a population of firms in one naturally and historically bounded area’ (ibid., p. 39), or what Piore and Sabel (1984) called the ‘fusion’ of economy and society. This socio-economic reconceptualization of Marshall’s industrial district has strengthened the non-economic, socio-territorial dimension of the concept and has provided valuable insights into the role of trust and co-operation as mechanisms of risk reduction and economic (relational) governance amongst local firms, and how a supportive form of local social capital aids the formation and success of industrial districts (Asheim, 2000). Another important strand of ‘geographical economics’, what its main architect Paul Krugman himself calls the ‘new economic geography’, also makes explicit reference to Marshall’s model of localization economies, and draws heavily on the role of geographical proximity as a source of increasing returns. But the treatment of industrial localization under this perspective differs markedly from

The rise of the cluster concept 7 that of the Italian literature, and that of Porter. In the work of Krugman and his associates, the approach has been to resurrect and build upon traditional location theory and regional science to develop highly formalized models of localized industrial specialization using recent advances in the mathematics of non-linear, multi-equilibrium models of imperfect competition. In this sense, in Fujita, Krugman and Venables (1999), Fujita and Thisse (2002), and Baldwin et al. (2003), the main statements of the ‘new economic geography’, the theoretical framework has moved far beyond Marshall’s simple model of economic localization. Indeed, what these authors seek to construct is nothing other than a universal micro-economic theory of spatial agglomeration that covers everything from intra-urban patterns of economic specialization and the spatial structure of cities right through industrial districts and clusters, to the North–South dualism in the world economy. As Fujita and Thisse (2002) boldly put it: ‘A few general principles seem to govern the formation of distinct agglomerations even though the content and intensity of the forces at work may vary with place and time’ (p. 3). There is little room in these mathematical models of economic localization for socio-institutional processes and factors of the sort stressed in Italian industrial district theory (for a critique of the ‘new economic geography’ models of specialized economic localization, see Martin and Sunley, 1996; Martin, 1999). The same might be said of those new endogenous growth theorists who have likewise discovered geography. These writers have used Romer-style and other ‘augmented’ and ‘endogenous’ production function models to suggest that the increasing returns to educated labour and R&D spending are highly localized, so that regional growth paths may diverge. But while this work has highlighted the importance of spatial proximity in the accumulation of human capital through knowledge spillovers, it fails to shed much light on why or how such endogenous growth processes become geographically concentrated in particular localities and not others (see Martin and Sunley, 1998). And, despite the focus on knowledge spillovers, these models have little to say on the local social, institutional, and inter-firm networks through which many such knowledge spillovers occur. Marshall himself placed great stress on the way in which industrial districts function as ‘knowledge communities’ (see Loasby, 1998). The geographical concentration of specialized and complementary firms in such districts increases the likelihood of exchanges of both formal and informal or tacit knowledge, so that local firms are connected to other firms’ knowledge in ways which allow both new ideas to diffuse easily within the locality and for the distinctiveness of each kind of knowledge to be maintained and enhanced (Marshall, 1930, p. 271): when an industry has thus chosen a locality for itself it is likely to stay there long . . . the mysteries of the trade become no mysteries; but they are as it were in the air . . . if one man [sic] starts a new idea, it is taken up by others and combined with suggestions of their own; thus it becomes the source of further new ideas.

8 Bjørn Asheim, Philip Cooke and Ron Martin This role of specialized economic localization in knowledge creation and innovation is the primary focus in a range of approaches that can be broadly defined as neo-Schumpeterian. A central theme in this literature is that innovation and entrepreneurship are quintessentially spatially embedded and localized processes. In the neo-Schumpeterian cluster literature, attention focuses on network theories of innovation and the emergence of ‘regional innovation systems’, localized ‘collective learning’, and local entrepreneurial milieux. For this reason, much of this literature is also heavily directed to successful ‘hightech’ districts and clusters (biotechnology being a particulary prominent case). How do these various perspectives on localized increasing returns compare to Porter’s approach? His concern is to contruct a general cluster theory applicable to a wide variety of specialized industrial agglomerations – from the industrial districts of the Third Italy, to high-tech regions, to geographical agglomerations of export specialization, to specialized inner-city enclaves – drawing on a wide array of conceptual notions, from Marshall’s ideas on localization economies, to Porter’s (1998, p. 207) own theory of business strategy, to ‘softer’ notions of institutional and social networks: A variety of bodies of literature have in some respects recognized and shed light on the phenomenon of clusters, including those on growth poles and backward and forward linkages, agglomeration economies, economic geography, urban regional economics, national innovation systems, regional science, industrial districts and social networks . . . Overall, most past theories address aspects of clusters or clusters of a particular type. Furthermore, in various footnotes Porter (1990) acknowledges a debt of gratitude, not to Czamanski but to Schumpeter and the Swedish economist Erik Dahmén for his notion of the ‘cluster’. From Schumpeter, Porter benefited (Porter 1990, p. 778) as follows: My fundamental perspective is more Schumpeterian . . . than neoclassical. Entrepreneurship and innovation prove central to national advantage. Why some firms and individuals innovate in particular industries, and why they are based in particular nations, will be the focus of much of what follows. Thereafter Porter makes links with the classic location theory asociated with Weber, Lösch, Hirschman, and Vernon, concluding that ‘The economic theory of location shows how firms will locate close to each other to gain access to the broadest array of customers. The rationale here is similar’ (ibid., p. 789). This leads to his acknowledgement of the importance for his idea of clusters of Dahmén’s work on the Swedish economy (ibid., p. 790): An early contribution is Dahmén’s 1950 . . . concept of development blocks. Dahmén stresses the necessary link between the ability of one sector to

The rise of the cluster concept 9 develop, and progress in another. In his examples, Dahmén often also talks of stages or vertical activities within a given industry. This interesting work is suggestive that connections among industries can be important to achieving advantage. Finally, after reviewing the relevance to ‘clusters’ of research on innovation and technological interdependencies, citing authors such as Lundvall, Rosenberg, Abernathy and Utterback, he concludes: ‘My theory integrates these sources and others into a broader framework . . . the local “diamond” bears centrally on the learning and diffusion process in a national industry’ (ibid., pp. 791–2). Hence at the stage of evolution in his thinking represented in The Competitive Advantage of Nations (1990), the die is more or less cast with all the flaws and confusions that remain in his later work and the diffusion, if not proselytization, of clusters and clustering to the policy world represented in Porter (1998). We can attempt to summarize Porter’s initial position as follows. First, although Porter’s cluster theory is more Schumpeterian than neoclassical, it retains some of the key features of the neoclassical perspective, namely that firms and individuals, not collectivities like networks or institutions, are the theoretical objects of interest in understanding economic growth dynamics. Schumpeter offers an important window upon innovation that, Porter might add, neoclassical economics traditionally assumed away. The point of analysing economic growth dynamics is to understand and secure national (competitive) advantage. Competitive advantage occurs more in some nations than others but the problem is not to universalize it but to retain and develop it. Schumpeter’s insights into innovation and entrepreneurship are important for this. Location theory also shows us that firms co-locate to maximize their market. Meanwhile, Dahmén stresses the way in which one sector can progress by selling into other sectors. This selling from one sector to another gives competitive advantage. Porter’s (1998, p. 208) distinctive aim has been one of Embedding clusters in a broader dynamic theory of competition that encompasses both cost differentiation and static efficiency and continuous improvement and innovation, and that recognizes a world of global factor product markets. What in effect he has done is to extend his previous work on competitiveness and competitive advantage by applying his well-established ‘competitive diamond’ (Porter, 1990), originally developed for firms, industries and nations, to locations and regions (Figure 1.2). In his earlier work, Porter argued that the (competitive) success of a nation’s export firms depends on a favourable national ‘competitive diamond’ of four sets of factors: firm strategy, structure and rivalry; factor input (supply) conditions; demand conditions; and related and supporting industries. The more developed and intense the interactions between these four sets of factors, the greater will be the productivity (‘competitiveness’) of the firms concerned

10 Bjørn Asheim, Philip Cooke and Ron Martin Geographical clustering of related industries/firms intensifies interactions within the ‘competitive diamond’

Competitive rivalry and knowledge spillovers within the cluster stimulate innovative activity

Firm rivalry and strategy

Clustering enhances innovation

Factor input conditions

Demand conditions

Related and supporting industries

Investment upgrading

Innovation enhances productivity High productivity raises competitive advantage of cluster, enables high wages and employment, which in turn attract skilled and educated labour

Figure 1.2 Porter’s cluster model

(Porter, 1990). He then argued, and this has since become his key theme and the basis of much of the cluster literature, that the intensity of interaction within the ‘competitive diamond’ is greatly enhanced if the firms concerned are also geographically localized or ‘clustered’. The externalities (increasing returns) associated with geographical clustering are enhanced by the development of a favourable ‘local context’, the local business, social, institutional and political environment that promotes appropriate forms of investment sustained upgrading of firms. In his view, the ‘geographic concentration of firms within the same industry is strikingly common around the world’ (1990, p. 120). More specifically he suggests that a nation’s most globally competitive industries are also likely to be geographically clustered within that nation. Hence what originally started out as a way of decomposing a national economy, the competitive diamond as a group of interlinked industries and associated activities, has become a spatial entity, the cluster as a geographically localized grouping of interlinked firms. The competitive diamond is the driving force making for cluster development, and simultaneously the cluster is the spatial manifestation of the competitive diamond. The systemic nature of the diamond produces a local concentration of the leading rival firms, which in its turn magnifies and intensifies the interactions between the diamond’s factors. Hence ‘the process of clustering, the intense

The rise of the cluster concept 11 interchange among industries in the cluster, also works best where the industries involved are geographically concentrated’ (Porter, 1990, p. 157). As we have noted, these ideas are not dissimilar to those Perroux presented in the early 1950s. Perroux argued that it was possible to talk about ‘growth poles’ (or ‘development poles’ at a later stage of his writing) in ‘abstract economic spaces’, where firms were linked together by an innovative ‘key industry’. According to Perroux, the growth potential and innovativeness of growth poles can be intensified by territorial agglomerations forming ‘concrete geographical spaces’ (Haraldsen, 1994; Perroux, 1955/1971). In Porterian parlance, this is achieved by the local ‘diamond’ (or cluster), with locally rival firms, demanding customers, but also certain collective assets, that propels the national industry forward. This seems a coherent viewpoint, but from only one highly specific position, which Porter discusses at some length (1990, pp. 210–25). This is also the only one of three industry accounts in his 1990 book that is substantially place-based. It is the world-renowned Italian ceramic tile industry, which in 1987 had 85 per cent of its production and 75 per cent of its employment in just one small town, Sassuolo, in Emilia-Romagna. Business school research is frequently criticized for its reliance upon case studies, and the Harvard Business School has one of the most extensive business case study libraries available. The obvious danger with case studies is that unless they are carefully selected, analytically and representatively, they run the risk of the journalistic ‘fallacy of composition’ whereby one sensational case of something is taken as representing some general condition in the wider population. It is otherwise known as the problem of ‘the sample of one’. Superficially, Sassuolo fits the local ‘competitive diamond’ notion almost too well, particularly in the way it accounts for so much of Italian national ceramic tile production and exports, has final firms in competition, with sub-contractors selling contractually to them, backward integration of the value chain from ceramics to the innovative design and global marketing of firing equipment needed to produce ceramics. But as a reading of Russo’s (1985) classic article on Sassuolo shows, locally and nationally contingent events and processes, like Italy’s ‘economic miracle’, its stupendous housing boom, and specific form of housing legislation, allowing subsidies favouring the use of ceramics, together with industrial relations conflicts, created a space from which hitherto ceramicsfree Sassuolo could benefit. Why Sassuolo? Because the appropriate clays existed only there and in the nearby Apennine mountains. As well as skirting with the fallacy of composition, Porter’s work betrays a major weakness in understanding the issue of geographical scale. The local ‘diamond’ is thus unproblematically capable of influencing national industry. Indeed throughout most of the cluster passages of Porter (1990) the cluster definition stretches alarmingly from the local to the national and back again with bewildering facility (Martin and Sunley, 2003). Of course, building on the previous point, where a local cluster has the inordinate impact that Sassuolo ceramics exerts upon the Italian national and export ceramics sector, that may just be acceptable logic. But most industry is nothing like that. So, initially at

12 Bjørn Asheim, Philip Cooke and Ron Martin least, Porter’s cluster notion caused enormous confusion until finally, in Porter (1998), he more or less put things right with his spatially more informed, and now widely cited definition of a cluster as ‘a geographically proximate group of interconnected companies and associated institutions in a particular field, linked by commonalities and complementarities’ (Porter, 1998, p. 199). This conforms to dictionary definitions of ‘cluster’ as ‘concentration’, ‘clump’, or ‘bunch’, all denoting gathering or huddling in geographic space. However, even though much of the conceptual confusion concerning the spatial delimination of clusters can be attributed to this fact that clusters can be seen as both industrial and spatial phenomena, i.e. either confined to industrial sectors defined from a functional (national) perspective (Porter in his 1990 definition), or defined by geographical (regional) boundaries (Porter in his 1998 definition), one should not regard these seemingly contradictory definitions of clusters as a problem. Instead, following Perroux one should use both conceptualizations of clusters (Malmberg, 2003). It is a quite normal situation to find regional clusters of specialized branches (i.e. concrete geographical space) being part of a national cluster of the same sector (i.e. abstract economic space). Yet, having snatched some coherence from the confusion that Porter (1990) himself had created, he undermines it again in the very next line, revealing a basic reluctance to dispense with the comfort blanket of universality: ‘The geographic scope of a cluster can range from a single city or state to a country or even a network of neighbouring countries’ (ibid., p. 199). Naturally, this only makes the conceptual confusion worse, something the Neo-Marshallian industrial district economists never once countenance in their carefully defined and meticulous studies. In Porter’s definition, the concept of the cluster is here elasticated into meaninglessness. For why bother troubling with a notion of geographic proximity when in the next line he stretches the definition to the limits of credulity? Even were one to allow that ‘a network of neighbouring countries’ expresses a possible geopolitical or georegional entity of sorts, the terminology merely betrays the crudest kind of geographical determinism. For why should the propinquity of shared borders denote either networking or, indeed, clustering? But this is not the only major problem with Porter’s cluster concept, for, whether or not there is a reality behind what Porter is trying to get at, we nevertheless argue, along with other contributors to this book, that clusters in the sense of industrial districts do, in fact, exist. Witness, for example, the detailed empirical evidence documented in Becattini’s (2001) or Dei Ottati’s (2004) studies of the woollen textile district of Prato near Florence. Dei Ottati (1994) examined in depth the distinctive kinds of relationships among individual firms, banks, and intermediaries; while in Becattini (Becattini, 2001, pp. 161–2) a detailed account is given of how and why the Chinese ‘districts within a district’ came about in Prato between 1989 and 1993: By December 1993 the Chinese community officially numbered over 1,400 residents, a negligible figure in itself for somewhere the size of Prato, but

The rise of the cluster concept 13 one that stands out noticeably if it is borne in mind that . . . in 1989 there were only 40 . . . the Prato Chinese are part of a galaxy of similar communities that have settled in the surrounding area (Campo Bisenzio, Signa, Lastra a Signa, etc.) that can be reckoned to number some tens of thousands . . . this community has rapidly and efficiently inserted itself into Prato’s manufacture of clothing goods, where it has carved out a number of niches (handbags, gloves, etc.) in which with several hundred small firms, it has a far from negligible importance. Porter’s methodology excludes this kind of empirical and analytical knowledge because he takes an operational definition of what counts as clustering, using secondary data that show high employment location quotients for specific sectors and sub-sectors. Occasionally, as in Porter (1998), diagrams of cluster ‘circuitry’ are drawn, but they are not based on firm interviews that would enable his researchers to acquire knowledge of precisely which inter-firm transactions are, say, preferential contracts with neighbours, which favour exchange relations with local significant others, whether a firm was a spin-out from another in the locality, what kinds of relationships exist, if any, with local business support agencies, lawyers, banks, venture capitalists, and the like. In other words, Porter rarely proves that anywhere or anything he asserts to be a cluster, actually is one. All that can be said from location quotient data and other aggregate measues of geographical localization and inferred inter-industry linkages is that there is a higher than average specific industrial agglomeration in location x or y, which actually does not get us very far beyond Weber or Lösch. Three final qualifications can be made. First, because of the static definition of cluster deployed by Porter (1998), none of the dynamism of spin-out, networking, and project-focused activity that the Neo-Marshallians routinely find in industrial districts is discernible from Porter’s methodology. Second, static cluster accounts give a foreshortened sense of the time it takes for clusters to form, or of their long-run evolutionary paths. Most of the Italian ones date to at least the early postwar years or the 1960s ‘economic miracle’ era, and many argue that some of them can be traced back to the Renaissance. Hence presenting clusters as a policy panacea is doubtful mis-selling (Martin and Sunley, 2003), a problem that policy organizations are beginning to discover, as we shall see. Third, Porter’s definition is, as we have noted, a market-driven one with reliance on rivalry and competitive advantage. However, arguably some of the most interesting new cluster forms, such as those that typify biotechnology, are science-driven, not primarily market-driven (for example, see Cooke, 2004a). Indeed, most firms in biotechnology clusters do not even make profits, but have discovered a new business model in which pharmaceutical firms and research councils co-invest in ideas. The affinity of Porter’s cluster concept with Marshall’s original industrial districts is, thus, in some ways obvious. Marshall’s triad of localization economies have become the four corners of Porter’s diamond. Just as Marshall emphasized the importance of a local ‘industrial atmosphere’ in the formation and

14 Bjørn Asheim, Philip Cooke and Ron Martin development of his districts, so Porter highlights the importance of ‘local context’ in the formation success of his clusters. While Marshall saw knowledge spillovers within his districts as a key externality or source of increasing returns for individual firms, so in his recent work Porter has argued that clusters enhance and promote innovation though competition and knowledge transfers. Where the two concepts differ, however, is that Porter situates his cluster concept within a more general framework concerned with the competitive strategy of firms, whereas Marshall’s industrial districts were a geographical manifestation of his view of economic evolution. Furthermore, Marshall’s perspective was on how the productivity and competitiveness of small firms was enhanced through economies of localization achieved by an extensive division of labour and strong product specialization between firms in industrial districts to match the internal economies of scale of large firms. This one-dimensional focus on efficiency and productivity has been criticized by the GREMI group (Groupement de Recherche Européenne sur les Milieux Innovateurs) for representing a static perspective which ‘considers the local relationships mainly in terms of locational efficiency’ (Camagni, 1991, p. 2), in contrast to looking at ‘the complex network of mainly informal social relationships in a limited geographical area, . . . which enhance local innovative capability through synergetic and collective learning processes’ (ibid. p. 3). Even if the specific Marshallian interpretation of agglomeration economies, defined as socially and territorially integrated (or embedded) properties of an area (such as mutual knowledge, trust, and industrial atmosphere) can be said to stimulate (incremental) innovations and innovation diffusion in districts, the major impact of such localization economies was viewed as securing the skills and social qualifications of the workforce, and only secondly the production of innovations (Asheim, 2000). Here Porter’s perspective clearly differs from Marshall’s with respect to the emphasis and focus on the way clustering promotes innovation as the main factor behind increased competitiveness. In their critical assessment of Porter’s work, Martin and Sunley (2003) raise several issues concerning the conceptual and theoretical underpinnings of the cluster notion. Porter sees his cluster concept as providing a theory of the ‘competitive advantage of locations’. He goes even further, arguing that ‘cluster theory bridges network theory and competition’ (1998, p. 226), and that ‘clusters offer a new way of exploring the mechanisms by which networks, social capital and civic engagment affect competition’ (ibid., p. 227). What is being proposed, here, it seems, is nothing less than a general theory of clusters and their social, institutional, and economic operation. As Martin and Sunley point out, three questions immediately arise. First, just how far can the full complexity of economic, social, and institutional factors and processes alleged to underpin cluster formation, development, and success be subsumed within a single overarching concept of ‘competition’? Second, to what extent is it possible in any case to construct a universal theory of cluster formation, dynamics and evolution capable of covering the wide range of cluster types and processes thought or argued to exist? Third, just how far does Porter’s theory really illuminate the

The rise of the cluster concept 15 localized externalities, knowledge spillovers and socio-institutional processes that are claimed to be so important in clusters? In fact, embedding cluster theory in a wider theory of ‘competitive advantage’, as Porter has sought to do, raises all sorts of problems concerning the meaning and interpretation of ‘competitive advantage’. For over the past few years, an intense debate has surounded the ideas of ‘competitiveness’ and ‘competitive advantage’. Indeed, Porter’s approach to competitiveness has itself been heavily criticized (for example, Jacobs and De Jong, 1992; O’Malley and Van Egeraat, 2000, Klein, 2001), and the very notion of ‘competitiveness’ has been challenged (for example, Krugman, 1994; 1996; Turner, 2001). In some quarters, the notion of ‘competitiveness’ is viewed with extreme caution, on the grounds that, while it may have meaning at the level of the individual firm, it is much more difficult to apply to localities, cities, regions, or nations (Kitson, Martin, and Tyler, 2004). Despite its attempt to link firm competitiveness with Marshallian-type localization economies on the one hand, and with local social networks and institutional arrangements on the other, Porter’s cluster theory does not really tell us much about either, nor about how spatial proximity actually enhances firm competitive performance, or about how local knowledge spillovers occur. It is a curiously eclectic and synthetic approach, certainly highly suggestive and seductively pedagogic, but ultimately lacking in theoretical rigour. This is to argue neither that Porter is totally wrong about his interpretation of competitive advantage and his extension of the notion to regions and cities, nor that other approaches to the conceptualization and interpretation of specialized economic localization – industrial district theory, new economic geography models, endogenous growth models, or evolutionary economics – are superior. Each has its particular limitations. They all share an emphasis on local externalities, or localization economies, but their interpretation of such externalities and the role they are assigned vary from one approach to another. Rather, the basic point is that it may not be possible to construct a single ‘best’ theory that fits all cases. Clusters vary considerably in type, size, origin, structure, organization, dynamics, and developmental trajectory. It seems most unlikely that different clusters can all be explained in the same way. We may well need different types of theory and explanation for different clusters (Markusen, 1996; Gordon and McCann, 2000; Martin and Sunley, 2003; Asheim and Gertler, 2005; Asheim and Coenen, 2006). Nevertheless, it is precisely its apparent universalism that is the attraction of Porter’s cluster concept to both academics and policy-makers. Porter’s cluster metaphor is highly generic in character, being deliberately vague and sufficiently indeterminate as to admit a very wide spectrum of industrial groupings and specializations (from footwear clusters to ceramics clusters to wine clusters to medical clusters to biotechnology clusters), demand–supply linkages, factor conditions, institutional set-ups, spatial scales, and so on, while at the same time claiming to be based on what are argued to be fundamental processes of business strategy, industrial organization and economic interaction. Rather than being a

16 Bjørn Asheim, Philip Cooke and Ron Martin model or theory to be rigorously tested and evaluated, the cluster idea has instead become accepted largely on faith as a valid and meaningful ‘way of thinking’ about national, regional, and local economies, as a template or procedure with which to decompose the economy into distinct industrial-geographic groupings for the purposes of understanding and promoting competitiveness and innovation. The very definitional incompleteness of the cluster concept has been an important reason for its popularity (Perry, 1999): clusters have ‘the discreet charm of obscure objects of desire’ (Steiner, 1998, p. 1). However, the concept has acquired such a variety of uses, connotations, and meanings that it has, in many respects, become a ‘chaotic concept’, in the sense of conflating and equating quite different types, processes, and spatial scales of economic localization under a single, all-embracing universalistic notion (Martin and Sunley, 2003). Such, of course, is the danger with a synthetic concept.

Problems and prospects for clusters in theory and practice We have seen that, since Marshall, observers have spoken of certain agglomerations of specialized economic activity, first as ‘industrial districts’, then as ‘new industrial spaces’, and more recently as ‘clusters’. When they do this they are mainly speaking of a concrete phenomenon rather than an abstraction. Usually they mean many firms in more or less one place, the firms competing but also having at least preferential relations with some others. These relations may range from ‘preferred supplier’ contractual status at one end of the spectrum of economic interaction, to high-trust, non-contractual collaboration, exchanging of favours and non-monetary transactions at the other. So a question that is of undoubted and abiding interest to economic geographers and others is why, in a world largely characterized by utilitarian, unsentimental buying and selling, do clusters exist? They ought to have disappeared with the medieval guilds, the source that Marshall, in effect, gave as the ultimate origin of industrial districts. They ought, certainly, to have been replaced by the modern business corporation with its advantages in resource and administrative capabilities, features that have exercised generations of industrial economists from Young (1928) to Coase (1937), Penrose (1959), and Richardson (1972), and modern authors such as Teece and Pisano (1994), Winter (2000), and Zollo and Winter (2002). Interestingly, Marshall himself, writing at the beginning of the twentieth century, seemed to anticipate the disappearance of the industrial district model of economic organization in the face of the emergence of the large vertically integrated firm, as internal economies of scale were assuming importance over external economies of scale (1919, p. 167): The importance of internal economies has increased steadily and fast; while some of the old external economies have declined in importance; and many of those which have arisen in their place are national, or even cosmopolitan, rather than local.

The rise of the cluster concept 17 But of course, in reality the smaller firms that often comprise industrial districts are nowadays far from being erased from the economic landscape. Indeed, we are in an era when ‘entrepreneurship’ is used as a policy injunction for practically everyone from the unemployed to star scientists in universities. Moreover, such is the perceived success of many small and medium-sized firms interacting profitably in confined geographical locations that many policymakers seek to replicate them. The likes of Como, Italy, where the world’s leading silk design capabilities are found, or Carpi in the same country, from where similarly high-quality knitwear originates, and Salling in Denmark with its numerous woodworking furniture factories at the centre of the high-quality Danish furniture industry, have all become policy role-models. Furthermore, while concentrations of such firms are competitive in design-intensive, quality markets that large corporations find it difficult or impossible to break into, such expertise is not confined to high-end production only. Thus Parma produces much of Italy’s mass-market cheese, as well as supermarket prosciutto, Montebelluna’s 500 firms produce 80 per cent of the world’s motorcycle boots, 75 per cent of its ski boots and 50 per cent of its track shoes; while Belluno produces two-thirds of the world’s spectacles, its turnover accounting for 85 per cent of Italian production with more than 70 per cent exported. Thus many of these significant market shares are not confined to luxury items but also involve niche production. Hence, for such proximate production formations of specialized smaller firms to be thriving, there must be problems with the conventional economic perspective that once predicted their demise. Therein lies an additional answer to the question why industrial districts and clusters exist: the inevitability of economic dominance by large-scale production and ‘trading with strangers’ were mistaken assumptions. It is nowadays becoming ever clearer that the large or giant firm is by no means universally the optimal organizational form for industrial production. In particular, it may well have reached its limits regarding innovation potential arising from in-house control of R&D, an historic core competence, as Chesbrough’s (2003) analysis of innovation outsourcing has convincingly shown. Sharing the missing spatial insight of most industrial economists, he nevertheless shows how private R&D investment in the US switched by 30 percentage points from large to small firms between 1981 and 1999. Overwhelmingly, such firms, especially in biotechnology and ICT, two of the most research-intensive fields, operate from cluster bases such as Silicon Valley and Greater Boston. Moreover, the theory of economic relationships that stressed utilitarian anonymity in production and consumption has also proved to be correct only in special circumstances such as ‘spot markets’. Indeed more general market relations are now understood to be based on varying, but often significant, magnitudes of ‘social capital’. Social capital is the concept of reciprocal, trust-based exchange of resources based on reputation, traditional or communally established norms ( Jacobs, 1961; Coleman, 1988). Itself a far from unproblematic concept (see Fine, 2001), the notion of social capital involves at one extreme, for example, exchange of favours

18 Bjørn Asheim, Philip Cooke and Ron Martin and, at the other, preferential economic practices among firms which are pecuniary rather than simply forms of ‘untraded interdependency’ (Dosi, 1988). This last is an important point as many interpretations of social capital stress the collectively minded, barter-like interactions neighbours, families, or communities might engage in. It is then a short, but mistaken, step to extend this notion to traditional industrial districts or modern clusters and to arrive at an idyllic non-monetary form of communal economic activity. This inclination towards polarization between ‘arm’s length exchange’ and collective interaction arises from the misleading way neoclassical economics rests on a massively undersocialized concept of humankind, a problem that exercised Marshall (1919) and especially later writers such as Granovetter (1973). Contemporary economic geography, written from a mostly non-neoclassical perspective, also rejects such an under-socialized view. Rather, it argues that ‘economic agents’ are profoundly social in their economic thought and action. Some research finds that ‘social capital’ is a pronounced feature of market activity in which even the nearest example to a caricature of a rational utility maximizer, the market ‘spot trader’, may nevertheless use ‘swift trust’ (Sako, 1992) to gain an edge (Cooke and Clifton, 2004). Interestingly, the most innovative firms are the least ‘neoclassical’, having considerably greater reliance on social capital networks, both local and global, than the average firm. This is not to sanction an uncritical use of the notion of social capital, which can all too easily become a vague term to account for the unexplained (see Fine, 2001). Nor is this to overstate the importance of the ‘local embeddedness’ of small firms. But there is sufficient evidence to attest to the role of local non-economic – that is social and cultural – factors and relationships in the production and reproduction of clusters. Other social norms motivating economic action by firms include those identified by Schumpeter (1975) as associated with innovation and entrepreneurship. Assume a discovery by the heroic innovator, whether the lone pioneer entrepreneur as in Schumpeter’s early work, or later the equally heroic but more team-based innovators in corporate R&D laboratories. If radical and commercial, the mere existence of an innovation attracts swarms of imitators seeking to appropriate an early surplus from ‘second-comer advantage’ (building on mistakes and learning from the innovator’s ‘first-mover’ advantage). Although Schumpeter failed to see the geographical implications of the quest for proximity to the innovators by the imitators, they are clear. Swarming is self-evidently one kind of spatial concentration activity that can give rise to clusters as entrepreneurs seek to ‘free-ride’ on waves of knowledge spillovers. Another, more common, variant of such a ‘spin-in’ model is ‘spin-out’, particularly from a founding firm. Among hundreds of cluster stories of this kind, the aforementioned Belluno spectacles district, for example, is illustrative. This local industry originated with a contract dated 1878 where a local man, Angelo Frescura, in collaboration with Giovanni Lozza and Leone Frescura, opened a craft workshop for the production of spectacles near Calalzo, in Cadore. Cadore remains the historical heartland of the Belluno spectacles district and is still the area with the highest concentration of companies.

The rise of the cluster concept 19 Hence, there is merit in trying to tie together the various reasons for (or interpretations of) spatial concentration of the sort found in clusters and districts. Markusen’s (1996) typology of industrial districts was one of the first to attempt to classify districts according to their structural, organizational and causative characteristics. In a related vein, Gordon and McCann (2000) suggest a threefold typology of reasons for clustering: •

•

•

The proximity (agglomeration) of firms induces a nexus of ‘Marshallian’ external economies from enhanced local skills supplies, cheap local infrastructure, specialized producer support services, and localized knowledge spillovers (see Caniëls and Romijn, 2005). Firms may be part of a regionalized or localized ‘industrial complex’ outsourcing system designed to generate ‘Toyotian’-style logistical and transaction costs reductions that enhance productivity and quality through preferred supplier interactions. Firms in proximity may seek to reap ‘associational’ economic benefits from systemic local and regional innovation and learning networks involving research institutes, industry associations, and governance measures (see also Cooke and Morgan, 1998).

It is important to recognize that there is a scale effect with each of these processes, contingent on both spatial and institutional factors. Thus ‘Marshallian’ districts tend to be small, even occupying highly localized ‘quarters’ of cities such as Birmingham’s and Arezzo’s ‘jewellery quarters’ (De Propris and Lazzereti, 2003; Lazzeretti, 2005), Florence’s ‘art restoration quarter’ (Lazzeretti, 2003), or Leipzig’s traditional ‘graphische Viertel’ or ‘graphics quarter’ (Bathelt, 2002). They are also highly specialized. Marshall himself noted the high specialization of specific small textile towns in northern England as between spinning and weaving, and coarse and fine weaving, as a characteristic feature of industrial districts (Marshall, 1919). A Toyotian cluster is more urban or subregional in scale and, while specialized, typically covers a wide range of supply sectors. An ‘associational’ system is likely to be regional in scale, and to contain more than a single cluster. For example, Baden-Württemberg contains at least two differently scaled and distinctive automotive clusters in Stuttgart (Porsche and Mercedes), a printing machinery cluster (in Heidelberg), a surgical instruments cluster (Pforzheim), and a machine tools cluster in the Black Forest. Institutionally speaking, co-ordination tends, according to Visser and Boschma (2004) to be cost-driven with ‘free-rider’ search characteristics prominent in ‘Marshallian’ systems, corporate hierarchical in ‘Toyotian’ systems, and interactive with quality, technical knowledge, and price equally prominent in ‘associational’ systems. One might, at the risk of over-simplification, characterize these governance regimes as typical, respectively, of liberal-market, developmental, and corporatist industrial systems more generally (Casper, Lehrer and Soskice, 1999). Similarly, Bottazzi, Dosi and Fagiolo (2002) suggest at least a fivefold categorization of localized industrial agglomeration or clusters:

20 Bjørn Asheim, Philip Cooke and Ron Martin •

•

• • •

Horizontally Diversified – ‘Made in Italy’ districts with small firms producing high-quality, design-intensive fashion products in traditional sectors such as clothing, ceramics, jewellery Vertically Disintegrated – A variant on the ‘Made in Italy’ type with a ‘Smithian’ division of labour in localized supply chains, specialized with local input–output linkages and user–producer knowledge exchange. Shoes, textiles, and some clothing are produced in such districts Local Hierarchical – an ‘oligopolistic core’ connected to subcontracting networks, as in transport equipment and white goods, for example Knowledge Complementarities – science-engineering-driven, as in Silicon Valley biotechnology clusters (absent in Italy) Path-dependent – spatially inert agglomerations without any particular advantage from agglomeration in itself, e.g. Detroit as described by Klepper (2000)

They point out that different types of agglomeration suggest different drivers for agglomeration (clustering) itself, taking into account also different sectoral specificities. Using Italian data, and employing econometric analysis, they find a statistically significant advantage, measured in export performance, for the industrial district categories of agglomeration over the rest. The authors ascribe this advantage to two phenomena theoretically and empirically pronounced in such districts: sectoral patterns of knowledge accumulation; and localized knowledge spillovers of various kinds, from innovation to labour. However, while such typologies are certainy useful, inevitably they also have limitations that need to be borne in mind. As Martin and Sunley (2003) point out in relation to Gordon and McCann’s scheme, such categorizations are always in one sense ideal or pure types, whereas reality is much more complex, in that a given cluster or district may be characterized by more than one set of clustering processes. For example, Marshallian and associational processes or features may well co-exist. And clusters may display both vertically integrated and horizontally integrated industrial organizational forms. This is not to undermine the need and search for such typologies, though it is to emphasize that these work best when they are causally based and when they admit of multiple causes and changes in cluster proceses and structures over time.

The implications for policy The marketing of the cluster idea to policy agencies has been remarkably successful. However, while the take-up of the concept was accompanied with unbridled enthusiasm in the 1990s, in the current decade signs of dissatisfaction are increasingly appearing. One major source of disappointment derives from different aspects of Porter’s cluster methodology, as implemented and promoted through his consultancy operation Monitor. Although Porter recognizes that not all clusters are alike, the way he differentiates them is discursive and superficial. Thus they are asserted to ‘vary in size, breadth, and state of development’ (1998,

The rise of the cluster concept 21 p. 204). ‘Some clusters consist of small- medium-sized firms . . . Other clusters involve both large and small firms’ (ibid., p. 204). Once again the reader is left wondering at the lack of critical reflection involved in such observations, and even more so the absence of scepticism from the eager clients of such a simplistic prospectus. As the first subheading of the ‘Clusters and Competition’ chapter in Porter (1998) displays with its question ‘What Is a Cluster?’, the answer, despite the obvious glosses regarding difference, is a ‘one-size fits all’ definition and policy formula. Several key problems have arisen from the implementation of Monitor-style cluster policies. Three recurring issues of cluster evaluation involve understanding the additional impact generated by a cluster approach at the macro level (impact on the regional or local economy), at the meso level (on the cluster or sector in question) and on the micro level (on the individual business). Unless the policy applies to a case like Sassuolo, which accounts for 85 per cent of Italian ceramics production, it is no simple matter to measure the impact of a cluster policy. Even so, given that clustering involves more than simple agglomeration, it is probably only at the very narrowest margin that the impact of typical cluster policy support for better networking, joint marketing, or common purchasing initiatives can be measured, and then only at the level of firm performance aggregated up to cluster performance. At the meso level some work has been conducted, most notably as reported by Brusco et al. (1996) on all industrial districts in Emilia-Romagna benchmarked against Italian sectoral data. According to this study, clusters performed better than their sector in terms of employee income, exports, and average firm size. However, the panel methodology utilized to derive these conclusions is open to question as many panel members were in fact closely associated with the clusters in question. Of course, separating out cluster policy effects from macro-economic effects such as a currency devaluation, or interest rate changes, that occurred during the timescale of the study is also no easy matter. At the firm level, what benefits do cluster policies bring to the firm that is located within the cluster in question? This is unclear, but it may be inferred from certain policy shifts observable among agencies that were early adopters of Monitor’s methodology, where the rate of growth of existing firms, even the rate of new firm creation, has been disappointing. Two cases of such shifts, involving the same cluster sector – biotechnology – may be cited in support of the thesis that cluster policy may have negligible effects on firm performance. These involve two early policy cases, Scotland and the Basque country. The latter hired Monitor to conduct the first scoping study for a Basque cluster strategy as early as 1990 (Cooke and Morgan, 1998). Scottish Enterprise hired Porter to map out its cluster strategy in 1993, although a pilot scheme involving promotion of biotechnology, semiconductors, food and drink, and energy clusters was not implemented until 1997. By 2003 both the Basque Ministry of Industry and Scottish Enterprise had launched new, non-cluster-based knowledge commercialization initiatives in these fields. The newly established BioBask initiative swiftly developed a basic research-led biotechnology strategy concentrating on

22 Bjørn Asheim, Philip Cooke and Ron Martin research niches in proteomics. In six months a planned multi-storey car park was converted into top-grade wet-labs, advertisements were placed in Nature Biotechnology to recruit proteomics biotechnologists worldwide, and from two hundred applicants forty were selected. Three linked spin-out firms occupy incubator space in the BioBask facility. In 2003 Scottish Enterprise announced three Intermediary Technology Institutes, funded at £450 million over ten years, one specializing in life sciences, the others in ICT and energy, to increase the rate of commercialization of basic bioscientific and medical research. New firm formation is among the potential vehicles for achieving this (see Cooke, 2004a, 2004b). In both cases disappointment at the long time-horizons of slow firm growth in these and the other cluster sectors prompted drastic policy reviews and changes of direction. A review of cluster policies in Scandinavia showed that a further four problematic issues have also arisen: • • • •

A lack of critical mass. How is it possible to sustain growth in a potential cluster? The cluster’s focus is primarily local. How can it be integrated into national and international networks? How can a new organization be structured to maintain the benefits of vertical co-ordination and co-operation, yet avoid becoming bureaucratic? How can the many initiatives be effectively co-ordinated, and diverse interests appropriately balanced?

The review claims that many cluster practitioners seem to be facing these same issues. One reason why the cluster development processes often appear more straightforward than they actually are is the abundance of so-called ‘best practice’ cases available. The descriptions are typically written with a focus on vision and strategy, but rarely on the basis of actual results, and they typically disregard the obstacles that are encountered. One cluster presented as a success story that was cross-checked by the reviewers had actually been terminated owing to organisational problems (Jensen, 2004). In addition, highly successful clusters can produce a different but equally problematic suite of policy problems. Rapid growth in a cluster can put severe strains on other aspects of the local economy, leading for example to high land, wage, and housing costs, and to pressure on local transport and public infrastructures (Martin and Sunley, 2003). Tensions may then arise between the needs of local firms and the wider set of imperatives facing local planners. And as some studies have shown, successful clusters can produce their own problems of social inequality between different groups of workers: not all sections of the local labour force necesarily benefit from a cluster’s growth, and even working conditions for many of those employed in the lead cluster firms may not be that favourable (Benner, 2002).

The rise of the cluster concept 23

Conclusion Clearly, clusters have been oversold. The nature of the marketing game and the appetite shown for new wonder growth strategies by economic policy-makers mean that there has long been the need for an objective assessment, since cluster marketing is a paradigm case of asymmetric knowledge. In his classic article, Nobel laureate George Akerlof (1970) explored market imperfections and market failure from the perspective of asymmetric information. Asymmetric knowledge is similar except that knowledge implies meaning for action whereas information has more passive, absorption connotations. Still Akerlof ’s key idea, which was to explore the nature of the market for ‘lemons’, is a useful metaphor in the cluster policy context. In American used-car business parlance, a ‘lemon’ is a second-hand car that breaks down. The parallels with cluster policies are almost exact. In the local economic development policy market, a new model from Italy arrives on the radar (the industrial district). The new model even seems to outpace older local economic development policies in nimbleness, flexibility, and performance. The model is examined, tested, and found to be cheap to implement, primarily requiring higher-quality, more interactive ‘drivers’. Adaptations are made to suit the model to the American market and a sales drive is successfully launched, not just in the US policy arena but internationally. However, neither the salesman (Porter) nor the customer (national governments and local, regional, city authorities almost everywhere) knows whether the model (cluster policy) will work outside its native conditions, but this crucial consideration is ignored. Needless to say, many policy-makers will have bought policy ‘lemons’, as we have seen. One important aspect of the cluster policy model that is infrequently focused upon is the labour dimension. An interesting example of just how problematic this aspect of clustering can be is the Hollywood film cluster (see Scott’s chapter in this volume). This is a specific type of cluster, and unlike the Italian industrial districts in displaying an oligopolistic industry structure where global firms such as Disney, Sony, AOL Time-Warner co-exist and contract, in a Smithian division of labour, to a multitude of small and medium-sized enterprises. On the back of this dominant-dependent industry structure, Hollywood spans the globe as the greatest purveyor of what Adorno (1991) dubbed ‘the culture industry’ the world has ever known. However, it floats on a raft of regulations that, as Scott shows, began with the Paramount Decree in 1948 that started the break-up of the old vertically integrated studio system and helped introduce more vulnerable, limited contract employment. Many fragmented worker organizations exist to seek to moderate the highly insecure and segmented working conditions that have developed in and around the Hollywood film cluster. Another example is provided by that classic exemplar of a cluster, Silicon Valley. Here, as Benner (2002) has shown, while some workers thrive and prosper in this paragon of specialized, high-technology-based local economic growth, others experience insecure work and inferior working conditions and arrangements. Of course as markets, and arenas of sunk economic and social costs, it is almost inevitable that clusters will also display the negativities of asymmetric knowledge;

24 Bjørn Asheim, Philip Cooke and Ron Martin they are, in other words, prone to problems of ‘lock-in’ and ‘hysteresis’ due to inadequate responsiveness to external shifts, changes, or threats from market processes (Grabher, 1993). The very mechanisms and structures of economic ‘inter-relatedness’ (Frankel, 1955) amongst the firms, activities, and institutions that characterize a cluster, and for a while imbue it with competitive advantage and successful performance, are likely later to become a source of rigidity, making it difficult for the cluster to adapt and adjust to changing conditions of competition, demand, and technology. Britian’s economic landscape, for example, is littered with the remnants of what were once thriving industrial districts, but which have long since lost their original advantages and dynamism: many of the districts celebrated by Marshall have either since disappeared or are today struggling under threat of extinction. Moreover, where clusters are dominated by oligopolistic producers, so clearly exemplified by the Hollywood example referred to above, those firms and individuals on the wrong side of the knowledge asymmetry scales must make double the effort to access fragmentary crucial knowledge of where a future project or contract opportunity might lie. Yet again, much may depend on the cultural foundations and social capital of a cluster (or district). For example, such are the strengths of Italian industrial districts as centres of asymmetric, design-intensive, neo-craft knowledge that they are able to attract global buyers from the world’s leading retail outlets as well as poor Chinese immigrants to access the key knowledge for designintensive textile and leather garment production. In short, different types of cluster (or district) will have different structural, organizational, cultural, and social characteristics, and will consequently have different capacities to respond to and cope with ‘shocks’ and changes from without. Building on these and other insights, and responding to the need for a reflective interrogation of the Porterian cluster concept and its role in thinking about and intervening in regional economic development, the essays that follow examine in more depth several of the issues raised in this chapter. As Martin and Sunley (2003, p. 5) have argued in relation to the cluster concept, ‘the mere popularity of a construct is by no means a guarantee of its profundity’. Influential and seductive though the cluster notion has become, there is much about it that is problematic: as is so often the case in the policy sphere, the rush to apply the cluster concept in national and regional development policies has run ahead of many theoretical, conceptual and empirical issues. Our aim in this book is not to reject the cluster idea outright – it undoubtedly provides one useful way in which to conceptualize, analyse, and intervene strategically in regional economic development. But collectively, the contributions that follow are concerned to address the problems that surround the definition of clusters, their theorization, the claims made for their economic benefits, and their use in policy-making. The cluster concept has both advantages but also important limitations. Thus far, the euphoria over the concept has tended to emphasize the former, without paying much attention to the latter. A more balanced account is required, one that acknowledges the limitations of the cluster concept as well as its usefulness. This is the underlying aim of this collection of essays,

The rise of the cluster concept 25 written by some of the leading authors in the field of cluster-based regional economic development.

References Adorno, T. (1991) The Culture Industry: Selected Essays on Mass Culture, London, Routledge Akerlof, G. (1970) The market for ‘lemons’: qualitative uncertainty in the market mechanism, Quarterly Journal of Economics, 84, 488–500 Asheim, B. (1996) Industrial districts as ‘learning regions’: a condition for prosperity? European Planning Studies, 4 (4), 379–400 Asheim, B. (1999) Interactive learning and localised knowledge in globalising learning economies, GeoJournal, 49 (4), 345–52 Asheim, B. (2000) Industrial districts: the contributions of Marshall and beyond, in Clark, G. L., Feldman, M. and Gertler, M. (eds) The Oxford Handbook of Economic Geography, Oxford, Oxford University Press, 413–31 Asheim, B. (2001) Learning regions as development coalitions: partnership as governance in European workfare states? Concepts and Transformation. International Journal of Action Research and OrganizationalRenewal, 6 (1), 73–101 Asheim, B. and Coenen, L. (2006) Contextualising regional innovation systems in a globalising learning economy: on knowledge bases and institutional frameworks, Journal of Technology Transfer, 31 (1), 163–73 Asheim, B. and Gertler, M. (2005) The geography of innovation: regional innovation systems, in Fagerberg, J., Mowery, D. and Nelson, R. (eds) The Oxford Handbook of Innovation, Oxford, Oxford University Press, 291–317. Baldwin, R., Forslid, R., Martin, P., Ottaviano, G. and Robert-Nicoud, F. (2003) Economic Geography and Public Policy, Princeton, Princeton University Press Bathelt, H. (2002) The re-emergence of a media industry cluster in Leipzig, European Planning Studies, 10, 583–612 Becattini, G. (1989) Sectors and/or districts: some remarks on the conceptual foundations of industrial economics, in Goodman, E. and Bamford, J. (eds) Small Firms and Industrial Districts in Italy, London, Routledge, 123–35 Becattini, G. (1990) The Marshallian industrial district as a socio-economic notion, in Pyke, F., Becattini, G. and Sengenberger, W. (eds) Industrial Districts and Inter-Firm Co-operation in Italy, Geneva, International Institute for Labour Studies, 37–51 Becattini, G. (2001) The Caterpillar and the Butterfly, Florence, Felice le Monnier Benner, C. (2002) Work in the New Economy: Flexible Labour Markets in Silicon Valley, Oxford, Blackwell Bottazzi, G, Dosi, G. and Fagiolo, G. (2002) On the ubiquitous nature of agglomeration economies and their diverse determinants: some notes, in Curzio, A. and Fortis, M. (eds) Complexity and Industrial Clusters, Heidelberg, Physica-Verlag Brackman, S., Garretsen, H. and van Marrewijk, C. (2001) An Introduction to Geographical Economics, Cambridge, Cambridge University Press Brusco, S. (1989) A policy for industrial districts, in Goodman, E. and Bamford, J. (eds) Small Firms and Industrial Districts in Italy. London, Routledge, 259–69 Brusco, S. (1990) The idea of the industrial district, in Pyke, F., Becattini, G. and Sengenberger, W. (eds) Industrial Districts and Inter-Firm Co-operation in Italy. Geneva, International Institute for Labour Studies, 10–19

26 Bjørn Asheim, Philip Cooke and Ron Martin Brusco, S., Cainelli, G., Forni, F., Franchi, M., Malusardi, A. and Righetti, R. (1996) The evolution of industrial districts in Emilia-Romagna, in Cossentino, F., Pyke, F. and Sengenberger, W. (eds) Local Regional Response to Global Pressure: The Case of Italy and its Industrial Districts, Geneva, International institute for Labour Studies Buchanan, J. and Yoon, Y. (eds) (1994) The Return of Increasing Returns, Ann Arbor, University of Michigan Press Cairncross, F. (1997) The Death of Distance: How the Communications Revolution Will Change our Lives, London, Orion Books Camagni, R. (1991) Introduction: From the local ‘milieu’ to innovation through cooperation networks, in Camagni, R. (ed.) Innovation Networks: Spatial Perspectives, London, Belhaven Press, 1–9 Caniëls, M. and Romijn, H. (2005) What drives innovativeness in industrial clusters? Transcending the debate, Cambridge Journal of Economics, 29, 497–515 Casper, S., Lehrer, M. and Soskice, D. (1999) Can high technology industries prosper in Germany? Institutional frameworks the evolution of the German software biotechnology industries, Industry and Innovation, 6, 5–24 Chesbrough, H. (2003) Open Innovation, Boston, Harvard Business School Press Coase, R. (1937) The nature of the firm, Economica, 4, 386–405 Coleman, J. (1988) Social capital and the creation of human capital, American Journal of Sociology, 94 (supplement), S95–S120 Cooke, P. (1998) Introduction: origins of the concept, in Braczyk, H., Cooke, P. and Heidenreich, M. (eds) Regional Innovation Systems, London, UCL Press, 2–25 Cooke, P. (2001) Regional innovation systems, clusters and the knowledge economy, Industrial and Corporate Change, 10 (4), 945–74 Cooke, P. (2004a) Regional knowledge capabilities, embeddedness of firms and industry organisation: bioscience megacentres and economic geography, European Planning Studies, 12, 625–42 Cooke, P. (2004b) Integrating global knowledge flows for generative growth in Scotland: life sciences as a knowledge economy exemplar, in OECD (ed.) Global Knowledge Flows and Economic Development, Paris, OECD Cooke, P. and Clifton, N. (2004) Spatial variation in social capital among UK small and medium-sized enterprises, in De Groot, H., Nijkamp, P. and Stough, R. (eds) Entrepreneurship and Regional Economic Development: A Spatial Perspective, Cheltenham, Edward Elgar Cooke, P. and Da Rosa Pires, A. (1985) Productive decentralisation in three European regions, Environment and Planning A, 17, 527–44 Cooke, P. and Morgan, K. (1998) The Associational Economy, Oxford, Oxford University Press Crouch, C., Le Gales, P., Toglia, C. and Voelzkow, C. (2001) Local Production Systems in Europe: Rise or Demise? Oxford, Oxford University Press Czamanski, S. (1974) Study of Clusters of Industries, Halifax, Dalhousie University, Institute of Public Affairs. Czamanski, S. and Ablas, L. (1979) Identification of industrial clusters and complexes: a comparison of methods and findings, Urban Studies, 16, 61–80 De Propris L. and Lazzeretti, L. (2003) The Birmingham Jewellery Quarter: A Marshallian industrial district, Working Paper Series 2003–17, Birmingham Business School, University of Birmingham Dei Ottati, G. (1994) Co-operation and competition in the industrial district as an organisational model, European Planning Studies, 2, 371–92

The rise of the cluster concept 27 Dei Ottati, G. (2004) The remarkable resilience of the industrial districts of Tuscany, in Cooke, P., Heidenreich, M. and Braczyk, H., (eds) Regional Innovation Systems, 2nd Edition, London, Routledge, 21–43 Doeringer, P. B. and Terkla, D. G. (1996) Why do industries cluster? In Staber, U., Schaefer, N. and Doeringer, P. B. (eds) Business Networks Prospects for Regional Development, Berlin, Walter de Gruyter, 175–89 Dosi, G. (1988) Sources, procedures and microeconomic effects of innovation, Journal of Economic Literature, 26, 1120–71 Fine, B. (2001) Social Capital versus Social Theory, London, Routledge Florida, R. (1995) Toward the learning region, Futures, 27, 527–36 Frankel, X. (1955) Obsolesence and technological change in a maturing economy, American Economic Review, 45, 296–319 Fujita, M., Krugman, P. and Venables, A. (1999) The Spatial Economy: Cities, Regions and International Trade, Cambridge, Mass., MIT Press Fujita, M. and Thisse, J.-F. (2002) Economics of Agglomeration: Cities, Industrial Location and Regional Growth, Cambridge, Cambridge University Press Gordon, I. and McCann, P. (2000) Industrial clusters: complexes, agglomeration or social networks? Urban Studies, 37, 513–32 Grabher, G. (1993) (ed.) The Embedded Firm: On the Socio-Economics of Industrial Networks, London, Routledge Granovetter, M. (1973) The strength of weak ties, American Journal of Sociology, 78, 1360–80 Gray, J. (1998) False Dawn: The Delusions of Global Capitalism, London, Granta Books Haraldsen, T. (1994) Teknologi, Økonomi og Rom – en Teoretisk Analyse av Relasjoner mellom Industrielle og Territorielle Endringsprosesser (Technology, Economy and Space – a Theoretical Analysis of Relations between Industrial and Territorial Development Processes), Doctoral Dissertation, Department of Social and Economic Geography, Lund University, Lund, Lund University Press Jacobs, D. and De Jong, M. (1992) Industrial clusters and the competitiveness of the Netherlands, De Economist, 140, 233–52 Jacobs, J. (1961) The Death and Life of Great American Cities, New York, Random House Jensen, B. (2004) Cluster Processes Typically Disregard Obstacles, Copenhagen, Oxford Research Keeble, D. and Wilkinson, F. (eds) (2000) High-tech Clusters: Networking and Collective Learning, Aldershot, Ashgate Kitson, M., Martin, R. L. and Tyler, P. (2004) Regional competitiveness: a key but elusive concept? Regional Studies, 38, 999–1000 Klein, J. (2001) A critique of competitive advantage. Paper presented at the Critical Management Studies Conference, University of Manchester, July Klepper, S. (2000) Entry By Spinoff, Pittsburgh, Carnegie Mellon University Krugman, P. (1991) Geography and Trade, Cambridge, Mass., MIT Press Krugman. P. (1994) Competitiveness: a dangerous obsession, Foreign Affairs, 73, 28–44 Krugman, P. (1996) Making sense of the competitiveness debate, Oxford Review of Economic Policy, 12, 17–35 Krugman, P. (1997) Pop Internationalism, Cambridge, Mass., MIT Press Lazzeretti, L. (2003) City of art as a high culture local system and cultural districtualisation process: the cluster of art restoration in Florence, International Journal of Urban and Regional Research, 27, 635–48

28 Bjørn Asheim, Philip Cooke and Ron Martin Lazzeretti, L. (2005) Density dependent dynamics in the Arezzo jewellery district, European Planning Studies (forthcoming) Loasby, B. J. (1998) Industrial districts as knowledge communities, in Bellet, M. and L’Harmet, C. (eds) Industry, Space and Competition, Cheltenham, Edward Elgar, 70–85 Lundvall, B.-Å. and Johnson, B. (1994) The learning economy, Journal of Industry Studies, 1, 23–42 Markusen, A. (1996) Sticky places in slippery space: a typology of industrial districts, Economic Geography, 72, 293–313 Marshall, A. (1919) Industry and Trade, London, Macmillan Marshall, A. (1930) Principles of Economics (8th edition; originally published 1890), London, Macmillan Marshall, A. and Marshall. M. (1879) Economics of Industry, London, Macmillan Martin, R. L. (1999) The new ‘geographical turn’ in economics: some critical reflections, Cambridge Journal of Economics, 23, 65–92 Martin, R. L. (2006) Alfred Marshall and economic geography, in Rafaelli, T., Darco, M. and Becattini, G. (eds) The Edward Elgar Companion to Alfred Marshall, London, Edward Elgar (in press) Martin R. L. and Sunley, P. (1996) Paul Krugman’s ‘geographical economics’ and its implications for regional development theory, Economic Geography, 72, 259–92 Martin, R. L. and Sunley, P. (1998) Slow convergence? The new endogenous growth theory and regional development, Economic Geography, 74, 201–27 Martin, R. L. and Sunley, P. (2003) Deconstructing clusters: chaotic concept or policy panacea? Journal of Economic Geography, 3, 5–35 Maskell, P. (2001) Towards a knowledge-based theory of the geographical cluster, Industrial and Corporate Change, 10, 4, 919–41 Morgan, K. (1997) The learning region: Institutions, innovation and regional renewal, Regional Studies, 31, 491–504 Morgan, K. (2004) The exaggerated death of geography: learning, proximity and territorial innovation systems, Journal of Economic Geography, 4, 3–22 Norton, R. D. (2001) Creating the New Economy: The Entrepreneur and US Resurgence, Cheltenham, Edward Elgar OECD (1999) Boosting Innovation, Paris, OECD OECD (2001) Innovative Clusters: Drivers of National Innovation Systems, Paris, OECD O’Malley, E. and Van Egeraat, C. (2000) Industry clusters and Irish indigenous manufacturing: limits of the Porter view, Economic and Social Review, 31, 55–80 Paniccia, I. (2002) Industrial Districts: Evolution and Competitiveness in Italian Firms, Cheltenham, Edward Elgar Penrose, E. (1959) The Theory of the Growth of the Firm, Oxford, Oxford University Press Perroux, F. (1955/1971) Note on the concept of growth poles, in Livingstone, T. (ed.) Economic Policy for Development: Selected Readings, Harmondsworth, Penguin Perry, M. (1999) Clusters last stand, Planning Practice and Research, 14, 149–52 Piore, M. and Sabel, C. (1984) The Second Industrial Divide, New York, Basic Books Porter, M. (1990) The Competitive Advantage of Nations, New York, The Free Press Porter, M. (1998) On Competition, Boston, Mass., Harvard Business School Press Porter, M. and Ackerman, F. D. (2001) Regional Clusters of Innovation, Washington, Council on Competitiveness Porter, M. and van Opstal, D. (2001) US Competitiveness 2001: Strengths, Vulnerability and Long-term Priorities, Washington, Council on Competitiveness

The rise of the cluster concept 29 Reich, R. B. (2001) The Future of Success: Work and Life in the New Economy, London: Random House Richardson, G. (1972) The organisation of industry, Economic Journal, 82, 883–96 Russo, M. (1985) Technical change and the industrial district: the role of interfirm relations in the growth and transformation of ceramic tile production in Italy, Research Policy, 14, 329–43 Sabel, C. (1989) Flexible specialisation and the re-emergence of regional economies, in Hirst, P. and Zeitlin, J. (eds) Reversing Industrial Decline: Industrial Structure and Policies in Britain and Her Competititors, Oxford, Berg, 17–70 Sako, M. (1992) Prices, Quality and Trust, Cambridge, Cambridge University Press Saxenian, A. (1994) Regional Advantage, Cambridge, Mass., Harvard University Press Schumpeter, J. (1975) Capitalism, Socialism and Democracy, New York, Harper Scott, A. (1988) New Industrial Spaces, London, Pion Scott, A. (2001) Global City Regions: Trends, Theory and Policy, Oxford, Oxford University Press Steiner, M. (ed.) (1998) Clusters and Regional Specialisation: On Geography, Technology and Networks, London, Pion Storper, M. (1995) Competitiveness policy options: the technology-regions connection, Growth and Change, Spring, 285–308 Storper, M. (1997) The Regional World: Territorial Development in a Global Economy, New York, Guilford Press Teece, D. and Pisano, G. (1994) The dynamic capabilites of firms: an introduction, Industrial Corporate Change, 3, 537–56 Trigilia, C. (1990) Work and politics in Third Italy’s industrial districts, in Pyke, F., Becattini, G. and Sengenberger, W. (eds) Industrial Districts and Inter-firm Co-operation in Italy, Geneva, International Institute for Labour Studies Turner, A. (2001) Just Capital: The Liberal Economy, London, Macmillan Varaldo, R. and Ferrucci, L. (1996) The evolutionary nature of the firm within industrial districts, European Planning Studies, 4, 27–34 Visser, E. and Boschma, R. (2004) Learning in districts: novelty lock-in in a regional context, European Planning Studies, 12, 793–808 Winter, S. (2000) The satisficing principle in capability learning, Strategic Management Journal, 21, 981–96 World Bank (2000) Electronic Conference on Clusters, Washington, World Bank Young, A. (1928) Increasing returns economic progress, Economic Journal, 38, 527–42 Zollo, M. and Winter, S. (2002) Deliberate learning the evolution of dynamic capabilities, Organisation Science, 21, 981–96

2

What qualifies as a cluster theory? Peter Maskell and Leïla Kebir

Introduction Clusters may be defined as non-random (Ellison and Glaeser, 1994) geographical agglomeration of firms with similar or closely complementary capabilities (Richardson, 1972).1 However, clusters have been scrutinized under many different labels. Some of the synonyms listed in Table 2.1 below may have the same essential meaning as conveyed by using the cluster concept as defined above but can differ in peripheral meaning by their implications (usually involving some minor idea or underlying assumptions in the meaning of the concept), connotations (usually including ideas that color the meaning of the concept often by providing historical or literary associations) or applications (usually the result of current idioms that have established restrictions on the use of a particular term). Marshall’s (1890) initially general or generic term of ‘the industrial district’ is, for instance, now often applied when wishing explicitly to emphasize the values and norms shared by collocated firms (see Brusco, 1982 among many others). In other cases singular academic contributions based on a particular term over time will have developed into distinct schools of thought while attracting rapidly increasing numbers of followers or critics.2 But even the present terminological diversity cannot hide the fact that the cluster phenomenon as such has attracted increasing attention during the last fifteen years. Many scientific journals have, consequentially, published one or more special issues on cluster theories, methods, or empirics while others now include a range of cluster analyses in each and every volume. The significant number of policy-oriented contributions reflects the fact that cluster building appeals to many policy-makers as the key to national, regional, and even local success (see Cooke, 1997, 2004). A simple quantitative illustration of the avalanche of recently published academic papers in this field is attempted in Table 2.1. The sudden surging interest in clusters is no unquestioned blessing. Few would, even at the outset, feel tempted to accuse the cluster literature at large for being overly concerned with precise definitions of important constructs or burdened with excessive specifications of the exact nature of the processes involved (Maskell, 2001b, Martin and Sunley 2003). The multitudes of existing contributions have mostly been concerned with making sense of empirical findings rather than with the

What qualifies as a cluster theory? 31 discovery of a common expanse of conceptual clarity. On the contrary one often finds an unfortunate habit of introducing novelty by making slight changes to the explicitly stated or implicitly applied definition of core concepts, or by importing constructs and variables from neighboring schools of thought without any impeding sensitivity towards the inherent theoretical and methodological tensions between what are, in essence, not completely parallel lines of inquiry. Courage and boldness in conceptual fusion are not always accompanied by careful reflections or in-depth discussions of the possible consequences for maintaining theoretical consistency. So while the cluster concept and its synonyms are still in great demand for current analytical and policy purposes, a scholarly practice has developed that might lead to their premature dismissal. Paraphrasing Reich (1990: 925) we run, perhaps, the risk that the cluster concept will join those rare terms of public discourse that have gone directly from obscurity to meaninglessness without any intervening period of coherence. In dejected moments it seems, furthermore, as if an increasing proportion of current cluster studies and cluster policy recommendations are at best only partly based on sets of formally connected statements, the final point of which is the explanation of an independent world, and discouragingly few empirical studies engage in thorough theory testing.3 It is, however, often easier to identify shortcomings than to fix them. What is needed is arguably some criteria or instrument whereby the theoretical core of the various lines of thought in this field can be made more explicit. In the next section we offer an attempt to contribute to this complex endeavor by explicitly considering the basic building blocks crucial for the construction of any comprehensive theory in this field. Within the limited space available it is, of course, possible only to provide a skeleton of an argument and we must rely on future work to assist in adding substance and supplying the contextual flesh needed for a fully satisfactory account. In the third section we provide hints on how such work may be conducted by applying our scheme to three currently significant approaches that take as their point of departure (1) externalities (illustrated by Marshall’s work on ‘Industrial district’); (2) competitiveness (illustrated by Michael Porter’s theory of cluster growth); or (3) territories (illustrated by the GREMI approach). In the fourth section we contemplate the type of policy advice likely to be offered from each of these three current archetypal lines of inquiry before, in the fifth section, we make a few concluding observations and identify the areas of particular ambiguity where further theoretical work is most urgently needed.

What constitutes a theory in the cluster field? Most academic disciplines harbor several competing epistemological positions, and this field is no exception.4 For the task at hand we find it helpful to apply the simple framework, cultivated by Whetten (1989), according to which a complete theory must address the questions of ‘what’, ‘how’, ‘why’ and usually

0 2 0 0 0 0 0 0 1 0 1

Cluster(s)/clustering of firm(s) Agglomeration – geographic(al) agglomeration(s) – spatial agglomeration(s) – agglomeration(s) of (same industry) firm(s) Geographic(al) concentration(s) Spatial concentration(s) Localized or localized industries or firms Growth pole Innovative milieu(s) Industrial district(s)

0 4 0 0 0 0 0 0 1 0 2

1960s 0 23 0 0 1 3 5 0 3 0 2

1970s 0 45 0 3 4 0 2 0 4 0 5

1980s 9 305 4 17 71 32 32 5 12 26 126

1990s

15 380 7 23 50 51 30 7 5 8 95

2000s

Source The table is based on a privately owned but publicly accessible database available through ISI Web of Knowledge . A search for each term for all years (1953–2004) was made for published articles only. An element of discretion was applied as entities were deleted from the list of results if deemed irrelevant for the purpose of this chapter.

Note 1950s: from 1 January 1953. 2000s: to 30 September 2004. Though the table is based on the best current representation of the development within scientific journal publishing, its inherent methodological problems should not be disregarded. Terms belonging to academic traditions that rely heavily on books are underrepresented by definition. The same applies to terms used mainly outside English-language journals. Many radical and experimenting journals are never included in the ISI database because of a common ‘curfew’ period of ten years from the time the first volume appears before it is considered for inclusion. The selection criteria used in the table is the presently best available, but admittedly imperfect in many ways and should, of course, not be taken to imply anything about quality, academic significance or the relevance of the terms included.

1950s

Term looked up in database

Table 2.1 Cluster publications 1953–2004: Number of articles, published in scholarly journals within the social sciences with the term ‘cluster’, its synonyms, or its more distant cousins in the title or in the abstract or among the keywords

What qualifies as a cluster theory? 33 also of ‘when/where/who’.5 The phrase ‘cultivated’ implies that this kind of framework can be found in numerous studies including, for example, the once much cherished Introduction to Regional Economics where Edgar M. Hoover summed up spatial economics in the question: ‘What is where, and why – and so what?’ (1971: 3). In Whetten’s later elaboration each of his questions will, when answered, provide a distinct building block, needed for the construction of a theory. First, confronting the question of ‘what’ will lead to identifying factors such as variables, concepts or constructs considered important for the explanation.6 Second, facing the question of ‘how’ provides causal links between these factors to form an ordered and explicit pattern of connections and relations. Taken together, the answers to questions of ‘what’ and ‘how’ constitute the field or subject of the theory. Third, it is when answering the question of ‘why’ that we find the core of the theory. In order to be convincing, the answer to the question of ‘why’ must offer logical and compelling justifications for the factors included (‘what’) and for the links suggested (‘how’). ‘Why’ also generates propositions that can establish new insights, challenge entrenched views and deepen our understanding of the phenomenon investigated. The explanatory and prescriptive quality and strength of any theory is often directly dependent on the solidity and novelty of the way in which ‘why’ is approached. Fourth and finally, ‘when/where/who’ adds contextual conditions and temporal or spatial limitations on the propositions generated and spell out the circumstances where the theory is unlikely to hold (Whetten 1989).7 According to this scheme, the profundity of the notion of clusters is thus conditional on the coherence of the reasoning when addressing the pivotal ‘why’.8 This was, by the way, also noted by Hoover a quarter of a century ago, who continued by pointing out, long before Krugman, how ‘[t]raditional geographers, though directly involved with what is where, lacked any real technique of explanation in terms of human behaviour and institutions to supply the why, and resorted to mere description and mapping’ (1971: 4, original italics). However, the task of addressing the crucial ‘why’ is complicated by the fact that it is less than totally satisfactory to provide even a very compelling account for the economic and social benefits that firms may accrue when collocating (the cluster existence argument) without also including a justification for the diseconomies encountered when exceeding certain geographical and sectoral thresholds (the extension argument).9 Without the latter we end up with a theory that claims that all kinds of activities from all corners of the world will ultimately end up at one single location because of the unrestricted benefits of collocating once the process gets started and the very first cluster is formed. A scrupulous treatment of ‘why’ would, in addition, also include an exhaustion argument that spelled out the internal or external conditions that made previous decisive collocation benefits turn sour during the lifecycle of the cluster. When the existence argument is undermined the clusters’ vitality becomes threatened and the final demise might become immanent.

34 Peter Maskell and Leïla Kebir In the next section we examine three different lines of inquiry within the cluster field in a deliberately broad sense. In each case we attempt to identify the theoretical elements discussed above in order to show their general applicability across particular lines of investigation.

Three sets of illustrations Contributions with a focus on local spillovers Since Marshall’s initial reflections on the cluster issue were published in 1890 they have formed the cornerstone of much subsequent thinking on this issue.10 Even though his ‘industrial districts’ occupied only a small fraction of the grand explanatory scheme developed in Principles, his specific interest in such cluster phenomena is about the (uneven) distribution of economic activity over space, and more specifically about the tendency for related firms to collocate at certain places over prolonged periods of time. The core variables are thus firms’ (different) location requirements and economies external to the individual firm but internal to the district (‘what’). Firms are linked directly by business (supply and purchase) relations and indirectly through the market for labor and for private or public services. Locational economies or ‘spillovers’ are initially perhaps unanticipated outcomes of a successful match between firms’ location requirements and the supply of location factors (‘how’). Additionally, the Marshallian framework has from the very beginning included the entire set of ‘why’ arguments set out in the previous section. The existence argument was balanced by an extension argument through the effect of the simultaneous centripetal and centrifugal forces that together determined the geographical pattern of firms’ location. The centripetal forces often consisted of cost advantages in transportation or when sharing an environment made particularly agreeable by, for instance, a dedicated infrastructure, a pool of notably skilled labor, an educational systems of distinctive relevance, a fine concentration of specialized suppliers etc. but went much further when including many of the less easily measured factors such as rivalry, search costs, institutional factors and various positive spillovers along both the vertical and horizontal dimensions of the cluster (Loasby, 1999, 2000). The extension argument of centrifugal forces was, in contrast, normally based on the costs of congestion, or the bidding-up of prices for land, labor or the services or goods provided, but could be extended to include negative spillovers when different industrial logics clashed. The exhaustion argument allowed for some vital factor (for instance a mineral deposit or a climate or transport benefit) to have been fully exploited or otherwise discontinued. It is, perhaps, at this stage worth noting how later generations of mainly Anglo-Saxon scholars by deliberate decision or by following the prevailing tradition in contemporary economic geography gradually turned to producing very descriptive, ideographic work.11 The crucial ‘why’ was, consequentially,

What qualifies as a cluster theory? 35 more often assumed or implied rather than carefully investigated and specified. It was almost as if all three arguments in ‘why’ – the existence, extension and exhaustion arguments – gradually became considered so self-evident that no discussion or investigation was required (Feser, 1999). When the new wave of interest in the cluster phenomenon started growing a decade or two ago the former coherent explanatory model had largely disappeared. With a few significant exceptions it was replaced by a one-sided model that addressed the existence argument in novel ways but almost totally disregarded the extension and exhaustion arguments. Many recent contributions stem, for instance, from the basically Marshallian belief (whether sustained by empirical evidence or not) that collocated firms benefit from the ease of identifying and communicating with suited partners in their vicinity (low search costs) or from their ability to bridge cognitive gaps that enabled them to understand motives and desires that under other circumstances would remain opaque. By reducing the costs of co-ordination and by overcoming problems of asymmetrical information, the process of clustering enables, it was maintained, a deepening of the local division of labor so that a higher level of specialization and knowledge creation may be within reach. The existence argument is thus based on the combined advantages of social coherence, relational flexibility, and ease of intracluster interaction, as well as the deepening of the local knowledge base that it occasions, while the extension and the exhaustion argument has attracted considerably less attention if any at all.12 One of the most notable exceptions to this trend is Michael E. Porter, to whom we shall now turn. Contributions with a focus on competitiveness Few contributions to the cluster literature have gelled the interests of a generation of scholars as Michael E. Porter’s work on competitive strategy. The commotion occasioned by his 1990 book and supported by his subsequent cluster-related papers has helped fulfill his prophecy that ‘economic geography must move from the periphery to the mainstream’ (Porter 1994: 38). Most bright second-year students of business, strategy, management or economic geography will now probably be able to reproduce his famous ‘diamond’ model where the boxes addresses the ‘what’ while the connecting arrows (and the accompanying explanation) take on the ‘how’ in his theory. They will most likely also be well aware of the fact that his main aim was to provide not a theory of the cluster but a ‘theory of national, state, and local competitiveness within the context of a global economy’ (Porter 1998: 197). As he initially explained (Porter, 1990a–1990b: 73): The basic unit of analysis for understanding national advantages is the industry. Nations succeed not in isolated industries, however, but in clusters of industries connected through vertical and horizontal relationships. A nation’s economy contains a mix of clusters, whose makeup and sources of

36 Peter Maskell and Leïla Kebir competitive advantage (or disadvantage) reflect the state of the economy’s development. In spite of not aiming at creating a theory of the cluster he does, nevertheless, address the ‘why’ in ways that meet the criteria set out in the second section. His cluster existence argument includes the way in which proximity (also in the form of shared culture and low transaction costs) makes Benefits flow forward, backward and horizontally. Aggressive rivalry in one industry tends to spread to others in the cluster . . . exchange of R&D and joint problem solving lead to faster and more efficient solutions . . . Suppliers also tend to be a conduit for transmitting information and innovation from firm to firm. Through this process, the pace of innovation within the entire national industry is accelerated. All these benefits are enhanced if suppliers are located in proximity to firms, shortening the communication lines. (Porter, 1990a: 151, 103) He later spelled out how Developing clusters also attract – and cluster participants seek out – people and ideas that reinforce the cluster. Growing clusters attract skilled people through offering greater opportunities. Entrepreneurs or individuals with ideas migrate to the cluster from other locations, as well, because growing cluster signals opportunity. A cluster’s success stories help attract the best talent . . . Cluster membership makes possible direct observation of other firms. (Porter, 1998: 241, 221) His extension argument is based on the old Marshallian duality of dispersing versus locating activities (see Porter, 1990a: 156–7) while his exhaustion argument includes a whole range of potentially important intra-cluster forces (such as ebbing domestic rivalry, the development of internal rigidities and regulatory inflexibilities) as well as a number of externally induced influences (such as technological discontinuities, deteriorating factor conditions and shifts in buyers’ needs (Porter 1990a: 166–9 and 1998: 243–4). Contributions with a focus on the region and its development In contrast to the Porterian ‘what’, with its focus on competitiveness and neglect of issues concerning uneven spatial development, the innovative milieu approach13 is concerned with technology, organization and, most significantly, with territory. Together these three elements are seen as constituting a localized initial context without frontiers in a strict sense, but presenting a certain degree of unity in terms of identifiable and specific behavior (Maillat, Quévit and Senn,

What qualifies as a cluster theory? 37 1993). The ‘how’, in turn, is addressed by introducing (1) a set of actors, who are independent enough to make strategic choices when managing material (infrastructures, machines, financial capacities etc.) and immaterial (know-how, institutions resources etc.) resources; (2) a learning dynamic that reveals the actors’ capacity for adapting to changes in the environment; (3) an organizational logic according to which actors co-operate to innovate and develop networks of interdependent commercial and non-commercial relationships (Maillat, Quévit and Senn, 1993). The innovative milieu approach is much broader in its scope than Marshall’s or Porter’s cluster theories but it does contain the three crucial elements of ‘why’ required to make a theoretical contribution within the field dealt with in this chapter. The existence argument is based on a set of relationships that develops spontaneously within a given geographical area and generates a localized dynamic process of collective learning. Together ‘they act as an uncertainty-reducing mechanism in the innovation process’ (Camagni, 1995: 320). An innovative milieu thus stimulates the development of know-how and the formation, development and vitality of innovation networks (Maillat, Quévit and Senn, 1993). ‘It facilitates mutual acquaintance, collaboration, dissemination and exchange of information, just as it allows for the development of trust-relations. It offers options for reciprocal openness and for disseminating know-how without any risk of unilateral appropriation, because the players share the same work ethic and a common will to cooperate’ (Maillat, 1998: 19). In short: Innovative milieux help the local actors conceive, devise and complete their joint projects.14 The extension argument is based on the proposition that co-operation between actors lead to the building of relational capital involving the mobilization of resources that are not necessarily of a monetary nature alone. Local sets of values (entrepreneurial, family, professional etc.) guide actors to contribute towards innovation and production while making social investments that permits them, ultimately, to co-operate on a basis of trust and reciprocity (Crevoisier, 2004: 371). The resulting networks help maintain and reproduce the boundary between the innovative milieu and the exterior in the sense that they define which actors constitute part of the local co-ordination system and which do not (ibid.). The exhaustion argument builds on the acknowledgement of how all milieux may lose cohesion because individual interests gain the upper hand over those of the community, if and when ‘opportunistic behavior causes defiance or the outward openness becomes inadequate to ensure the enlargement of new co-operative relations or the replacement of technologies’ (Maillat, 1998: 15). Consequently, territories differentiate and old industrial fortresses disintegrate.

Public policy options Within the common field of cluster theory each of the three specific lines of investigation briefly introduced in the third section has a distinct perspective on the cluster phenomenon and provides dissimilar justifications for the existence,

38 Peter Maskell and Leïla Kebir extension and exhaustion of clusters that could, perhaps, be developed into new propositions and testable hypothesis. Our aim with the following is less ambitious: namely to point out some of the prescriptive differences that stem from how theory is constructed in each of the three examples. In doing so we restrict ourselves for reasons of space to concentrating on the realm of advice aimed at public policy-makers while acknowledging that cluster theory at large may also forward potentially influential pieces of advice useful mainly or exclusively for managers or stake holders in the business community. Some pieces of advice are universal in the sense that they are rooted in all cluster theories when implicitly accepting the currently prevailing division of labor between a private and a public sector, with the latter in charge of providing infrastructure, education and similar elements of general demand and applicability. Other pieces of policy advice are more intimately associated with one or more specific streams of theories, and it is to these we shall now turn. Policy contributions with a focus on local spillovers If for the sake of brevity we allow ourselves to assume that Marshall’s industrial districts follow a lifecycle model, where stages of infancy are succeeded by increasing maturity and subsequent stagnation or decline, the framework developed in the previous sections may be taken to suggest how each such stage can merit a set of specific public policies distinctively different from what will be generally beneficial at the other stages. At the infant stage, when firms with complementary or similar capabilities have started to reap some of the benefits of collocation through experimentation or conjecture, the relevant public policy options are mainly market-conformist by supporting what is already in the making and by helping to provide inputs in short supply. Targeted labor-mobility-improving measures, specific educational efforts and vocational training programs, dedicated initiatives to enhance creativity and collaboration, physical infrastructure improvements, actions to develop competent seed and venture capital sensitive to the particular requirements and structure of the local firms through taxation relief or by redirecting public funds, all belong to this set of infant stage policies. Most of the abundant policy ambitions and initiatives to support and develop clusters in recent years have been concerned with the next, mature, stage of cluster development. The existence arguments spelled out above lead us to question the value of these initiatives and view them as being at least partly misdirected. The important point is that precisely because the Marshallian cluster theory is about self-organized, market-led dynamics, the factors and processes emphasized by the theory largely take care of themselves. Finally, when clusters for one reason or another reach the exhaustion stage the policy challenge shifts from being supportive to becoming creatively destructive by actively disjoining present means–ends designates and by dismantling institutions molded to accommodate and support yesterday’s economic struc-

What qualifies as a cluster theory? 39 tures. By assisting communities when faced with the need to unlearn previously successful routines such policies provide cognitive and economic space for new waves of entrepreneurial activity that might subsequently help put the cluster on a new and promising track. Interestingly, there appears to be a great local variation in the ability to unlearn. Some clusters can inaugurate novel institutions and simultaneously dissolve obsolete ones while similar or stronger efforts in other clusters are unsuccessful. In the volatile environment of the current globalizing exchange economy such ‘unlearning’ capabilities might turn out to be of paramount significance for the ability of clusters (as well as the larger entities of regions or even nations) to attract firms and participate in sustaining their competitiveness in an already established industry, or to rebuild competitiveness by developing new industries. In areas under less fortunate circumstances, making an appropriate policy response is an even more formidable task. In such difficult situations a successful outcome of even the most energetically pursued and cleverly designed policies may appear so late that little remains to be saved. More than one initially enthusiastic development agency has over the years come to a complete standstill owing to the numerous and complex difficulties that emerge when renewal implies jeopardizing the interests of individuals or larger groups with the incentive and power to prevent or impede the process in spite of the cost of their actions to the overall society. Openness and competition among different political entities provide what is arguably the best check. While local co-ordinated action is usually a blessing, local power groups that are too close-knit are thus an unquestioned evil when uncomfortable decisions have to be made. Policy contributions with a focus on competitiveness Given the overall and explicit aim of Michael E. Porter’s cluster theory it is no wonder that the overarching emphasis is placed on establishing and maintaining competition. There is a strong current running through the theory that the business sector, if left to itself, might easily slip into unhealthy practices of collision that will, ultimately, prevent rivalry and thereby undermine the cluster existence argument. Securing competition is therefore undoubtedly the single most important policy recommendation emerging from his theory and perhaps the only area where strong and constant public presence and intervention are required. The second and optional role remaining for public policy-makers is demanding. The authorities must become catalysts and challengers that encourage and push enterprises to raise their performance even though this can be unpleasant and even painful for the firms involved. It is to this end that Porter advocated the enforcement of strict product safety and environmental standards long before such thoughts became fashionable and very long before social responsibility became part of managerial rhetoric and practice. He realistically notes that Most of the policies that would make a real difference either are too slow and require too much patience for politicians or, even worse, carry with

40 Peter Maskell and Leïla Kebir them the sting of short-term pain. Deregulating a protected industry, for example, will lead to bankruptcies sooner and to stronger, more competitive companies only later. (Porter, 1990b: 87) Other kinds of governmental action have the limited prospect of being partially successful only when working in tandem with favorable underlying conditions in the diamond model (the theory’s ‘what’ and ‘how’, see previous section). Finally, Porter argues consistently that in addition to the three areas mentioned above the main recommendation for policy-makers is of a hands-off nature. He frequently warns against intervening in factor and currency markets and stresses that ‘in politics, a decade is an eternity. Consequently, most governments favor policies that offer easily perceived short-term benefits, such as subsidies, protection, arranged mergers – the very policies that retard innovation’ (Porter, 1990b: 87). Porter’s disregard for possible damaging consequences of successful clusters for a just and even regional development is a direct consequence of his theory’s main focus on national competitiveness. Cluster growth is a means towards this end and no mercy should be shown towards policy-makers who wish to restrict this process owing to some misdirected passion for regions and places that bleed while fuelling the process. Contributions with a focus on the region and its development The completely opposite focus and recommendations can be found in the ‘innovative milieu’ approach that views the role of government and semi-public forms of governance in a much more positive light. The approach is not at all based on a belief in the blessing or basically advantageous functioning of the market economy. Firms are not the only important actors in the approach and local synergy could and should in some circumstances be enhanced through the creation of a local ‘agent d’animation’ or cross-firm organizer. Another area where local firms need help to develop in the overall interest of the region has to do with trans-border activities. Through public policy measures such activities can be encouraged to secure the inflow of ideas and exchange of resources beneficial for an immediate and long-term growth trajectory. This line of argument is somewhat in parallel with the recent thinking developed within the knowledge-based view of the cluster. Both approaches point to areas of interaction with the outside world where the local dynamics are not necessarily sufficient to maintain long-term sustainability and where explicit policy measures may be warranted even during the mature stage. While collocated firms thus usually have a good sense for and understanding of the relevant global technological frontiers they are often less well equipped to monitor and grasp categorically new knowledge especially if organized differently. The reason is that the clustered firm’s international network of contacts to

What qualifies as a cluster theory? 41 suppliers, customers and immediate competitors does not automatically include novel developments along parallel technological or commercial trajectories even when they are pertinent to future competitive positions. Rather than making extensive efforts in generating and promoting local buzz (Bathelt et al., 2004) through various forms of social engineering the main emphasis should therefore be placed on external communication policies directed towards widening the horizon and extending the reach of the local actors by confronting them with other equally competitive or superior ways of how to organize and develop well-known local products or services. It is quite characteristic of the ‘innovative milieu’ approach that the exhaustion process is seen as containing the seeds for revival and new periods of flourishing. The innovative milieu is envisioned as potentially able to utilize the tensions that emerge during the process of change by guiding the localized production system towards a new state in which the territorial logic continues to manifest itself. ‘The result is a milieu that possesses specific resources, rules for functioning, its own territory, and on a deeper level a technical culture and one of interdependencies . . . The territory is thus both the imprint of the former functioning of the milieu and the matrix of its transformation’ (Crevoisier, 2004: 374). If it does not succeed in this turn-over process, ‘the localised production system disappears and the territorial logic gives way to the functional logic’ (Maillat, 1998: 21). Common features across lines of investigation Taken together, it is to some extent striking that, regardless of the rather deep theoretical differences between the three approaches discussed above, they all share the same view when coping with derelict clusters. All argue that the main policy target in the post-exhaustion restructuring process is to create room for novel private sector initiatives as swiftly and effectively as possible rather than to pursue some governmental strategy of picking the winner by applying a range of top-down measures. In this they are probably very wise. Countless well intentioned but ineffectual cluster policies from all parts of the world seem to highlight the limits of the nation state, or any other political authority, in creating economically sustainable competitive advantages by design from above. No kind of vogue phrasings or remolded instrument packages can apparently alter the fact that the role of policy in the development of cluster advantages can only be marginal, indirect and long-term. Results are measured in decades if measurable at all.

A few final comments The chapter attempts to make four basic points. The first is the simple claim that the theoretical underpinning of the recent deluge of cluster studies is often less than totally clear. It is especially maintained that, while the cluster concept and its synonyms or more distant cousins are still in great demand for current

42 Peter Maskell and Leïla Kebir analytical and policy purposes, a scholarly practice has developed that might lead to their premature dismissal. The second point suggests that theory-development authorities might, in fact, have provided tools helpful for identifying and highlighting the factors (‘what’), relations (‘how’) and justifications (‘why’) that together constitute the core of a theory. It is argued that by paying more attention to disentangling what each theory is about or not about some insights might also emerge regarding its possible range of applicability. The third point concludes that, in order to be convincing, a theory of the cluster must do more than provide even a very compelling account for the particular benefits of collocation. It must also include an explanation for the balancing forces that prevent unconstrained cluster growth and for the conditions that may lead to the decline or extinction of the cluster. The fourth point concerns the difference in theory-constituting elements that emerge when a closer look is taken at some commonly used approaches, and the dissimilar public policy options that follow. No thorough or fully satisfying account of any of these points has, of course, been possible within the limited space available and much work remains to be done. Nevertheless, the chapter does illustrate how the three specific approaches investigated provide very dissimilar justifications for the existence, extension and exhaustion of clusters that could, if given sufficient care, be developed into new propositions and testable hypotheses. The usefulness of many current cluster studies would be vastly improved if more effort was directed towards developing critical hypotheses and using the empirical material, often painfully assembled, for cautiously testing such hypotheses before prescriptive advice was offered. What is urgently needed is further work aimed at distilling basic explanatory elements of some of the many theories currently in play in the field and bringing them to trial by confronting them with real-world data. There are several reasons for this. First, such work is much needed in order to get rid of explanations that simply do not hold true and to understand more precisely the ‘when/where/ who’ of the survivors. Second, such work is, of course, the only way in which reliable prescriptive advice can be developed and offered to managers and policymakers. Third, and most important for this field in academic terms, it is by engaging in such processes that a new stage of theoretical progress may become possible.

Notes 1 Whether the firms are functionally connected or not is thus an empirical question, not a feature of the core definition. See Chapter 3 below for a brief account of the confusion occasioned when nesting concepts of functionality and proximity. 2 Such schools of thought include the GREMI approach (Aydalot, 1986, Maillat, 1991, Maillat and Perrin, 1992, Maillat et al., 1993, Camagni, 1995, Ratti et al., 1997, Maillat, 1998, Crevoisier and Camagni, 2000, Camagni et al. 2004), the many largely Marshallian studies of the Italian industrial districts (Becattini, 1990, Brusco, 1986, 1999, Brusco and Righi, 1989, Dei Ottati, 1994, 1996, Bellandi, 1996,

What qualifies as a cluster theory? 43

3 4

5 6

7 8

9 10

Garofoli 1992a, 1992b, 1993, Gottardi, 1996, Belussi, 1999a, 1999b), the local production system school (Benko and Lipietz 1992, Courlet and Pecqueur, 1992, Courlet, 2001), the French ‘proximité’ tradition (Kirat and Lung, 1999, Gilly and Torre, 2000, Rallet 2000), an econometric type of cluster analysis (Swann et al., 1998), different ‘systemic’ analysis (Markusen et al., 1986, Saxenian, 1994, Malecki, 1991), some of which have focused explicitly on the geography of innovation (Feldman, 1994, Stenberg, 1999, Breschi, 2000), as well as the cherished approach applied by Porter (1990a). Most recently a broader ‘knowledge school’ has emerged (Maskell, 1998, 2001a, Maskell et al., 1998, Maskell and Malmberg 1999, Loasby, 1999; see Brenner, 2000, and Moulaert and Sekia, 2003 for overviews) as well as the learning region approach (Florida, 1995, Morgan, 1997, Simmie, 1997, Asheim, 1997, Maillat and Kebir, 2001). Overviews can be found in Harrison (1992), Norton (1992), Malmberg (1997, 1996), Baptista (1998), Storper (1995), Bianchi (1998), and Yeung (2000). Yet some of the scholars included in this book belong to the selected congregation well versed in this important activity. See also Hanson (2001) for an overview over recent attempts to test important features highlighted in the cluster literature. See Barnes (2001, 2002) for recent reflections on two such positions. Considerations of space force us to provide an admittedly rudimentary account of complex issues of theory building that under other and better circumstances would require a full paper or more. We acknowledge recent attempts within economic geography to reinterpret what theory is and to apply a definition much looser than that outlined above, but have not yet come across published contributions that have applied this novel interpretation within the field under scrutiny in this chapter. See further Dubin (1978), Gagliardi (1999), Lengnick-Hall and Wolff (1999) and Sutton and Staw (1995). Other epistemological positions may argue that conceptualization is theory building, and thus protest against lumping variables, concepts and constructs into one single group. We assert that conceptualization represents a step towards theory-building by providing important building blocks, but that concepts as such have no explanatory nor predictive power. In Hoover’s phrasing ‘[w]here’ referred to the location in relation to other economic activities involving ‘questions of proximity, concentration, dispersion and similarity or disparity of spatial patterns’ (Hoover, 1971: 3). However, some theoretical contributions become known mostly for the specific and often novel way by which they confront the ‘when/where/who’. The ‘when’ is, for instance, prominent in Piore and Sabel’s (1984) much quoted book on The Second Industrial Divide, in which a significant theoretical contribution to the cluster literature is placed within an explicit timeframe of Fordist versus post-Fordist modes of production, while Steven Klepper’s path-breaking paper (2002) on the formation of Detroit as a durable cluster of car manufacturing highlights ‘who’ by emphasizing the pivotal role of very knowledgeable entrepreneurs who spun-off from the incumbents that survived the first dramatic shakeout in the industry. Klepper’s cluster contribution is thus also an important supplement to mainstream economic assumptions where prices a whisker above competitive levels attract new entrants. The extension arguments are usually based on some specific combination of diseconomies of scale (see further below). Two current traditions derive directly from this rich source. One is the predominantly Italian Industrial District literature (see Chapters 3 and 4 in the present volume for an in-depth presentation) that focus on the way in which large firms may be matched or even out-competed by flexible arrangements of a myriad of independent, small firms working together within a confined area i.e. a Marshallian Industrial District. The other tradition with direct roots in Marshall’s initial

44 Peter Maskell and Leïla Kebir

11

12 13

14

contribution is the Knowledge-Based Theory of the cluster (see Maskell and Malmberg, 1997, Maskell, 2001a, Malmberg and Maskell, 2002, Bathelt et al., 2004) that emphasizes the role of knowledge formation when addressing the ‘why’ and its three components: the existence, the extension and the exhaustion argument. The descriptive tradition was apparently less pronounced on the European continent perhaps because of the strong economic and later also sociological tradition for abstract reasoning and theoretical contributions within the field (e.g. Weber, 1909). In contrast, the research tradition set out by George Goudie Chisholm in the UK and furthered by Dudley Stamp became known for its focus on the assembly and ordering of concrete facts and detesting the abstract theorizing about deeper reasons or chains of causality (ridiculed in Robinson’s (1908: 251) remark that it was so kind of the good Lord to have made the great rivers run through the great cities). Yet some of Chisholm’s surviving notebooks cast a somewhat different light on his interests and research agenda (Chisholm, 1868). The extension argument is, however, explicitly addressed in, for instance, Maskell (2001b). The concept was coined by the GREMI group. GREMI is an acronym for ‘Groupe de Recherche Européen sur les Milieux Innovateurs’ or the European Research Group on Innovative Milieux, which was formed in 1986 to study the interaction between innovations and localized factors (together termed ‘territory’) in particular in the new French spatial dynamics – the retournement spatial – in order to explain different regional development trajectories and the diffusion of new technologies (Aydalot, 1986). Scholars working with this line of inquiry are careful in stressing that no fully fledged theory has yet been formulated (Crevoisier, 2004: 368). This research agenda represents a clear break from former French traditions in which large firms, with the capacity to shape the territory and to generate innovation, used to occupy centre stage in the investigation of spatial development issues (Matteaccioli and Tabariés, 2002).

References Asheim, B. T. (1997) ‘“Learning regions” in a globalised world economy: towards a new competitive advantage of industrial districts’, in Taylor, M. and Conti, S. (eds) Interdependent and Uneven Development: Global–local Perspectives, Aldershot, pp.143–76. Aydalot, P. (1986) Milieux innovateurs en Europe [Innovative milieux in Europe], Paris: GREMI. Baptista, R. (1998) ‘Clusters, innovation and growth: A survey of the Literature’, in Swann, P. G. M., Prevezer, M. and Stout, D. (eds) The Dynamics of Industrial Clustering: International Comparisons in Computing and Biotechnology, pp. 13–51. Oxford: Oxford University Press. Barnes, T. J. (2001) On theory, history, and anoraks, Antipode 33: 162–68. Barnes, T. J. (2002) Critical notes on economic geography from an aging radical. Or radical notes on economic geography from a critical age, ACME: An International E-Journal for Critical Geographies (www.acme-journal.org), 1: 8–14. Bathelt, H., Malmberg, A. and Maskell, P. (2004) Clusters and knowledge: Local buzz, global pipelines and the process of knowledge creation, Progress in Human Geography 28 (1): 31–56. Baumard, P. (1996) Organizations in the fog: An investigation into the dynamics of knowledge, in Moingeon, B. and Edmondson, A. (eds) Organizational Learning and Competitive Advantage. London: Sage. Becattini, G. (1990) The Marshallian industrial districts as a socio-economic notion, in

What qualifies as a cluster theory? 45 Pyke, F., Becattini, G. and Sengenberger, W. (eds) Industrial Districts and Inter-Firm Co-operation in Italy, pp. 37–51. Geneva: International Institute for Labour Studies. Bellandi, M. (1996) Innovation and change in the Marshallian industrial district, European Planning Studies 4 (3): 357–68. Belussi, F. (1999a) Path-dependency vs. industrial dynamics: an analysis of two heterogeneous districts, Human Systems Management 18: 161–74. Belussi, F. (1999b) Policies for the development of knowledge-intensive local production systems, Cambridge Journal of Economics 23: 729–47. Benko, G. and Lipietz, A. (eds) (1992) Les régions qui gagnent. Districts et réseaux: les nouveaux paradigmes de la géographie économique [The winning regions – Districts and networks: the new paradigms of economic geography] Paris, Presses Universitaires de France. Bianchi, G. (1998) Requiem for the Third Italy? Rise and fall of a too successful concept, Entrepreneurship and Regional Development 10: 93–116. Brenner, T. (2000) The evolution of localised industrial clusters: identifying the processes of self-organisation, Papers on economics & evolution, no 0011, Jena, Max PlankInstitute. Breschi, S. (2000) The geography of innovation: A cross-sector analysis, Regional Studies 34 (3): 213–30. Brusco, S. (1982) The Emilian model: Productive decentralisation and social integration, Cambridge Journal of Economics 6 (2): 167–84. Brusco, S. (1986) ‘Small firms and industrial districts: The experience of Italy’, in Keeble, D. and Wever, E. (eds) New Firms and Regional Development in Europe, pp. 184–202. London: Croom Helm. Brusco, S. (1999) ‘The rules of the game in industrial districts’, in Grandori, A. (ed.) Interfirm Networks: Organization and Industrial Competitiveness, pp. 17–40. London: Routledge. Brusco, S. and Righi, E. (1989) Local government, industrial policy and social consensus: the case of Modena, Economy and Society 18 (4): 405–24. Camagni, R. P. (1995) The concept of ‘innovative milieu’ and its relevance for public policies in European lagging regions, Papers in Regional Science 74 (4): 317–40. Camagni, R., Maillat, D. and Matteaccioli, A. (eds) (2004) Ressources naturelles et culturelles, milieux et développement local [Natural and cultural resources, milieux and local development], Neuchâtel: GREMI, EDES. Chisholm, G. G. (1868) Essay on the Nature and Origin of our Knowledge of Extension and Space. Edinburgh: University of Edinburgh (handwritten notebook). Chisholm, G. G. (1889) Handbook of Commercial Geography, 12th edition 1932. London: Longmans, Green & Co. Coase, R. H. (1960) The problem of social cost, Journal of Law and Economics 3: 1–44. Cooke, P. (1997) Regions in a global market: The experiences of Wales and BadenWürttemberg, Review of International Political Economy 4 (2): 349–81. Cooke, P. (2004) Complexity and industrial clusters: Dynamics and models in theory and practice, European Planning Studies 12 (4): 595–9. Courlet, C. (2001) Les systèmes productifs locaux: de la définition au modèle [Local production systems: From the definition to the model], in DATAR (ed.) Réseaux d’entreprises et territoires: Regards sur les systèmes productifs locaux [Networks of firms and territories: A Look at the Local Production Systems], Paris, La Documentation Française, pp. 17–61.

46 Peter Maskell and Leïla Kebir Courlet, C. and Pecqueur, B. (1992) Les systèmes industriels localisés en France: un nouveau modèle de développement [Localised production systems in France: a new model of development], in Benko, G. and Lipietz, A. (eds), Les régions qui gagnent – Districts et réseaux: les nouveaux paradigmes de la géographie économique [The Winning Regions – Districts and Networks: The New Paradigms of Economic Geography]. Paris: Presses Universitaires de France, pp. 81–102. Crevoisier, O. (2000) Les milieux innovateurs et la ville [Urban innovative milieux: innovation, production systems and anchorage], in Crevoisier, O. and Camagni, R. (eds), Les milieux urbains: innovation, systèmes de production et ancrage [Urban Innovative Milieux: Innovation, Production Systems and Anchorage] Neuchâtel: GREMI, EDES, pp. 7–32. Crevoisier, O. (2004) The innovative milieu approach: Towards a territorialised understanding of the economy? Economic Geography 80 (4): 367–79. Dei Ottati, G. (1994) Co-operation and competition in the industrial district as an organisational model, European Planning Studies 2: 463–83. Dei Ottati, G. (1996) Trust, interlinking transactions and credit in the industrial districts, Cambridge Journal of Economics 18: 529–46. Dubin, R. (1978) Theory Development, New York: Free Press. Ellison, G. and Glaeser, E. L. (1994) Geographical Concentration in the US Manufacturing Industries: A Dartboard Approach. Working Paper no. 4840, Cambridge, Mass.: National Bureau of Economic Research (NBER). Feldmann, M. P. (1994) The Geography of Innovation, Dordrecht: Kluwer. Feser, E. J. (1999) Old and new theories of industrial clusters, in Steiner, M. (ed.) Clusters and Regional Specialisation: On Geography, Technology and Networks, pp. 19–40. London: Pion. Florida, R. (1995) Toward the learning region, Futures 27 (5), 527–36. Gagliardi, P. (1999) Theories empowering for action, Journal of Management Inquiry 8 (2): 143–7. Garofoli, G. (1992a) Industrial districts: structure and transformation, in Garofoli, G. (ed.) Endogenous Development and Southern Europe pp. 49–60. Aldershot: Avebury. Garofoli, G. (1992b) The Italian model of spatial development in the 1970s and 1980s, in Benko, G. and Dunford, M. (eds) Industrial Change and Regional Development, pp. 85–101. London: Bellhaven Press. Garofoli, G. (1993) Economic development, organization of production and territory, Revue d’Économie Industrielle 64 (2e trimestre 1993): 22–37. Gilly, J. P. and Torre, A. (eds) (2000), Dynamiques de proximité [Proximity Dynamics], Paris: L’Harmattan. Gottardi, G. (1996) Technology strategies, Innovation without R&D and the creation of knowledge within industrial districts, Journal of Industry Studies 3 (2): 119–34. Hanson, G. H. (2001) Scale economies and the geographic concentration of industry. Journal of Economic Geography 1 (1): 255–76. Harrison, B. (1992) Industrial districts. Old wine in new bottles? Regional Studies 26: 469–83. Hoover, E. M. (1971) An Introduction to Regional Economics, New York: Alfred A. Knopf. Kirat, T. and Lung, Y. (1999) Innovation and proximity: Territories as loci of collective learning processes, European Urban and Regional Studies 6 (1): 27–38. Klepper, S. (2002) The capabilities of new firms and the evolution of the US automobile industry. Industrial and Corporate Change 11 (4): 645–66.

What qualifies as a cluster theory? 47 Lengnick-Hall, C. A. and Wolff, J. A. (1999) Similarities and contradictions in the core logic of three strategy research streams, Strategic Management Journal 20 (12): 1109–32. Loasby, B. J. (1999) Industrial districts as knowledge communities, in Bellet, M. and L’Harmet, C. (eds) Industry, Space and Competition: The Contribution of Economists of the Past, pp. 70–85. Cheltenham: Edward Elgar. Loasby, B. J. (2000) Organisations as Interpretative Systems. Paper presented at the DRUID Summer Conference, Rebild, Denmark (www.druid.dk) Maillat, D. (1991) Local dynamism, milieu and innovative enterprises, in Brotchie, J., Batty, M., Hall, P. and Newton, P. (eds) Cities of the 21st Century, London: Longman. Maillat, D. and Perrin, J.-C. (eds) (1992) Entreprises innovatrices et développement territorial [Innovative Firms and Territorial Development], Neuchâtel: GREMI, EDES. Maillat, D. (1998) From the industrial district to the innovative milieu: Contribution to an analysis of territorialised productive organisations, Recherches Economiques de Louvain 64: 111–29. Maillat, D. and Kebir, L. (2001) The learning region and territorial production systems, in Johansson, B., Karlsson, C. and Stough, R. R. (eds), Theories of Endogenous Regional Growth, Lessons for Regional Policies, Berlin: Springer-Verlag, pp. 255–77. Maillat, D., Quévit, M. and Senn, L. (eds) (1993) Réseaux d’innovation et milieux innovateurs: un pari pour le développement régional [Innovative networks and innovative milieus: A stake for regional development], Neuchâtel: GREMI, EDES. Malecki, E. J. (1991) Technology and Economic Development: The Dynamics of Local, Regional and National Change, Harlow: Longman. Malmberg, A. (1996) Industrial geography: Agglomerations and local milieu, Progress in Human Geography 20: 392–403. Malmberg, A. (1997) Industrial geography: location and learning, Progress in Human Geography 21: 573–82. Malmberg, A. and Maskell, P. (1997) Towards an explanation of industry agglomeration and regional specialization, European Planning Studies 5 (1): 25–41. Malmberg, A. and Maskell, P. (2002) The elusive concept of localization economies: Towards a knowledge-based theory of spatial clustering, Environment and Planning A 34 (3): 429–49. Markusen, A., Hall, P. and Glasmeier, A. K. (1986) High Tech America, Boston: Allen and Unwin. Marshall, A. (1890) Principles of Economics, London: Macmillan. Marshall, A. (1919) Industry and Trade: A Study of Industrial Technique and Business Organization, and of Their Influences on the Condition of Various Classes and Nations, London: Macmillan. Martin, R. and Sunley, P (2003) Deconstructing clusters: chaotic concept or policy panacea? Journal of Economic Geography 3: 5–35. Maskell, P. (1998) Successful low-tech industries in high-cost environments: The case of the Danish furniture industry, European Urban and Regional Studies 5 (2): 99–118. Maskell, P. (2001a) ‘Knowledge creation and diffusion in geographic clusters: Regional development implications’, in Felsenstein, D., McQuaid, R., McCann, D. and Shefer, D. (eds), Public Investment and Regional Economic Development, pp. 59–76, London: Edward Elgar. Maskell, P. (2001b) Towards a knowledge-based theory of the geographical cluster, Industrial and Corporate Change 10 (4): 919–41. Maskell, P. and Malmberg, A. (1999) Localised learning and industrial competitiveness, Cambridge Journal of Economics 23 (2): 167–86.

48 Peter Maskell and Leïla Kebir Maskell, P., Eskelinen, H., Hannibalsson, I., Malmberg, A. and Vatne, E. (1998) Competitiveness, Localised Learning and Regional Development: Specialisation and Prosperity in Small Open Economies, London: Routledge. Matteaccioli, A. and Tabariés, M. (2002) Historique du GREMI [A History of the GREMI], URL: www.unine.ch/irer/gremi/accueil.htm. Morgan, K. (1997) The learning region: Institutions, innovation and regional renewal, Regional Studies 31 (5): 491–503. Moulaert, F. and Sekia, F. (2003) Territorial innovation models: A critical survey, Regional Studies 37 (3): 289–302. Norton, R. D. (1992) Agglomeration and competitiveness: From Marshall to Chinitz, Urban Studies 29 (2): 155–70. Piore, M. and Sabel, C. (1984) The Second Industrial Divide, New York: Basic Books. Porter, M. E. (1990a) The Competitive Advantage of Nations, London: Macmillan. Porter, M. E. (1990b) The competitive advantage of nations, Harvard Business Review, March–April: 73–93. Porter, M. E. (1994) The role of location in competition, Journal of the Economics of Business 1: 35–9. Porter, M. E. (1998) On Competition, Boston: Harvard Business School Publishing. Rallet, A. (2000) De la globalisation à la proximité géographique: pour un programme de recherches [From Globalisation to Geographical Proximity: For a Research Program], in Gilly, J. P. and Torre, A. (eds) (2000), Dynamiques de proximité [Proximity Dynamics], Paris: L’Harmattan, pp. 37–57. Ratti, R., Bramanti, A. and Gordon, R. (eds) (1997) The Dynamics of Innovative Regions: The GREMI Approach. Aldershot: Ashgate. Reich, R. (1990) But now we’re global, The Times Literary Supplement 31: 925–6. Richardson, G. B. (1972) The organisation of industry, Economic Journal 82: 883–96. Robinson, E. van Dyke (1908), Economic Geography: An Attempt to State What It Is and What It Is Not. Papers of the 21st meeting, American Economic Association, pp. 247–57. Saxenian, A. (1994) Regional Advantage: Culture and Competition in Silicon Valley and Route 128, Cambridge, Mass.: Harvard University Press. Simmie, J. (ed.) (1997) Innovation, Networks and Learning Regions? London: Jessica Kinsley Publisher. Stenberg, R. (1999) Innovative linkages and proximity: Empirical results from recent surveys of small and medium sized firms in German regions, Regional Studies 33: 529–40. Storper, M. (1995) The resurgence of regional economies, ten years later: the region as a nexus of untraded interdependencies, European Urban and Regional Studies 3 (2): 191–221. Sutton, R. I. and Staw, B. M. (1995) What theory is not, Administrative Science Quarterly 40 (3): 371–84). Comments by Karl Weick and by Paul DiMaggio pp. 385–97. Swann, P. G. M., Prevezer, M. and Stout, D. (1998) The Dynamics of Industrial Clustering: International Comparisons in Computing and Biotechnology, Oxford: Oxford University Press. Weber, A. (1909) Über den Standort der Industrien. Tübingen: J. C. B. Mohr. Translated by Carl Joachim Freidrich and published 1929 under the title: Theory of the location of industries, Chicago: The University of Chicago Press.

What qualifies as a cluster theory? 49 Whetten, D. A. (1989) What constitutes a theoretical contribution? Academy of Management Review 14 (4): 490–5. Yeung, H. W. (2000) Organising ‘the firm’ in industrial geography I: Networks, institutions and regional development, Progress in Human Geography 24. Young, A. (1928) Increasing returns and economic progress, Economic Journal 38: 527–42.

3

True clusters A severe case of conceptual headache Anders Malmberg and Dominic Power

Cluster n. A group of the same or similar elements gathered or occurring closely together; a bunch. (The American Heritage Dictionary of the English Language, third edition. Boston and New York: Houghton Mifflin Company, 1992) Clusters are geographic concentrations of interconnected companies, specialised suppliers, service producers, firms in related industries, and associated institutions (for example, universities, standard agencies, and trade associations) in particular fields that compete but also cooperate. Critical masses of unusual competitive success in particular business areas, clusters are a striking feature of virtually every national, state, and even metropolitan economy, especially those of more economically advanced nations. (Porter 1998a, 197f) Cluster headache n. A severe recurring headache . . . characterized by sudden sharp pain, watering of the eye, and runny nose on one side of the head. (The American Heritage Dictionary of the English Language, third edition. Boston and New York: Houghton Mifflin Company, 1992)

Introduction The cluster approach has, since its appearance in academic and policy scenes in the early 1990s, had an enormous impact. As an analytical approach it is undoubtedly persuasive and has contributed to substantial progress in the analysis of several of the classical issues dealt with by economic geographers. At the same time it is an elusive, and at times confusing, concept open to multiple interpretations and understandings. So whilst it has been a powerful rallying call and focus point for debate on issues of regional competitiveness and adjustment, the cluster approach has equally caused ‘recurring headaches’ for many of us active in the field of cluster research – and cluster-based industrial, regional or innovation policy formulation.

True clusters 51 The aim of this chapter is to analyse why this is the case and, it is hoped, to contribute to a somewhat clearer idea of what the cluster concept can and can not do for us.1 Our point of departure is ambivalent. On the one hand, we believe that the cluster movement has in recent years meant a lot to the revitalization of research in economic geography (broadly defined) and to progressive reformulation of agendas in regional and industrial policy. On the other hand, one must wonder if something has not gone seriously wrong along the way. The cluster concept and the associated approaches or models it has given rise to have arguably come to embrace and stand for too much, such that now it has become increasingly unclear what they represent and what they can help us achieve. We start the chapter with a discussion of what we see as the main contributions of the cluster approach. Following this we suggest that despite the important contributions the concept has made it has also been dogged by considerable conceptual confusion. We discuss the origins and dimensions of this conceptual confusion and go on to suggest that the core problem is that the concept has been elevated to the status of an ideal type: a persuasive theoretical construct but one ill-suited to empirical investigation and policy formulation. This development has, we argue, helped to side-track empirical research and has led to the domination of a series of implicit assumptions that seem to guide empirical research. These assumptions are hard to verify from the empirical work available and we suggest research needs to rethink the hypotheses it is working from. We conclude by arguing for greater degrees of conceptual flexibility and suggest that if the cluster debate is to move forward we need to sidestep cluster puritanism and concentrate our efforts on understanding why clusters have a role in knowledge and innovation.

The contribution of the cluster approach Some basic points of departure first. In a knowledge-based economy, the ability to innovate is more important than cost efficiency in determining the long-term ability of firms – and regions – to prosper. Innovations, defined broadly, occur predominantly as a result of interactions between various actors, rather than resulting from the creative act of a single individual or firm (Håkansson 1987; Von Hippel 1988; Lundvall 1992). From this, it follows that the level of analysis for understanding the processes of industrial innovation and change should involve some notion of an industrial system or of a network of actors interacting while carrying out similar and related economic activity. There are a number of reasons why interactive learning and innovation processes are not aspatial or universal, but on the contrary unfold in such ways that geographical space plays an active role. Territorially delimited institutional and cultural traits impact upon the direction and speed of innovation processes. Spatial proximity carries with it, among other things, the potential for intensified face-to-face interaction, short cognitive distance, common language, trustful relations between various actors, easy observation, and immediate comparison (Malmberg and Maskell 2002; Storper and Venables 2002). In short, spatial

52 Anders Malmberg and Dominic Power proximity seems to enhance processes of interactive learning and innovation, and therefore industrial systems should be assumed to have a distinctly localized component. The cluster concept, and associated models and lines of argument, offer a neat response to all the above assertions: the cluster promises to produce innovation and competitiveness via a series of interactive processes within systems of actors assembled in a milieu defined through some form of spatial proximity. Phrased in such a way it is perhaps little surprise that the approach as presented by Porter (1990) and subsequently developed by himself, his associates and others (Porter 1994, 1998a, 2000; Malmberg et al. 1996; Enright 1998; Malmberg and Maskell 2002) has caught the imagination of social scientists and brought some genuine contributions to the analysis of key issues of economic geography. The cluster approach provides a way to describe the systemic nature of an economy: i.e. how various types of industrial activity are related. Beginning with the firms in the industry where we find the main producers of the primary goods, the cluster also embraces supplier firms and industries providing various types of specialized inputs, technology, machinery and associated services, as well as certain important customers and more indirectly related industries. There is much to be said in favour of this way of approaching the systemic nature of economic activity. It opens up a scope for analysing interactions and interdependencies between firms and industries across a wide spectrum of economic activity. In addition it contributes to the bridging of a number of more or less artificial and chaotic conceptual divides that characterize so much work in economic geography and related disciplines. These include, for example, manufacturing versus services, high-tech versus low-tech, large companies versus SMEs, public versus private activities, etc. A single cluster, defined as a functional industrial system, may embrace firms, actors and activities on both sides of these divides (see also Dicken and Malmberg 2001). Furthermore, Porter’s model of the determinants of competitiveness in clusters, the ‘diamond model’, identifies a number of mechanisms proposed to foster industrial dynamism, innovations and long-term growth. Essentially, the model is built around four sets of intertwined forces related to (1) factor conditions; (2) demand conditions; (3) related and supporting industries; (4) and firm structure, strategy and rivalry. The treatment of these factors includes several points that are indeed novel. First, in relation to factor conditions the emphasis on the role of specialized factors and factor-upgrading redirects our focus from the very general classical notions of the availability and cost of capital, labour and land towards a much more nuanced understanding that stresses the type of specialized factor conditions – smart money, specialized skills, dedicated and advanced infrastructures – which are developed historically to fit the needs of a particular economic activity. These are important location factors since they are difficult to move and difficult to imitate in other regions (cf. Maskell et al. 1998). Another, perhaps more original, idea is that of the roles of selective factor disadvantages in promoting dynamism and long-term growth: a regional economist’s version

True clusters 53 of the old idea that ‘necessity is the mother of invention’. Arguably, no previous account in economic geography and related fields has so explicitly made the point that shortcomings in factor conditions (such as labour shortages and high wages, scarce natural resources, expensive electricity, strict public regulations, etc.) can actually trigger technological and institutional innovations that will in the longer term be more important contributors to the competitive success of firms in specific places. Second, the treatment of the demand side as a primarily qualitative factor is original. Most previous models have emphasized access to a large market as an important locational advantage. Porter’s account, in contrast, alerts us to the fact that it is the sophistication of demand that matters if we are interested in innovation and long-term competitiveness. According to this view, the locationally advantaged firm is the one that is in a position to receive and react to signals from sophisticated demand, rather than simply the one that is blessed with ‘many customers’ in the local market. This idea is also present in other recent approaches to the dynamics of industrial systems; for instance, in Eliasson’s (2000) notion of the competence bloc the ‘competent customer’ plays a key role (Malmberg and Power 2005a). Third, the importance of local rivalry is made much more explicit than in previous models of spatial agglomeration. That a firm may gain advantages from being located close to other firms in the same industry is, of course, a key insight in classical agglomeration theory. Rarely, though, has this advantage been attributed to the fact that spatial proximity between rivals will trigger dynamism and growth. The idea here is that local rivalry adds an intensity and emotional dimension to competition that can be harder for actors to perceive in dispersed global markets. The firm down the road is often seen as the ‘prime enemy’, a bit like the rivalry between neighbouring football clubs. Firms in a local milieu tend to develop relations of rivalry, where benchmarking in relation to the neighbours is more direct, partly for reasons of local prestige, and partly, presumably, simply because direct comparison is simplified (it is much easier to see if your neighbour has a better car than you). One could speculate that there are several reasons for the latter. It is easier to monitor the performance of a neighbouring firm than a competitor far away. In addition, if one firm displays superior performance, it is obvious to everyone that this cannot be ‘blamed’ on different external conditions, since these are, in principle, identical for all firms in the local milieu (cf. Malmberg and Maskell 2002). Furthermore, as numerous sociologists and social psychologists from Thorstein Veblen to today have pointed out, self-esteem, personal comparisons, competitiveness and locally accrued social status are powerful human emotions and motivators. On these points, at least, it should be acknowledged that the cluster approach has contributed to genuine progress: the role of specialized production factors and selective factor disadvantages, sophisticated customer demand, and local rivalry are novel and innovative proposals that have enriched our understanding of why conditions in a local milieu in general, and agglomerations of similar and related firms in particular, might promote superior firm performance.

54 Anders Malmberg and Dominic Power

Porter’s contribution to conceptual confusion Arguments regarding the persuasiveness and competitiveness of clusters have become widely circulated in academic as well as in policy circles since the early 1990s. In a recent paper, Martin and Sunley (2003) scrutinize the cluster concept and the broader ‘cluster trend’ in economic geography and related disciplines and advance a number of more or less justified points of critique. Indeed there is a growing concern that there is a good deal of fuzziness surrounding the cluster concept (Markusen 1999; Martin and Sunley 2003). In our view, the really disturbing lack of clarity is at the most basic level: what is meant by the terms ‘cluster’ and ‘clustering’? This seemingly trivial question is causing continuing and increasing problems. We are not thinking here of subtle definitional issues relating to the scales, boundaries and criteria for the identification of clusters. Rather, we think that the main confusion is related to whether clusters and clustering should be seen to be primarily functional or spatial phenomena. On this particular issue, Porter himself has contributed to the conceptual mess by presenting quite different basic definitions in various texts since 1990. Compare, for example, the following quotations. In the original cluster account, Porter writes: The competitive industries in a nation will not be evenly distributed across the economy . . . A nation’s successful industries are usually linked through vertical (buyer/supplier) or horizontal (common customers, technology, channels, etc.) relationships . . . The reasons for clustering grow directly out of the determinants of national advantage and are a manifestation of their systemic character. One competitive industry helps to create another in a mutually reinforcing process. (Porter 1990, pp. 148–9, emphasis added) It is only after saying that clusters are sets of functionally interrelated industries (within the spatial context of a nation) that Porter goes on to discuss spatial aspects: Geographic concentration of firms in internationally successful industries often occurs because the influence of the individual determinants in the ‘diamond’ and their mutual reinforcement are heightened by close geographic proximity within a nation. (Porter 1990, pp. 156–7, emphasis added) Thus in the 1990 book, it is obvious that Porter regarded clusters as functionally related industries, while at the same time observing that such functional clusters ‘often’ seemed to be prone to ‘geographic concentration’ since spatial proximity amplifies the mechanisms that make clusters of industries dynamic and innovative. Then, throughout the 1990s, Porter adopted a view according to which geographic concentration gradually became an integral part of the definition of the cluster. Thus, in a recent paper, Porter (2000) writes:

True clusters 55 A cluster is a geographically proximate group of interconnected companies and associated institutions in a particular field, linked by commonalities and complementarities. (Porter 2000, p. 254, emphasis added) Now it seems clusters are defined by geographical proximity; even though the precise scale of this geographic concentration is left to the imagination. This gradual slide in the definition of the cluster concept is unfortunate and, we think, a main source of confusion. Indeed it would be practical if we could collectively strive to establish a terminology that is as free as possible from basic confusion. It is deeply unsatisfactory to develop a scholarly conversation around a core concept the meaning of which various participants in the conversation have different opinions about; not just differences over details but differences at the level of basic definitions. We certainly need one concept that brings out the idea of functionally linked economic activities. ‘Industrial system’ would seem to be an appropriate generic term, but since the ‘C-word’ is presumably going to be around for a while, ‘industry cluster’ could be a useful alternative. When, on the other hand, we face geographical concentrations of similar or related economic activity, we could preferably use the traditional term ‘agglomeration’, or possibly ‘spatial (or localized) cluster’, in order to avoid some of the confusion. This is more than a question of terminology for it seems to us that there needs to be some flexibility built into the way we use, define, delimit and debate the cluster concept in spatial settings (i.e. when we look to the real world). Perhaps a key issue here, to which we will return later in the chapter, rests on the observation that it seems obvious that (functional) industry clusters will not normally be confined to, or contained within, any narrowly defined and spatially bounded scale (Malmberg and Power 2005a). On the contrary, most industry clusters will have widespread global connections and, if we were able to identify their boundaries in spatial terms, the spatial scale would in most cases certainly not be just an urban region. For instance, dynamic and innovative high-tech firms (for instance the pharmaceuticals giants) will most likely look to find the best technological and scientific partners irrespective of where they are located. By making spatial configuration (i.e. the degree of agglomeration) an attribute of an industrial cluster, rather than part of its definition, one could establish a platform for more fruitful analyses of how ‘geography’ comes into play in the overall process of industrial competitiveness, growth and transformation. In other words, rather than trying to squeeze ‘cluster charts’ into narrowly defined regions (where they rarely will fit in), we should research hypotheses such as those found in the diamond model regarding the role of proximity and local milieus in the proposed mechanisms leading to competitiveness. When it comes to spatial agglomerations of similar and related economic activities, i.e. localized clusters in the terminology proposed here, there are also reasons to believe that firms in such settings are less interrelated than Porter and others have led us to believe (Malmberg and Power 2005b).


Clusters: strictly defined ideal types? What we have just described leads on to a more general set of problems and questions. Should the cluster concept be very strictly defined, and in consequence be applied very selectively to deal with a limited number of exclusive real-world cases? Or should we adopt a more flexible stance where the precise definition of the cluster concept is left more fluid, such that the concept can be used to research – and act upon – a much larger number of real world cases (cases where only one or a few ‘cluster characteristics’ are in evidence)? The way the cluster concept tends to be used today indicates that there are at least four different dimensions or defining criteria that should be present for a true, fully fledged, cluster to be said to exist. The first two were discussed in the previous section: those of spatial proximity and functional inter-linkage. Thus, according to the spatial agglomeration criterion, a cluster is defined as a geographical concentration of similar and related economic activity. This criterion brings two major problems. The first is, as we have seen, that extreme flexibility prevails when it comes to determining what is meant by geographical concentration: are we talking about an industrial estate, a city, a region or even a nation? In principle, we are left free to make our own judgements about the balance between space and systems, where we draw the lines and what we include.2 The second criterion of the cluster concept in action is, again, the idea of the cluster as a functionally defined industrial system. The point here is that a cluster is not limited to an individual industry but embraces all the actors, resources and activities that come together to develop, produce and market various types of goods and services. One problem with this is that there are no theoretical reasons to argue that such system should be defined or delimited in either a narrower or a broader sense. The context of the analysis or the policy action determines whether it is more appropriate to focus on, for instance, a broader automotive cluster or instead to choose more narrowly defined clusters that make only trucks, buses, private cars or even particular components. Another problem relates to how much spatially agglomerated activities should compete and/or collaborate with each other in order to be conceptualized as related. In practice the spatial extension of most functional systems is much lager than what we normally think of as functional regions (daily urban regions, local labour market regions, etc.). As the cluster approach has become increasingly popular as a policy tool and found itself being adapted to practical purposes a third criterion of what a cluster is has become prominent in both policy initiatives and academic research. This criterion is based on the existence and links between identity, self-awareness and policy action. According to some observers, the institutionalization of some common idea or purpose is a necessary ingredient of a true cluster. For a cluster to be said to exist, some actor (often employed by a public institution rather than a private company) has to identify it as a cluster, whether existing or ‘dormant’ (or ‘potential’, or ‘emerging’), give it a name (preferably one that refers

True clusters 57 to a low lying area surrounded by hills: ‘XXXX-Valley’), and start acting in order to consciously develop it (Rosenfeld 1997; Raines 2001; Lundequist and Power 2002). Thus, in policy circles clusters have become more or less synonymous with the existence of a policy programme and a number of more or less concerted policy actions. This could be seen as a discursive definition of the cluster concept where a cluster has come to refer to a specific policy initiative. Such clusters might or might not have a resemblance with the functional and geographical dimensions already discussed. In our view, cluster-based policy programmes could preferably be referred to simply as cluster initiatives; as indeed more policy-oriented work is already doing (see for instance Sölvell et al. 2003). A fourth criterion of the cluster concept is one in which the cluster idea becomes synonymous with competitive success. This is the idea that the cluster is not just a system or a geographical concentration but one that is also dynamic, innovative and competitive – doing things that ‘distant rivals cannot match’ (Porter 1998b). In this view the cluster is always a success story and in essence an end-state. It is a concept that describes and elevates particular states of achievement. As such it is a concept that is not entirely applicable to those who wish to describe or apply generalized developmental trajectories or processes. How then should we regard this situation where there are obviously (at least) four sets of meanings attached to the cluster notion? One refers to functional systems and interaction, another to spatial agglomeration, one to self-identity and policy action, and another to already proven successful examples. Porter himself, as we saw, seems to believe that the first two more or less coincide and could thus be treated as one. In relation to the third, in his view the issue of self-awareness and policy action is subordinate. He states that government agencies that significantly influence a cluster can be considered part of it, but it is obvious that, at least in his earlier writings, there could well be dynamic clusters without such agencies. In the case of the fourth criterion he also seems very open to thinking that success is an essential dimension of a true cluster. If one takes cluster theory seriously, then a strict definition of a true cluster is thus based on the criteria that: • • •

•

There should be a spatial agglomeration of similar and related economic activity. These activities should be interlinked by relations and interactions of local collaboration and competition. There should be some form of self-awareness among the cluster participants and some joint policy action (‘we are a cluster and we are determined to develop together’). The cluster should be, in one way or another, successful (innovative, competitive).

As we mentioned above, it is often hard to tell in the existing literature whether the use of the cluster concept refers to any of the above four criteria or indeed

58 Anders Malmberg and Dominic Power some imagined ‘ideal type’ where all, or at least the first three, coincide. As we understand the debate to date, what has emerged is a series of implicit assumptions and models that have in essence lifted the idea of the cluster to the level of an ‘ideal type’. Max Weber (1947, 1949) suggested that the ideal type was a part of researchers’ attempts to wrestle with the problem of the ambiguities presented to us by the empirical world: an attempt which always involves the imposition of order, the emphasizing and perhaps exaggeration, and even elimination, of certain elements of the reality presented to us. The construction of an ideal type is an attempt to arrive at a unified definition but for those interested in direct empirical description it is inherently problematic since it is essentially the construction of a gross stylization (Aron 1970). Indeed Weber himself viewed the ideal type as a heuristic device, a mental concept, that whilst conceptually pure ‘cannot be found empirically anywhere in reality’ (Weber 1949; Weber 1947: 110). It appears to us that cluster thinking has become preoccupied with the development of pure ideal types. The problem introduced above is that the more strictly we define clusters, the smaller the number of real world cases that conform to the definition. In a strict, or puritan, view of clusters then we should find a mixture of certain degrees of all four criteria contained in the ideal type, and definitely all of the first three, before a cluster can be said to exist. Indeed despite a current abundance of clusters and cluster initiatives there are few places in the world (at whatever spatial scale one looks) where we can find such ‘true’ or ‘real’ clusters. Porter’s often cited claim that ‘clusters are a striking feature of virtually every national, state, and even metropolitan economy, especially those of more economically advanced nations’ (1998a: 198) becomes increasingly doubtful when stricter definitions of a cluster are applied. The introduction of the success criterion as part of the definition is problematic for another reason. In much cluster research, there is an implicit model that looks something like: Competitive Success = f(Interaction, Agglomeration, Policy institutions) This too carries problems with it. The first, again, is that of exclusiveness. How, for instance, do we deal with cases where everything seems to be in place – an agglomeration, an interactive system and a policy framework – but there is still no innovativeness or the core product is simply no longer competitive? Is this a cluster? Alternatively how do we deal with a case where there are the beginnings of an agglomeration, the initial bubbling of a creative milieu and a well organized, high-profile cluster organization but as yet nothing much more than lots of investment in a promise and a dream? Is this a cluster? But the more serious problem is that the model, when combined with the ideal type definition, leads to circular reasoning: clustering is claimed to lead to competitiveness, while at the same time clusters are partly defined on the basis of their competitive success.

True clusters 59

Researching clustering and knowledge creation? The above has implications for how economic geographers and others should approach the clustering issue. A preoccupation with ideal type reasoning on the cluster concept has contributed to sidetracking empirical research on clustering. The introduction of the cluster concept could have triggered lots of research on the fruitful issue of how industrial transformation occurs as a result of interactions within and across industrial systems (i.e. clusters defined in the functional sense) and the role of spatial proximity (concentration or agglomeration, i.e. clustering in the spatial sense) in such processes. Instead, we would argue, there has been far too much focus on interaction between firms within geographically defined spaces and numerous rather pointless attempts at trying to assess to what degree there is actual interaction going on locally and thus whether a specific region can indeed be said to contain a ‘fully fledged’ or ‘true’ cluster or not (Martin and Sunley 2003). We are at the stage then when there is a lot of confusion about what the concept actually involves, with the effect that research (and policy) has become far too based on a number of ideal types and criteria that may not offer us the most solid conceptual basis for scholarly conversations and real-world interventions. One possible solution to the resulting conceptual and empirical patchwork may be to go back to the underlying theoretical assumptions upon which the various cluster approaches seem to rest. After having spent recent years immersed in cluster literature, conferences and case studies, it seems to us that the cluster concept is essentially a concept attempting to reconcile the fact that most economic activities (and workers) occur in localized clumps with the reality that the products and services they produce (whether they are new or rather old offerings) have to find a market within a globalized knowledge economy. In our view, the cluster approaches’ popularity rests on a deeper understanding of a more macro-economic nature: that the leading-edge competitive forces dominant in the Fordist period – mass production, cost cutting, price competition, and product standardization – have given way, at least in the rich countries, to a stress on how added-value can be created through harnessing the knowledge, flexibility, adaptability and innovativeness of our firms and populations. If we are right in thinking this way, then the cluster approach is less about maintaining competitiveness through collective control of resources and agglomeration economies and rather more about findings ways in which knowledge and innovation can be given a supportive environment. Going back to basics means trying to understand the cluster movement less as a tool for developing regional competitiveness and rather more as a conceptual framework for analysing the fundamental dynamics of knowledge creation and innovation in industrial settings (which we all roughly agree is the ground upon which competitiveness grows). In the rest of this chapter we will attempt to work with this understanding of clusters and clustering and suggest that to get at the core of the cluster approach we could do a lot worse than to look at how clusters have been said to create knowledge and learning at a ‘local’ level.

60 Anders Malmberg and Dominic Power Clusters, localization and knowledge creation: three received hypotheses Drawing on extensive literature reviews we have undertaken (the results of which also appear in Malmberg and Power 2005b), we suggest here that it is possible to identify certain broad areas of agreement in the literature – about how knowledge is created in clusters – that could be seen as basic underlying hypotheses driving current research. The following three hypothetical propositions are, we argue, those that underpin the majority of cluster research we are aware of: (1a) Knowledge in clusters is created through various forms of local inter-organizational collaborative interaction. This hypothesis is grounded in the proposal that firms that collaborate more on technology with other firms and actors (e.g. universities) in the local milieu will innovate more, and in the idea that firms that meet sophisticated demand from demanding customers in the local milieu will be forced to innovate at a higher pace than other firms. (2a) Knowledge in clusters is created through increased competition and intensified rivalry. The claim here is that rivalry between similar firms in a local milieu will be more intense, almost emotional, and this will create a pressure to innovate in order to outsmart the local rival. In part, this is related to the fact that firms in a localized cluster are more visible to each other, and thus that observation, monitoring and benchmarking are easier and more efficient. Therefore, firms with nearby rivals will be more innovative than firms that have their main competitors located elsewhere. (3a) Knowledge in clusters is created through spill-over effects following from the local mobility and sociability of individuals. This hypothesis is based on the idea that knowledge diffusion will be more rapid among local firms than among globally dispersed firms, owing to the intensity of informal interaction in the local milieu as well as through flows of people in the local labour market. Here the cluster is seen to rely on underlying localized factor conditions. In particular, there have recently been a number of studies that propose a version of the cluster concept that stresses the centrality of local labour market processes to the innovative capacity, competitiveness and indeed existence of clusters. It is the dynamism of the local labour market that is held to account for the associated clusters’ dynamism. These, we would argue, are all interesting and researchable hypotheses that could be deduced from the cluster literature, based on the underlying argument that the forces that enhance the dynamism of an industry cluster are strengthened by geographical proximity, via a series of mechanisms. Clusters, localization and knowledge creation: empirical evaluation However, the empirical validation of the propositions advanced in the cluster literature leaves a lot to be desired. This is partly due to the fact that there has been a general reluctance to spell out the theoretical propositions made in a form that would make it possible to subject them to systematic empirical

True clusters 61 validation. After a systematic review of the empirical literature on clusters and clustering we found that the empirical basis for the above three hypotheses is such that the three could be recast as follows: (1b) Knowledge in clusters is seldom created through local inter-organizational collaborative interaction. A distinctly mixed picture emerges from the literature on inter-firm transactions (such as buyer–supplier relations, etc.) and inter-firm collaboration (such as joint product development, etc.). It appears that intense collaborative interaction with similar and related firms in the localized cluster does not come out as a major knowledge creating mechanism (Angel and Engstrom 1995; Hendry et al. 2000). In addition, though examples do exist, they are few, and there tend to be modest commercial relations between firms within spatial clusters. Furthermore in a localized cluster the majority of firms tend to have most of their important suppliers and customers somewhere else (Larsson and Lundmark 1991; Angel and Engstrom 1995; Larsson 1998; Markgren 2001) and innovation and knowledge creation tend to follow value chains that are most often global (Fuellhart 1999; Zeller 2001; Owen-Smith and Powell 2002; MacKinnon et al. 2003). University–industry collaborative links do exist in some places ( Jaffe 1989; Jaffe et al. 1993; Anselin et al. 1997; Narin et al. 1997; Zucker et al. 1998; Howells 2002; Rodríguez-Pose and Refolo 2003) but they tend to be more appropriate to some industries than others and are in general much more global than local. Finally whilst other, more informal, types of collaboration are more common locally and temporally (Keeble et al. 1999; Wallsten 2001; De Propris 2002; Isaksen 2003) it appears that such relations also normally extend far beyond the confines of narrowly defined regions. (2b) Knowledge in clusters may well be created through increased competition and intensified rivalry, though we are not sure yet. If Porter is correct in thinking that local rivalry is crucial to motivating and driving knowledge creation and innovation (Porter 1990, 1998a, 2001) then it is surprising that the extent to which local rivalry occurs and the effect it has have not been well studied empirically. Some evidence does exist (Sakakibara and Porter 2001; Power and Hallencreutz 2002; Boari et al. 2003; King et al. 2003) but the extent to which local rivalry effects knowledge creation in this limited evidence varies. Furthermore there seem to be many firms and sectors that see themselves as relatively isolated from competitors, in that they have very few and these are often located very far away (Glaeser et al. 1992; Audretsch and Feldman 1996; Baptista and Swann 1996, 1998; Malmberg et al. 2000). While it is too early to dismiss this hypothesis, we would not at this stage propose that rivalry is a more important booster of knowledge creation than various forms of collaboration. (3b) The creation of knowledge in clusters is probably helped by spillover effects following from the sociability of individuals and almost certainly by labour mobility. What we found in the studies we reviewed is that informal knowledge exchanges do seem to occur across groups of professionals and specialized individuals in clusters (Thrift and Leyshon 1994; Coe 2000; Bennett et al. 2001; Lissoni 2001; Grabher 2002; Benner 2003; Welz 2003). There is also mounting evidence that local

62 Anders Malmberg and Dominic Power labour mobility plays an important role in rates of innovation and that localized clusters that are relatively successful tend to have higher rates of labour mobility into the cluster, within the cluster and within the cluster firms (Angel 1991; Almeida and Kogut 1999; Gilson 1999; Breschi and Lissoni 2001; Cooper 2001; Fosfuri et al. 2001; Dahl 2002; Dahl and Pedersen 2003; Lewis and Yao 2003; Madsen et al. 2003; Rosenkopf and Almeida 2003; Song et al. 2003; Power and Lundmark 2004). Of course what is good for the cluster’s overall knowledge creation and spreading might not be good for all firms, and there is evidence suggesting that firms in clusters with high labour mobility view such ‘dynamism’ as a considerable cost (Almeida and Kogut 1999; Lawson 1999; Dahl 2002) and even a potential threat to the trade secrets they hope to commercialize (Ronde 2001; Fosfuri and Ronde 2004). To summarize very broadly, the available evidence suggests to us that, if we are interested in knowledge creation and knowledge-based innovation, localized clusters seldom appear to be the localized systems of interrelated firms bound together by tightly knit organized inter-firm transactions and collaborations that many academics and policy-makers seem to want them to be. Perhaps the image of cosy collaborations and friendly groups of scientists developing wonderful products after a short drive from home is fatally flawed. The evidence suggests that, for instance, these scientists might be better off driving to the nearest airport than the local business park (cf. Bathelt et al. 2004), and that if they are staying home they are more likely to be innovative if they are enviously keeping watch on their competitors’ achievements than if they are collaborating with them. On the basis of the evidence it seems that localized clusters are perhaps best understood as sites of informal social interaction and as arenas for flexible and well-functioning markets for specialized and skilled labour. In short, there seems to be little evidence that organized inter-firm transactions and co-operation characterizes successful firms. At the same time, there is growing evidence that labour market dynamics and social interaction at the level of the individual can play important roles in firms’ and clusters’ knowledge creation processes.

Conclusion and implications In this chapter we have suggested that the state of play in cluster research is currently rather unsatisfying. We have a situation where considerable conceptual confusion reigns and this confusion presents academics and practitioners with real ‘cluster headaches’. In particular, we have pointed to the idea that the cluster literature has become far too concerned with conceptual puritanism that aims to identify certain attributes and qualities that earmark a set of industrial activities as a ‘cluster’. Indeed we have suggested that an implicit set of ideal types and an implicit model of cluster competitiveness have emerged to dominate many aspects of the cluster debate. However, for those of us concerned with using what is in essence a very interesting and positive concept to help us better understand the economic world and understand how policies can support it, the current situation presents us with a real set of problems.

True clusters 63 Not least of these problems is that we in academia have developed a conceptual discourse that is increasingly poorly attuned to understanding the objects of study – actual ‘clusters’ – and to helping policy interventions that are rapidly shaping it into a patchwork of clusters. By concentrating on conceptual purity and modelling ideal types we are faced with the fact that very few true clusters exist. This presents us with some problems: what do we do with the rest? What do we do with those clusters that do not fully conform to the ideal type but seem to be there anyway? What do we do if we see something dynamic in a certain place but find out that it is not a system, an agglomeration or institutionalized? We argue here that there seems to be an expedient value in being less categorical and in realizing that there is a gap between the realities of competitiveness on the ground and the models of competitiveness that the regional development and economics literature put forward. We should then be realistic and study competitiveness everywhere it occurs and not get too stuck on the definitional and categorical details. Getting stuck in the modelling and categorizations usually, in our limited experience, involves recourse to policy that simply points out what is lacking from the ‘cluster’ and tries to work out where that could be inserted; rather than building on competitive strengths or thinking about whether the dynamics of the case in question are best served by less severe cluster bandages. We might also argue that another possible solution to the problems arising from the existence of few true localized clusters is that we begin to be less focused on the role of the local in competitiveness and innovation studies and instead focus more energy on the links between the extra-local and the local in competitiveness and innovation, which after all usually involves being competitive or innovative in markets that are not in the same place as the ‘cluster’. Thus we should be careful not to fetishize ideal types and the local in our attempts to understand and support regional development. Finally, whilst we have suggested there is a need for more conceptual flexibility and that the debate has become somewhat stilted, we do believe that, by refocusing on certain essential elements, rigorous and analytically challenging debate and progress are possible. We argue that a sensible way forward is to take a step back from the cluster puritans’ attempts to delimit the concept strictly and instead go back to basics and start asking questions that are based on the underlying reason why cluster approaches have in fact become popular. We think that this involves taking seriously the idea that what we are really interested in is how new knowledge (and innovation and learning) come about in localized clusters. However, if we review available empirical studies of clusters we find that certain assumptions have driven much research and that the evidence for these findings is often somewhat mixed and limited. This suggests to us that it is high time that cluster research rethinks its underlying assumptions and tries to move forward on the basis of new questions and hypotheses; albeit ones that are based on the same fundamental search for answers to questions about knowledge creation that gave the cluster approach its appeal in the first place.


Notes 1

2

This chapter builds upon and develops arguments presented in a series of papers, co-written by the authors in various constellations, over the last few years, notably Dicken and Malmberg 2001; Malmberg 2003; Malmberg and Maskell 2002; Malmberg and Power 2005a, 2005b; and Lundequist and Power 2002. In practice, there are of course conventions according to which we normally tend to adopt some notion of a functionally defined (‘daily urban’) region as the basis for defining the spatial boundries of localized cluster.

References Almeida, P. and B. Kogut (1999). ‘Localization of knowledge and the mobility of engineers in regional networks.’ Management Science 45 (7): 905–17. Angel, D. (1991). ‘High-technology agglomeration and the labor market: The case of Silicon Valley.’ Environment-and-Planning A 23 (10): 1501–16. Angel, D. P. and J. Engstrom (1995). ‘Manufacturing systems and technological change: The US personal computer industry.’ Economic Geography 71 (1): 79–102. Anselin, L., A. Varga and Z. Acs (1997). ‘Local geographic spillovers between university research and high technology innovations.’ Journal of Urban Economics 42: 422–48. Aron, R. (1970). ‘The Logic of the Social Sciences’ in Max Weber, ed. D. Wrong. Upper Saddle River, NJ, Prentice Hall. Audretsch, D. B. and M. P. Feldman (1996). ‘Innovative clusters and the industry life cycle.’ Review of Industrial Organization 11 (2): 253–73. Baptista, R. and P. Swann (1996). The Dynamics of Industrial Clusters: A Comparative Study of the US and UK Computer Industries. Working Paper 165. London, London Business School, Centre for Business Strategy. Baptista, R. and P. Swann (1998). ‘Do firms in clusters innovate more?’ Research Policy 27: 527–42. Bathelt, H., A. Malmberg and P. Maskell (2004) ‘Clusters and knowledge: Local buzz, global pipelines and the process of knowledge creation.’ Progress in Human Geography 28 (1): 31–56. Benner, C. (2003). ‘Learning communities in a learning region: The soft infrastructure of cross-firm learning networks in Silicon Valley.’ Environment and Planning A 35 (10): 1809–30. Bennett, R., P. Robson and W. Bratton (2001). ‘The influence of location on the use by SMEs of external advice and collaboration.’ Urban Studies 38 (9): 1531–57. Boari, C., V. Odorici and M. Zamarian (2003). ‘Clusters and rivalry: does localization really matter?’ Scandinavian Journal of Management 19 (4): 467–89. Breschi, S. and F. Lissoni (2001). ‘Localised knowledge spillovers vs. innovative milieux: Knowledge ‘tacitness’ reconsidered.’ Papers in Regional Science 80 (3): 255–73. Coe, N. (2000). ‘The view from out West: embeddedness, inter-personal relations and the development of an indigenous film industry in Vancouver.’ Geoforum 31 (4): 391–407. Cooper, D. (2001). ‘Innovation and reciprocal externalities: information transmission via job mobility.’ Journal of Economic Behavior and Organization 45 (4): 403–25. Dahl, M. (2002). Embedded Knowledge Flows through Labor Mobility in Regional Clusters in Denmark. DRUID’s New Economy Conference, Copenhagen. Dahl, M. and C. Pedersen (2003). Knowledge Flows through Informal Contacts in Industrial Clusters: Myths or Realities? DRUID Working Paper 03–01. Copenhagen, DRUID.

True clusters 65 De Propris, L. (2002). ‘Types of innovation and inter-firm co-operation.’ Entrepreneurship and Regional Development 14 (4): 337–53. Dicken, P. and A. Malmberg (2001). ‘Firms in territories: A relational perspective.’ Economic Geography 77 (4): 345–63. Eliasson, G. (2000) ‘Industrial policy, competence blocks and the role of science in economic development’. Journal of Evolutionary Economics 10: 217–41. Enright, M. (1998). ‘Regional clusters and firm strategy’, in A. Chandler, P. Hagström and Ö. Sölvell (eds) The Dynamic Firm: The Role of Technology, Strategy, Organization and Regions. Oxford, Oxford University Press. Fosfuri, A. and T. Ronde (2004). ‘High-tech clusters, technology spillovers, and trade secret laws.’ International Journal of Industrial Organization 22 (1): 45–65. Fosfuri, A., M. Motta and T. Ronde (2001). ‘Foreign direct investment and spillovers through workers’ mobility.’ Journal of International Economics 51 (1): 204–22. Fuellhart, K. (1999). ‘Localization and the use of information sources: The case of the carpet industry.’ European Urban and Regional Studies 6: 39–58. Gilson, R. J. (1999). ‘The legal infrastructure of high technology industrial districts: Silicon Valley, Route 128, and covenants not to compete.’ New York University Law Review 74 (3): 575–629. Glaeser, E., H. Kallal, J. Scheinkman and A. Shleifer (1992). ‘Growth in Cities.’ Journal of Political Economy 100: 1126–52. Grabher, G. (2002). ‘The project ecology of advertising: Tasks, talents and teams.’ Regional Studies 36 (3): 245–62. Hendry, C., J. Brown and R. Defillippi (2000). ‘Regional clustering of high technologybased firms: Opto-electronics in three countries.’ Regional Studies 34 (2): 129–44. Howells, J. (2002). ‘Tacit knowledge, innovation and economic geography.’ Urban Studies 39 (5–6): 871–84. Håkansson, H. (1987). Corporate Technological Behaviour: Co-operation and Networks. London, Routledge. Jaffe, A. (1989). ‘Real effects of academic research.’ American Economic Review 79: 957–70. Jaffe, A., M. Trajtenberg and R. Henderson (1993). ‘Geographic localization of knowledge spillovers as evidenced by patent citations.’ Quarterly Journal of Economics 108: 577–98. Keeble, D., C. Lawson, B. Moore and F. Wilkinson (1999). ‘Collective learning processes, networking and “institutional thickness” in the Cambridge region.’ Regional Studies 33 (4): 319–32. King, C., A. Silk and N. Ketelhohn (2003). ‘Knowledge spillovers and growth in the disagglomeration of the US advertising-agency industry.’ Journal of Economics and Management Strategy 12 (3): 327–62. Larsson, S. (1998). Lokal förankring och global räckvidd: En studie av teknikutveckling i svensk maskinindustri [Local Embeddedness and Global Reach: A Study of Technological Development in the Swedish Machinery Industry]. Uppsala, Geografiska Regionstudier nr. 35. Uppsala Universitet. Larsson, S. and M. Lundmark (1991). Kista – företag i nätverk eller statusadress? En studie av Kistaföretagens länkningar. Forskningsrapport nr. 100. Uppsala, Kulturgeografiska Institutionen, Uppsala Universitet. Lawson, C. (1999). ‘Towards a competence theory of the region.’ Cambridge Journal of Economics 23: 151–66. Lewis, T. and D. Yao (2003). Innovation, knowledge flow, and worker mobility. (2003)

66 Anders Malmberg and Dominic Power http://rider.wharton.upenn.edu/~yao/SVsubmitjune03.pdf (web resource, accessed 10 December 2005). Lissoni, F. (2001). ‘Knowledge codification and the geography of innovation: The case of Brescia mechanical cluster.’ Research Policy 30 (9): 1479–500. Lundequist, P. and D. Power (2002). ‘Putting Porter into practice? Practices of regional cluster building: Evidence from Sweden.’ European Planning Studies 10 (6): 685–704. Lundvall, B.-Å. (1992). National Systems of Innovation: Towards a Theory of Innovation and Interactive Learning. London, Pinter. MacKinnon, D., A. Cumbers and K. Chapman (2004). ‘Networks, learning and embeddedness amongst SMEs in the Aberdeen oil complex.’ Entrepreneurship and Regional Development 16 (2): 87–106. Madsen, T., E. Mosakowski and S. Zaheer (2003). ‘Knowledge retention and personnel mobility: The nondisruptive effects of inflows of experience.’ Organization Science 14 (2): 173–91. Malmberg, A. (2003). ‘Beyond the cluster: Local milieus and global connections’, in J. Peck and H. Yeung (eds) Remaking the Global Economy. London, Sage. Malmberg, A. and P. Maskell (2002). ‘The elusive concept of localization economies: Towards a knowledge-based theory of spatial clustering.’ Environment and Planning A 34: 429–49. Malmberg, A. and D. Power (2005a). ‘On the role of global demand in local innovation processes’, in P. Shapiro and G. Fuchs (eds) Rethinking Regional Innovation and Change. New York, Springer. Malmberg, A. and D. Power (2005b). ‘(How) Do (Firms In) Clusters Create Knowledge?’ Industry and Innovation 12 (4): 409–31. Malmberg, A., B. Malmberg and P. Lundequist (2000). ‘Agglomeration and firm performance: Economies of scale, localisation and urbanisation among Swedish export firms.’ Environment and Planning A 32: 305–21. Malmberg, A., Ö. Sölvell and I. Zander (1996). ‘Spatial Clustering, local accumulation of knowledge and firm competitiveness.’ Geografiska Annaler Series B 78B (2): 85–97. Markgren, B. (2001). Är närhet en geografisk fråga? Företags affärsverksamhet och geografi – en studie av beroenden mellan företag och lokaliserings betydelse. [Is Proximity a Geographical Question in Business Relationships]. Uppsala, Uppsala University, Department of Business Studies. Markusen, A. (1999). ‘Fuzzy concepts, scanty evidence, policy distance: The case for rigour and policy relevance in critical regional studies.’ Regional Studies 33: 869–84. Martin, R. and P. Sunley (2003). ‘Deconstructing clusters: chaotic concept or policy panacea?’ Journal of Economic Geography 3 (1): 5–35. Maskell, P., H. Eskelinen, I. Hannibalsson, A. Malmberg and E. Vatne (eds) (1998). Competitiveness, Localised Learning and Regional Development. London, Routledge. Narin, F., K. Hamilton and D. Olivastro (1997). ‘The increasing linkage between US technology and public science.’ Research Policy 26: 317–30. Owen-Smith, J. and W. Powell (2002). Knowledge Networks in the Boston Biotechnology Community, unpublished working paper. Stanford, Stanford University. Porter, M. (1990). The Competitive Advantage of Nations. New York, The Free Press. Porter, M. (1994). ‘The role of location in competition.’ Journal of the Economics of Business 1: 35–9. Porter, M. (1998a). Competitive Advantage: Creating and Sustaining Superior Performance. Boston MA, Harvard Business School Publishing.

True clusters 67 Porter M (1998b) ‘Clusters and the new economics of competition.’ Harvard Business Review (November–December). Porter, M. (2000). ‘Locations, clusters and company strategies’, in G. Clark, M. Feldman and M. Gertler (eds) The Oxford Handbook of Economic Geography. Oxford, Oxford University Press. Porter, M. (2001). Clusters of Innovation: Regional Foundations of U.S. Competitiveness. Washington, Council on Competitiveness. Power, D. and D. Hallencreutz (2002). ‘Profiting from creativity? The music industry in Stockholm, Sweden and Kingston, Jamaica.’ Environment and Planning A 34 (10): 1833–54. Power, D. and M. Lundmark (2004). ‘Working through knowledge pools: labour market dynamics, the transference of knowledge and ideas, and industrial clusters.’ Urban Studies 41 (5/6): 1025–44. Raines, P. (2001). The Cluster Approach and the Dynamics of Regional Policy-Making. Regional and Industrial Policy Papers No. 47. Glasgow, European Policies Research Centre, University of Strathclyde. Rodríguez-Pose, A. and M. Refolo (2003). ‘The link between local production systems and public and university research in Italy.’ Environment and Planning A 35: 1477–92. Ronde, T. (2001). ‘Trade secrets and information sharing.’ Journal of Economics and Management Strategy 10 (3): 391–417. Rosenfeld, S. (1997). ‘Bringing business clusters into the mainstream of economic development.’ European Planning Studies 5 (1): 3–23. Rosenkopf, L. and P. Almeida (2003). ‘Overcoming local search through alliances and mobility.’ Management Science 49 (6): 751–66. Sakakibara, M. and M. Porter (2001). ‘Competing at home to win abroad: Evidence from Japanese industry.’ Review of Economics and Statistics 83 (2): 310–22. Sölvell, Ö., G. Lindqvist and C. Ketels (2003) The Cluster Initiative Greenbook. Stockholm: Ivory Tower. Available at http://www.ivorytower.se/greenbook/ general.html. Song, J., P. Almeida and G. Wu (2003). ‘Learning-by-hiring: When is mobility more likely to facilitate interfirm knowledge transfer?’ Management Science 49 (4): 351–65. Storper, M. and A. Venables (2002). Buzz: The Economic Force of the City. DRUID Summer Conference on ‘Industrial Dynamics of the New and Old Economy – who is embracing whom?’ Copenhagen/Elsinore. Thrift, N. and A. Leyshon (1994). ‘A phantom state? the de-traditionalisation of money, the international financial system and international financial centres.’ Political Geography 13: 299–327. Von Hippel, E. (1988). The Sources of Innovation. Oxford, Oxford University Press. Wallsten, S. (2001). ‘An empirical test of geographic knowledge spillovers using geographic information systems and firm-level data.’ Regional Science and Urban Economics 31 (5): 571–99. Weber, M. (1947). Max Weber: The Theory of Social and Economic Organization. New York, The Free Press. Weber, M. (1949). The Methodology of the Social Sciences. New York, The Free Press. Welz, G. (2003). ‘The cultural swirl: anthropological perspectives on innovation.’ Global Networks – A Journal of Transnational Affairs 3 (3): 255–70.

68 Anders Malmberg and Dominic Power Zeller, C. (2001). ‘Clustering biotech: A recipe for success? Spatial patterns of growth of biotechnology in Munich, Rhineland and Hamburg.’ Small Business Economics 17 (1–2): 123–41. Zucker, L., M. Darby and J. Armstrong (1998). ‘Geographically localized knowledge: Spillovers or markets?’ Economic Inquiry 36 (1): 65–86.

4

In search of a useful theory of spatial clustering Agglomeration versus active clustering Fiorenza Belussi

Introduction1 Over the past two decades or so there has been a veritable flood of interest, within economic geography, economics, and business studies, in industrial localization, the observed tendency for many industries to form specialized concentrations in particular locations. Industrial localization of course is nothing new. Interest in geographical concentrations of firms in the same or related industries has grown considerably in the last decades because of the ‘success’ of well-known industrial districts or clusters such as Silicon Valley and Route 128 in the USA (Dorfman, 1983; Saxenian, 1994). The contribution that industrial districts have made to the Italian economy (Becattini, 1979, 1987; Becattini et al., 2003) have also stimulated a wide international debate about the competitive performance of these specific geographical concentrations of ‘localized industries,’ (Pyke et al., 1990; Harrison, 1992; Cossentino, Pyke and Sengenberger, 1996; Biggero, 1999; Belussi and Gottardi, 2000; Guerrieri, Iammarino and Pietrobelli, 2001; Paniccia, 2002; Belussi, Gottardi and Rullani, 2003). Studies revealed that industrial districts and clusters are widespread in many different countries and industries (Enright, 2001), and that they seem to be becoming more important in the response to the globalization process (Dunning and Wymbs, 1999; Narula, 2003). In recent years even numerous international organizations have developed ad hoc recommendations and specific policies for encouraging firms to form and build up industrial districts or clusters (for example, European Commission, 2001; OECD, 1999 and 2001; DTI, 2001, and Harvard Business School, 2002). All started with the rediscovery of the seminal writings of Alfred Marshall, which – particularly in Italy – gave rise to a new stream of empirical studies on the issue of small firms, industrial decentralization, peripheral development, and a “Third Italy model” (which can be reflected in the flex-spec hypothesis, carried on by Piore and Sabel during the 1980s). However, the highest diffusion of these themes was reached thanks to the work of Porter on cluster competitiveness (Porter, 1998a and 2000), where he enlightened the positive effects played by geographical concentrations of similar and interrelated firms, supported by

70 Fiorenza Belussi active specific local institutions. Thus, in the 1990s, the interest in a model of industrial organization, based on the agglomeration of small and mediumsized enterprises, widely expanded in many branches of business studies, economic sociology, industrial economics, regional economics, and development economics. The issue of spatial agglomeration, at a more abstract and meso-economic level, has also been extensively treated in several other fields of research: • •

•

firstly, by the so-called new economic geography (Krugman, 1991, 1995), which used the Marshallian framework to demonstrate the tendency toward urban agglomeration secondly, by some evolutionary economists (Arthur, 1990), where the concepts of positive externalities and increasing returns to scale were integrated to deliver a theoretical explanation for the rise and evolution of agglomeration per se, without referring to any specific industrial organizational model, as is done, on the contrary, in the Marshallian or Porterian tradition (for a review see Baptista, 1998) thirdly, the influence of agglomeration has been studied in several ‘innovation studies’ in relation to the importance of technological spillovers deriving from the geographical co-localization of agents and firms, from the early studies of Jaffe (1986) to the recent works on the geography of innovation of Feldman and Audretsch (1999), or because the presence of ‘learning mechanisms’ was spatially bounded, and thus considered crucial to determine the parameters of efficiency of a set of interactive agents, within spontaneous or policy-guided ‘local’ or ‘regional innovation systems’ (Acs, 2000; Cooke, 2002; Braczyk, Cooke and Heidenreich, 1998).

The obvious difference is that in the former approaches we are studying specific systems whose boundaries are given (or assumed ex ante, in relation to the theoretical scheme of analysis), and whose mechanism of evolution represents the focus of the analysis, while in the latter case we are theorizing a simple correlation of ample applicability among certain variables (agglomeration, R&D spillovers, patenting activity, etc.), inserted in variously defined spatial contexts (regions, nations, large conurbations, etc.). These two approaches appear to be quite different, for the theoretical assumptions initially adopted, and for the methodologies of verification applied.

Districts and industrial clusters: a semantic ambiguity or an umbrella for a variety of phenomena? There is a semantic ambiguity in the literature on industrial districts and/or clusters. Do districts and clusters indicate exactly the same phenomenon? If so, why do we use two words with the same meaning? A possible explanation may lie in the fact that two streams of the literature during the 1980s, the Marshallian one and the business literature influenced by Porter’s work, merged together. General confusion is in addition generated by the problem that, with few

In search of a useful theory 71 exceptions, the majority of researchers using the concept of the industrial district, or the alternative notion of the cluster, have avoided the troublesome effort of precisely defining the object of their analysis. Such a wide number of contributions on the issue of enterprise clustering have created a Babel of connotations and meanings. The intriguing matter of this story that can confuse an inattentive reader is that the two terms have been in some cases used as near-substitutes, but in other cases they have ceased to denote a large variety of phenomena, because under the umbrella of industrial district and/or cluster, extensively used in economics, business, regional economics, industrial economics, economic geography and sociology, different models of (local) development and inter-firm arrangements can be recognized. Most of the literature lacks sufficient power of theoretical generalization; and testable regularities are not precisely identified. Quantitative analyses have used spurious data bases, and research conducted on individual case studies confuses the two levels of analysis: the idiosyncratic component, related to the obvious specificity of the territory examined, and the studied ‘clustering’ effect (Signorini, 1994), deriving exactly from the features of the phenomenon treated. In order to settle this controversy, we have no other alternative than to return to the origin of our discussion. Are industrial districts or clusters simply high concentrations of (small) firms specializing in one main industry? Are industrial districts and clusters significantly competitive simply because they refer to the phenomena of local agglomeration? And when does the positive effect of agglomeration cease to play a role (in other words, when do diseconomies of agglomeration emerge)? Is co-operation a necessary ingredient for the definition of the clustering phenomenon: a sine qua non condition, whose absence invalidates the existence of a district or a cluster? Do co-operative behaviours exist ex ante, because they are endogenously embedded in a given territory, or are they a frequency-dependent phenomenon, which may arise in various patterns with different intensities? Another relevant theme is the role of institutions. Porter (and not directly Marshall) introduced the presence of institutions in his paradigmatic description of the idea of the cluster. But do local institutions always intervene as supportive agents for the local systems? Or can we also register cases in which institutions fail to play a positive role, and where there clearly emerge ‘institutional market failures’ (Provasi, 2002)? And, finally, if clusters and industrial districts really represent two distinct models of territorial clustering, do we possess a solid block of theory to differentiate these cases in the literature? Let us consider an example of territorial clustering that has been much studied in recent years and has attracted the efforts of many researchers, who have investigated the Silicon Valley phenomenon. I have the impression that, if Silicon Valley were located in Italy, it would have been described by the international literature as a high-tech ‘industrial district’, because the Marshallian echo retains influence. On theoretical grounds, the critical issue is to understand the relationship between the empirical forms of specific industrial districts and/or clusters and

72 Fiorenza Belussi their ‘normative’ model. This requires the isolation of the industrial district and/or cluster rationale, whether it be rooted in economic or in socio-economic theory.

The origin of the concept of the industrial district Let us start with Marshall’s concept of industrial districts (Marshall, 1890), based on the importance of external economies, to understand the development of agglomerated clusters of small and medium-sized firms. His work focuses on the benefits of external economies emerging from the close proximity of actors in the process of economic activity. Particularly relevant is Marshall’s concept of ‘industrial atmosphere’, which informs much of the literature on Italian industrial districts, and is related to a business and social environment conducive to the acquisition of the benefits of proximity deriving from imitation, vicarious learning, quick adoption, technical change and innovation introduced thanks to the generation of collective or individual new knowledge. Other characteristics are: (1) the concentration of many small factories specializing in different phases of the same production processes; (2) the gradual accumulation in the area of a skilled labour force; and (3) the creation of subsidiary industries and specialized suppliers. Marshall advocated that, at least for certain types of production, there were two efficient manufacturing systems: the large vertically integrated production unit, and the industrial district. He came to the conclusion that the same economies that benefit the large factories can sometimes be secured by small factories placed in the same locality. He called these economies ‘external economies’, depending on ‘the aggregate volume of production of the kind of neighbourhood’ (Marshall, 1890, p. 265) juxtaposing them to the ‘internal economies’ related to the co-ordination of activities under the vertically integrated factory. We can thus arrive at the following conclusion: 1 2 3 4

5

The Marshallian industrial district is a specific locality where there is a certain type of productive specialization. The district is characterized by a high density (prevalence, but not absolute dominance) of small to medium-size firms. Firms co-operate along the supply chain because there is an extended interfirm division of labour. Typically the district derives leadership in a special industry (Marshall, 1919, p. 287) from the ‘industrial atmosphere’ if, as Marshall emphasized, obstinacy or inertia of firm behaviours in changing times will not ‘ruin it’. We have an industrial district if in the same area there is a high variety of similar producers and this stimulates high creativity and, sometimes, intercommunication of ideas between machine makers and machine users (Marshall, 1919, p. 603).

Using the Marshallian approach we can derive some analytical consequences: 1

The industrial district is not a universalistic model of firm clustering.

In search of a useful theory 73 2

3 4

5

The industrial district is a specific organizational model, ceteris paribus equally as efficient – in the condition of technical or economic divisibility of activities – as the large firm. This does not always imply that a bunch of similar small firms specializing in a particular activity, and clustered in an area, are per se efficient: for instance, they could adopt technologies inferior to those in use by large organizations. A district cannot be formed by a single firm network: the same definition of industrial district recalls the concept of a large population of firms and introduces the concept of a sizeable ‘threshold’ of local firms. In the industrial district we can clearly find a mechanism for increasing returns embedded in the territory, but Marshall never denied the possibility of contemporary increasing returns bound to the individual organization; in fact, he spoke also of large firms and plants located in industrial districts (Marshall, 1919, p. 285); in districts there is still room for increasing organizational efficiency depending on individual firms’ strategies. From Marshallian theory we assume that some external efficiencies in industrial districts are related to the volume of activities (scale efficiency), but also other forms of efficiency (dynamic efficiency): they are conditioned by the stage of evolution of each industrial district.

In other words, as Becattini (2003) observed, one thing is to look at the main descriptive components of an industrial district in abstract terms, and another is to analyse the process of ‘districtualization’ of a territory within a historical path-dependent dynamic process (p. 15).2 This in some ways is opposed to the idea of static efficiencies that can be reached by increasing size and output volumes.

Districts as functional clusters A cluster is defined in the Concise Oxford Dictionary as ‘a group of similar things growing together’. This definition implies the existence of some integration of closely related things, which may be involved in a dynamic process: either only through spatial proximity or via a space less functional relatedness, or again through a process which implies both. The cluster is a functional concept but it can also describe a process of agglomeration. This ambiguity is intrinsically related to the term ‘cluster’, and it is at the basis of the reasons that probably drew Porter to adopt this term in the 1990s (instead of going back to the solid Marshallian conceptualization, which at the time was already being adopted by numerous researchers, especially in Italy). Porter was above all interested in the development of his theory on competitiveness (Porter, 1980). His strategy of research was to find some linkages between the spatial dynamic of some productive systems and his famous diamond (firm rivalry, new entry of competitors, power of upstream suppliers and machinery producers, threat of substitutive products, and factors-demand conditions): thus, the five factors of competitiveness, which he originally applied in a macro context (Porter, 1990) to explain the ‘competitive advantage of nations.’

74 Fiorenza Belussi In order to do this, Porter needed a flexible concept which could be used for ‘sectors’ and ‘areas’. So he came across the term ‘cluster’ (Porter, 1998a, pp. 197–287): a functional concept, similar to the French idea of filière, applicable in a variety of contexts. Although the term ‘cluster’ began to be understood as a source of competitive advantage deriving from localization, it does not necessarily reveal a mechanism whereby the sources of efficiencies are exclusively embedded at a territorial level, because they can also be related to the existence of various types of synergies belonging to the nodes of a network, or to a national system. This double nature – both territorial and functional – in the usage of the term ‘cluster’ can be easily traced in Porter’s writings. For Porter ‘A cluster is a geographically proximate group of interconnected companies and associated institutions in a particular field, linked by commonalities and complementarities’ (1998a, p. 199; 2000, p. 16). However, he forgets to tell us the spatial boundaries of his territorial system. A cluster is not just a small portion of a territory, a piece of identifiable ‘localised industry’. Porter (1998a, chapter 7) repeatedly mentions the existence of ‘regional clusters’, mapping them in the US (for instance the Californian wine cluster), Portugal, Sweden, and Italy, but in his book he also presents ‘national clusters’ like the Italian (my italics) cluster of the fashion and shoe industry. Clusters are geographic concentrations of interconnected companies, specialised suppliers, service providers, firms in related industries, and associated institutions (for example, universities, standards agencies, trade associations) in a particular field that compete but also cooperate. Clusters, or critical masses of unusually competitive success in particular business areas, are a striking feature of virtually every national, regional, state and even metropolitan economy, especially in more advanced nations. (Porter, 1998a, pp. 197–8) Indeed, Porter provides more than one definition of a cluster, and this gives rise to conceptual confusion. Perhaps there is a too ample idea of what the phenomenon can embrace. According to circumstances the Porter analytical exercise can be ‘sectoral’ or ‘spatial’, ‘regional’ or ‘national’. In his seminal work Porter also specifies that tracing the boundaries of a cluster is often a matter of degrees, and that this is a creative process, related to the understanding of links and complementarities existing among industries and institutions. In fact, he argues, the ‘external effects’ significantly related with competitiveness and productivity determine the very boundaries of a cluster. Cluster boundaries do not correspond to conventional industry classification, because they are formed by a combination of final producers, intermediate goods, and machinery industries. The recipe is apparently very simple: in order to identify a cluster we may start from a large firm or from a concentration of similar firms and then look upstream and downstream, horizontally and vertically along the chain of firms

In search of a useful theory 75 and institutions. The aim is to individuate factors and complementarities and to underline the role of the related institutions, which provide infrastructures, norms, and public goods. But if we study the ‘economic interrelatedness’ of a cluster, we can shift easily from spatial interconnections, which are defined by the geographical proximity, to virtual connections (Rallet and Torre, 2004), which are related to the many external linkages that each local organization activates with the external world: in other terms, by doing so, we put at risk the same possibility of defining a ‘given system’, and we lose the limiting boundary conditions between what is inside and outside our model.3 The Porter analytical exercise, with its indeterminacy of geographical boundaries, is more a tool for a ‘systemic strategic analysis’ (Porter, 1998b) than a sound piece of theory: escaping from the devil of ‘objective indicators’, available data, and measurements, it enters the paradise of pure ‘subjective’ – and each time changeable – definition.

Clusters versus districts: the Italian tradition It has been maintained that the concept of cluster is elusive and heterogeneous, and the literature on clusters a patchy constellation of ideas (Martin and Sunley, 2003): a concept that has been marked as a ‘brand’, rather than as another intellectual product. In my opinion this is certainly true: Porter’s cluster concept contains many unresolved ambiguities. However, Porter himself makes ample recourse to the Marshallian tradition, and in his recent empirical work he appears to be aware of the necessity of introducing a more coherent spatial approach (Porter, 2000, p. 16). The concept of cluster developed by Porter clearly differs from the Italian ‘industrial district’ research tradition, as examined by Tessieri (2001). First of all, this Italian industrial district’s stream of literature is deep-seated in the neo-Marshallian revival. Secondly, one of the main objectives of the analysis, different from the alleged indeterminacy of the Porterian cluster, is the precise definition of the geographical boundary of the local system under examination (although achieved with difficulty also in Italy at statistical level by the various attempts to map the Italian classification grid, as shown by IPI, 2002). The main problem encountered by empirical research conducted mainly in Italy is related to the fact that very few districts arise in a local context in a ‘pure’ form as in the case of Prato, where the local textile industry is the dominant feature of the local manufacturing sector, and the whole community has been engaged since the Middle Ages. Many Italian districts specializing in light or medium high-tech sectors are smaller than Prato, and less spatially concentrated, as is the case for instance of several in the Veneto region. Some industrial districts are diluted in urban conurbations, so they do not ‘emerge’ distinctly from the statistical data. But also the Italian line of enquiry shows a ‘shortage’: an obsession with ‘true’ objective statistical indicators for a phenomenon that is not simple to investigate.

76 Fiorenza Belussi This has rendered the international applicability of the theory of the industrial district less usable, especially for emerging or embryonic forms of industrial districts, where much lies in the intuition and experience of the researcher, because data are not consistent, or simply do not exist. The empirical determination of the Italian ‘localized industries’ on which numerous researchers have worked in the last decade has followed mainly three avenues: (1) the construction of objective statistical parameters related to the labour market and to territorially self-contained flows of commuting (with the identification of the so called 199 IRPET-ISTAT local labour systems, Sforzi, 1987), and (2) the utilization of measurements of the industry localization index (Anastasia, Corò, and Crestanello, 1995) and (3) a more qualitative research strategy focused on empirical comparisons of individual case studies (Falzoni, Onida, and Viesti, 1992; Club dei distretti industriali, 2003). The first stream of research has been slowly abandoned: numerous selected local systems were not really industrial districts but only areas of territorial specialization (they lack the necessary connotation of Marshallian ‘leadership’, in terms of firm autonomy, competitive capability, and threshold). In addition the restrictive condition imposed by the IRPET-ISTAT definition of the exclusive presence of small firms rather conflicted with the history of numerous Italian districts, where we register the growth of enterprise groups or of large firms. In this sense, the experience realized in more than 20 years of research suggests considering a testable pluralistic multi-sequences approach, which is a combination of avenues 2 and 3. The conceptual boundaries of a vital organizational system are decisively an important issue, but they are not confined to a simple statistical algorithm. Thirdly, in the Italian variant of the analysis of the industrial district, a special emphasis on the social features was introduced. The Italian industrial district has been portrayed as a system embedded in the larger social community. The definition of the industrial district includes the idea of a socio-territorial entity which is characterised by the active presence of both a community of people and a population of firms in one naturally and historically bounded area. In the district, unlike in other environments, such as manufacturing towns, community and firms tend to merge . . . The fact that there is a dominant activity differentiates the district from a generic ‘economic region’. The self-containment and the progressive process of division of labour, together with the realisation of productive specialisation, produces a growing surplus of products that cannot be sold in the district. (Becattini, 1990, pp. 38, 52–3) Following Becattini, most Italian economists, also supported by several authors such as Piore and Sabel in US and Schmitz in UK (1992), would claim that the mere agglomeration of specialized firms is not enough to denote an industrial district, but other conditions attaining to the attitudes and values and identity

In search of a useful theory 77 of local population are also important to determine the existence of an industrial district. In this view, industrial districts are ‘socio-economic systems’ joining together a community of people with common values or culture. Most of the social features of this type of industrial district include: (1) equilibrium between cooperation and competition among rival firms (You and Wilkinson, 1994; Asheim, 1996), (2) social integration (Brusco, 1982, 1990), (3) the existence of trust which enforces co-operation, economizes on transaction costs and fosters flexibility and innovation (Dei Ottati, 1996), and (4) extended inter-firm division of labour with no asymmetries of power among the clustered enterprises (Sforzi, 2003). Co-operation can be seen as a rule of governance of the industrial districts, and qualifies them as social networks or, according to some organization theorists, as an ideal-typical organization model between market and hierarchy. This also resembles what in sociology has been termed the ‘communitarian’ model (Grabher, 1993). This view raises the question of whether co-operation – as a mechanism producing trust – is at the same time a necessary and pervasive rule of governance of industrial district, and how we can ‘operationalise’ the measurement of cooperation, and finally, whether it is possible to conceive (or empirically observe) industrial districts without significant levels of co-operation. A similar question concerns institutions that are considered an important ingredient for the efficiency or dynamism of industrial districts. Do we admit equi-functional alternatives? How does each local institution deal with the different national contexts that we can observe?

In search of a more agnostic theory of spatial clustering It follows from our reflections that we must pursue a kind of integrated perspective, which mixes the widely known (but sometime too elusive) cluster approach, in which the subjective analysis of cluster competitiveness has a role, with the more solid Marshallian perspective, with the caution of escaping from idiosyncratic ‘pure forms’, for considering a large variety of industrial district evolution. This allows us to take into account different evolutionary stages (from embryonic to maturity-decline), and different district typologies. The concept of industrial district must be operationalized, avoiding an excessive qualification with precise socio-economic features, for example horizontal and vertical networking, innovativeness, learning, co-operation, trust, and so on (Paniccia, 2002, p. 6), that could be integrated in a second stage of the analysis, in the phase of typology-building (see Figure 4.1).4 This implies a theoretical fusion of three existing ramifications of the literature as argued in Figure 4.2: (1) the socio-economic concept, (2) the cluster approach (amended by a more conscious awareness of cluster/district borders), and (3) the more mixed contributions for which the concept of cluster has become

78 Fiorenza Belussi NO INTERACTIONS

INTERACTIONS

Areas with agglomeration (and productive specialization)

Marshallian districts or industrial clusters (interconnected systems) with: • large firm population (density) • small and medium-size firms • linkages between final producers and subcontractors • identity and ad hoc institutions (vocational training; R&D research centres, and financial institutions) • local entrepreneurship

Figure 4.1 Agglomeration versus active clustering

synonymous with the term ‘industrial district’, but in a way in which the definition of the industrial district has become more elastic than in the Marshallian tradition; let us think for instance of clusters or districts focused on non-manufacturing goods, like media clusters, financial service clusters, cultural clusters (Cooke, Pandit, and Swann, 2000), etc. The ‘agnostic’ definition of industrial district or cluster must be comprehensive enough to include areas showing different organizational arrangements and to denote the agglomeration of small and medium-sized firms specializing in one or a few industries in a bounded area. The geographical extent of the industrial district or cluster is one of the still open questions of this type of literature. While the Italian tradition focuses generally on small localities (formed by a high density of firms localized in few municipalities where contiguity is crucial to define the community and the related institutions), at the international level clusters or districts are often thought of as localized industries based on larger geographical areas. Industrial districts and clusters lead to competitive advantages by generating a number of benefits that are not available to firms that are not located in geographical concentrations: • • • •

increasing returns driven by the systemic properties embedded within the local systems reductions in transaction costs innovation and technological development depending on local interactions reduced costs via effective learning (learning by imitation and emulation)

In search of a useful theory 79 •

• •

benefits provided by localised external economies (specialized labour market, specialization led by the increased local division of labour, and existence of specialized competent suppliers) first-mover advantages from the initial territorial specialization advantages related to being customer-driven organizations and to product diversification.

Marshall emphasized the role of industrial districts as models conducive to an ‘industrial atmosphere’ where the spread of innovation is the consequence of decentralized industrial creativity. These benefits appear to be strong in hightech areas (Swann et al., 1998). However, they are also prevalent in many Italian industrial districts that are not directly involved with high-tech products or processes. Within the industrial district or cluster model the creation of new knowledge appears to be the output of agents’ interactions. This is more the result of search strategies and random interactions, an ‘open innovation’ strategy, rather than a planned and deliberate effort in which R&D activities are involved as described by the standard model. The generation of new knowledge (innovation) occurs via numerous sources: design and engineering activity, learning processes coming from production departments, interactions with clients and suppliers, re-use of existing local knowledge (Belussi, 2002). Firms belonging to industrial districts are often not only innovative leaders but also fast adopters. Rarely is innovation truly protected. Imitation occurs via observational learning and rival emulation. Original innovations, during their diffusion within the industrial district, are typically improved (in their costs, or in performance, or in other important intrinsic characteristics). The higher the population of potential adopters, the higher the generation of variety, and the higher the probability that some modifications are introduced along the diffusion cycle. This appears to be the first advantage that district firms can achieve. A second advantage is the existence of an ample range of locally differentiated capabilities and resources. This preserves the versatility of the local structure. A third advantage is inter-firm linkages with the suppliers of machinery localized in close proximity. A fourth advantage derives from easy access to informational channels (Antonelli, 2000). Areas where there exists just a ‘pure’ agglomeration of firms must be analytically separated by clearly interconnected localized systems as districts or clusters, which have followed a pattern of districtualization. (At the beginning, embryonic clusters or districts are formed only by a few local networks of firms, and this situation is analytically confused, but in retrospect the emergence of the process of ‘districtualization’ suits the indicators.) Taken alone, without any specification, the term ‘cluster’ identifies a very large class of phenomena related to territorial specialization. The ‘cluster’ category developed by Porter does not contain the above listed (Figure 4.2) specifications or restrictions (a Porterian ‘cluster’ has no minimum threshold of agglomeration, three firms are not a district in a Marshallian sense, but they can belong to a cluster). However, if we

80 Fiorenza Belussi INDUSTRIAL DISTRICT AS A SOCIO-ECONOMIC CONCEPT Becattini (1987, p. 47): ‘The Marshallian industrial district constitutes, thus, a localized thickening (in this spatial determination we find its weakness and its strength) of interindustry relationships, that show a consistent stability during time.’ Pyke, Becattini and Senberger (1990, pp. 16–17): ‘Industrial districts are geographical defined systems, characterised by a high number of firms active in different stages and in different modes of the production of a homogeneous product. A significant characteristic is that a large part of these firms are small firms or very small firms . . . The various districts are specialised in different products with various degrees of complexity and with different final uses . . . A characteristic of the industrial district is that it has to be thought as a unique unity, a social and economic system . . . Important is the fundamental role played by the various forms of cooperation among the firms which is communitarian.’ CLUSTERS AS VAGUE SPATIAL SYSTEMS Rosenfeld (1997, p. 4) ‘A cluster is very simply used to represent concentrations of firms that are able to produce synergy because of their geographical proximity and interdependence, even though their scale of employment may not be pronounced or prominent.’ Feser (1998, p. 26): ‘Economic clusters are not just related and supporting industries and institutions, but rather related and supporting institutions that are more competitive by virtue of their relationships.’ Roelandt and Den Hertog (1999, p. 9): ‘Clusters can be characterised as networks of producers of strongly interdependent firms (including specialised suppliers), linked to each other in a value-adding production chain.’ Enright (1996, p. 191): ‘A regional cluster is an industrial cluster in which member firms are in close proximity to each other.’ Lundvall and Borras (1997, p. 39): ‘The region is increasingly the level at which innovation is produced through regional networks of innovators, local clusters and the crossfertilising effects of research institutions.’ CLUSTERS AS CLOSE SUBSTITUTE FOR THE CONCEPT OF INDUSTRIAL DISTRICT Maskell (2001, p. 925): ‘The term cluster is used synonymously in the literature together with industrial agglomeration or localisation, while the term industrial district . . . is often applied when wishing explicitly to emphasise to values and norms shared by co-localised firms.’ Asheim and Isaksen (2002, p. 77): ‘The crux of the regionalisation argument is that the regional level, and specific local and regional resources may still be important in firms’ effort to obtain global competitiveness . . .firms in the cluster rely on unique regional resources and local cooperation when innovating.’ Cooke and Huggins (2002, p. 4): ‘Clusters are geographically proximate firms in vertical and horizontal relationships, involving a localised enterprise support infrastructure with shared developmental vision for business growth, based on competition and cooperation in a specific market field.’

Figure 4.2 Main definitions of industrial districts or clusters Source: partially based on Martin and Sunley (2003) and other sources

In search of a useful theory 81 decide to introduce them, we build a convergence between the two terminologies and between the two analytical approaches: thus, while a generic cluster is not a district, a cluster with the above specified characteristic is also a Marshallian district. A local district or cluster is a specific local system, based on a given technological filière, with an objective density (a conspicuous number of firms), local entrepreneurship, identity, competitiveness, ad-hoc created institutions based on the district specialization, and life-cycle evolutionary pattern (from embryonic to maturity-decline stage). However, the argumentation presented by Porter, and the empirical cases discussed, push in the same direction of considering crucial for the analysis not just the level of geographical proximity (agglomeration) but the bounded social and economic interactions identifiable in a local system (supported by spontaneous and planned institutions and by specific economic interdependence). Industrial districts or clusters differ in terms of competitiveness, industry structure, size of firms and organizational arrangements. In addition, many ‘local’ districts contain ‘global’ key firms and other organizations or agencies that play a national role in their country. Similarly, local competitive industrial districts possess complex connections to global networks and markets (Amin and Thrift, 1994; Becattini and Rullani, 1996; Asheim and Cooke, 1999). A number of possible types of industrial districts or clusters could be defined by gathering data on knowledge and information flows (types of innovations, forms of learning, patents, R&D expenditures, etc.). Using graph analysis we can provide a classification of the district flows. If the mapping exercise is conducted over a period, the evolution of the industrial district can be assessed. Mapping flows of goods, services, innovation and learning makes possible a comparison with other districts. Comparison of the flows of goods and services particularly for final output and exports should be possible. A hypothetical example of such a map is given in Figure 4.3 where an industrial district with strong interorganizational linkages and a well developed innovation and learning system is highlighted.

The path of evolution of industrial districts and clusters From the existing literature we can extract a stylized development path (Belussi, 2000; Klink and De Langen, 2001; Brenner, 2001), of development, expansion, maturation and transition. Locations that have some favourable starting conditions initiate growth, thanks to the rooting in the territory of some founder firms. The development of extensive external economies allows the starting of a recursive trend towards the consolidation of the industrial district or cluster. The initial development of district or cluster capabilities (Figure 4.4) can be enforced by the development of an institutional system conducive to growth, where network benefits will emerge, and extensive co-operation is promoted among the local agents. One of the driving forces of industrial districts and

Final firm 4

Final firm 2

Subcontractor

Distribution and Logistics

Subcontractor

Subcontractor

Final firm 3

Final firm 1

Market Research

Figure 4.3 Map of industrial district or cluster

R&D Agency

University

Training Agencies

Leading final firms

Subcontractor

Supplier 4

Supplier 3

Supplier 2

Supplier 1

Knowledge flows

Flows of information and learning

Flows of goods and services

In search of a useful theory 83 Geographical Factors • Natural resources • Location infrastructure conditions • Demographic factors leading to large markets for sales and inputs

Historical Factors • Accidents of history • Triggering factors

Cost advantages from clustering Innovation capabilities in new products and services

• Low level external economies • Low level of technological dynamism • Development of an institutional system not conducive to growth)

• High level of external economies • High level of technological dynamism • Development of an institutional system Initial development of cluster capabilities

Stability or decline Division of labour

Acquisition of external knowledge and production of internal knowledge

Incremental innovations Possible institutional changes

High level of learning Acquisition of market shares

Possible genesis of a new type of industrial districts/cluster reconfiguration

Development of high level networking – joint production, marketing, R&D and innovation activities Development of infrastructure based on effective learning

Initial saturation. Relocation activities to gain benefits from other locations, for example, access to cheap labour or access to new markets, etc.

Figure 4.4 Evolution of industrial districts or clusters

84 Fiorenza Belussi clusters is the existence of an efficient institutional context (Jacobs and De Man, 1996). However, there is a danger of seeking to replicate elsewhere what has been successful in one location, because of the existence of territorial specificity. This is not a linear process, because, in some localities, the path-dependent nature of local institutions and a weak innovative system might bring about a stationary situation, and the industrial district or cluster during the time may decline, and concentrations might stop. Looser systems, declining mature localized areas or moribund industrial districts or clusters have been less studied by the business literature than successful industrial districts or clusters (Belussi, 1999; Provasi, 2002). The emergence of negative externalities is a common experience of the industrial districts or clusters ‘story’. These negative externalities include congestion, cut-throat competition in final markets among local firms, increased prices for inputs and property, too much embeddedness of the institutional context, and locking-in into obsolete and/or ineffective innovation and learning systems. The district or cluster may decline also for positive reasons, because the activation of centrifugal forces disperses the local concentration of firms and pushes towards the re-location of some local firms in other countries. The reasons for changing localization include: (1) the need to access a different pool of knowledge, extending and developing a new innovation and learning systems by embracing firms and organizations situated in other locations, (2) the development of new markets, (3) the desire to gaining access to valuable assets that are embedded in other locations, and (4) the use of wage differentials. Often one of the advantages of re-location is to enter already highly developed industrial districts or clusters in order to rapidly gain access to local knowledge and capabilities. In this way, the concentration effects can be multiplied. It can also be proved that well developed districts or clusters, beyond a certain threshold, act as centripetal systems, which attract external firms and MNCs. The process of internationalization of industrial districts or clusters has been studied mainly in the light of considering the entry of multinationals (Dunning and Wymbs, 1999; Cantwell and Iammarino, 2000; Lundan, 2002; Brown and McNaughton, 2002). However, small and medium-sized enterprises (SMEs) based in industrial districts/clusters using their networks can also enter global chains (Enright and Ffowcs-Williams, 2000; Corò and Rullani, 1998). The beneficial aspects of foreign leading firms located in districts and clusters have been found in France (Dunford, 1994) and in high-tech clusters in Canada, Sweden, and the UK (Cooke and Huggins, 2002; Swann et al., 1998).

Conclusions There is no agreement in the literature about the ways in which to define and classify industrial districts and clusters. In this chapter we have suggested a certain level of convergence. In his latest writings Porter seems to be more interested in ‘territorial clusters’ than before. On the other hand, Italian economists are

In search of a useful theory 85 also now exploring new cognitive approaches to the study of industrial districts and focusing on the interdependence between the local and the global. Becattini himself, in a recent article (2002), seems to favour a terminological integrated approach: ‘Having determined that Prato was an industrial district, perhaps the archetype of all industrial districts . . . What did we find in this industrial cluster?’ (p. 97). For practical purposes it is necessary to provide a classification that is broad enough to permit the inclusion of a variety of cases (excluding cases without a significant threshold). It is also necessary to focus on geographical concentrations that involve a large number of SMEs (please note: numerical dominance and not economical dominance), on local entrepreneurship, and on industry agglomerations that are not exclusively formed by non-interactive large MNCs.

Notes 1

2 3

4

This chapter uses some material already prepared for a previous work written with F. McDonald for the paper ‘The evolution of industrial districts and policies towards them: developing policies to help enlargement of the European Union by using the experience of Western European counties – Industrial Districts’ Relocation Processes’, contract no. HPSE-CT2001-0009 – related to an EU project organized by the Tagliacarne Institute ìn Rome. Many ideas presented in the chapter have been deeply discussed in our frequent meetings in Rome together with Lucio Biggiero, Alessia Sammarra, Ivana Panniccia, Gera Sarcina, and Debora Giannini. A longer version of this chapter was presented at the 2004 Druid Summer Conference in Copenhagen. I have had the opportunity to discuss many ‘sensitive’ issues regarding the approach to the industrial district concept directly with Professor Becattini, whose illuminating vision has helped me to see ‘order through chaos’. See also Lazzeretti (2003). Clearly, there also exist many different ways to classify industrial clusters and the process of local development (Bergman, 1998; Rosenfeld, 1997; Belussi and Arcangeli, 1998; Viesti, 2000). There are also at least two major country studies that have sought to define and measure clusters (DTI, 2001; Harvard Business School, 2002), and numerous international organizations have launched a new policy area of intervention (European Commission, 2001; UNIDO, 2001; OECD, 1999, 2001). The advantage of the proposed approach is to avoid fanciful typologies, like the one proposed by Markusen (1996) of ‘new industrial districts’, where Marshallian districts are opposed, for instance, to hub-and-spoke districts (a local system dominated by one vertically integrated firm is not by definition a district!).

References Acs, Z. (ed.) (2000) Regional Innovation, Knowledge and Global Change, London: Pinter Amin, A. and Thrift, N. (1994) Globalisation, Institutions and Regional Development in Europe, Oxford: Oxford University Press Anastasia, B. Corò, G., and Crestanello, P. (1995) Problemi di individuazione dei distretti industriali: esperienze e rapporti con le politiche, Oltre il Ponte, XIII, (52), pp. 63–109 Antonelli, C. (2000) Collective knowledge, communication and innovation: The evidence of technological districts, Regional Studies, 34, pp. 535–47

86 Fiorenza Belussi Arthur, W. (1990) Silicon Valley locational clusters: Do increasing returns imply monopoly, Mathematical Social Sciences, 19, pp. 235–51 Asheim, B. (1996) Industrial districts as ‘learning regions’: A condition for prosperity, European Planning Studies, 4 (4), pp. 379–400 Asheim, B. and Cooke, P. (1999) Local learning and interactive innovation networks in a global economy, in Malecki, J. and Oinas, P. (eds), Making Connections: Technological Learning and Regional Economic Change, Aldershot: Ashgate, pp. 145–78 Asheim, B. and Isaksen, A. (2002), Regional innovation systems: The integration of local sticky and global ubiquitous knowledge, Journal of Technology Transfer, 27, (1), pp. 77–86 Baptista, R. (1998) Clusters, innovation, and growth: A survey of the literature, in Swann, P., Pervezer, M. and Stout, D. (eds), The Dynamics of Industrial Clustering: International Comparisons in Computing and Biotechnology, Oxford: Oxford University Press Becattini, G. (1979) Dal ‘settore industriale’ al ‘distretto industriale’: alcune considerazioni sull’unità d’indagine dell’economia industriale, L’industria, 1 Becattini, G. (1987) L’unità di Indagine, in Becattini, G. (ed.), Mercato e forze locali: il distretto industriale, Bologna: Il Mulino Becattini, G. (1990) The Marshallian industrial district as a socio-economic notion, in Pyke, F., Becattini, G. and Sengenberger, W. (eds), Industrial Districts and Inter-firm Co-operation in Italy, Geneva: International Institute for Labour Studies Becattini, G. (2002) From Marshall to the Italian ‘industrial districts’: A brief reconsideration, mimeo Becattini, G. (2003) From the industrial district to the district districtualisation of production activity: some considerations, in Belussi, F., Gottardi, G. and Rullani, E. (eds), The Technological Evolution of Industrial Districts, Boston: Kluwer Becattini, G. and Rullani, E. (1996) Local systems and global connections: The role of knowledge, in Cossentino, F., Pyke, F. and Sengenberger, W. (eds), Local and Regional Response to Global Pressure: The Case of Italy and Its Industrial Districts, Geneva: Ilo Becattini, G. et al. (2003) From Industrial Districts to Local Development, Cheltenham: Edward Elgar Belussi, F. (1999) Path-dependency vs. industrial dynamics: The analysis of two heterogeneous Italian districts specialised in leather upholstery, Human System Management, 18, pp. 161–74 Belussi, F. (ed.) (2000) Tacchi a spillo: il distretto calzaturiero della Riviera del Brenta come forma organizzata di capitale sociale, Padua: Cooperativa Libraria Editrice Università di Padova Belussi, F. (2002) The Italian system of innovation: The gradual transition from a weak ‘mission-oriented’ system to a regionalized learning system, in Borras, S. and Biegelbauer, P. (eds), Innovation Policies in Europe and Us: The new Agenda, Aldershot: Ashgate Belussi, F. and Arcangeli, F. (1998) A typology of networks: Flexible and evolutionary firms, Research Policy, 27, pp. 415–28 Belussi, F. and Gottardi, G. (2000) Evolutionary Patterns of Local Industrial Systems: Towards a Cognitive Approach to the Industrial District, Aldershot: Ashgate Belussi, F., Gottardi, G. and Rullani, E. (2003) (eds), The Technological Evolution of Industrial Districts, Boston: Kluwer Bergman, E. (1998) Industrial trade clusters in action: Seeing regional economies whole, in Steiner, M. (ed.) Clusters and Regional Specialisation, London: Pion

In search of a useful theory 87 Biggiero, L. (1999) Markets, hierarchies, networks, districts: A cybernetic approach, Human Systems Management, 18, pp. 71–86 Braczyk, H., Cooke, P. and Heidenreich, M. (eds) (1998) Regional Innovation Systems, 2nd edition [2004], Cambridge, UCL Press Brenner, T. (2001) Self-organisation, local symbiosis of firms and the life-cycle of localised industrial clusters, paper presented at the conference Localised Industrial Clusters, Jena, 31 January Brown, P. and McNaughton, R. (2002) Global competitiveness and local networks: A review of the literature, in McNaughton, R. and Green, M. (eds), Global Competition and Local Networks, London: Gower Brusco, S. (1982) The Emilian model: Productive decentralisation and social integration, Cambridge Journal of Economics, 6, pp. 167–84 Brusco, S. (1990) The idea of the industrial district: its genesis, in Pyke, F., Becattini, G. and Senbenberger, W. (eds), Industrial Districts and Inter-firm Co-operation in Italy, Geneva: International Institute for Labour Studies Cantwell, J. and Iammarino, S. (2000) Multinational corporations and the location of technological innovations in the UK regions, Regional Studies, 34, pp. 317–22 Club dei distretti industriali (2003) Guide to the Italian Industrial Districts, Montebelluna, Treviso, Fondazione del Museo dello scarpone Cooke, G., Pandit, N. and Swann, P. (2000) The dynamics of industrial clustering in UK broadcasting, Working Paper of the Department of Economics, University of Loughborough, Loughborough Cooke, P. (2002), Knowledge Economies: Clusters, Learning and Cooperative Advantage, London: Routledge Cooke, P. and Huggins, R. (2002) High technology clustering in Cambridge, in Amin, A., Goglio, S. and Sforzi, F. (eds), The Institutions of Local Development, London: IGU Corò, G. and Rullani, E. (eds) (1998) Percorsi locali di internazionalizzazione, Milan: Angeli Cossentino, F., Pyke, F. and Sengenberger, W. (eds) (1996) Local and Regional Response to Global Pressure: The Case of Italy and Its Industrial Districts, Geneva: Ilo Dei Ottati, G. (1996) Trust, interlinking transactions and credit in industrial districts, Cambridge Journal of Economics, 18, pp. 529–46 Dorfman, N. (1983) Route 128: The development of a regional high-technology economy, Research Policy, 12, pp. 299–316 DTI (2001) Business clusters in the UK – A first assessment, London: DTI Dunford, M. (1994) French technopoles: Cathedrals in the desert or industrial complexes for the 21st century, in Krumbein, W. (ed.), Ökonomische und politische Netwerke in der Region: Beitrage aus der internationalen Debatte, Hamburg: Lit-Verlag Munster Dunning, J. and Wymbs, C (1999) The geographical sourcing of technology-based assets by multinational enterprises, in Archibugi, D., Howells, J. and Michie, J. (eds), Innovation Policy in a Global Economy, Cambridge: Cambridge University Press Enright, M. (1996) Why local clusters are the way to win the game, World Link, 5, pp. 24–5 Enright, M. (2001) Regional clusters: What we know and what we should know, Kiel Institute International Workshop on Innovation Clusters and Interregional Competition, Kiel Enright, M. and Ffowcs-Williams, I. (2000) Local Partnership, Clusters and SME Globalisation, Paris: OECD

88 Fiorenza Belussi European Commission (2001) Methodology for regional and transnational technology clusters: Learning with European best practices, Brussels, Enterprise Directorate General Falzoni, A., Onida, F. and Viesti, G. (eds) (1992) I distretti industriali: crisi o evoluzione?, Milan: Egea Feldman, M. and Audretsch, D. (1999) Innovation in cities: Science-based diversity, specialisation and localized competition, European Economic Review, 43, pp. 409–29 Feser, E. (1998) Old and new theories of industrial clusters, in Steiner, M. (ed.) Clusters and Regional Specialisation, London: Pion Grabher, G. (1993) The Embedded Firm: On the Socio-economics of Industrial Networks, London: Routledge Guerrieri, P., Iammarino, S. and Pietrobelli, C. (2001) The Global Challenge to Industrial Districts: Small and Medium-sized Enterprises in Italy and Taiwan, Cheltenham, Edward Elgar Harrison, B. (1992) Industrial districts: Old wine in new bottles, Regional Studies, 26, pp. 469–83 Harvard Business School (2002) Cluster mapping project, Institute for Strategy and Competitiveness, Cambridge, MA, Harvard Business School IPI (2002) L’esperienza italiana dei distretti industriali, Rome, Ministero delle Attività Produttive Jacobs, D. and De Man, A. (1996) Clusters, industrial policy and firm strategy: A menu approach, Technology Analysis and Strategic Management, 8, pp. 12–22 Jaffe, A. (1986) Technological opportunity and spillovers of R&D: Evidence from firms’ patents, profits, and market value, American Economic Review, 76, pp. 984–1001 Klink, A. and De Langen, P. (2001) Cycles in industrial clusters: The case of the shipbuilding industry in the northern Netherlands, Tijdschrift voor Economische en Sociale Geografie, 92 (4), pp. 449–63 Krugman, P. (1991) ‘Increasing returns and economic geography’, Journal of Political Economy, 99, pp. 484–99 Krugman, P. (1995) Development, Geography and Economic Theory, Cambridge, MA: MIT Press Lazzeretti L. (2003) City of art as a high culture local system and cultural districtualisation processes: The cluster of art restoration in Florence, International Journal of Urban and Regional Research, 27 (3), pp. 635–48 Lundan, S. (2002) Network Knowledge in International Business, Cheltenham: Edward Elgar Lundvall, B. and Borras, S. (1997) The globalising learning economy: Implications for innovation policy, Report from DG XII, Commission of the European Union Markusen, A. (1996) Sticky places in slippery space: A typology on industrial districts, Economic Geography, 72, pp. 293–313 Marshall, A. (1890) Principles of Economics, 1st ed., London: Macmillan Marshall, A. (1919) Industry and Trade, 1st ed., London: Macmillan Martin, R. and Sunley, P. (2003) Deconstructing clusters: Chaotic concept or policy panacea? Journal of Economic Geography, 1 (3), pp. 5–35 Maskell, P. (2001) Towards a knowledge based theory of the geographical cluster, Industrial and Corporate Change, 10 (4), pp. 921–43 Narula, R. (2003) Globalisation & Technology, London: Polity OECD (1999) Boosting Innovation: The Cluster Approach, Paris: OECD OECD (2001) World Congress on Local Clusters, Paris: OECD

In search of a useful theory 89 Paniccia, I. (2002) A critical review of the literature on industrial districts: In search of a theory, in Paniccia, I., Industrial Districts: Evolution and Competitiveness in Italian Firms, Cheltenham: Edward Elgar Porter, M. (1980) Competitive Strategy: Techniques for Analyzing Industries and Competitors, New York: Free Press Porter, M. (1990) The Competitive Advantage of Nations, New York: The Free Press Porter, M. (1998a) On Competition, Boston: Harvard Business School Press Porter, M. (1998b) Clusters and the new economics of competition, Harvard Business Review, 76, pp. 77–90 Porter, M. (2000) Location, competition and economic development, Economic Development Quarterly, 14, pp. 23–34 Provasi, G. (2002) Le istituzioni dello sviluppo, Rome: Donzelli Pyke, F, Becattini, G. and Senbenberger, W. (eds) (1990) Industrial Districts and Interfirm Co-operation in Italy, Geneva: International Institute for Labour Studies Rallet, A. and Torre, A. (2004) Proximité et localisation, Economie Rurale, 280, pp. 25–41 Roelandt, T. and Den Hertog, P. (1999) Cluster analysis and cluster-based policy making in the OECD, in Boosting Innovation: The Cluster Approach, Paris: OECD Rosenfeld, S. (1997) Bringing business clusters into the mainstream of economic development, European Planning Studies, 5, pp. 3–23 Saxenian, A. (1994) Regional Advantage: Culture and Competition in Silicon Valley and Route 128, Cambridge, MA: Harvard University Press Schmitz, H. (1992) On the clustering of small firms, Industrial Districts and Clusters Bulletin, 23, pp. 64–9 Sforzi, F. (1987) L’identificazione spaziale, in Becattini, G. (ed.) Mercato e forze locali: il distretto industriale, Bologna: Il Mulino Sforzi, F. (2003) The industrial district and the new Italian economic geography, in Belussi, F., Rullani, E. and Gottardi, G. (eds), The Technological Evolution of Industrial Districts, Boston: Kluwer Signorini L. (1994) Una verifica quantitativa dell’effetto distretto, Sviluppo Locale, XXVI (1), pp. 117–39 Swann, P., Prevezer, M. and Stout, D. (1998) The Dynamics of Industrial Clustering: International Comparisons in Computing and Biotechnology, Oxford: Oxford University Press Tessieri, N. (2001) Rassegna bibliografica sullo sviluppo locale e sui sistemi locali di piccola e media imprese in Italia, in IPI, L’esperienza italiana dei distretti industriali, Rome: Ministero delle Attività Produttive UNIDO (2001) Development of Clusters and Networks of SMEs. The UNIDO programme. http://www.unido.org/doc/331111.htmls Viesti, G. (2000) Come nascono i distretti industriali, Bari: Laterza You, J. and Wilkinson, F. (1994) Competition and cooperation: Towards understanding industrial districts, Review of Political Economy, 6, pp. 259–78 Zollo, M. and Winter, S. (2002) Deliberate learning the evolution of dynamic capabilities, Organisation Science, 21, pp. 981–96

5

Cutting through the chaos Towards a new typology of industrial districts and clusters Ivana Paniccia

In search of an encompassing theory of industrial districts and clusters After more than a decade of studies on industrial districts (IDs) and clusters carried out by different scholars in various disciplines and regions of the world, the question of what precisely constitutes an ID and a cluster remains only partly answered. The conceptualization of the ID provided by Becattini – the pioneer of studies on Italian IDs – was somewhat restrictive (Pyke et al., 1990), but the term has subsequently been used by various authors with rather conflicting meanings, so that the umbrella term ‘ID’, now extensively used in various disciplinary fields, covers many different forms of the organization of work and many different socio-cultural patterns. The concept of cluster has suffered an even worse fate, owing to the fact that its main proponent – Michael Porter (1990, 1996) – put forward a concept which lacked precise conceptual boundaries, so that the term, according to the critical appraisal by Martin and Sunley (2003), has acquired such a variety of uses, connotations and meanings that it has, in many respects, become a ‘chaotic concept’, in the sense of conflating and equating quite different types, processes and spatial scales of economic localization under a single, all-embracing universalistic notion. The empirical literature on IDs, largely based on case studies, has highlighted a multifaceted world of local agglomeration economies, although it is far from including a significant number of rigorous comparative studies. On the one hand, specific and non-replicable features of single IDs have been associated with a general model albeit one not always specified; on the other, general features of the ‘model’ delivered by the literature are loosely applied – without rigorous empirical assessment – to areas which simply display a high concentration of (small) firms. This trap has ensnared both those non-Italian scholars who have used such generalizations as ‘Third Italy’ to find an empirical referent for their models of ‘flexible specialization’, or ‘new competition’, or whatever, and those Italian researchers who have too easily attributed the performances of a list of ‘official’ IDs – defined thus solely because they consist of agglomerations of SMEs – to the distinctive features of Becattini’s IDs – such as, for example, cooperation.

Cutting through the chaos 91 There is a tendency in the cluster literature to assume, on the basis of certain examples of success such as Silicon Valley, that agglomeration invariably brings with it innovation, ‘competitiveness’ or other advantages. Yet this is to overlook the fact that these cases are distinguished by numerous locational, industrial and cultural factors that interrelate and intertwine. One should instead disentangle these various factors and determine under what conditions they explain, jointly or separately, the performances observed. The approach developed here aims at unravelling the relationship between proximity and efficiency, flexibility, innovation, growth and cooperation by emphasizing the conditions under which agglomeration accompanies these features. I do not propose to reduce the complexity of ‘real spaces’ and territories to a few stylized relations of causality assembled in one all-encompassing and general model. I believe instead that it is more appropriate, given the multidisciplinary nature of the concepts under scrutiny, to adopt an eclectic approach which encompasses different currents of theory and delivers not a model but a set of models; a typology within a coherent analytical framework. In this sense, it is very far from Porter’s concern to provide a universalistic theory of clusters based on the competitive advantage theory, as a passe-partout approach. This chapter proceeds as follows. The following two subsections introduce the complexity of building a typology and provide two workable definitions of IDs and clusters. The second section illustrates the structural features, that is, the criteria used to draw up the typology, while the third section describes the typology proposed in detail, giving some empirical references for each type. The conclusions tackle the issue of the evolution of IDs and clusters and discuss the expected performance results for each type with only a short account of policy implications, given length constraints. The approach to a typology of clusters and industrial districts Two features are required for a classification to have scientific value (and not to be a metaphor or an idea) and therefore to be a typology (which is an ex ante classification, as compared to a taxonomy, which is an ex post one). First, the resulting ‘types’ or ‘clusters’ must have predictive and explanatory capacity. In other words, what is important from a scientific point of view is being able to say that if an object belongs to a cluster then other common features not already included in the clustering must derive from it. The second requirement of a good typology is that it must specify the conditions (efficiency, equilibrium, and so on) under which the classification has value (Grandori, 1990). Moreover, the types must exhibit a self-reproducing quality and some internal coherence, or an internal logic of development such as to give them sufficient stability; otherwise they will be only contingent phenomena. My typology also considers a few other attempts at classification in the international literature, in which Markusen’s (1996) and Gordon and McCann’s (2000) typologies represent a milestone, but given length constraints it does not discuss the related differences.

92 Ivana Paniccia The typology proposed is intended to encompass the largest possible number of real forms in different socio-cultural contexts, but it does not claim to be exhaustive; rather, the intention is to give a methodological account which highlights the variety of factors at work in industrial agglomeration and their combinative rules. Some preliminary definitions The typology presented is based on operational criteria, since it derives from a previous quantitative assessment of Italian IDs by the present author.1 Further work to test it on a sample of areas selected in various developed countries is currently in progress and will render it more solid. Given length constraints, operational indicators are not discussed here, while only some workable definitions are given. The term ID is used generically to denote an agglomeration of small to medium-sized enterprises (SMEs) specializing in one or a few industries in a bounded area. If the requisites of the prevalence of SMEs and small scale are relaxed, the terms ‘geographical agglomerations’ (GAs) or ‘economies of agglomeration’ may be more appropriate, and under these rather awkward headings so-called ‘clusters of firms’ are included as well. These ‘agnostic’ definitions are sufficiently comprehensive to include areas with different organizational structure, clustering processes (interfirm linkages, knowledge spillovers, rivalry, business and social networks) or institutional settings, and they prevent the attribution to the ID of precise socio-economic features (for example, horizontal and vertical networking, innovativeness, trust, and so on), which are conversely implied in the concept of the ‘canonical’ ID (Becattini 1990).2

The classification criteria By ‘deconstructing’ the various definitions of GAs put forward in the literature into their constitutive elements or factors, various dimensions of analysis are unravelled (Paniccia, 2002). There follows a brief discussion of the following classification criteria, that is, the structural features of GAs: • • • • • • • •

size of the industry degree of specialization or diversification industrial specialization (degree of decomposability and nature of technology) spatial scale range of horizontally related industries. These include industries related by common technologies, end-users, distribution channels and other nonvertical relationships range of vertically related industries. These include all the stages of activity along a supply chain type of interdependence size distribution of firms

Cutting through the chaos 93 • • • • •

competencies activity set urban setting or environment ownership and governance economic and social institutions.

The spatial concentration of firms is a prerequisite for external economies. When it comes to the level of measurement, a number of related concepts and new dimensions of analysis emerge. The concept of an agglomeration of firms seems to come closer to the idea of a large number of small firms than to that of the greatest concentration in a given region. In fact, in an area poor in firms, an index of specialization, like the location quotient for example, may yield high values relative to the presence of other industries in the area, but the number of firms may in any case be too small to generate significant external economies. This is the reason for including the number of firms and employees as measures for the industry size. The degree of specialization (or, symmetrically, diversification) of the industrial structure is another proxy for the agglomeration process and for the extent of the division of labour. A concentrated industry displays an index of specialization higher than the average. An extended process of labour division is generally associated with the formation of numerous small firms each specializing in a different task. The specialization of firms may differ according to whether a vertical or horizontal division of labour is prevalent. As explained by Brusco (1990) and Maskell (2001), the horizontal dimension triggers innovative behaviours which ultimately affect the competitiveness of districts. Further implications for effectiveness may derive from the vertical division of labour. Firstly, this produces a flexible production system, because each task can be reorganized with a different mix of specialized producers. Flexibility, in its turn, has two further effects: it enables rapid responses to be made to variations in the degree and quantity of final demand, and it accelerates innovative processes. Secondly, the extensive presence of a myriad small firms produces an even or democratic model of society (Perrow, 1992; Piore and Sabel, 1984). Given certain conditions, a vertical division of labour is able to contain the social effects (unemployment) of a reduction in demand. Once the most appropriate measure(s) of agglomeration have been decided, it is then necessary to establish at which level of geographical scale and industrial classification they should be calculated. Should the scale be that of a global network, a country, a region, a county, a city, a town or even a city-quarter? The typology presented here determines endogenously the geographical scale for each type.3 A continuous and substantial flow of goods and information between firms or the search for skilled labour may well incur lower transportation and transaction costs if they take place at a small geographical scale, which could be approximated by the Travel-To-Work-Area concept (Coombes et al. 1986) or the metropolitan area.

94 Ivana Paniccia More controversial is the relationship between knowledge and proximity. Technological innovation, in evolutionary theories, is strongly influenced by spatial proximity mechanisms that favour processes of polarization and cumulativeness (Lundvall, 1988). More recently, other authors have argued against the ‘proximity matters’ view that the mechanisms supporting knowledge flows differ greatly, and that they do not necessarily entail physical proximity as a prerequisite for knowledge exchanges (Breschi and Lissoni, 2001). Communication, learning and coordination invariably require a common language, code, routine or standard (Amin, 2000, cited in Morgan, 2004), depending on the nature of the exchange and organization being studied. Therefore, frequent interactions involving the exchange of information may also take place between distant actors provided these have a shared system of standards and codes. This juxtaposition reflects the recent debate on the role of physical or spatial proximity as opposed to that of relational or organizational proximity. The two views may be reconciled by taking an evolutionary perspective and acknowledging that physical and organizational proximity interact in a complex fashion. Organizational proximity may well be a surrogate for physical proximity in the case of exchanges that involve formal (codified) information; but cannot be used to explain, for example, unforeseeable contingencies, and at the stage when a cultural or technological standard has to be developed (Morgan, 2004). In an evolutionary perspective, codes, standards or routines and the other forms of communication and coordination mentioned develop not from scratch but through a process of socialization which is necessarily localized. Local space must in fact be viewed not in abstract or impersonal terms but in its relational nature as a receptacle, or better a medium, for exchanges of physical and intangible assets such as knowledge, information, and affective relations. This explains the inherent dynamic logic of GAs and why the scale may differ depending on the age of development of the area or industry. Concerning the level of industrial classification, reference to the industry code as it is assigned by official statistics may often be erroneous. Conversely, fruitful use can be made of the concept of filière or of interrelatedness (see below), which will have to be reconstructed ex ante on the basis of the industries’ technology base. The range of horizontally related and vertically related industries indicates the extent to which the localization of an industry stimulates the formation of complementary industries and the extent of the labour division process. The structural dimensions of the degree of specialization, the size of the industry, and the range of vertically and horizontally related industries have a direct and positive impact on the performance of the labour division process of the industries (efficiency, competitiveness, growth). However, the effect exerted on performance by the broadness of the industry is less straightforward. According to the studies on technological innovation, the relationship between users and producers and in general the integration of different technologies, may foster innovation if the industries concerned co-evolve. For example, the user industry may cease to be a stimulus for the producer industry if the former has

Cutting through the chaos 95 exigencies tied to an overly small market with specific characteristics unsuited to a global market or uses distant from those that prevail at the global level of the technology. The number and the nature of auxiliary industries is strictly related to the technological regime (appropriability, technological opportunities, degree of cumulativeness and characteristics of the technological base) of an industry and of its market. With a given technology and state of the market, it is possible to use taxonomies of industrial sectors to configure the relationships between firms on a territorial scale. Generally, studies on clusters and IDs underestimate the technological trajectory (technological opportunities) and technical constraints of an industry (Martin and Sunley, 2003). Conversely they are important predictors of an agglomeration’s innovative capacity, of different forms of organization of labour within firms, and of industry-specific location factors. Moreover, the industrial sector also controls for exogenous factors related to demand growth. It is assumed here that certain types of GAs are compatible with certain industries’ specificities in terms of degree of decomposability (Brusco, 1982), modularity, ‘fungeability’, and integrative versus diffusive complexity (Antonelli, 2003). In general it is assumed that the GAs or IDs considered are all based on decomposable industries: otherwise interconnection or interdependence – although with significant differences in degree – would not be significant, and the agglomeration would find its rationale only in traditional location factors. Moreover, since specific location factors may be distinguished for each industry, owing to the nature of their production technologies and/or marketing or consumption processes, this variable controls for the sectoral effect on clustering and enables distinction to be made between the types where the traditional location factors may or may be not accompanied by the other features discussed here. The degree of interrelatedness between technologies, the competencies required to govern specific processes of goods and knowledge production, and the model of labour division configure different interdependence linkages among the activities of the filière. Organizational studies, and particularly the ‘networks’ literature, provide some important categories for analysing interdependence by highlighting the rules of governance of relationships among firms, among actors, and between firms and social actors. Thompson (1967, cited in Grandori, 1997) identifies a parallelism between types of interdependence – ‘pooled’, ‘sequential’, ‘reciprocal’ (or twoway), and ‘intensive’ (or ‘organic’) – and types of coordination. Grandori (1997) uses this scheme to predict that the most efficient and effective mode of coordination will depend on the type of interdependence, in conjunction with two other explanatory factors: the allocation of property rights (structure of interests) and informational complexity. Vertical and horizontal relationships among firms specializing in different or similar activities in the same supply chain, or complementary industries, may be classified according to the kind of (transactional) interdependence that exists between the different activities. The

96 Ivana Paniccia kind of interdependence also enables distinct subcontracting relationships to be identified.4 As said, transactional interdependence is related to and also depends on the technical constraints of the activity concerned, and therefore on the industry’s characteristics as specified above. The more a technology is complex, the more heterogeneous and numerous are the activities and the more interdependent the actors. The mechanisms that can effectively underpin the relations between nodes of complex knowledge are based on direct interaction, certification of knowledge, and joint problem-solving that may involve proximity at different scales. The degree of labour division and the nature of the industry determine the prevailing size of firms at each stage of the filière. The ‘canonical’ literature on IDs does not specify the size of the production unit at each individual stage, but it admits a limited range of variability. It assumes, however, that the multiplicity of phases and fierce competition will keep optimal technical dimensions fairly small. Strictly speaking, this will not exclude even large firms, particularly if the district produces phase products for the outside world as well (Brusco, 1990).5 Size is also an important feature because of its implications for the endowment of competencies at firm level. The distribution of competencies among the population of a GA configures different types of proximity: that is, the degree of homogeneity of its population (including the population of firms) and the viable or compatible relationships with other ‘populations’. The competence theory of the firm (Wernerfeld, 1984), as applied to firms or regions or IDs (Lawson, 1998), adds an unexplored dimension to analysis of local agglomeration. Drawing on Penrose (1959), Teece et al. (1994) distinguish between: •

•

technical competencies, which concern the ability to develop and implement technological solutions, design products and processes, and to operate machines and facilities effectively. They are related to the technology base of the localized industry or industries. organizational or economic competencies, which concern the ability to find economically efficient solutions to activities carried out in business organizations (allocative, transactional, organizational and financial competencies). Competencies must be specified for the different actors operating in an area.

The distribution of competencies between firms may be treated in highly operational manner in order to measure their relevance to competitiveness, doing so à la Porter by looking at firms’ commercial independence in terms of the property rights on design and technical conception implied in the framework just discussed. The greater the number of firms with commercial autonomy and property rights on design and technical conception located in an area, the more added value activities are performed locally. The ‘location of strategic functions’ (or ‘activity set’, Enright, 1998), is therefore assumed to be a distinctive feature which differentiates among types according to the extent to which it is distributed among the local population of firms.

Cutting through the chaos 97 The mechanisms of formation of competencies have a pronounced territorial specification. In terms of my typology there is a clear nexus between certain kind of competencies, their place of formation and the urban and cultural specificities of places in terms of the economic or social functions (e.g. tertiary, cultural or political-administrative), size (population), local and international connections, patterns of interaction between people, and so on, that have been summarized in the classification criterion of the urban environment (Borja and Castells, 1997; Borja, 2001; Florida, 2002; Derudder et al., 2003).6 So-called ‘buzz’ cities (Storper and Venables, 2004) are a particular combination of (1) creative and cultural functions; (2) finance and business services; (3) science, technology research; (4) power and influence (government, headquarters, trade associations, and international agencies) and are deemed to be great attractors of technological competencies and creativity. From the foregoing discussion it emerges that the picture of forms of agglomerations is not complete if patterns of firms’ ownership and governance are not included among the defining features in our framework. Size depends also on a bundle of institutional factors such as the prevailing models of governance as well as other regulatory aspects (regulation of labour, bankruptcy rules, law of succession, level of taxation and administrative burdens, etc.). The source of capital, whether local, national or international, state-supported or private, is illustrative of the mechanisms responsible for the formation of the agglomeration area, whether locally embedded or externally driven, spontaneous or institution-driven. The prevailing nature of capital in an area also typifies its mechanisms of capital accumulation. As regards ownership, located firms are distinguished among local, domestic, and foreign according to the region of origin of the majority owner or headquarters location. The owners of a domestic firm originate from an area external to the district but within the same national boundaries. A classification of subsidiaries drawn from international studies can be used as regards foreign-owned companies. These studies provide important insights into the organizational structure of transnational corporations (TNCs) and their impact on recipient regions (Birkinshaw and Hood, 1998; Phelps et al., 2003). With respect to governance, a distinction is drawn among family, entrepreneurial, and managerial firms. Family firms are those where ownership and management coincide: the owner corresponds to a family, and some of its members are involved in manufacturing operations. Entrepreneurial firms are those where ownership and management coincide, but the owner is not involved in manufacturing operations and has mainly supervisory or strategic tasks; external (non-family) members may be involved in managerial roles. Managerial firms are those where ownership and management are disjoint. Ownership and governance features are to be considered jointly and have to be matched with data on the size distribution of firms. A fragmented industrial structure may actually conceal a high concentration of ownership via either equity sharing or informal forms of control (e.g. ‘groups’ of firms). The governance mechanisms of firms are a particular expression of the more general institutional environment of a place. I include economic and social

98 Ivana Paniccia institutions among the factors, along with the urban environment, that far from having a biunique relationship with the types may ensure a certain degree of variability within them. Economic institutions (local development agencies, trade associations and unions, service centres, banks and financial institutions, consortia, etc.; universities, private and public R&D laboratories, etc.) are more explicitly oriented to complementing and sustaining the motivations, cognitive schemes, and actions of economic actors than are social institutions, which perform the same role, but without having being created for economic purposes (associations, families, churches, political parties, etc.). In the ‘canonical’ literature, social and economic institutions are essential ingredients of an ID’s success (Becattini, 1990; Piore, 1990, 1992; Trigilia, 1990), as well as in Porter’s approach.

The Typology The following typology can be constructed by considering all the above criteria as having the same explanatory value:7 1 2 3 4 5 6

(Semi)canonical IDs Diversified or urban IDs Satellite platforms or hub and spoke agglomerations Co-location areas Concentrated or integrated agglomerations or IDs Science-based or technology agglomerations

The setting of the criteria in each type is illustrated schematically in Table 5.1. The following description of the types seeks to infer some other common features not already included in the criteria of classification, such as the patterns of internationalization and evolution, which are discussed in greater detail in the conclusions. A life-cycle for each type is admitted, but the typology is built with consideration taken only of the growth stage of each type. Examples of the various types are given by referring to my previous works or to other authors’ case studies. The canonical type 8 This type is characterized by an extended division of labour mostly based on family-owned and family-run firms with few employees and which rely on family members or relatives. The structure of the industry is highly fragmented, in that it consists of independently owned firms scattered among small units. The structure of the industry often derives from an old manufacturing tradition and the industry is highly decomposable and not complex (generally fashion or design-intensive industries). The network of relationships among firms is dense (star structure) and is characterized by a two-way sequential interdependence whereby firms exchange

Cutting through the chaos 99 information (bi-directional flows) and goods (one-way or bi-directional flows). This does not exclude, however, the existence of more integrated firms, networks of firms or a few (non-native) subsidiaries’ TNCs with a ‘world mandate’ to produce specialized intermediate or final products, but their presence does not alter – at least in their maturity stage – the flat model of labour division. If the internal ties are numerous, the opposite is the case as regards external ties, since these IDs have a limited degree of international openness in terms of global chains of production and transmission and reproduction of knowledge. Relationships with external markets (information and goods) and technologies are ensured by few specialized actors: wholesalers, traders or suppliers of instrumental goods or, in some cases, by final firms or even individual companies such as the ‘impannatore’ in Prato (Becattini, 1989), which act as co-ordinators at the head of an entire chain of operations. The large number of firms and the long-standing manufacturing tradition jointly ensure a high degree of self-containment, since all the specific activities in the specialization industry’s filière are carried out locally. These include the manufacturing of investment goods (machinery) and the provision of auxiliary services (transport, banks, etc.) specifically dedicated to the leading industry. Because of the pervasiveness of the dominant specialization, a marked sense of belonging is displayed by the entire community, including institutions. Governance agendas are dominated by the needs of the specialization industry. Local institutions, like local banks, trade unions, trade associations and technical schools, tend to offer services tailored to the needs of local firms and workers, but their performance, activism and extent of membership (or in other words the local ‘institutional thickness’) is not homogeneous owing to a variety of factors including the past history and the national and local political context. Entrepreneurship is local and endowed with organizational, technical and financial competencies and has generally a self-employment background. Workers are on average highly skilled. Learning occurs at the firm level, but also through inter-firm relationships, such as the sharing of common suppliers and labour mobility. The complete range of co-operative agreements, along vertical and horizontal lines, can be observed while this type is in the growth stage. In its maturity, when demand shrinks and competitive pressures increase, horizontal cooperation in terms of the sharing of orders becomes rarer, and there arise opportunistic behaviours such as labour poaching, unfair fees for subcontractors and freeriding on quality (Bertini and Forlai, 1989; Maselli, 1993; Balestri and Toccafondi, 1994; Paniccia, 1998). Cultural and income disparities are compressed because people live and work locally. This also explains why canonical IDs cannot arise in metropolitan cities with their large peripheries and often conflicting subcultures. The Marshallian triad of external economies is contained within a single urban area, which also explains their limited extent. The typical urban setting is that of a town of smallto-medium size, with low mobility (at exit) of the population since people tend to stay in the area for a long time, and which do not host significant tertiary

Located

Two-way Mixed two-way interdependence: and one-way bi-directional flows sequential of information, knowledge and goods

Activity set

Vertical and horizontal relationships (direction and content)

Located

Limited

Extended

Small

Range of vertically related industries

Small

Spatial scale

Highly decomposable; fashion-led or design industries

Limited

Highly decomposable; fashion-led or design industries

Type of industry

Small to large

Small

Satellite

Low

Small

Small

Areas of co-location

One-way sequential interdependence: subcontractors work on commissioners’ specifications and designs

Not-located

Internal interdependence

Located

Limited

Limited

Limited

Very limited

Small

Large

Varied Varied (decomposable or modular industries)

High-to-moderate Low diversification

Range of Moderately horizontally related extended industries

Low

Small to large

Large

Degree of diversification

Small

Large

Size of the industry: • number of firms • number of employees

Diversified

Canonical

Factors

Table 5.1 Distinctive features of the generic typology

Large

Large

Science-based

Two-way sequential and reciprocal interdependence: subcontractors work on their own designs; horizontal

Located

Moderately extended

Extended

Small to large

Moderately decomposable and industries complexity

Two-way sequential and reciprocal interdependence: subcontractors work on their own designs; horizontal

Located

Moderately extended

Extended

Small to large

Moderate or highly decomposable, complexity and/or fungeability

High-to-moderate High diversification in diversification service industries

Large

Small

Concentrated

Local/endogenous Local/endogenous Mixed: local small Mixed: local and or exogenous firms; mediumexogenous sized and large firms are exogenous (TNCs)

Firms’ ownership

Varied

Varied

Limited entrepreneurial/ managerial competencies and worker skills

Small to Medium or large medium-sized urban areas with urban areas. Weak tertiary functions tertiary functions

Limited entrepreneurial/ managerial competencies and worker skills

Urban setting

Diffused entrepreneurial competencies and worker skills

Size generally above the national average; in sub-type: polarized

Diffused entrepreneurial competencies and worker skills

Average size depending on the type of industry: Size distribution differentiated by subtypes: (1) even distribution or (2) polarized

Type of competencies

Mostly micro and small firms

Mostly small firms

Size distribution of firms

Mixed local and multinational

Medium-sized urban areas with tertiary functions

Diffused entrepreneurial/ managerial competencies, managerial abilities and worker skills

Mixed proportion of small and medium-sized firms. A few large firms

co-operation among firms on RTD & R&D centres

continued

Mixed local and multinational

Medium-sized or large urban areas with tertiary functions and international projection

Mainly managerial and entrepreneurial competencies, scientists

Mixed proportion of small-, medium- and large-sized firms

co-operation among firms on R&D projects

(Very) active role

Family and technical schools, political parties or voluntary associations

Economic institutions

Social institutions

Like canonical and knowledge institutions (schools of design, TV, studios)

Low-tointermediate role

Family (mostly) or Family or entrepreneurial entrepreneurial firms firms

Firms’ governance

Diversified

Canonical

Factors

Table 5.1 continued

Family for endogenous entrepreneurs; dismissed large firms

Very weak role of local institutions; national institutions may be strong

Mixed ‘Rationalised product subsidiaries’

Satellite

Family for endogenous entrepreneurship; dismissed large firms

Very weak role of local institutions; national institutions may be strong

Mixed

Areas of co-location

Family and technical schools, trade associations

Active local, and national institutions; supporting role of knowledge institutions

Small firms are either entrepreneurial or family-based and large or leading firms are managerial; some ‘product mandate subsidiaries’

Concentrated

Universities (community of scientists) ex alumni associations

Very active role of knowledge institutions; venture capital

Entrepreneurial small firms and managerial large or leading firms; ‘product mandate subsidiaries’

Science-based

Cutting through the chaos 103 activities. A distinctive feature of these towns is the large number of places where the local population meets and exchanges information and ideas (e.g. the Italian piazze). The most prominent examples of this type are to be found in Italy, and they include the cases of Prato, Porto Sant’Elpidio and Santa Croce, at least in the 1990s. Diversified IDs or urban IDs This type can be considered a variant of the canonical type, but it has its own logic of existence that impedes a view of it as merely a preliminary stage in the development of the canonical type: it may in fact evolve into a different type. Diversified or urban IDs are characterized by a smaller number of firms, mainly final firms surrounded by a variable high proportion of subcontractors with commercial autonomy: a feature which also explains their less extended labour division and their less extended range of complementary activities or industries (like those providing investment goods), and a high degree of diversification in other unrelated manufacturing and service industries. Related to the less extensive division of labour is a slightly larger average size of firms, ceteris paribus. As a consequence, firms are related by mixed two-way and one-way sequential interdependence. Their urban settlement offers business services and expertise, larger and more diversified labour markets and cultural amenities, and it is also well equipped to foster design and craft-based activities.9 Final production has distinguishable qualities and most of the final firms have their own trademarks and trade in the national and more often international markets. The diversification of economic activities is an asset for the area either in the ‘high road’ perspective of converging technological trajectories, which may push the leading sector towards a pattern of rejuvenation and innovation, or in the much more frequent ‘low road’, or non-convergent perspective, that emerges in the case of a sectoral crisis, which may smooth the conversion of people and assets in the diverse localized sectors, including tourism. These two perspectives depend critically on the nature of the industry, the degree of interrelatdness of industries and the urban setting. In this sense, embeddedness in an urban setting is a constitutive feature for this type, and, differently from the canonical type, there is no need to have a coincident scale for the urban economies with agglomeration economies. This type is settled either in urban contexts, both small towns – such as for example the Italian case of Pesaro (a region town) – or large cities, as for example the audiovisual district of Prague (WEID, 2004), or in large city boroughs like London’s Clerkenwell in long-ago 1861; or they may well be compatible with the new so-called ‘intermediate locations’ all of which have access to the pecuniary economies of urban settlements (Phelps, 2004). Therefore, urban IDs are those industrial agglomerations that combine or are able to reap the benefits of a specialization of inputs within a moderate division of labour, with the opportunities for access to a diversified set of activities and far-flung relationships.

104 Ivana Paniccia This, in its turn, also affects the formation pattern of entrepreneurship, whether linked to an endogenous (often artisan) or to an exogenous background. The institutional setting of these areas is not strictly functional to the features of the leading industry and the local policy agenda is not determined by its needs. The satellite type(s) Satellite GAs generally comprise a limited number of small firms (compared to the ‘canonical type’, but not in absolute terms and not compared to the diversified type) usually operating as subcontractors to larger commissioning firms, which may be localized in the same area or outside. The degree of labour division is therefore not particularly marked, and a horizontal pattern tends to prevail. Firms’ size, however, is not necessarily small, but reflects contingent factors or a higher degree of inefficiency (given the technical constraints of the industry). Sequential interdependence with only one-way flows of goods characterizes the relationships between suppliers or subcontractors and final firms. There is little evidence of horizontal cooperation, while there may be cases of vertical co-operation between the commissioning firms and the local suppliers or subcontractors which entail supplier quality upgrading, the shortening of delivery times and inventory control. Spin-offs may occur from the few leading commissioning firms, in some cases directly supported by the former, which loan machinery or ensure the stability of orders for a limited period, no differently from what occurs in canonical (and diversified) IDs in their formation stage. From this it follows that the subtype with commissioning firms, either national or TNCs (depending on the subsidiary’s mandate), located therein may have a major impact on the development of these areas. The ownership of commissioning firms may be local; but for a very small proportion of them, it may be national-domestic or foreign-owned. According to the presence of larger firms, their ownership, and the amount of work subcontracted locally, there may be two variants of this type resulting in different degrees of self-containment: First, where the local commissioning firms are foreign-owned, a sub-type of ‘TNCs-led’ (satellite) GAs may be distinguished: a few TNCs supported by clustered supply chains or rather dependent SMEs specializing in modular and complex industries, for example automotive.10 The role of subsidiaries is that of a rationalized product subsidiary, specializing in the production of a particular intermediate input to be used elsewhere in the group. These TNCs lack strategic functions, such as R&D, marketing or design. In most cases, the non-local entrepreneurship at the helm of larger firms has been attracted locally by the availability of financial or fiscal incentives, or by specific attraction policies, as in reconversion economies such as Wales, Birmingham and North-RhineWestphalia. Second, when the presence of large firms is negligible and the industry is not modular or complex, these areas may be better defined as external subcontracting

Cutting through the chaos 105 areas. The subcontractors are dependent on external commissioning firms and specialize in one or a few specialized phases of the production cycle, those with more labour content. In the areas specializing in final consumption goods, a metalworking or mechanical engineering tradition, and, more generally the development of auxiliary industries, does not occur. Examples are provided by many southern Italian cases, a few regions specializing in clothing in Slovakia and other Central and Eastern European IDs (Smith, 2003; WEID, 2004), Toulose and Sophia-Antipolis in France in the 1960s (Longhi, 1999). The rationale of agglomeration resides in the reduction of costs and delivery times and the availability of a specialized labour market in an endogenous scenario of location, but also the availability of incentives in an exogenous scenario. The scale of industrial agglomeration is independent of the urban one. Areas of co-location This type is characterized by the co-location of firms specializing in similar activities, performing most of the activities of the filière and producing for the final market generally through the intermediation of buyers. A recent manufacturing tradition and the scant availability of seed capital explain both why firms undertake most of the stages of the filière internally, and why few of them have a large investment base. The majority of firms do not have their own trademark and supply local or regional markets with non-tailored goods, in contrast for example to diversified IDs. The range of horizontally related industries is very limited, because agglomeration economies are not yet fully developed. A substantial difference with respect to satellite types is that the strategic functions, although not very advanced and often typical of exogenous firms, are located in the area. Whether companies’ prevailing ownership is local, this type may be seen as a preliminary stage of the canonical or urban/diversified types, but its resilience suggests considering it as a distinct model. In this case they arise from a dismissed incubator firm (whose existence may have derived from the location of specific inputs), the upgrading of an artisan tradition or the transformation of a different industrial activity. On these ‘pre-existences’ depend the characteristics of the urban setting, whether this is the suburb of a metropolitan city or a small town, which generally occupies a small space. Traditional urbanization economies do not exert a specific influence on the firms’ performance, although they may affect their enfolding evolutionary pattern, among other factors, with particular reference to access to critical technical and managerial competencies. These circumstances are to be found in the early development of IDs in certain Italian regions (e.g. the Marche region) back in the 1960s or 1970s, and currently in Eastern and Central Europe where large (integrated) privatized firms have been established by managers of the previous state-owned enterprises (Smith, 2003; WEID, 2004). Whether these areas have attracted greenfield investments or partnerships in similarly vertically integrated firms (‘miniature replicas’), the endogenous nature of the model is weakened and its evolution become subject

106 Ivana Paniccia to the risks of footloose investments. In this variant, the attraction factors may have been the availability of a competent workforce or premises or specific incentives. Romania offers examples of the genuine endogenous types to be found in textiles and wood processing, respectively, in Timis and Mures counties (INCLUD). A hybrid between the subcontracting and the co-location types may be imagined, consisting in the coexistence of generally SMEs, but also a few large firms producing for the final market, with locally owned subcontractors. Final firms may either contract out a significant proportion of their production to local small firms or not do so if they are integrated. In the latter case, they subcontract only a small fraction of their production locally. We may term this variant a dualistic co-location area. Examples may be found in southern Italy and in the Romanian counties of Arad, Banat and Crisana, specializing in shoe manufacturing (WEID, 2004), where integrated firms are often TNCs. As in the case of the satellite type, economic institutions are very rare because the local specialization is too weak to generate demand for specialized public services or active trade unions and employers’ associations. The concentrated or integrated type This type is characterized by two groups of firms which interact with each other. On the one hand, there are relatively larger and vertically integrated firms, often at the head of ‘formal’ or ‘informal’ group of firms and with a market or technological leadership; on the other, there are small firms specializing in more specific tasks and supplying components or services to the former group of firms, or operating as final firms. The leading firms are either national (local) or multinational. In the latter case, subsidiaries have a ‘world’ or ‘regional product mandate’ and they oversee the entire range of activities (production, R&D, finance, marketing, etc.) related to one line of production or one product which is then marketed on the global market or in macroregional markets. The predominant specialization, whether in fashion or design-led and specialized industries (in Italian cases), is integrated with a strongly developed mechanical industry which makes the tools and machines used in the leading industry’s production process. Unlike the canonical type, where a wide range of complementary activities similarly exist, these have developed their final markets. In this sense, concentrated types are an evolution of the canonical type. Moreover, integration extends to other complementary industries vertically or laterally. Reciprocal interdependence is common in these districts: it may come about between firms, but it is much more frequent among the internal R&D departments of more innovative firms. The IDs specializing in mechanical engineering industries also include cases of ‘intensive’ interdependence, which involves the joint application of complementary resources to a common activity in an integrated manner. This typology also comprises cases of user–producer

Cutting through the chaos 107 interactions for innovative activities, which are related to the local diffusion of technological and managerial competencies. Owing also to the complex nature of the industry of specialization, these areas foster the development of high-level technological competencies and related knowledge institutions. Like the urban type, these areas benefit from a diversified economic structure and have already set off on the ‘high road’ of a converging technological trajectory which pushes the leading sector towards a pattern of innovation. This pattern, which requires a rupture in the self-contained nature of these areas, is closely dependent on the quality of the urban setting to attract young talents. The web of economic institutions is quite dense in the Italian case, where chambers of commerce, employers’ associations and trade unions are particularly active. Mirandola (biomedical industry), Montebelluna (sportswear and shoes) and Sassuolo (ceramics and ceramic tiles) are among the most prominent examples of this type in Italy (Biggiero, 2002; Corò et al., 1998; WEID, 2004). The technology or science-based ID/GA type This type may be distinguished as a variant of integrated IDs when the specialization industry by convention can be labelled ‘science-based’ according to Pavitt’s classification. These industries are characterized by rich technological opportunities, complexity and/or fungeability, and firms generate a range of products whose product life-cycles tend to be short. The organization of labour is of network type. Small and large firms coexist. TNCs with a ‘world’ or ‘regional product mandate’ are present. Entrepreneurship is not native on average but depends on the agglomeration of scientists or engineers and specialized suppliers of materials, equipment and services. These places are the locations for important scientific and communication knowledge infrastructures, such as universities and private and public research centres. Together with alumni associations and others, universities also act as social institutions: that is, as ambits of socialization and as arenas for the exchange of ideas and reputation building. Similarly, the internal organization of private firms may foster experimentation and non-conventional views which facilitate innovation. Common codes of communication and work ethics do not develop among natives who absorb them from infancy; rather, they are absorbed and sustained by people with different social and national backgrounds, who join these ‘knowledge’ communities after apprenticeship on undergraduate courses. Besides ‘knowledge’ institutions, another distinctive institution of this type is venture capital, which is the main sources of finance, while the equivalent institutions in canonical IDs were local banks and families. Leading firms focus on what they do best, acquiring the rest of their inputs from the dense local infrastructure of suppliers, and partly from outside, and they collaborate with key suppliers to define and manufacture new products or processes. But co-operation also takes place between firms and institutions. Inter-firm networks spread the costs and risks of developing new technologies

108 Ivana Paniccia and foster reciprocal innovation among specialized firms. Organic interdependence applies. These districts are located in generally very pleasant urban and metropolitan areas, which act as magnets for talent; environments which provide far more positive settings for communication (city ‘buzz’) and hence more opportunities to foster the rate of technological change. The density of interaction among people and among firms in metropolitan areas is not dissimilar to that which occur in canonical IDs, but the interactions take place among heterogeneous and yet complementary agents. Moreover, because of the nature of interactions and different embedding of economic actors, these GAs tend to be dispersed over a large spatial scale (even a region), after the first stage of development of the industry. As regards the degree of specialization, this type is characterized by a wide multisectoral range of economic activities including both manufacturing and service industries. Examples are provided by computer manufacturing (Silicon Valley, USA), ICTs (Sophia-Antipolis, France; Madrid, Spain), biotechnology (Cambridge, UK) and aerospace (Toulouse, France), to cite just a few (Saxenian, 1994; Longhi, 1999; Garnsey, 2002; Rama et al., 2003; Cooke and Huggins, 2002).

Conclusions For the proposed typology to have predictive capacity, it is necessary to associate some performance results with each type. Overall, efficiency and market effectiveness effects vary according to the intensity of the agglomeration. Canonical, integrated and technology IDs or GAs are therefore the most competitive types. In this regard, my typology is also an attempt to render discrete the continuous relations that may be depicted by a logistic function between agglomeration and efficiency (or certain performances), although the effects of agglomeration per se are combined with other structural factors intended to represent the control variables for agglomeration. Another way to analyse the ‘normative’ value of the various types is to evaluate the impact of exogenous conditions on the evolutionary pattern of ID types. Canonical IDs are very able to absorb fluctuations in the economic cycle. The rate of growth of employment in a sample of IDs has been found to be higher than the national average, while the rate of unemployment was systematically lower than the corresponding regional average (Paniccia, 2002). These IDs have similarly been able to cope with changes in tastes or with pressures on costs, while they are vulnerable to structural changes in technology or final markets, but also new standards or organizational procedures. These IDs are likely to experience lock-in phenomena and decline, which hamper their successful reconversion in cases of radical change, unless unpredictable events occur or effective policies are enacted. The major impediments to the upgrading of these areas reside in their entrepreneurial model (competencies and governance structure of firms) and in the weak attraction of new resources exerted by their

Cutting through the chaos 109 urban hosting environment. Since local firms are mainly based on families, the survival of a system of this kind depends on familial channels for the intergenerational transmission of the corporate assets; channels which are necessarily subject to random factors. Other challenges for these districts are the depletion of natural resources (shortage of land, pollution of air and water) and the social integration of migrants. The role of traditional social institutions which characterized Italian IDs at least until the 1990s tends to weaken, and needs to be replaced by that of other less local institutions. Compared to canonical IDs, diversified areas are less able to adapt to short-run market and non-radical technological changes. Their lower level of internal variety (Maskell, 2001) impedes experimentation with new organizational or technical solutions within the dominant specialization industry. On the other hand, these IDs may benefit from their greater diversification when a long-term crisis hits the dominant manufacturing industry along what has been called the high or low road. Satellite GAs are among the weakest forms of agglomerations, although they may experience high growth rates of employment and exports. They are in competition with less favoured regions (LFRs), given that their typical production is close to mass or standardized production and that they often have low-cost locational advantages. If relocation occurs in these areas – given the footloose nature of local FDIs – it may impair their further development, whether this concerns high or low added value activities. In the former case, it impairs the local formation of the competencies required. In the latter case, relocation may trigger disruptive competition among subcontractors if it is used as a threat against local subcontractors in order to keep the prices of intermediate inputs low. The disruption of employment, the disappearance of specialized firms, even if low-level ones, together with a strengthening of the larger firms, may destroy the nascent identity of the district. Conversely, when these areas offer better economic conditions together with a higher level of competence than, for example, LFRs do, relocation is less likely to occur, and they may prosper. Co-location areas are less exposed to international competition because of the protected nature of the final market for local firms. Similarly weaker is the risk of relocation for the endogenous variant, but – on the other hand – they may not benefit from a potential propelling role of larger firms, if they are not localized. Moreover, their dependence on local markets is questionable given the increasing integration of international markets. Concentrated IDs are better equipped than canonical IDs to deal with most external challenges. The presence of more closely integrated firms linked to suppliers specialized in different industries, and generally with other centres of knowledge production, may reduce the risk of not being able to tap technological and market changes in advance. Moreover, concentration moves in the upstream and downstream industries may be more easily countered (than, for example, in canonical or diversified IDs) because of the bargaining power

110 Ivana Paniccia of firms. However, they may similarly experience problems of resources shortage, migrant integration, disaffection of natives (particularly the younger generation) with their living places, and despoilment of the natural environment, as documented in the Italian cases. Competitive pressures may nevertheless induce a shrinkage in those suppliers or subcontractors (generally small firms) that do not interact reciprocally with leading firms. To the extent that concentrated IDs are characterized by a dense network of economic and social relationships, they tend to be ‘sticky’ to the places where they are localized. Investments (greenfield, alliances, and other forms) by TNCs in these areas are motivated by an endeavour to tap context-specific knowledge. This does not rule out, however, that some phases of production may not be delocalized in the search for non-locally available resources (related to the shortage of personnel) or for low-cost intermediate inputs, with detrimental effects on local subcontractors. Technology GAs are apparently able to adapt to new events and to revitalize themselves even beyond the limits of their own industry’s life-cycle. They cannot be considered immune from risks of decline, in that, for example, the mechanisms of knowledge production and reproduction or the selection mechanisms may be hampered (due to the risk of ‘elitism’, for example – as Cooke and Huggins, 2002, noted). Another challenge resides in the intrinsic relationship of dependency of this type with the hosting environment: rent-seeking behaviours, interorganizational rivalries or the lack of integration with the new professional communities that inevitably develop in such a dynamic environments, and with the other communities that live in the same space. Some typical patterns of evolution from one type to another can be imagined, but the evolutionary processes do not follow a given sequence, nor should they be taken as deterministic. The structural features identified mark out a space of probability where certain outcomes are more likely than others. It may be noticed, however, that the evolutionary patterns of the types so far discussed, owing to the increasing complexity of co-ordination, all converge on the strengthening of horizontal and vertical relationships, which also induces a shrinkage of ‘fringe’ subcontractors. Moreover, these relationships tend increasingly to traverse the boundaries of the GAs or IDs, because of enhanced inward or outward internationalisation. Generally speaking, when national and international demand shifts towards a greater degree of complexity, variability and volatility and when production is carried out in different locations, more formal co-ordination mechanisms or integrated organizations (including ‘networks’ of firms) tend to emerge. This process, far from being a mere return to hierarchy, represents a strengthening of the ‘buzz’ within organizations which may well interact with the ‘buzz’ of the city. However, the final configuration that the ‘new’ IDs and GAs will assume at the end of this process of deep change is not easily detectable at this stage and would require further analytical effort.

Cutting through the chaos 111 Policy implications This contribution has offered a concise account of the variety of GAs and of the factors that explain their formation, evolution and performance. The combination of factors that describe the most successful types do not easily occur. These often emerge as the outcome of an idiosyncratic historical process of formation matched by a favourable external demand scenario. Pursuing clusters’ competitiveness ultimately requires a rich basket of policy tools, which may range from urban planning to technology policy, to mention only some. Moreover, certain tools are difficult to manage for local or regional policymakers, because they require a higher level of governance. Neglecting the complexity of GAs and IDs may lead to ineffective and wasteful cluster policies. When the reference model of policy-making is too distant from reality, the risk is to use unfitting tools (e.g. service centres that local firms will never use) or target fictitious objectives (e.g. enhancing alleged externalities which are not relevant to the history of specialization).

Notes 1 My typology and its underlying operational criteria have some similarities with that used in a research project funded by the European Commission (WEID), to which I contributed as an external consultant. The results of this project did not make possible a sound and complete test of my typology. However some of the case studies included in this project are taken as illustrative of my typology and are referenced as WEID, 2004. 2 Some Italian researchers on IDs (e.g. Belussi in this book) may be unsympathetic to my use of the term ID, by retaining this term indissolubly linked to the conceptualization provided by Becattini and his co-authors. This is only a terminological question. 3 From an operational point of view, this fact creates a problem of circularity, since the point of departure of the investigation process coincides with one of its outcomes. It is possible, however, to use certain identification methods that yield a continuous space of agglomerated areas, by starting from a small scale (e.g. Iuzzolino, 2004). 4 In a recent paper (Rama et al., 2003), the subcontracting relationships are measured in terms of incidence (of subcontracting), directionality, durability of relationships, producers’ motivations for externalizing production, and other producer characteristics. 5 On an operational ground, the variable concerning size should be weighted against the industry average in order to smooth effects due to inefficiencies. 6 This factor, which I have not considered in my previous work, may be operationalized by using different taxonomies of the urban space available in the literature (for example, Scott, 2001, or Derudder et al., 2003, who provide a classification of ‘urban arenas’ by calculating an index of city connectivity) or by calculating the rates of specialization or density of different activities, including amenities, business services, and the flows of people mobility. 7 This does not exclude that by imposing a hierarchy over the factors, yielding high and low-rank factors, a smaller typology with a higher number of variants may result. 8 For the sake of simplicity, I use the label ‘canonical’ instead of ‘semi-canonical’, which would be more appropriate since the former actually denotes the theoretical conceptualization of Becattini (1990), as explained in my previous work.

112 Ivana Paniccia 9 This feature also explains why I defined the Italian IDs belonging to this type as ‘craft-based’ in my previous works (Paniccia, 1998, 2002). 10 The geography of the automotive industry in Europe is characterized by several GAs in different regions. The WEID project, for example, covers four cases of clustering in the automotive industry located in Slovenia, Saxony (Germany), West Midlands (UK) and in Mlada Boleslav (Czech Republic). Given the organization of this industry into different levels or ‘tiers’ of suppliers through to final car producers, the classification of the different GAs in this type, or rather in the ‘integrated’ one, depends critically on which ‘tier’ of suppliers is overrepresented.

References Antonelli, C. (2003) Knowledge Complementarity and Fungeability: Implications For Regional Strategy. Regional Studies, 37 (6/7), pp. 595–606. Balestri, A. and Toccafondi, D. (1994) Imprenditori e distretti industriali. Prato: Edizioni Pratofutura. Becattini, G. (1989) Modelli locali di Sviluppo. Bologna: Il Mulino. Becattini, G. (1990) The Marshallian Industrial District as a Socio-Economic notion. In F. Pyke, G. Becattini and W. Sengenberger (eds) Industrial Districts and Inter-Firm Co-Operation in Italy. Geneva: ILO, pp. 37–51. Bertini, S. and Forlai L. (1989) Prospettive del contoterzismo nel quadro dell’evoluzione del modello Pratese. Bologna: Nomisma. Biggiero, L. (2002) The Location of Multinationals in Industrial Districts: Knowledge Transfer in Biomedicals. Journal of Technology Transfer, 27 (1), pp. 111–22. Birkinshaw, J. and Hood, N. (1998) Multinational Corporate Evolution and Subsidiary Development. Houndmills: Macmillan. Borja, J. and Castells, M. (1997) Local and Global: The Management of Cities in the Information Age. London: Earthscan. Borja, J. with Muxi, Z. (2001) L’espai public: Ciutat i ciutadania. Barcelona: Diputacio De Barcelona. Breschi, S. and Lissoni, F. (2001) Localised Knowledge Spillovers vs. Innovative Milieux: Knowledge ‘Tacitness’ Reconsidered. Papers in Regional Science, 803, pp. 255–73. Brusco, S. (1982) The Emilian model: Productive decentralisation and social integration. Cambridge Journal of Economics, 6 (2), pp. 167–84. Brusco, S. (1990) The Idea of the Industrial District: Its genesis. In F. Pyke, G. Becattini, and W. Sengenberger (eds) Industrial Districts and Inter-Firm Co-Operation in Italy. Geneva: ILO, pp. 10–19. Cooke, P. and Huggins, R. (2002) High Technology Clustering in Cambridge. In A. Amin, S. Goglio and F. Sforzi (eds) The Institutions of Local Development. London: IGU. Coombes, M. G., Green, A.E. and Openshaw S. (1986) An Efficient Algorithm to Generate Official Statistical Reporting Areas: The Case of 1984 Travel-to-Work Areas Revision in Britain. Journal of the Operational Research Society, 10, pp. 943–53. Corò, G., Gurisatti, P. and Rossi, A. (1998) Il distretto sport system di Montebelluna. In G. Corò and E. Rullani (eds) Percorsi locali di internazionalizzazione. Competenze e autorganizzazione nei distretti industriali del nord-est. Milan: F. Angeli. Derudder, B., Taylor, P.J., Witlow, F. and Catalano, G. (2003) Hierarchical Tendencies and Regional Patterns in the World City Network: A Global Urban Analysis of 234 Cities. Regional Studies, 37 (9), pp. 875–86.

Cutting through the chaos 113 Enright, M. (1998) Regional Clusters and Firm Strategy. In A. Chandler, P. Hagstrom and O. Solvell (eds) The Dynamic Firm: The Role of Technology, Strategy, and Regions. Oxford: Oxford University Press. Florida, R. (2002) The Rise of the Creative Class: And How It’s Transforming Work, Leisure, Community and Everyday Life. New York: Basic Book. Garnsey, E. (2002) Il caso Inglese: L’impresa business e civic dell’high tech di Cambridge. Studi Organizzativi, pp. 9–31. Gordon, I.R. and McCann, P. (2000). Industrial Clusters: Complexes, Agglomeration and/or Social Networks? Urban Studies, 37, pp. 513–32. Grandori, A. (1997) An Organizational Assessment of Interfirm Co-Ordination Modes. Organization Studies, 18 (6), pp. 897–925. Grandori, A. (1990) Epistemologia ed economia aziendale: Note per un dibattito. L’industria, XI (4), pp. 76–103. INCLUD (Industrial Cluster Development) Project, for the period 2000–2006. Interreg III B CADSES Programme (the Community Initiative for the European Regional Development Fund), http://www.includ.net/. Iuzzolino, G. (2004) Costruzione di un algoritmo di identificazione delle realtà distrettuali sulla base di indicatori di agglomerazione settoriali. In Economie locali. Modelli di agglomerazione e apertura internazionale. Nuove ricerche della Banca D’Italia sullo sviluppo territoriale. Atti Del Convegno. Bologna. 20 Novembre 2003. Banca D’Italia. Marzo 2004. Lawson, C. (1998) Towards a Competence Theory of the Region. Cambridge Journal of Economics, 23, pp. 151–66. Longhi, C. (1999) Networks: Collective Learning and Technology Development in Innovative High Technology Regions: The Case of Sophia-Antipolis. Regional Studies, 33, pp. 333–42. Lundvall, B.Å. (1988) Innovation as an Interactive Process: From User-Producer Interaction to the National System of Innovation. In G. Dosi, C. Freeman, R. Nelson, G. Silverberg and L. Soete (eds) Technical Change and Economic Theory. London: Pinter Publishers, pp. 349–69. Markusen, A. (1996) Sticky Places in Slippery Space: A Typology of Industrial Districts. Economic Geography, 72 (3), pp. 293–313. Martin, R. and Sunley, P. (2003) Deconstructing clusters: Chaotic concept or Policy panacea? Journal of Economic Geography, 3 (1), pp. 5–35. Maselli, M. (1993) Mercato Terzisti E Tariffe. Pratofutura, 51, 1–3. Maskell, P. (2001) Towards a Knowledge-Based Theory of the Geographical Cluster. Industrial And Corporate Change, 10 (4), pp. 921–43. Morgan, K. (2004) The Exaggerated Death of Geography: Learning, Proximity and Territorial Innovation Systems. Journal of Economic Geography, 4, 3–21. Paniccia, I. (1998) One, a Hundred, Thousands Industrial Districts: Organizational Variety in Local Networks of Small and Medium Sized Enterprises. Organization Studies, 19 (4), pp. 667–99. Paniccia, I. (2002) Industrial Districts: Evolution and Competitiveness in Italian Firms. Cheltenham and Northampton, MA: Edward Elgar Publishing. Penrose, E. (1959) Theory of the Growth of the Firm. Oxford: Oxford University Press. Perrow, C. (1992) Small-Firms Networks. In N. Nohria and R.G. Eccles (eds) Networks and Organizations. Boston, MA: Harvard Business School Press, pp. 445–70. Phelps, N.A. (2004) Clusters, Dispersion and the Spaces in Between: For an Economic Geography of the Banal. Urban Studies, 41 (5/6), pp. 971–89.

114 Ivana Paniccia Phelps, N.A., Mackinnon, D., Stone, I. and Braidford, P. (2003) Embedding the Multinationals? Institutions and the Development of Overseas Manufacturing Affiliates in Wales and North East England. Regional Studies, 37 (1), pp. 27–40. Piore, M.J. (1990) Work, Labour and Action: Work Experience in a System of Flexible Production. In F. Pyke, G. Becattini and W. Sengenberger (eds) Industrial Districts and Inter-Firm Co-Operation In Italy. Geneva: International Institute For Labour Studies, pp. 52–74. Piore, M.J. and Sabel, C.F. (1984) The Second Industrial Divide. New York: Basic Books. Porter, M. (1990) The Competitive Advantage of Nations. London: Macmillan. Porter, M. (1996) Competitive Advantage: Agglomeration Economies and Regional Policy. International Regional Science Review, 19 (1), pp. 85–94. Pyke, F., Becattini, G. and Sengenberger, W. (eds) (1990) Industrial Districts and InterFirm Co-Operation In Italy. Geneva: ILO. Rama, R., Ferguson, D. and Melero, A. (2003) Subcontracting Networks in Industrial Districts: The Electronics Industries of Madrid. Regional Studies, 37 (1), pp. 71–88. Saxenian, A. (1994) Regional Advantage: Culture and Competition in Silicon Valley And Route 128. Cambridge, MA: Harvard University Press. Scott, A.J. (2001) Global City Regions: Trends, Theory, Policy. Oxford: Oxford University Press. Smith, A. (2003) Power Relations, Industrial Clusters, and Regional Transformations: Pan-European Integration and Outward Processing in the Slovak Clothing Industry. Economic Geography, 79 (1), pp. 17–40. Storper, M. and Venables, A. J. (2004) Buzz: Face-to-face contact and the urban economy. Journal of Economic Geography, 4 (4), pp. 351–70. Teece, D., Rumelt, R., Dosi, G. and S. Winter (1994) Understanding corporate coherence. Theory and evidence. Journal of Economic Behaviour and Organisation, 23, pp. 1–30. Trigilia, C. (1990) Work and Politics in the Third Italy’s Industrial Districts. In F. Pyke, G. Becattini and W. Sengenberger, Industrial Districts and Inter-Firm Co-Operation In Italy. Geneva: International Institute For Labour Studies, pp. 160–84. WEID (2004) ‘West-East Industrial Districts. Industrial Districts’ Relocation Processes: Identifying Policies in the Perspective of EU Enlargement, project funded by the European Commission under the 5th Framework Programme (1998–2002) Key Action ‘Improving the Socio-economic Knowledge Base’. www.west-east-id.net. Wernerfeld, B. (1984) A Resource-Based View of the Firm. Strategic Management Journal, 5 (2), pp. 171–80.

6

Entrepreneurs as agents in the formation of industrial clusters Maryann Feldman and Johanna L. Francis

Introduction Entrepreneurs play a special role in cluster formation. Entrepreneurs start firms that capitalize on technological opportunities, adapting scientific breakthroughs and generic technologies to create new product markets and reorganize economic activity. Through their individual decisions in creating and developing new companies, entrepreneurs may collectively also spark regional industrial transformation. Cluster formation is a complex and self-organizing process that occurs in developmental stages. Agglomeration economies emerge over time from the activities of individual entrepreneurs and the institutions that co-evolve to support them. The attributes observed in a mature and fully functioning cluster are artifacts of the formation process and reflect attributes and relationships formed as the cluster developed, rather than preconditions for cluster development. Cortright and Mayer (2001) conclude that no general set of conditions was responsible for particular industrial clusters in the United States; instead, unique factors appeared to be associated with each. This is not very satisfying from either an academic or a policy perspective. An alternative view is that cluster formation is a process predicated on the actions of entrepreneurs and their symbiotic relationships with their local environments. The cluster and its characteristics therefore emerge over time from the individual activities of the entrepreneurs and the organizations and institutions that evolve to support them. Markusen (1996) draws a distinction between places that are sticky and able to hold on to new ideas and translate them into industrial clusters and places that are slippery and not able to benefit in the long term from innovation and investment. This provocative characterization does not address the process by which regions change their status, transforming from slippery to sticky places creating a sustainable industrial cluster. Moreover, its focus on the region ignores the way in which regions change and evolve. Specifically, the role of individual economic agents and their optimizing decisions in shaping local institutions and conditions have not been accorded much attention in this type of analysis. A large part of the economic development discourse is informed by attempts to replicate the characteristics associated with a fully functioning regional system:

116 Maryann Feldman and Johanna L. Francis including attributes such as a pro-technology-transfer local research university, an active venture capital industry, social networks and adequate support services. However, the role typically ascribed to these factors appears to lag rather than to lead cluster formation (Feldman, 2001). In addition, the literature usually treats institutions as static exogenous factors rather than as evolving, adaptive social constructs (Nelson and Winter, 1982). The specific configurations and relationships between institutions, as shaped by evolving economic interests, may matter more than simply their presence. More importantly, this perspective ignores the importance of entrepreneurs as economic-change agents, able to create or attract the necessary resources and institutions to support their ventures, as well as draw on the rich historical and regional context in which they operate. Models of regional economic development largely ignore the role of the individual change-agent in the development of regional economies (Appold, 2000), and have not incorporated the way in which entrepreneurs actively interact with and shape their local environments (Boschma and Lambooy, 1999). The main perspective we advance in this chapter is that entrepreneurs spark cluster formation and regional competitive advantage. Entrepreneurs, in the process of furthering their individual interests, may act collectively to shape their local environments by building institutions that further the interests of their emerging industry and, in this way, form innovative industrial clusters. This conceptualization draws from the literature on complex systems, emphasizing that systems of innovation are not due to predictable linear processes but instead rely on the adaptive, self-organizing behavior of entrepreneurs, who in turn rely on support from their local environments, including government resources. Thus, while many seek to emulate the sustained competitive advantage that an industrial cluster represents, these dynamic systems cannot be simply imitated but require the temporal development of unique and not easily replicable assets and capabilities (Feldman and Martin, 2004). Many of the unique factors associated with cluster genesis determined by Cortright and Mayer (2001) are also instrumental in shaping the general form and evolution of the cluster. Institutions, including universities and anchor firms, government policy and historical experience or path dependency are still important factors in cluster evolution and may explain cluster location and timing (Feldman and Desrochers, 2003; Feldman and Francis, 2003). Entrepreneurial activity – through active learning and experimentation, the reinvestment of profits and expertise, the extension of relationships with universities and government labs, the building of local institutions such as industry networking associations and the subsequent pull of a new group of actors to the region – shapes the local environment. This model is abstracted from empirical observation of the U.S. Capitol region experience and a comparison of this experience with the experiences of several other American high technology clusters. The Capitol region initially lacked attributes associated with an entrepreneurial environment; yet, despite this history, over the past fifteen years innovative clusters in biotechnology and the Internet have become established (Feldman, 2001; Feldman and Francis, 2004). Clearly clusters that are initiated

Entrepreneurs as agents 117 or encouraged by government, such as the cluster in Oulu, Finland, or the Dae-Duk Science Park in South Korea, did not have the same entrepreneurial driven origins that Silicon Valley or the Capitol cluster had (Hyry et al., 2003; Sripaipan, 1993). Their origin was carefully planned and their growth nurtured through co-operation among firms, local government and area universities. Whether such clusters take hold (in Markusen’s typology ‘become sticky’) and grow relative to the size of their public subsidy depends to a large degree on the decisions of individuals – whether entrepreneurs at the helm of newly formed small firms or chief executives of long established multinationals – and the actions they take to build locally grounded institutions that sustain and enhance their individual firm-level activity. The model we present here is based on appreciative history-friendly theorizing and is evolutionary in spirit (Boschma and Lambooy, 1999; Malerba et al., 1999; Teubal and Andersen, 2000). The intention is to provide a model of cluster formation that does not rely on trying to replicate wholesale the conditions of existing clusters but instead focuses on cluster genesis as a process that moves through common stages with defined inflection points. While the history, early conditions and individuals associated with every cluster may be unique, there are policy prescriptions that can be discerned from considering commonalities in the path of their development. This chapter emphasizes the ability of entrepreneurial individuals to create a cluster while building their firms, attracting and creating resources and community. Certainly, the literature would be enhanced with more comparative case studies that considered cluster genesis and provided nuanced typologies of the variations in cluster formation processes and dynamics (for example, Kenney and Patton, 2004). At the end of the chapter, we also discuss how our theory of entrepreneurial-driven cluster formation and growth coheres with planned clusters such as many of the Continental European and Asian clusters.

The life-cycle of cluster formation, development and maturation Case studies on the formation of industrial clusters suggest that a complex, self-organizing process is at work (Chiles, Meyer, and Hench, 2002; Feldman, 2000; Feldman and Francis, 2001); the typical formation of an entrepreneurial environment appears to be characterized by three general stages. In the initial stage, the region is inert: few, if any, entrepreneurial start-up companies exist within the industry of interest. The region may have assets in terms of universities, government labs, and large companies, but it does not have significant entrepreneurial activity. The movement from latent to active entrepreneurship appears to be in response to some exogenous shock. For some regions, the exogenous shock may be corporate mergers and acquisitions, such as occurred with the New Jersey electronics industry (Leslie and Kargon, 1994). In Washington, DC, government downsizing and budgetary stringency made self-employment a

118 Maryann Feldman and Johanna L. Francis viable and attractive option. These exogenous and unanticipated factors lowered the opportunity cost of entrepreneurship.1 Potential entrepreneurs may also be particularly sensitive to changes in capital gains tax rates because the majority of their future income will come from capital appreciation on their company equity (Poterba, 1989; Gompers and Lerner, 1999). In this simple model, the second stage is the formation of the cluster. At this stage, learning and adapting to new events and responding to changes in the policy environment are important in the development of the cluster. Entrepreneurial activity is creative and innovative and therefore the specific needs of the entrepreneur cannot be predicted a priori, but are developed over time as start-ups are created and their needs recognized. Maskell and Malmberg (1999) conceived of industrial clusters as ecologies of mutually dependent firms and institutions. Looking at clusters such as Silicon Valley, the Texas conurbation, Research Triangle Park, and the Capitol region, we see that entrepreneurship responded to each unique environment differently, creating clusters with their own signature characteristics, as well as different abilities to withstand external shocks (Engelking, 1999; Feldman, 2001; Leslie and Kargon, 1994; Link, 1995; Saxenian, 1994). The ultimate result is a fully functioning entrepreneurial environment within an innovative and adaptable industrial cluster. Networks of entrepreneurs, policy makers, and secondary industry contractors spring up; universities, colleges, and technical centers recognize the need for high-tech personnel and offer training programs to satisfy that demand. The final stage of cluster development is the establishment of a critical mass of resources. At this point, the location has established a reputation as the place to be for a particular technology. Consider the case of Cambridge, Massachusetts, where massive amounts of new investment from large pharmaceutical companies have been attracted to the biotech cluster. In the next sections, we will consider each of the phases in cluster formation and development, focusing on the role of the entrepreneur. The first phase, the emergent phase, occurs when entrepreneurial innovation is ignited by a confluence of exogenous events. The self-organization of the cluster and the deepening of self-reinforcing feedbacks among entrepreneurs, enterprises, institutions and resources characterize the second phase. The third phase is the maturation of the industry into the well functioning, rich innovative and entrepreneurial system.

Entrepreneurs as cluster sparks: what creates a cluster? One critical concern in studying industrial clusters is the choice of the initial conditions or the situation prior to the emergence of a cluster. For example, Kargon, Leslie, and Schoenberger (1992), emphasized the early input of Frederick Terman as the founder of Silicon Valley: He orchestrated the creation of a world-class research institute with strong ties to the business community and an environment that encouraged students to become entrepreneurs or at least be actively involved in corporate research programs. Recent work moreover

Entrepreneurs as agents 119 documents a strong preexisting tradition of university–industry interaction and a leveraging of government contract work (Sturgeon, 2001). Daniel Patrick Moynihan (1927–2003) once said that no one without 40 years to spare should get involved in urban renewal, and this may be extended to considerations of regional economic transformations. For example, Link (1995, 2002) found that the genesis of Research Triangle Park was predicated on some 70 years of patient government investment. In the U.S. Capitol region, changes in government policy toward employment and outsourcing created a frenzy of start-up activity, which began rather humbly with systems integrators and biological services firms. Over 30 years, entrepreneurs reinvested in the region and created conditions that drew resources to the cluster (Feldman, 2001). Even in Silicon Valley, for example, a small group of people with a vision for the development of the region championed the aeronautical and electronics industries and the region was poised to benefit from technological advance in computers (Sturgeon, 2001). The factors that promote the creation of a new cluster are different from those that encourage its growth. In many of the U.S. clusters we have discussed, such ‘old economy’ factors as managerial skill, latent entrepreneurial talent and skilled labor were critical determinants for their origin. Several of the clusters – notably Silicon Valley and Boston–Cambridge – became quickly known at their outset as ‘hotbeds’, attracting entrepreneurial skill from other parts of the country (Bresnahan, Gambardella and Saxenian, 2001). These experiences are very different from the origins of several European and Asian clusters, such as the high-technology clusters in Finland and Sweden, the Hsingchu Science Park in Taiwan and the DaeDuk Science Park in South Korea. In each of those cases, careful planning by local governments, firms, and universities was the proximal cause of their creation. These were attempts at urban renewal, planned from the ground up to transform regions that had exhausted their early potential – whether in natural resources or in cheap manufacturing products. Cluster formation in these cases had a dual purpose: to solve unemployment and other local social and economic problems and secondly to derive a competitive advantage (Porter, 2000). Our approach can be placed within the complex adaptive systems model, where innovation is a non-linear process. It is complementary to the triple helix model of cluster formation (Etzkowitz, 2000). In the triple helix model, clusters form from the interaction among university/research, private firm/industry and policy partners. Here the emphasis is on the interactive as opposed to the linear model of innovation. But not all of the clusters we discuss fit the triple helix model, where the constituent actors are not only interlinked but switch roles over time. In the triple helix approach, the innovative firms become research hubs, taking over some of the innovative activity typically associated with the university, and universities become incubators of science entrepreneurs, taking over a role characterstic of large firms. Research Triangle Park in North Carolina is an example of interlinked, multiple role actors where some firms over time oriented toward pure research.

120 Maryann Feldman and Johanna L. Francis Innovation is a complex process predicated on the actions of individuals Classical work on entrepreneurial activity, such as Schumpeter (1939), Knight (1921), and Kirzner (1973), suggests that entrepreneurs have a greater ability to perceive opportunity, accept challenges, and organize resources. Blanchflower, Oswald, and Stutzer (2001), drawing on this tradition, theorized that the differential ability to perceive opportunities and subsequently act them is the most significant factor affecting the decision to become an entrepreneur and is a good predictor of the success of the venture. Any geographic region can have latent entrepreneurship, that is, individuals who prefer to be self-employed or who desire to become entrepreneurs but who do not act, for a variety of reasons, which may include risk aversion, insufficient start-up capital, lack of opportunity, lack of innovative ideas to develop, and barriers to new firm creation, among others. Further, entrepreneurial ability (or the entrepreneur’s marginal product) is unevenly distributed across individuals, and, more importantly, particular skills are unequally developed. To be a biotech entrepreneur, for example, a background in bioscience is typically required; moreover, most of the current biotech entrepreneurs also took innovations or licensed innovations from their own labs to create their companies. Clearly these skills are not evenly distributed across the country. The policy question is, therefore, how to translate latent entrepreneurship into active entrepreneurs and how to provide potential entrepreneurs with the skills they need to become high-tech entrepreneurs. Schumpeter (1976: 82–3), in his emphasis on the evolutionary nature of capitalism, proposed that a shock to the system of production is required for technological change to occur and that it is this crisis or opportunity to which individuals respond. The destruction of the old means of production through the creation of a new means of production was an important part of Schumpeter’s innovation (capitalist) cycle. Bygrave and Hofer (1991: 19) suggested, ‘The essence of the entrepreneurial process is a fundamental discontinuity in the industry involved.’ However, just as Schmookler’s (1966) scissors metaphor posited that innovation is simultaneously the product of supply and demand, entrepreneurship may similarly require the convergence of technological opportunities and a perceived reduction in the risk or opportunity cost of starting a new venture. Carroll (2002) and Hurst and Lusardi (2002) found that the entrepreneurial decision is correlated with the receipt of an inheritance or some economic windfall. Feldman (2001) determined that entrepreneurship might also be motivated by a decrease in job security or career advancement possibilities. Our conceptualization posits that some initial change, whether a crisis, a discontinuity in an industry, or a technological opportunity, creates the impetus for latent entrepreneurs to become active and engage in starting companies. If the entrepreneurial decision is sufficiently sensitive to exogenous factors, rather than merely a function of preferences, then it can be influenced by government policy.2 Government policy may either facilitate this transference or inhibit its realization.

Entrepreneurs as agents 121 Entrepreneurship is local Entrepreneurship is inherently a local phenomenon; individuals start companies in the location where they have formed business networks and have access to resources (Delaney, 1993; Feldman, 2001; Romanelli and Feldman, 2003; Stuart and Sorenson, 2003). Individuals start companies on the basis of their prior experience and interests, typically fulfilling some niche that a larger corporation may judge too small, exploiting a new opportunity that may have a risk profile unsuited to a larger corporation, or using a unique set of skills and knowledge to develop applications from licensed patents. In building their companies, entrepreneurs rely on their local contacts, connections, and knowledge of the business environment. Many individuals have location inertia because of such reasons as family mobility constraints, location preferences, familiarity with the environment, costs associated with changing residence, or the cost of establishing a new company in a thickly populated environment, where office and housing costs tend to be higher. As one interviewee rhetorically asked, ‘If you are changing your job, would you also want to complicate your life by changing your residence?’ Those few who do relocate tend to move to a location of some prior attachment, such as where they went to school or received training or where they have family or some other social connection. This contrasts with entrepreneurial enterprises attracted to a region where they have neither the connections nor the attendant loyalty that roots them in the local community. For a region attempting to create a cluster, understanding what factors inhibit potential entrepreneurs from starting companies may suggest policy venues to address. For example, Germany recognized that its bankruptcy laws created a barrier to the formation of new companies; this recognition could potentially be far more important for developing entrepreneurial ventures than tax incentives. We use the term innovative cluster to refer to a geographically defined collection of related firms. While many industries exhibit a tendency to cluster spatially to a resource requirement or by historical accident, we focus on innovative activity that involves the production of novel products and processes. The ability to innovate provides long-run sustainable advantage for a firm or a region (Porter, 1989). Innovative industries are knowledge-intensive and incorporate new advances that may originate in scientific discoveries, such as in the biotech or nanotech industries, or in the application of know-how developed through practice as in industrial equipment manufacturing or specialty foods. Innovative firms often defy classification by standard schemes as they create an industry or industry segment by responding to market opportunities typically operating in niches not profitable for larger or more established firms. The nature of innovation makes it difficult to plan industrial clusters. At its earliest stages, before technological breakthroughs are generally appreciated and potential applications are known, locating at the center of innovative activity may provide critical competitive advantage. Realizing the potential of a technology requires a sophisticated understanding of consumer needs, existing markets for product innovation and factor inputs, and prevailing production technology. Co-location

122 Maryann Feldman and Johanna L. Francis increases awareness of emerging trends and reduces uncertainty for firms: innovation clusters spatially in locations where knowledge externalities reduce the costs of discovery and commercialization. By the time an industry is sufficiently well known as to be targeted for economic development, those jurisdictions where the technology was first developed have probably already captured the lion’s share of the benefits and are positioned for greater advantage. The path of emerging industries is difficult to predict and extremely fluid, and planning efforts based on current assumptions will never be able to anticipate future scientific developments or the direction that a technology may take (Lambooy and Boschma, 2001). The decision to start a firm can be viewed from a variety of perspectives (Bhide, 1999, provides a review). Neoclassical economics assumes that risk aversion and uncertainty limit the supply of entrepreneurs (see, for example, Holtz-Eakin et al., 1996). Interestingly, although entrepreneurial households report higher risk tolerance than non-entrepreneurs in surveys, there is uncertainty as to whether causality runs between risk-taking and subsequent success or if successful entrepreneurship leads to lower risk aversion (Carroll, 2002; Blanchflower and Oswald, 1998). We build on the Schumpeterian tradition by assuming that the entrepreneurial decision is a complex mix of individual preferences, opportunities, and access to resources. Our model posits some initial change – whether a crisis, a discontinuity in industry, or an opportunity – that creates the impetuous for latent entrepreneurs to start companies. This sets into motion a series of interactions within the institutional, economic, and policy environments that influence the success of a region in maintaining these start-ups and furthering the maturation of the cluster to create stickiness in Markusen’s (1996) typology. The dynamics between the types and quality of resources, the networks and institutions that provide support and further business interests ultimately affect the sustainability of the start-ups. This model is in contrast to the typical stories describing the evolution of Silicon Valley, Route 128 or Research Triangle Park – a story that highlights the role of strong directed efforts. For example, Kargon, Leslie and Schoenberger (1992) emphasize the early input of Frederick Terman as the founder of Silicon Valley who orchestrated the creation of a world-class research institute with strong ties to the business community and an environment that encouraged students to become entrepreneurs or at least to take active involvement in corporate research programs. Similarly, Route 128 may be traced to a long tradition of university-industry interaction and leveraging of government contract work (Kenney and von Burg, 1999) while Research Triangle Park was built on dedicated public sector efforts (Link, 1995). The stories may not be easily adapted to other regions. In contrast, we posit that the underlying formation of a cluster is predicated on an entrepreneurial environment that is more organic and self-organizing and progresses through cumulative stages. But even for the clusters whose origin can be traced to strong leadership from industry or academe, resolute entrepreneurial initiative was what moved the cluster from an idea to a reality.

Entrepreneurs as agents 123 The characteristics of the cluster therefore emerge from the individual activities of entrepreneurs, and organizations and institutions that co-evolve to support them. Once the first ventures have been started, the process of entrepreneurship is a classic trial and error or learning-by-doing process (Zaltman et al., 1973). In this sense, the ability of local firms to learn and adapt to new events are important determinants in the development of the cluster. Maskell and Malmberg (1999) propose that technology clusters may be conceptualized as ecologies of mutually dependent firms and institutions dedicated to learning and knowledge creation. In the first stages, these relationships begin to form.

Entrepreneurs as cluster shapers: cluster evolution The second phase is dominated by increased entrepreneurial activity. During this stage, entrepreneurs defined resources to promote and protect their interests. In this way, the independent actions of entrepreneurs are catalytic components of a self-organizing system. Commercializing technology requires a vision of how that technology will be used, who the consumers will be, what attributes they might value, and how to introduce the product to the market. The direction of the cluster’s evolutionary path and whether the cluster successfully matures are a function of the collaboration and shared vision of this community. At the second stage, the nascent cluster enters a critical time in which networks and community form as essential factors in the development of the cluster. Having the experience and example of the initial start-ups, the successful cluster becomes self-sustaining: entrepreneurs attract physical and human capital to the area, public and private networks are built up to support and facilitate the ventures, relevant infrastructure is created through public and private initiatives and services grow up to feed these companies. It is in the creative response of the entrepreneur to his or her environment that the nature and stability of the cluster are determined. Looking at clusters such as Silicon Valley, the Texas conurbation, Research Triangle Park, and the Capitol region, we see that entrepreneurs responded to their unique environments, creating clusters with their own signature characteristics and with differing abilities to withstand external shocks (Leslie and Kargon, 1996; Saxenian, 1994; Link, 1995; Engelking, 1999; Feldman, 2001). In other regions, what started as a promising industrial cluster may re-direct toward research and development where only one or two large firms establish themselves – as in the case of the New Jersey electronics industry – rather than promoting new firm formation (Leslie and Kargon, 1994). The ultimate result is a fully functioning entrepreneurial environment within an innovative and adaptable industrial cluster. The success of the initial startups and the synergy between them generates new possibilities for further startups and new spin-offs. Networks of entrepreneurs, policy makers, and secondary industry contractors spring up and universities, colleges and technical centers recognize the need for high-tech trained personnel and offer programs to satisfy that demand. The success and experience of the initial activity further generates local recognition of the nascent industry. Local recognition, a reduction in risk,

124 Maryann Feldman and Johanna L. Francis and increased opportunities created by the first wave of companies, contribute to more start-up activities. At this stage, we see the creation of regional public sector financing and grantgiving programs. Government policy creates further incentives for investment. Incubators and other technology partnerships are created to promote growth of the industry, and mergers and acquisitions begin to thin out the companies. Successful entrepreneurs also move from their initial start-up to start other companies, becoming serial entrepreneurs with deep roots in the community. Additionally, venture capitalists re-locate to the area or open branch offices signifying that the region has succeeded in one of the factors the literature cites as important. The factors described above, when added together, represent a whole that is larger than its parts and a system which may be self-sustaining and reinforcing. The mature state of this cluster is not very different from Silicon Valley in terms of venture capital, major research university involvement, university–technology transfer practices, and the types of social capital and entrepreneurial efforts observed. The developing cluster is often described as an ‘innovative milieu’. The concept of the innovative milieu (e.g., Camagni, 1995) focuses on the rich co-operation among firms and the capacity of the environment to sustain and promote this interaction (Camagni, 1995; Hyry et al., 2003). Although brilliantly successful clusters highlight high technology in the U.S., there are also examples of cluster failures – clusters that were not able to adapt to shocks and whose entrepreneurs and start-up activity folded or went elsewhere when the environment changed (see Leslie and Kargon, 1994, for an example or Wallsten, 2004). The ability to promote and sustain entrepreneurship, to offer incentives to reconfigure resources and to adapt to changing circumstances yields long-term economic growth, regardless of geographic scale. When entrepreneurs engage in building external resources they certainly further their own interests but to the extent that these resources benefit others locally they also further the development of a cluster. Technological change is path-dependent One striking fact that emerges is that the history of each cluster is unique, suggesting that cluster development is path-dependent or heavily influenced by chance historical events (Kenney and Von Burg, 1999). Although technological innovation is radical and disruptive, the earliest automobile manufacturers adapted techniques from carriage makers and served the same functions and markets. Similarly, both Silicon Valley and Route 128 built on their prior expertise in electronics. Historical events matter in determining the success of future industries in particular locations. Technological progress involves adapting general-purpose breakthroughs to serve existing markets and consumer needs. General-purpose technologies are adapted to commercialization by those who have experience with current product markets. This process of co-invention is highly localized owing to the

Entrepreneurs as agents 125 nature of knowledge creation and its initial application. As soon as a technology reaches a stage where it can be codified, it easily transfers across geographic space. However, at its earliest stages, before it is capable of codification, locating near the center of innovative activity provides critical competitive business advantage. This is one of the reasons innovations cluster geographically. Evidence on the location of the biotech industry highlights the importance of the location of the chemical and pharmaceutical industries, especially their headquarters and research and development laboratories (Gray and Parker, 1998; Orsenigo, 2001: 81–2; Zeller, 2001). Orsenigo (2001) suggested, ‘The pre-existence of a strong pharmaceutical national industry, with some large internationalized companies may have been a fundamental prerequisite for the rapid adoption of molecular biology.’ He also noted: The strength of the local science base is important but may not be the only factor in accounting for the development of the biotech industry. The biotech industry in Italy developed in Milan, which did not have the toprated academic research while Naples, an important academic center, did not develop a biotech industry. (Orsenigo, 2001) By the time a technology is known to the economic development community, it is probably too late for state governments to begin investing with the intent of pulling companies out of an established cluster to relocate in their jurisdiction: the established centers have an advantage (Cortright & Mayer, 2001). To the extent that society values the overall impact on development of an industry, state and local incentives may actually impede economic growth by creating bidding wars. Moreover, to the extent that these incentives redistribute resources from those locations where they would be most productive, the overall innovation system is compromised.

Critical mass and system maturation The development of high-technology clusters is not a deterministic process. However, there are several factors associated with cluster maturation and stability: strong industry networks, supportive local culture, and the ability to withstand re-configuration (Andersen and Teubal, 1999) or adverse shocks (Saxenian, 1998). Critical to whether an initially promising cluster matures or declines is entrepreneurial ‘spawning’. Gompers, Lerner and Scharfstein (2003) focus on entrepreneurial spawning in the venture-capital innovation hubs of Silicon Valley and Boston Route 128 and find that entrepreneurial experience and networks are critical factors in the creation of new firms. Younger venture-capital-backed firms are an important source of entrepreneurs for new venture-capitalbacked startups. The results suggest that working in entrepreneurial firms

126 Maryann Feldman and Johanna L. Francis exposes potential entrepreneurs to relevant networks of suppliers and customers, and provides information about starting companies and attracting venture capital backing. These factors lower the barrier for individuals to spinoff their own firms. In effect, these firms act as incubators for entrepreneurial human capital. This result is consistent with phase three in our model where the entrepreneurial human capital developed within earlier start-ups provides entrepreneurs with the ability to leave their initial firms to create spinoffs. In addition, this research points to the paramount importance of developing venture capital within a cluster as a means to foster more successful spinoffs and prevent the cluster from stagnating or dissipating. There are no quantitative measures to indicate when a mature cluster exists, but there are two important characteristics. The first relates to labor market thickness. A mature industrial cluster offers a myriad of job opportunities in an industry, making it possible for an employee to change jobs without changing residence. The second characteristic relates to the ability of the region to withstand economic downturns. As industries evolve and change, we expect that clusters would be able to remake themselves through continual reinvestment and reconfiguration of resources. Glaeser (2003) offers the example of Boston, which had remade its economy three times since the Colonial era, owing to the availability of locally skilled capital. See also Boschma (1999). Of course, these transitions are costly to individuals and their families when skill sets become obsolete and jobs disappear. This reiterates the importance of social policy as a backbone of industrial competitiveness and economic growth.

The entrepreneur and the cluster We have presented a model of cluster genesis that suggests clusters have their origins in a confluence of events: opportunity, existence of raw materials, including ideas and skilled human capital, and the reduction of risk. But, how common is this experience in the creation and evolution of other technology-intensive clusters? Examination of other U.S. clusters reveals that their early genesis was pathdependent and idiosyncratic – with entrepreneurial activity and firm strategy playing a decisive role. Feldman and Schreuder (1996) examine the historical origins of the pharmaceutical industry in the mid-Atlantic region and find that a series of serendipitous events created an early industry concentration. However, over time an infrastructure developed that provided firms in the region with competitive advantage. As a result, more firms and resources were attracted to the region. Indeed, the pharmaceutical industry still exhibits a strong geographic concentration in the state of New Jersey. Klepper (2002, 2004) considers the emergence of Detroit as the leading automobile cluster in the U.S. His work highlights the successive generations of entrepreneurs working in the industry. Most importantly, he finds that the pedigree and experience entrepreneurs acquired from working for Olds Motor Works, a leading innovator at the time, was instrumental to their firm success and also to the growth of the region. Scott

Entrepreneurs as agents 127 (2004) examines the early genesis of the Hollywood motion picture industry. He finds that a highly successful business model was developed and diffused there – an example of the co-evolution of industry and technology. While each of these clusters had very different origins, over time they developed the supporting conditions that the literature associates with successful entrepreneurial environments. Moreover, these regions are so successful that they are identified with the industries that are concentrated there. Similarly, the genesis of Silicon Valley high-tech clusters, while often traced to the efforts of Frederick Terman, the visionary dean of engineering at Stanford University who developed close industry–university partnerships, has roots that extend back to the early twentieth century. Sturgeon (2001) provides an historical view of the development of firms well in advance of the renowned spinoffs originating from Fairchild Semiconductor and argues that the strength of the aeronautical and electronics industries championed by a small group of people with a vision for the development of the region created the hightechnology conurbation. The clusters we examine here are exclusively within the neo-liberal AngloSaxon tradition, where individual initiative, minimal public intervention, in the form of taxation or subsidies, and private enterprise are the traditional driving forces of economic activity. The formation of innovative clusters is certainly not unique to this context. In this section, we consider two clusters that formed under very different initial conditions – the ICT cluster in Finland and the Hsinchu Science Park in Taiwan. The high-technology cluster in Oulu, Finland, formed around the Technopolis technology park created in 1982 and grew from linkages among Nokia, as an anchor firm, government planning (in particular, local city councils), and the Oulu University. Although the cluster formed out of a planning initiative by the University and the city of Oulu, 19 local firms participated in the Park’s formation and shared the risk of its failure (Hyry et al. 2003). In this case, we can point to a specific planning initiative promoted by academe, business, and government as the origin of the Oulu ICT cluster. Unlike the Capitol region, which had no similar planning initiative, Oulu’s earliest stage was mapped out. Subsequent growth also occurred by design as a new association was formed (The Northern Light Association) to develop the business environment, including executive education, and aid (credit and business advice) to young firms. However, the input of entrepreneurs was critical from the genesis to the maturation of the cluster. It was the early entrepreneurial firms who guided the development process and directed the creation of resources that they required – resources that benefited more than the firms that lobbied for them. In this way a community of common interest emerged that coalesced into a cluster. But without the tri-partite approach to regional transformation spurred on by national and local government desires to overcome the economic and social problems facing northern Finland, it is unlikely that a high technology cluster would have been created or grown in the area long associated with natural resource extraction.

128 Maryann Feldman and Johanna L. Francis Taiwan’s Hsinchu science based Industrial Park is another example of a government-executed plan for high-technology development. Hsinchu was chosen as a location for a science park on the basis of proximity to Tsing Hua University, Chiao Tung University, and the Industry Technology Research Institute. Within the Hsinchu Science Park, there are six high-technology industries actively promoted by government policies. The Taiwan government pursues a highly activist policy in promoting high-technology growth through facilitating human capital development, providing start-up capital, preferential land tax and other taxes, and providing a supportive management environment (Wang 2001). Unlike even the experience in the Oulu cluster, the Taiwanese government actively intervened in determining technologies to promote, where to put R&D funding, and in encouraging movement of inventions from the laboratory to industry. In addition, Taiwan pursued a policy of attracting overseas skilled labor, particularly scientists with experience in Silicon Valley. The Hsinchu Science Park is therefore at the opposite extreme, in terms of central planning, from the Capitol region cluster. Although entrepreneurial activity is encouraged with subsidies, tax breaks, and credit, it could not be viewed as the critical factor in the Hsinchu cluster formation or evolution. In this case, entrepreneurs provide another input to the national innovation system that is almost as ‘inert’ as the innovations themselves. In this section, we have considered two cases of cluster formation whose evolutionary trajectory is very different from the American clusters we examined. This is not to suggest that institutional factors, particularly the role of the state in promoting economic growth through regional transformation, are unimportant. Other theories of cluster formation focus on political-institutional factors such as whether the political environment is neoliberal or social corporatist (see for example Chang, 1994). According to these neo-institutional theories, clusters in countries with Anglo-Saxon models of minimal public intervention in economic activity (which would include the U.S., U.K., Canada, and Australia) may respond to shocks or structural change differently from clusters in countries with an interventionist paradigm. Although the political environment is certainly important in determining the way the cluster forms and grows, we consider how the individual entrepreneur acts within his environment. While we have focused on ‘how do clusters form and grow?’ other theories of cluster formation focus on the question: ‘why do industries cluster?’ Porter (1989, 2000) posits that industries cluster for competitive advantage created by the knowledge spillovers more easily captured by the rich ‘neighborhood’ of firms, universities, and their service industries located in close proximity. According to Porter, clustering provides firms and regions (as well as nations) with a competitive edge because they increase the productivity of the constituent firms and provide an environment that stimulates innovation. Our theory of entrepreneurial-led cluster formation and development is a complement to Porter’s theory: firms cluster because of the economic advantages clustering provides. The entrepreneur chooses to locate his or her new firm within the cluster to take advantage of these spillovers.

Entrepreneurs as agents 129 Similarly Maskell (2001) contends that clusters form because closely located firms in the same or related industries enhance the ability to create knowledge by variation and a more specialized division of labor. According to Maskell, clusters exist as a result of the improved knowledge creation that takes place horizontally across firms engaged in similar processes and vertically through the input/output-linked firms in the cluster. This learning by doing (and observing) or imitation allows firms located in the cluster to grow at a faster rate and to become more efficient. There is some evidence that, particularly within certain industries, clustered firms have better success rates than those outside the cluster (Orlando, 2000). Undoubtedly these reasons support the entrepreneur’s choice of location. Particularly new firms with inexperienced entrepreneurs would benefit from being ‘nurtured’ within the supportive environment of the cluster. Venture capitalists are also more able to identify and screen new projects or firms when they are located in areas known for that productive activity.

Conclusions The economic success of Silicon Valley – in terms of individual wealth creation, corporate profits, and job creation – has been so impressive that government officials in locations across the world try to imitate or replicate its success. Government policies aimed at replicating the conditions that exist in the Silicon Valley region today are based on the belief that other local areas may capture the benefits of new high-technology firm formation with the attendant economic growth. Conditions that we associate with an entrepreneurial environment are the result of a functioning entrepreneurial environment and do not illuminate the early efforts by which such entrepreneurship first took hold and allowed the cluster to develop. We know a great deal about the characteristics and functioning of mature industrial clusters such as Silicon Valley, California and Route 128 near Boston, Massachusetts (see, for example, Saxenian, 1994; Roberts, 1991). Critically important from a policy perspective is the question of how a cluster is started in a region that previously would not be characterized as innovative. We have a limited understanding of how such clusters develop and why they occur in certain areas and not others. For example, looking at the electronics industry, Leslie and Kargon (1994: 217) wonder, ‘Why should the electronics industry evolve so differently in one place than in another, despite common technologies and national markets? Why, for instance, should Silicon Valley be located in northern California rather than in northern New Jersey?’ We develop the argument that the location of entrepreneurs with the skills and opportunity to capitalize on an emerging technology significantly affects where hightechnology clusters emerge. Is replication of a mature entrepreneurial environment sufficient to foster entrepreneurship? Notable failures in such places as San Antonio suggest it is not (Wallsten, 2004). Saxenian (1994) analyses Silicon Valley from the perspective of how it adapted to restructuring in the semiconductor and computer

130 Maryann Feldman and Johanna L. Francis industry and establishes the importance of social relationships in defining the capacity of the region to evolve, adapt to shocks and accommodate to new demands. In this chapter, we have examined how one region, initially lacking an entrepreneurial tradition, accomplished the transformation into a fully functioning rich regional system. Such a transformation entails a fundamental shift or phase change from an inert innovative system to an active system. Certainly the Capitol region was the site of a large government research infrastructure, classified as a State-anchored region using Markusen’s (1996) typology. In this regard, the concentration of resources and highly skilled labor, plus access to sophisticated, demanding technology users, were pre-existing conditions. The transformation to private-sector entrepreneurial growth did not appear to represent movement along a technological trajectory (Kenney and von Burg, 1999) but rather was a product of cumulative capacity building brought on by exogenous shocks and involved human agency, adaptation and evolution. Currently, a myriad of economic-development policies attempt to encourage entrepreneurship. Rather than being actively promoted and encouraged by economic development policies, some clusters, such as the Capitol region clusters, had much more humble and pedestrian beginnings. The conditions that we associate with entrepreneurship developed over time. In the early stage of these new technologies, the trajectory of their further development was unclear and it would have been difficult to anticipate the types of specific assistance entrepreneurs needed. Individual entrepreneurs were in the best position to move the technology, the industry and the region forward. The role of local government policy in promoting entrepreneurship is unclear, and, as no early examples presented themselves in the Capitol region, we have not directly examined this question here. It is interesting to note that the Silicon Valley success is viewed as an outgrowth of intense technology transfer and interaction between industry and universities in that region. Local government policies played a role, but these tend to be implemented and effective in the later stages of cluster development. Are there general lessons to be learned from the development of the Capitol region or is this case unique? Is every case unique? Certainly, the Capitol region benefited from above-average household income and higher-than-average education levels, giving it very different resource endowments compared to other underdeveloped regions lacking an entrepreneurial culture. The general lesson, then, is that entrepreneurs adapt and, when they are successful, they build the types of resources that support their activities: Over time a coherent system develops. A distinction should be drawn between the conditions that support innovation and the conditions that support entrepreneurship. The two concepts are related: entrepreneurship facilitates the realization of innovation, as firms are formed to commercialize and advance new ideas. Conducive external environments and resources make innovative activity easier but may not be sufficient to induce new firm formation: this is where the concepts diverge. The critical condition for entrepreneurial enterprise is opportunity. Even if the regional conditions do not match those of successful clusters, incentive to develop locational

Entrepreneurs as agents 131 opportunities welcomes entrepreneurial activity. Our understanding of regional economic systems may be enhanced by a consideration of entrepreneurs as economic agents who actively interact with their local environments, adapt to new situations, crises or opportunities using location-specific assets, and finally, build and augment local institutions. Certainly, this is not the last word on this topic. It is our hope that this historically informed appreciative theorizing will inspire others to take a more detailed look. It is only through an appreciation of the nuances of cluster development that we might begin to adequately inform policy.

Notes 1

2

The literature on the entrepreneurial decision often focuses on the change in the opportunity cost of entrepreneurship as a significant factor in the decision to leave wage employment. In particular, increases in the volatility of wage income, such as what would be expected during an economic recession, may be associated with more individuals switching from worker to entrepreneur. It is unclear whether liquidity constraints are a barrier to new firm creation. Hurst and Lusardi (2002) determine that they are not: on average, they found that individuals who wish to begin a business were not constrained financially in their attempt to do so. Other studies (Evans and Jovanovic, 1989; Evans and Leighton, 1989) suggested that liquidity constraints do matter and that wealthier individuals have a higher propensity to begin a business.

References Andersen, E. and Teubal, M. (1999) High tech cluster creation and cluster re-configuration – a systems and policy perspective, Industrial and Corporate Change 9, 87–111. Appold, S.A. (2000) The control of high-skill labor and entrepreneurship in the early US semiconductor industry, Environment and Planning, A 32 (12), 2133–60. Bartik, T.J. (1991) Who Benefits from State and Local Economic Development Policies? Kalamazoo, MI: Upjohn Institute for Employment Research. Bhide, A.V. (1999) The Origin and Evolution of New Businesses. Oxford: Oxford University Press. Blanchflower, D. and Oswald, A. (1998) What makes an entrepreneur?, Journal of Labor Economics 16, 26–60. Blanchflower, D.G., Oswald, A. and Stutzer, A. (2001) Latent entrepreneurship across nations, European Economic Review 45, 680–91. Blaug, M. (1985) Economic Theory in Retrospect. Fourth Edition. Cambridge: Cambridge University Press. Boschma, R.A. (1999) The rise of clusters of innovative industries in Belgium during the industrial epoch, Research Policy 28, 853–71. Boschma, R.A. and Lambooy, J.G. (1999) Evolutionary economics and economic geography, Journal of Evolutionary Economics 9 (4), 411–29. Bresnahan, T. Gambardella, A. and Saxenian, A.L. (2001) Old economy inputs for new economy outcome: Cluster formation in the New Silicon Valleys. Industrial and Corporate Change 10 (4), 835–60.

132 Maryann Feldman and Johanna L. Francis Bygrave, W. and Hofer, C. (1991) Theorizing about entrepreneurship, Entrepreneurship: Theory and Practice 16 (2), 13–22. Camagni, R. (1995) The concept of innovative milieu and its relevance for public policies in European lagging regions, The Journal of the RSAI 74 (4): 317–40. Carroll, C.D. (2002) Portfolios of the rich, in L. Guiso, M. Haliassos and T. Jappelli (eds) Household Portfolios: Theory and Evidence. Cambridge, MA: MIT Press, 389–429. Chang, H.-J. (1994) State, institutions and structural change, Structural Change and Economic Dynamics 5, 293–313. Chiles, T., Meyer, A.D. and Hench, T.J. (2004, Sept/Oct). Organizational Emergence: The Origin and Transformation of Branson, Missouri’s Musical Theaters, Organization Science 15 (5), 499–519. Contract Pharma (2002) Top pharmaceutical and bio-pharmaceutical companies report, http://www.contractpharma.com. Accessed July 19, 2004. Cortright, J. and Mayer, H. (2001) High tech specialization: A comparison of high technology centers, The Brookings Institution Survey Series, January 1–13. De Vol, R., Wong, P., Ki, J., Bedroussian, A. and Koepp R. (2004) America’s biotech and life science clusters: San Diego’s position and economic contributions. Research Report, Milken Institute, Santa Monica, CA. Delaney, E. (1993) Technology search and firm bounds in biotechnology: New firms as agents of change, Growth and Change 24, spring, 206–28. Department for Trade and Industry (DTI) (1998) Our Competitive Future: Building the Knowledge Driven Economy. White Paper, HMSO, CM4176, London. Engelking, S. (1999) Austin’s economic growth: A case study in futuristic planning, Economic Development Review 16, 21–4. Etzkowitz, H. (2000) Academia interrupted: Normative change in science, Science Policy (Fall), 11 (5). Etzkowitz, H. (2003) Innovation in innovation: The triple helix of university– industry–government relations, Social Science Information 42 (3), 293–337. Evans, D.S. and Leighton, L. (1989) Some empirical aspects of entrepreneurship, American Economic Review 79, 519–35. Evans, D.S. and Jovanovic, B. (1989) An estimated model of entrepreneurial choice under liquidity constraints, Journal of Political Economy 97 (4), 808–27. Feldman, M.P. (2000) Where science comes to life: University biosciences, commercial spin-offs and regional economic development, The Journal of Comparative Policy Analysis, Research and Practice, 2, 345–61. Feldman, M.P. (2001) The entrepreneurial event revisited: An examination of new firm formation in the regional context, Industrial and Corporate Change 10, 861–91. Feldman, M.P. (2003) The locational dynamics of the US biotech industry: Knowledge externalities and the anchor hypothesis, Industry and Innovation 10 (3): 311–28. Feldman, M.P. and Desrochers, P. (2003) The evolving of research universities in technology transfer: Lessons from the history of Johns Hopkins University, Industry and Innovation 10, 5–24. Feldman, M.P. and Francis, J.L. (2001) Entrepreneurs and the formation of innovative clusters. Unpublished manuscript, Johns Hopkins University. Feldman, M.P. and Francis, J.L. (2002) The entrepreneurial spark: Individual agents and the formation of innovative clusters. In A.Q. Curzio and M. Fortis (eds) Complexity and Industrial Clusters. Heidelberg: Springer-Verlag. Feldman, M.P. and Francis, J. (2003) Fortune favors the prepared region: The case of

Entrepreneurs as agents 133 entrepreneurship and the Capitol Region biotechnology cluster, European Planning Studies, 11, 765–88. Feldman, M.P. and Francis, J. (2004) Homegrown solutions: Fostering cluster formation, Economic Development Quarterly, 18, 127–37. Feldman, M.P. and Francis, J.L. (forthcoming) Fortune favors the prepared region: The case of entrepreneurship and the Capitol region biotechnology cluster, European Planning Studies 11: 765–88. Feldman, M.P. and Martin R. (2004) Jurisdictional advantage. In Jaffe, A.B., Lerner, J. and Stern, S. (eds) Innovation Policy and the Economy, Volume 5. NBER, National Bureau of Economic Research Working Paper, Cambridge, MA. Feldman, M.P. and Schreuder, Y. (1996) Initial advantage: The origins of the geographic concentration of the pharmaceutical industry in the mid-Atlantic region, Industrial and Corporate Change 5, 839–62. Glaeser, E. (2003) Reinventing Boston: 1640 to 2003. NBER Working Paper 10166. Gompers, P. and Lerner J. (1999) What drives venture capital fundraising? NBER Working Paper 6906. Gompers, P.A. and Lerner, J. (1999) The Venture Capital Cycle. Cambridge, MA: MIT University Press. Gompers, P.A., Lerner, J. and Scharfstein, D. (2003) Entrepreneurial spawning: Public corporations and the genesis of new ventures, 1986–1999. NBER Working Paper 9816. Gray, M. and Parker, E. (1998) Industrial Change and Regional Development: The Case of the U.S. Biotechnology and Pharmaceutical Industries. Cambridge: University of Cambridge, Economic and Social Research Council Center for Business Research. Guidera, M. (1996) Start-ups finding it easier to get paired with ‘angels’: University of Maryland’s Dingman Center joins Investor Network in matchmaking effort, The Baltimore Sun August 21, 1C. Haynes, K., Fuller S.S. and Qiangsheng, L. (1997) The northern Virginia economy: The changing role of federal spending, Research in Urban Economics 11, 145–61. Holtz-Eakin, D., Penrod J.R., and Rosen, H.S. (1996) Health insurance and the supply of entrepreneurs. Journal of Public Economics 62, 209–35. Hurst, E. and Lusardi, A. (2002) Liquidity constraints, wealth accumulation and entrepreneurship. Unpublished manuscript, Chicago Business School. Hyry, M., Haapasalo, H. and T. Donnelly. (2003) The role of social networking in developing regional business: The case of Oulu. Paper prepared for the conference: Reinventing Regions in Global Economy April 2003. Kahn, R.E. and Cerf, V.G. (1999) What is the internet (and what makes it work)? Paper prepared for the Internet Policy Institute, December. Kargon, R., Leslie, S. and Shoenberger, E. (1992) Far beyond big science: Science regions and the organization of research and development. In Galison, P. and Hevly, B. (eds) Big Science: The Growth of Large-Scale Research. Stanford CA: Stanford University Press. Kenney, M. and Patton, D. (2004). Supporting the high-technology entrepreneur: Support network geographies for semiconductor, telecommunications equipment, and biotechnology startups. Working paper. Kenney, M. and Von Burg, U. (1999) Technology, entrepreneurship and path dependence: Industrial clustering in Silicon Valley and Route 128. Industrial and Corporate Change, 8, 67–103.

134 Maryann Feldman and Johanna L. Francis Kirzner, I. M. (1973) Competition and Entrepreneurship. Chicago: University of Chicago Press. Klepper, S. (2002) The capabilities of new firms and the evolution of the US automobile industry, Industrial and Corporate Change 11 (4), 645–66. Klepper, S. (2004) Agglomeration through spinoffs: How Detroit became the capital of the US automobile industry, Working paper, Carnegie Mellon University. Knight, F.H. (1921) Risk, Uncertainty and Profit. New York: Houghton Mifflin. Lambooy, J.G. and Boschma, R.A. (2001) Evolutionary economics and regional policy, Annals of Regional Science 35, 113–31. Lerner, J. (1996) The government as venture capitalist: The long-run effects of the SBIR Program, NBER Working Paper W5753, Washington. Leslie, S. and Kargon, R. (1994) Electronics and the geography of innovation in postwar America, History and Technology 11, 217–31. Leslie, S. and Kargon, R. (1996) Selling Silicon Valley: Frederick Terman’s model for regional advantage, Business History Review 70 (4), 435–72. Link, A.N. (1995) A Generosity of Spirit: The Early History of the Research Triangle Park. Research Triangle Park: The Research Triangle Foundation of North Carolina. Link, A.N. (2002) From Seed to Harvest: The History of the Growth of Research Triangle Park. Chapel Hill: University of North Carolina Press. MacRae, M. (2002) ‘Biotech growth: Will your city be a leader or a loser, Pharmaceutical Technology, July, 106. Malerba, F., Nelson, R., Orsenigo, L. and Winter, S. (1999) History-friendly models of industry evolution: The computer industry, Industrial and Corporate Change 8, 3–40. Markusen, A. (1996) Sticky places in slippery space: A typology of industrial districts, Economic Geography, 72, 293–313. Maskell, P. (2001) Towards a knowledge-based theory of the geographic cluster, Industrial and Corporate Change, 10 (4): 921–36. Maskell, P. and Malmberg, A. (1999) Localised learning and industrial competitiveness, Cambridge Journal of Economics 23, 167–85. Nelson, R. (1998) The agenda for growth theory: A different point of view, Cambridge Journal of Economics 22, 497–520. Nelson, R. and Winter, S. (1982) An Evolutionary Model of Economic Change. Cambridge, MA: Harvard University Press. O’Harrow, R., Jr and Lipton, E. (1996) George Mason U. embroiled in debate about its mission, Washington Post February 14, A1. OECD (1993) Small and Medium Sized Enterprises: Technology and Competitiveness. Paris: OECD. Orlando, M.J. (2000) On the importance of geographic and technological proximity for R&D spillovers: An empirical investigation. Federal Reserve Bank of Kansas City Research Working Paper. Orsenigo, L. (2001) The (failed) development of a biotechnology cluster: The case of Lombardy Small Business Economics 17 (1/2), 77–92. Owen-Smith, J. and Powell, W. (2004) Accounting for emergence and novelty in Boston and Bay Area biotechnology. Draft prepared for 2004 Genesis of Clusters meeting in Waxholm, Sweden. Papke, L.E. (1991) Interstate business tax differentials and new firm location: Evidence from panel data, Journal of Public Economics, 45, 47–68. Piore, M. and Sabel, C. (1984) The Second Industrial Divide: Possibilities for Prosperity. New York: Basic Books.

Entrepreneurs as agents 135 Porter, M. (1989) The Competitive Advantage of Nations. New York: Free Press. Porter, M. (2000) Location, competition and economic development: Local clusters in a global economy, Economic Development Quarterly 14 (1), 15–34. Poterba, J. (1989) Venture capital and capital gains taxation. In L. Summers (ed.) Tax Policy and the Economy. Cambridge, MA: Massachusetts Institute of Technology Press. PriceWaterhouseCoopers (1998) Toward a New Economy: Merging Heritage with Vision in the Greater Washington Region. Potomac Knowledge Ways Project, Washington, D.C., http://knowledgeway.org/voice/newecon/homepage.html. Roberts, E.B. (1991) Entrepreneurs in High Technology. New York: Oxford University Press. Romanelli, E. and Feldman, M.P. (2003) The endogenous production of entrepreneurs and clusters. Draft paper, October. Romanelli, E. and Feldman, M.P. (2004) The anatomy of cluster development: The case of U.S. biotherapeutics, 1976–2002. Mimeo. Saxenian, A. (1988) The Cheshire Cat’s grin: Innovation and regional development in England, Technology Review, February/March, 67–75. Saxenian, A. (1994) Regional Advantage. Cambridge, MA: Harvard University Press. Saxenian, A. (1998) Keynote speech, IT Conference, USIS American Embassy in Sweden, June 9, http://www.usis.usemb.se/Silicon/text/saxenian.html. Schmookler, J. (1966) Invention and Economic Growth. Cambridge, MA: Harvard University Press. Schumpeter, J.A. (1934) Theory of Economic Development (1911, in German; tr. 1934). Cambridge, MA: Harvard University Press. Schumpeter, J.A. (1939) Business Cycles: A Theoretical, Historical and Statistical Analysis of the Capitalist Process. 2 volumes. New York: McGraw-Hill. Schumpeter, J.A. (1942/1976) Capitalism, Socialism, and Democracy. New York: Harper & Row. Scott, A.J. (2004) Origins and growth of the Hollywood motion-picture industry: the first three decades. Working paper. Small Business Administration (2004) SBIR Award Statistics. Accessed July 2004. http://www.sba.gov/sbir/indexsbir-sttr.html#sbirstats. Sough, R.R., Campbell, H. and Haynes, K.E. (1998) Small business entrepreneurship in the high technology services sector: An assessment of edge cities of the U.S. National Capital Region, Small Business Economics 10, 61–74. Sripaipan, C. (1993) Financing industry–university interaction: A Southeast Asian perspective, TDRI Quarterly Review 8 (2), 8–16. Stuart, T. and Sorenson, O. (2003) The geography of opportunity: Spatial heterogeneity in founding rates and the performance of biotechnology firms, Research Policy 32, 229–53. Sturgeon, T.J. (2001) How Silicon Valley came to be. In Kenney, M. (ed.) Understanding Silicon Valley: The Anatomy of an Entrepreneurial Region. Stanford: Stanford University Press. Tether B. and Storey, D. (1998) Smaller firms and Europe’s high technology sectors: A framework for analysis and some statistics evidence, Research Policy 26, 947–71. Teubal, M. and Andersen, E. (2000) Enterprise restructuring and embeddedness: A policy and systems perspective, Industrial and Corporate Change 9, 87–111. Wallsten, S. (2004) Do science parks generate regional economic growth? An empirical analysis of their effects on job growth and venture capital. Working Paper 04–04. Washington DC: AEI-Brookings Joint Center for Regulatory Studies.

136 Maryann Feldman and Johanna L. Francis Wang, K. (2001) Innovation in Taiwan. Working paper. Taiwan: Industrial Economics & Knowledge Center Industrial Technology Research Institute. Zaltman, J.I., Duncan, R. and Holbek, J. (1973) Innovations and Organizations. New York: Wiley Press. Zeller, C. (2001) Cluster biotech: A recipe for success? Spatial patterns of growth of biotechnology in Munich, Rhineland and Hamburg, Small Business Economics, 17 (1/2), 123–41.

7

Innovation, learning and cluster dynamics Bart Nooteboom

Introduction This chapter focuses on cluster dynamics, in two meanings: the evolution or development of clusters, and their contribution to innovation. There are three reasons for this. First, although much has been written about clusters, we still know little, theoretically and empirically, about how clusters develop and evolve. Second, while it is often claimed that clusters yield innovation, it has not been made clear how that works, in terms of firms acting in clusters, and it is not clear that clusters always do contribute to innovation. Indeed, this chapter will argue that sometimes they inhibit rather than promote innovation. Third, shifts of activity due to globalization demand attention to the effects on the location and structure of clusters. Recent studies show that the famous Italian industrial districts are becoming locally disembedded, and shift some activities, especially in production, to emerging, proximate, lower-wage countries such as Romania. These issues have a certain urgency from the perspective of public policy. Policy makers have caught on to the fashion of cluster thinking, or perhaps we should say ideology, and contemplate active cluster policy to promote innovation. If the ‘buzz’ of clusters is based on myth more than facts, this can have serious adverse effects. Some studies have broached issues of cluster development under globalization and innovation (Asheim and Isaksen, 2002; Boschma and Lambooy, 2002; Oinas and Malecki, 2002; Zuchella, 2003). The purpose of the present chapter is to extend that work. It is of dubious validity to claim that regional features have a direct effect on innovation. The challenge is to show how they affect the activities of firms, in their conduct of innovative activities, in processes of invention, innovation, spillover, division of labour, allocation of resources and collaboration. For this purpose, the notion that firms are locally ‘embedded’ is in need of clarification, at least partly in terms of the linkages between institutions and firms, and between firms. I do not exclude the possibility that some regional features impact directly on firm behaviour, without their being traceable to ties between organizations. Indeed, I will argue that there is something like ‘regional culture’ that affects firm behaviour fairly directly. But we have to be explicit in showing how it does so. Much embedding, on the other hand, entails ties between organizations, which also have to be made explicit, at least in part, to explain

138 Bart Nooteboom regional effects. Here, good use can be made of the sociological and management literatures on networks of firms (for example, Granovetter,1973; Coleman, 1988; Burt, 1992; Krackhardt, 1999; Nooteboom and Gilsing, 2003). While local embedding may contribute to innovation, and it remains to be seen how exactly that works, in terms of the operations of firms, an escape from local embedding may also be needed for innovation. Oinas and Malecki (2002) proposed that in the study of regional systems of innovation we should recognize the need for linkages outside a region, and it may be better to speak of ‘spatial systems of innovation’. Embedding, in the sense of linkages between activities, need not always be tied to location, and may occur also in ‘communities’ that are to some extent virtual, with communication at a distance. This issue of local and other types of embedding is a central theme of the present chapter. It will be argued that processes of innovation and learning have different stages, with different characteristics. In particular, in early development there may be a relatively greater need for local embedding, while later development requires disembedding, as suggested by Asheim and Isaksen (2002). The chapter proceeds as follows. First, it briefly considers the definition of clusters and innovation networks, and proposes three kinds of embedding: institutional embedding, structural embedding (structure of ties between firms) and relational embedding (type and strength of ties). Second, for an analysis of innovation, insights are derived from the literature on innovation and learning, in particular the notion of cognitive distance (Nooteboom, 1999) and the distinction between exploitation and exploration (March, 1991). Exploitation refers to the efficient employment of current assets, including intangible assets such as capabilities, while exploration refers to the development of new capabilities. Innovation typically starts with exploration and then moves on to exploitation. A crucial problem is how one next moves out again into exploration. Here, clusters may fail and may get locked into exploitation. Third, next to issues of competence, in innovation inter-firm relations also entail issues of governance, i.e. the management of relational risk. Effects of localization may lie more in governance than in competence. Fourth, an analysis is given of the requirements for exploitation and exploration networks, from a perspective of both competence and governance, concerning the three types of embedding: institutions, network structure and strength of ties. Some empirical illustrations are provided. Finally, conclusions are drawn concerning the development of clusters and the implications for cluster policy.

Clusters, networks and embedding Cooke and Huggins (2002) defined clusters as follows: Geographically proximate firms in vertical and horizontal relationships involving a localized enterprise support infrastructure with shared developmental vision for business growth, based on competition and cooperation in a specific market field.

Innovation, learning and cluster dynamics 139 Apparently, clusters and industrial districts are more or less synonymous. The concept of a network is more general, and does not necessarily entail local embedding, a shared objective, or a specific market. Thus, a cluster is a network but not necessarily vice versa. This chapter focuses on clusters in the context of innovation. Presumably, the notions of ‘innovation clusters’ and ‘regional systems of innovation’ are close synonyms. As noted above, Oinas and Malecki (2002) proposed that in the study of regional systems of innovation we should recognize the need for linkages outside a region, and it may be better to speak of ‘spatial systems of innovation’. These considerations raise the question whether local embedding and geographical proximity should be retained as defining characteristics of clusters. If they are, we may have to say that in their development clusters are transformed into other types of networks. To avoid this definitional issue, I will at times speak of innovative networks, rather than clusters, which may or may not be strongly locally embedded. In the introduction, we noted that the notion of ‘embedding’ requires specification. If we are to explain the causality of regional characteristics, it seems that we have to make explicit how regional level variables affect firms, and this is likely to operate in large measure through ties between firms and other organizations, such as schools, universities, intermediaries of many kinds and bodies of public administration, and between firms themselves, in relations of supply, demand, alliances and other forms of collaboration. Here, use can be made of the sociological and business network literature. Three kinds of embedding were suggested above: institutional embedding, structural embedding and relational embedding. Institutional embedding relates to the impact of regulation and norms of conduct, taxes, subsidies, legal system, infrastructure, schooling, research, labour market etc. Structural embedding derives from the social network literature. Structural features of networks are size (number of participants), density (actual number of direct ties as a ratio of the maximum possible number), centrality (of which there are several forms) and stability of structure (rate of entry and exit). Relational embedding appears in the social network literature in the notion of the ‘strength of ties’, but is developed in more detail in the literature on alliances or inter-organizational relations (IORs). In other words, I propose that an adequate understanding of clusters requires a combination of geography, social networks and interorganizational relations.

Innovation and cognitive distance For innovation, diversity is a crucial condition, to produce Schumpeterian ‘novel combinations’, as demonstrated, in particular, in evolutionary economics (Nelson and Winter, 1982). Diversity is associated with the number of agents (people, firms) who are involved in a process of learning or innovation by interaction. Next to the number of agents involved, a second dimension of diversity is the degree to which their knowledge and skills are different. This entails the

140 Bart Nooteboom notion of cognitive distance (Nooteboom, 1992, 1999). Note that here cognition is seen in a broad sense, including not only rational evaluation but also emotionladen value judgements, and mental heuristics of attribution, inference and decision making. The notion of cognitive distance derives from a social constructivist view of knowledge, according to which perception, interpretation, understanding and value judgment entail mental constructions on the basis of mental categories that are developed in interaction with the physical and social world. As a result, different people, and different organizations, which have developed their cognition along different paths of development, in different conditions, will perceive, interpret and evaluate the world differently. A central task of organizations is to sufficiently reduce cognitive distance, in an organizational focus, including epistemic as well as moral categories, to make possible the achievement of joint purpose. Note that such cognitive categories serve not only for guiding cognition in the narrow, epistemic sense of attention, perception and interpretation but also for setting behavioural values, in a moral order, to facilitate collaboration, constrain opportunism, build trust, and limit and resolve conflicts of interest (governance). Such categories tend to be internalized, to a greater or lesser extent, by people, as part of tacit knowledge, assimilated though socialization and habituation. Existing cognitive structures constitute absorptive capacity. On the level of organizations, this was recognized by Cohen and Levinthal (1990). Here, absorptive capacity includes organizational capabilities to assimilate information, internally distribute it and implement knowledge in design, development, production and marketing. It depends, among other things, on R&D. The notion of absorptive capacity is crucial in the analysis of spillovers, which play a central part in the analysis of clusters and regional innovation systems. Spillover depends on absorptive capacity. Absorptive capacity, connected with organizational focus, not only enables but also constrains organizational cognition, yielding organizational myopia, which needs to be compensated for by engaging in outside relations with other organizations, with different, complementary foci, at some cognitive distance. This produces a new purpose for inter-organizational alliances, next to the usual considerations, known from the alliance literature (for an elaboration, see Nooteboom, 2000). Firms need to make a trade-off between organizational identity in the form of a clear focus and wide scope of internal competencies. A wide scope, with a wide focus, entails limited identity. A limited scope, with a narrow focus and stronger identity, can be compensated for by alliances. The notion of organization as a focusing device is relevant for two reasons. First, it is needed to understand the functioning of firms in networks. Second, it may have implications for the notion and role of clusters. Perhaps a defining characteristic of a cluster is that it entails a shared culture with corresponding cognitive focus, in an epistemic and moral order. This is certainly facilitated by local embedding and geographical proximity, but those may not be necessary. Perhaps there is a viable and fruitful combination of geographical distance complemented by frequent meetings to build and maintain the shared focus.

Innovation, learning and cluster dynamics 141 In processes of learning and innovation, in interaction between firms, cognitive distance between firms introduces both an opportunity and a problem. The opportunity lies in diversity: the novelty value of a relation increases with cognitive distance. However, mutual understanding (absorptive capacity) decreases with cognitive distance. If learning performance from interaction is the mathematical product of novelty value and understandability, the result is an inverse-U shaped relation with cognitive distance. Optimal cognitive distance lies at the maximum of the curve (For an empirical test of optimal cognitive distance, see Nooteboom et al., 2005). The analysis of optimal cognitive distance has several implications for cluster dynamics. One is that firms should seek optimal distance for innovation. Another implication concerns the duration of relationships. Between firms, cognitive distance may be reduced to the extent that they have engaged in continued interaction, especially when that interaction was exclusive. In other words, their foci start to overlap, in a shared epistemological and normative framework. This reduces the novelty value of a partner’s cognition, with a reduction of innovation performance. This suggests that while familiarity breeds trust (Gulati, 1995), it may also reduce learning potential, so that for the purpose of learning ties should not be too strong in terms of duration. In sum, next to optimal cognitive distance there is also something like an optimal duration of ties for learning: long enough to build mutual understanding and trust, but not so long as to run out of innovative steam (for an empirical test of optimal duration, see Wuyts et al., 2005). The point has important implications for cluster dynamics, and for any cluster policy. Too durable, local embedding, particularly when it is cut off from outside contacts, may reduce cognitive distance too much. It may be good for trust but bad for learning. As will be argued in more detail later, this suggests one way in which clusters may inhibit rather than promote innovation. Note that the claim of a negative effect of duration of a relation was qualified by the condition that the relation is exclusive, i.e. the firms involved have no contacts outside the relation, pertaining to the subject of collaboration. When, on the other hand, both sides of a relationship tap into outside, non-overlapping networks, they may be continually re-charged with novel insights that keep their relation vibrant. This connects with Burt’s (1992) notion of the advantages in bridging ‘structural holes’: the relation just described provides such a bridge. This illustrates why clusters, to the extent that they entail stable relations, may need outside ties for ongoing innovation. In sum, regional variables that are relevant for innovation in cluster dynamics include variables that affect absorptive capacity of firms, such as educational facilities, R&D in firms, R&D in public organizations and the transfer of outcomes to firms. They also include variety and cognitive distance in the cluster, the duration of linkages and outside linkages that replenish variety.

142 Bart Nooteboom

Exploration and exploitation The economic success of regions and clusters requires success in both exploitation and exploration (March, 1991). Exploitation, that is the efficient employment of current assets and capabilities, is needed to survive in the short term. Exploration, the development of novel capabilities, is needed to survive in the long term. Thus, to survive in the short and long term, firms must somehow combine the two. That is a paradoxical task. Exploitation often requires the maintenance of a stable organizational structure, and division of labour, with unambiguous terms and clear standards, in a narrow organizational focus; while exploration requires the reverse: loosening of structure for novel reconfigurations, shifting meanings and deviation from existing standards, in a wide focus. A key problem is how exploration may be based on experience in exploitation, and how to ensure that the outcome of exploration will be exploitable. How do exploitation and exploration build on each other? What path of development can we think of that maintains exploitation while at the same time it also encourages exploration? Here, use is made of the ‘cycle of discovery’ proposed by Nooteboom (2000). It plays a key role in this chapter, in the search of a model of cluster development. In particular, this model will show how clusters can both support and inhibit innovation, at different stages of the cycle of development. The cycle of development proposes several stages of evolution, in which there is an alternation of variety of content and variety of context. First, variety of content (of a concept or practice) that emerges from exploration is reduced, via consolidation into a dominant design, as suggested in the innovation literature. As a result of reduced uncertainty, demand increases, and new producers jump on the bandwagon. Related industries and existing distribution channels follow suit, and adapt, from fear of missing the new boat. The new technology– product–market combination develops into a dominant design or ‘dominant logic’ (Bettis and Prahalad, 1995) of organization, including network structure and ‘industry recipes’ (Spender, 1989), with pressures to conform, that is ‘organizational isomorphism’ (Dimaggio and Powell, 1983). New entrants exert pressure on price, and for the sake of efficient production increase of scale, a division of labour and associated specializations emerge. So far, this is nothing new. However, the question next is how one gets away from the dominant designs in technology and organization, in a next round of (radical) innovation. This is a crucial point in cluster dynamics: how does a cluster get away from possible inertia as the outcome of previous development? Here, the proposal is that for exploration to arise from exploitation, one next needs to open up to a new variety of contexts of application, that is through generalization. These novel contexts of application may be sought voluntarily, in an expansion of activity. Voluntary new applications of established capabilities appear to be based on an instinctive drive that among psychologists is known as a principle of ‘overconfidence’. In economics, there is also a pressure to extend the market, as growth in the original market stagnates. With entry into new

Innovation, learning and cluster dynamics 143 markets, one needs to adapt products and organization, which requires knowledge from outside. Clusters may need MNCs as a vehicle for this reach outside existing boundaries. This is in line with Boschma and Lambooy’s (2002) analysis of developments in Italian industrial districts, where they identified the role of MNCs as ‘bridging enterprises’, to carry activities into international markets and to access outside sources of knowledge, and with Asheim and Isaksen’s (2002) analysis of how Norwegian clusters had to make a shift from local to global operations. However, new conditions of market, technology and institutions may also be imposed from outside one’s familiar niche. An illustration of this, in the development of multi-media, is that publishers finally adopted digitalization and electronic distribution of text, for fear of losing their market position (Gilsing, 2005). For clusters, this may arise from an invasion of multinationals. A novel context is needed for three reasons. The first is that established capabilities arise and consolidate in a given niche, and therefore perform well there, and are taken for granted, so that new conditions of technology, demand, infrastructure and institutions are needed to gain new insights in limits of validity. The second reason is to build insight into novel goals and motivation for change. The third reason is to yield insight into potential novel content of practice, inspiration for which is found in the novel context. First, to maintain exploitation as much as possible, there is an attempt to make minor, incremental adjustments to established practice, a process we may call differentiation. Insight for this may come from previous experience, in novel selections from familiar repertoires, which are retrieved in an attempt to improve fit in the novel context. Next, when this fails, experiments are conducted with novel elements, adopted from the novel context, which seem to be successful where familiar practice fails, in hybrids of old and new elements, what may be termed reciprocation. This yields an opening up to new variety of content. The function of this is twofold. First, it still allows for ongoing exploitation, albeit in new forms. Second, it allows for experimentation with new elements, to test their potential, without sacrificing existing basic design principles. Next, when such potential emerges, there is more willingness to make more radical changes in organization or technological architecture, when that is needed for the novelty to realize its full potential, that is the process of accommodation. Here, rigidities of established structures, which may have initially offered an advantage for exploitation, become a liability. Emerging novelties cannot achieve their potential under the systemic limitations imposed by existing structures, practices and ways of thinking. If the cluster or network is unable to cope with this, it may need to be broken up, so that different elements have more scope to adapt, in different ways, to new conditions. Accommodation, then, leads to a new beginning, under radical uncertainty, in search of novel dominant designs, in consolidation, and we are back at the beginning of the cycle. In sum, in the efficient exploitation of previous innovations, clusters are in danger of getting stuck in inertia. To escape from this, they need to break out of existing structures, or they must allow and indeed invite entry into the cluster,

144 Bart Nooteboom to generate novel insights into the limits of current practice, and into new needs and opportunities, even if, or precisely because, this may lead to a restructuring and reconfiguration of the cluster around the next radical innovation.

Effects of embedding The opportunities and problems for combining exploitation and exploration depend on institutional, structural and relational embedding. The difficulty of combining exploitation and exploration, in differentiation and reciprocation, depends on three structural features of the exploitation system of the cluster, which determine its rigidity: 1 2

3

The complexity of division of labour, defined as the number of component activities and the density of direct ties of dependence between them. Structure is simple when complexity is low. The modularity of the system, on the basis of clear and stable constraints on activities along such ties of dependence, in the form of standards, needed to maintain systemic integrity. The opposite would be ambiguity and variability of constraints, by which activities need to be continually coordinated. The tightness of constraints, that is the scope for variety in contributions from component activities. Structure is loose when tightness is low.

Exploitation is systemic when it has features (1) and (3) (complex and tight), and stand-alone in the opposite case (simple and loose). In the case of feature (2) (modularity), component activities can be autonomous, and can be replaced, as long as they satisfy the constraints on interfaces in the position they take in the structure. If exploitation entails a systemic structure, simultaneous exploration is constrained by the many and tight constraints on component activities. Exploration would soon yield a breaking of constraints on interfaces, resulting in many unknown repercussions in the dense structure of dependencies, such as change of content of linked activities, which may in turn trigger change elsewhere, possibly resulting in wide-ranging architectural change (Henderson and Clark, 1990). Management, or cluster policy, would rightly be wary of accepting that risk and cost, unless there were a clear and proven potential of the novelty that would justify them. One possibility is to see to it that the exploitation structure is not systemic but modular or stand-alone, even if this entails loss of efficiency, in order to maintain options for exploration by reconfiguration. Here, it matters what options for reconfiguration are at hand. Here, perhaps, we encounter the notion of ‘Jacobs externalities’ (Boschma and Lambooy, 2002). In urban regions with a large variety of different activities, and a rich, varied, complex infrastructure, with a wide scope of spillovers, new ideas, and activities that become complementary in new ways, there is more scope for new exploration. When exploitation is irremediably systemic, there are two basic options to combine exploitation and exploration: separation in place and separation in time.

Innovation, learning and cluster dynamics 145 With separation in place within a region, exploitation would occur in one part, and exploration elsewhere, in different clusters, perhaps,. This includes the classic separation between production (exploitation) and R&D (exploration). A familiar case is the pharmaceutical industry, where small biotech firms explore novel medically active substances, and, when those are found to be ripe for exploitation, large pharmaceutical firms take over to carry out the lengthy process of clinical testing, efficient large-scale production and distribution. A problem arises in ensuring that exploration is based on inspiration from exploitation and that exploitation follows exploration. This may be achieved by mobility of firms or people between the two sectors. For example, researchers in R&D organizations are encouraged to try to apply their inventions in firms, and workers in firms are encouraged to reflect on their experience and new alternatives in an R&D environment. Exploration often requires disintegration: new elements that do not fit in existing structures (of production, supply, market, distribution channels, institutions) need to shield themselves off in a niche where deviation from established structures and processes is feasible. This often requires the emergence of new firms that are not ‘imprisoned’ in existing structures and interests. This may entail entry from outside, or spinoffs from existing firms or networks, of entrepreneurs who escape from these ‘locked in’ organizational or network arrangements. Alternatively, a region or cluster may specialize in either exploitation or exploration, and seek the other in relations with other regions with complementary specialization. This is a form of separation in place, but between rather than within regions. Here, outside linkages are especially important. In a region that specializes in exploitation, when exploitation is highly systemic, an option is to encourage entrepreneurs that cannot find the leeway to innovate within the region to ‘spin off ’ into a region that is more oriented at exploration, and come back when results of exploration are ready for exploitation.

The risks of lock-in and spillover Inter-organizational relations, and networks, have a competence and a governance side (Nooteboom, 2004b). Competence includes, in particular, innovative competence, that is the ability to generate, efficiently employ and diffuse novelty, in technology, products, production, organization, distribution and other aspects of marketing. This was discussed in the previous sections. Governance refers to relational risks and ways to manage them. Generally, in the literature, the focus has been on competence, to the neglect of governance. It is important to include governance because regional effects may lie at least as much, and perhaps more, in the management of relational risk than in the development and support of competence. For the analysis of governance, the focus is on risks of lock-in and spillover. In relational embedding, risk of lock-in includes the ‘hold-up’ risk from transaction cost economics This is relevant here for several reasons, one of which is that it has implications for the strength of a tie, in terms of its duration and

146 Bart Nooteboom frequency of interaction. Hold-up risk results from dependence as a result of relation-specific investments, defined as investments that have value only (or largely) in a specific relation. Specific investments entail switching costs: when the relation breaks, the investments have to be made anew in a new relation. For such investments to be made, the relation should be expected to last sufficiently long, and be sufficiently intensive, to recoup specific investments. Lock-in yields a temptation for the partner to expropriate value in opportunistic behaviour. Specific investments include the usual types, offered by transaction cost economics: location specificity of facilities, physical asset specificity (installations, tools, instruments), human asset specificity (training), dedicated capacity, brand name specificity, and time specificity. The specificity of investments depends on the flexibility of technology: the more an investment allows for a variety of products or production processes, the less specific it is. It also depends on the availability of shared standards: the more different firms share standards, the less specific investments are. Here, in particular, lies a task for public bodies, in the provision of standards and in the stimulation, and perhaps in the policing, of their use. An example of specific investments that is of particular relevance in the present context of cluster dynamics is associated with the earlier discussion of optimal cognitive distance. One can increase mutual understanding, and thereby increase optimal cognitive distance and raise innovative performance, but this entails an investment that may be largely or partially specific to a relation, so that by the logic of transaction cost theory the relation would need to last sufficiently long to make that investment worthwhile. Note that now we have the elements of an important trade-off in cluster dynamics. Here we find that a certain stability of relations may be needed to promote mutual understanding needed to utilize opportunities from cognitive distance. In the previous section we found that, when relations become too durable, they may run out of innovative steam. Investment in relation-specific trust may also constitute a relation-specific investment. This is important especially in the present context of cluster dynamics, since under the uncertainty of innovation, contracts, as a means of governance, are difficult to specify. Hence, for governance one must fall back on other instruments of governance, such as reputation and relation-specific trust (see Nooteboom 1999, 2004b). Lock-in may also arise from structural embedding. Here one is locked into a community by constraining coalitions of members of the community. There might also be network-specific investments, which can be used within but not outside the network. Some of these may indeed be related to the network level (structural embeddedness) rather than to the level of individual relations (relational embeddedness). Network-specific investments may also arise in finding out ‘who is who’ in the network, and in getting embedded in local reputation systems. Lock-in may also arise from institutional embedding. Here, one is locked into the regional ‘focus’ of shared understandings and moral precepts or customs,

Innovation, learning and cluster dynamics 147 local obligations of loyalty and conformity, and lack of cognitive distance, in ‘group think’. Now I turn to spillover risk. Linkages with other actors (firms, other organizations, individuals) provides access to variety of knowledge, and this is the positive side of spillover, emphasized in cluster studies. However, firms may also see spillover as a risk, in that knowledge that is part of one’s ‘core competence’, which constitutes competitive advantage, may be used in competition, either by a direct contact (relational embedding) or indirectly, elsewhere in the network, through a sequence of direct contacts (structural embedding). Note that the assessment of spillover risk requires a trade-off between knowledge adopted by others and knowledge gained from them. The risk is potentially serious only when there is a net loss rather than a gain. In relational embedding, spillover risk depends on how tacit or documented knowledge is, with the latter spilling over more easily than the former. It also depends on the absorptive capacity of potential competitors, that is their ability to effectively understand and implement knowledge spillovers. That in turn depends on the ‘cognitive distance’ between actors, on the differences in their ability to perceive, understand and evaluate relevant information. Finally, spillover risk depends on the speed with which knowledge changes: if it is obsolete by the time it has spilled over and has been absorbed and imitated by potential competitors, spillover risk disappears. In structural embedding, spillover risk depends on the density of the network, and the centrality of one’s position within it. In sum, regional variables that are relevant for relational risk include the type of industry and technology, in particular their implications for specificity of investments; the tacitness of the knowledge involved; and the speed at which that knowledge changes. They also include the availability and use of technical standards, the density of networks and the features that affect absorptive capacity, as indicated above.

Instruments of governance Relational risks require governance to limit them and to create trust. Counter to transaction costs economics (Williamson, 1993), I hold that trust can go beyond calculative self-interest, in loyalty and benevolence, and yet be viable in markets, although I acknowledge that such trust should not be unconditional, and is subject to limits (Nooteboom, 2002). In the notion of trust, we need to distinguish between ‘competence trust’, in the ability of people and firms to satisfy expectations, and ‘intentional trust’, in the commitment of people to perform to the best of their abilities, and not to engage in opportunistic behaviour. A survey of instruments for the governance of intentional risk, as ‘sources of collaboration’, is given in Table 7.1. Here, a distinction is made between macro and micro, and between self-interested and other-directed sources of collaboration. The distinction between macro and micro sources of collaboration, in Table 7.1, is also known as the distinction between ‘universalistic’ or ‘generalized’

148 Bart Nooteboom Table 7.1 Sources of collaboration Macro; universalistic

Micro; particularistic, relation-specific

Self-interest Opportunity control

Contracts, legal enforcement

Hierarchy, managerial ‘fiat’,

Incentive control

Reputation

Dependence: unique partner value, switching costs

Altruism

Values, social norms of proper conduct, moral obligation, sense of duty, bonds of kinship

Empathy, routinization, benevolence, identification, affect, friendship

Source Adapted from Nooteboom (2002)

sources versus ‘particularistic’ sources, made by Deutsch (1973, p. 55), and between impersonal and personalized sources made by Shapiro (1987). The first arises partly from institutional embedding, in laws, norms, values, standards, and agencies for their enforcement; and partly in the structural embedding of relations. The former is associated with ‘institution-based trust’. This kind of trust requires that we trust those institutions to support trustworthiness of people and organizations. Structural embedding includes opportunities for coalitions and reputation mechanisms. The ‘micro’ sources arise in specific relations, in relational embedding, and are often personalized. The table further distinguishes between self-interested and altruistic or ‘otherdirected’ sources of co-operation. The self-interested sources are associated with the notions of deterrence and ‘calculus-based trust’ (McAllister, 1995; Lewicki and Bunker, 1996). In the present reconstruction, this includes opportunity control and incentive control (Nooteboom, 1999). Opportunity control entails that the space of feasible action is constrained. Incentive control affects the choice of opportunities, in the space of feasible actions. Within organizations, opportunity control entails control by hierarchy, and in inter-firm relations it entails control by contract. Contracts are useful only to the extent that one is able to specify them adequately and monitor conformance to them. Even under the best of institutional conditions, legal ordering cannot be closed, including all relevant future contingencies. This is problematic especially in innovation, with its unknowable future contingencies of contract execution. Also, in innovation knowledge is sometimes highly tacit, which would also inhibit the specification of contracts. In incentive control, partner B behaves well towards A because he or she is dependent on A for one or more of the following reasons: A has a unique value to B, B faces switching costs as a result of relation-specific investments, partner A holds some form of ‘hostage’ from B, or B has to protect his or her reputation. Reputation mechanisms depend on the structure of the network, in particular network density, and on the presence of intermediaries who take up positions

Innovation, learning and cluster dynamics 149 of centrality and act as selectors of gossip (to test legitimacy of complaints), and as amplifiers and broadcasters. This is especially important when contracts are not feasible, as in innovation. Intermediaries have several other roles to play in the governance of inter-firm relations, such as to aid in mutual understanding, guarding spillovers, providing intermediation or arbitration in conflicts, building relation-specific trust and helping to end relations with a minimum of conflict (for further discussion, see Nooteboom, 2002, and, in the context of learning regions, Nooteboom, 2003). The latter is important in view of the need to maintain sufficient flexibility of relations for the sake of innovation, as discussed before. The notion of hostage is also taken from transaction costs economics. In business, ‘hostages’ often take the form of information or knowledge that is sensitive, in the sense that it could cause great damage when leaked to competitors. It can also take the form of cross-participation, or the borrowing of staff, with the threat of poaching them. Now we turn to the other-directed sources of collaboration, in trust that goes beyond calculative self-interest. On the macro level, they lie in established, socially inculcated norms and values (macro). They include pressures of allegiance to groups one belongs to, and values and norms inculcated by socialization into those groups. Of course, one can never be sure ex ante to what extent a stranger without reputation has actually internalized such norms and values. On the micro level of specific relationships, trust may be based on empathy. This entails that one knows and understands how partners think and feel. It is connected with mutual openness, and acceptance of control by others, which are crucial for the build-up of trust (Zand, 1972). While trust can go beyond calculative self-interest, it has, and should have, limits. Empathy allows one to assess strengths and weaknesses in competence and intentions, to determine limits of trustworthiness under different conditions. Identification-based trust goes further: it entails that people think and feel in the same way, sharing views of the world and norms of behaviour. This may lead to affect- and friendship-based trust. Routine-based trust, proposed by Nooteboom (1999), entails that when a relation has been satisfactory for a while, awareness of opportunities of opportunism, for oneself and for the partner, is relegated to ‘subsidiary awareness’ (Polanyi, 1962). One takes the relation for granted and does not continuously think about opportunities to gain extra advantage. As Herbert Simon taught us long ago, routinized behaviour is rational in view of bounded rationality, since it allows us to focus our limited capacity for attention and rational evaluation on matters that are new and have priority. Routines are rational also in the sense that they are based on proven success in past behaviour. On the other hand, the problem with routines is that they may no longer be adequate when conditions change. However, when results or perceived events exceed certain tolerance levels, routines are often summoned back from subsidiary into focal awareness, to be subjected to rational scrutiny. Empathy, identification and routinization may be enhanced by joint membership of clubs, such as sports clubs, community centres, Rotary clubs etc.

150 Bart Nooteboom In sum, regional variables that are relevant for instruments of governance include legal systems, norms and values of conduct, as part of regional culture, intermediaries that support reputation mechanisms, arbitration and intermediation, building of relation-specific trust and the ending of relations (institutional embedding). They also include the effect that network structure has on reputation mechanisms (structural embedding), and on whatever social conditions that may affect the building of relation-specific trust (relational embedding).

Synthesis: structure and strength of ties A key question arises from this discussion: what features of embedding are relevant, from the perspective of both competence and governance, for an analysis of differences between networks for exploitation and networks for exploration? First, consider the structure of ties in a network, in structural embedding. From a perspective of competence, recall that cognitive diversity has two dimensions: the number of actors and ties between them, and cognitive distance in the ties. The first is determined by the size and density of the network. As analysed above, in the context of innovation, large network size and density allow for more access to different sources of information, but by the same token also increase possibilities of spillover. A third feature of network structure is network stability, which has implications for how variety develops in time, in entry and exit of new members. This is important to maintain variety for the sake of exploration. High stability may be good for mutual trust and efficiency of exploitation, but bad for exploration. Another well-known feature of structure is centrality, of which there are several types. Here, I focus on degree centrality, which is the degree to which some nodes have more direct ties than other nodes. An extreme case of centrality is a hub-and-spoke production structure. Centrality may be needed for the co-ordination of activities. A central position yields power, but possibly also constraints on behaviour, in view of the many possibly divergent interests involved (Krackhardt, 1999), and in its task of centralized co-ordination it may suffer from information overload. In view of the latter, structure may need to be hierarchical. As noted before, from a perspective of governance, size and density affect possibilities of lock-in by coalitions, reputation mechanisms and shared norms of ethical behaviour. In relational embedding, seven dimensions are proposed (Bogenrieder and Nooteboom, 2004; Nooteboom and Gilsing, 2003). The first four arise from considerations of competence, and the other three from considerations of governance. 1

scope, defined as the range of activities involved in the tie. Does it involve only knowledge on the location and relevance of knowledge, anywhere in the network, or also the actual exchange or joint production of new knowledge (cf. Hansen, 1999)? Does it involve knowledge only on a small number of issues, or on a wider range of issues, concerning technology, markets, organization, and reputation of players in the network?

Innovation, learning and cluster dynamics 151 2

3 4

investment in mutual understanding, needed to build mutual understanding, for crossing cognitive distance. To the extent that this investment takes time and is specific, ties need to entail sufficient: frequency and/or duration of interaction.

While investment, frequency and duration facilitate learning, they also facilitate spillover. As argued earlier, long duration of a tie may lead to identification, which enhances mutual understanding and trust, but may reduce learning potential, particularly if the tie is exclusive, that is, in the areas of collaboration (in the scope of the relation) there are no direct ties with others. From the perspective of governance, ties require instruments for the management of relational risk of lock-in and spillover, specified in Table 7.1. This yields the following three dimensions of tie strength: 5 6 7

opportunity control, by contract incentive control, by mutual dependence, reputation or hostages trust and mutual openness, beyond control.

All dimensions of governance depend on institutional embedding. This is one important area where regional variables impact on inter-firm relationships. Contracts depend on the availability of a legal infrastructure. As indicated in Table 7.1, trust may be relation-specific, on the basis of empathy, identification, affect and routinization, but also taps into shared values and norms, in a given community. As analysed above, reputation depends not only on the structure (density) of the network, but also on the availability of other social groups that facilitate, filter and guide gossip, such as professional and industry associations, clubs and the like. Such institutions as well as personal acquaintances may also fulfil other roles of intermediaries for the building of trust. Trustworthiness of (potential) partners may be attributed not only on the basis of experience in transactions but also from chance meetings and observations in a rich variety of social settings that are most readily available in local embedding. Local embedding may be needed more for governance than for competence. This is related to the notion of optimal cognitive distance. Cognition here is a wide notion, and includes mental categories concerning both morality, in ways of dealing with each other, and cognition in the narrower sense of substantive knowledge and skills, concerning technology, markets etc. For governance one may need more proximity in morality and for competence one may need more distance in substantive knowledge. This effect is reinforced by the fact that morality is often more tacit, and requires more face-to-face interaction, gossip and chance interactions, in local ‘buzz’, than substantive knowledge does. This important point is elaborated below.

152 Bart Nooteboom

Networks for exploration and exploitation Differences between networks for exploration and networks for exploitation may now be specified in terms of the features of embedding set out above. In networks for exploration, there is uncertainty concerning future dominant designs, in both technology and organization, which yields structural uncertainty concerning the configuration of future networks for exploitation. One needs access to actors who might offer complementary knowledge, but one does not know what elements of knowledge will turn out to be relevant when a dominant design develops. Also, one does not know what actors will survive by that time. Therefore, the network has to be dense. Later, it will be argued that dense structure is also needed for a reputation mechanism. Here, we start to diverge from the thesis of the ‘strength of weak ties’ proposed by Granovetter (1973) and Burt (1992), according to which structure should not be dense, and ties should be weak. They assumed, implicitly, that one knows: • • • •

what knowledge will be relevant who has what knowledge who will survive to provide knowledge, and that one is able to absorb that knowledge.

In exploration, however, one does not yet have such knowledge and absorptive capacity, and therefore one has to hedge relational bets. One does not yet know what ties will turn out to be redundant, since one does not know who will develop what knowledge and what the configuration of relevant elements of knowledge will be. One has to maintain direct linkages even if they may later turn out to be redundant, to keep options of access open, covering for the risk that some ties will drop out and thereby eliminate indirect access to other sources. Even if a tie is already known to be redundant for access to a known source of knowledge, it may be needed to assess, understand and absorb that knowledge. More precisely, if A remains linked to both B and C, even if there is also a link between B and C, this may help A to understand C by comparing what A understands from C with what B understands from C. In other words, a dense structure enables firms to ‘triangulate’ among their multiple sources and thus better assess their value, and to better absorb knowledge from them. This role of third parties for the sake of competence is extended by roles of intermediaries for governance, discussed before. The argument against redundant relations, from Granovetter and Burt, was that their set-up and maintenance yield excess costs. However, relevant costs are only those of relation-specific investments in mutual understanding, since other, more generic investment would be useful also in other ties. In exploration, in contrast with exploitation, specific investments other than in mutual understanding are often limited in size, in activities such as prototyping rather than large outlays for efficient production, marketing, distribution, and servicing. Furthermore, in exploration, costs are less of an issue, since competition focuses

Innovation, learning and cluster dynamics 153 on form, in connecting complementary competencies in the fast development of prototypes, rather than on the price of a ready product, as in exploitation. In sum, we need to recognise the trade-off between costs and benefits of redundancy. In exploration: • • •

the relation-specific costs of setting up and maintaining ties may not be high, or at least not as high as in exploitation such costs may not have priority and redundancy may be needed – to hedge structural bets and bets on knowledge content – for triangulating knowledge content and reliability, and – for aiding the absorption of knowledge.

To maintain the variety of cognition needed for exploration, network stability is expected to be generally low, allowing for entry and exit. Exploration is facilitated by volatility of interaction, allowing for chance meetings, to discover interesting potential partners. Here, local embedding may be needed for reasons of competence. Under conditions of radical innovation, with uncertainty concerning what elements will emerge and survive in what configuration, centrality is likely to be low, especially in stand-alone technology. Centrality might yield an obstacle, from attempts to maintain the power invested in an established centralized position. With respect to the strength of ties, in exploration, uncertainty is diffuse and wide-ranging, so that interaction entails many issues, including technology, organization and perhaps also future market demand, the availability of competent suppliers and so on, and as a result ties tend to be strong in the dimension of scope. It was already noted above that building mutual understanding might require a relation-specific investment, which requires sufficient frequency of interaction and/or duration, to make such investment worthwhile. However, since knowledge changes fast, in exploration, the economic life of the investment is short, so that it should be recouped in a short time, in frequent contacts, and duration need not be long. How long duration should be depends, among other things, on the size of specific investment for mutual understanding, which depends on the depth and level of specialization of knowledge, and the degree to which it is tacit. Duration should not be too long, for two reasons. The first is that it would prevent novel architectures of configurations. This is particularly relevant under systemic conditions, where innovation often takes the form of frequent and rapid architectural change. Here, one might think of the car industry, for example. The second reason is that too durable relations may yield identification that goes so far, in an excess of familiarity, as to reduce innovative potential. However, this depends on how exclusive the relation is. If A and B have a tie, on a certain subject, and both A and B also have other ties, on the same subject, to different nodes, apparently unafraid of spillover risks, then their mutual value as sources of knowledge may be replenished from those outside contacts, so that a long duration does not necessarily kill learning potential.

154 Bart Nooteboom From a governance perspective, in exploration the use of opportunity control by contracts tends to be problematic. In exploration much knowledge is tacit, which complicates the specification of contracts. Uncertainty about contingencies, even in the very near future, may also preclude their detailed specification. Since change is rapid, the content of contracts would have to change frequently, which makes them less cost-effective. In view of new and not yet dispersed knowledge, it would be difficult to monitor and assess conformance to contracts. In exploration, governance is likely to be based on incentive control, with a balance of mutual dependence, ‘hostages’ in the form of sensitive information, a reputation mechanism, and/or on trust and mutual openness. A reputation mechanism is especially strong here, in view of the uncertainty about possible future configurations of relations. Since it is impossible to assess who may and who may not in the future yield an important connection, one has to be careful in all relations. As analysed above, a reputation mechanism requires density of relations, and is facilitated by institutional embedding. The institutional basis for trust typically lies in professional values, norms, and standards, guarded by professional associations, which also play an important role in reputation mechanisms. Typically, in exploration trust initially is competence trust, in professional knowledge and skill, and this establishes a basis for intentional trust to develop, on the basis of pre-existing professional empathy. Here we find a second argument for frequency of interaction, needed for the build-up of trust, in empathy, identification and routinization (Table 7.1). Such relation-specific, personalized trust entails, and requires, a great deal of mutual openness (Zand, 1972). It is known from the trust literature that trust is stimulated by mutual dependence. In exploration, such mutual need is high in order to search for complementary knowledge, in the race for a viable prototype. Thus, the hypothesis is that in exploration ties tend to be strong in terms of scope, frequency and trust/mutual openness, high in terms of relation-specific investments, depending on the complexity and tacitness of knowledge, and of some duration, depending on how systemic the technology is. A potential problem now is that density of the network, investment in mutual understanding, frequency of interaction, and trust and openness may all produce a high risk of knowledge or technology spillover. However, at this stage, in exploration, in which there is considerable uncertainty as to what dominant design will emerge, and what products it will lead to, and in what markets, knowledge often is ‘pre-competitive’, so that spillover risk may be limited. Also, it may be difficult to assess who will in future turn out to be a potential competitor. Restricting relations for fear of spillover would soon entail no relations at all. In any case, knowledge may change so fast as to eliminate serious spillover risk. Another potential danger is that the network becomes too tight and stable, with too durable relations between members of an ‘in-crowd’, in a tight ‘clan’ (Ouchi, 1980), which reduces diversity in terms of both people involved and cognitive distance, thereby encouraging innovative stagnation. To counter this, as discussed above, ties should not last too long, especially when technology is

Innovation, learning and cluster dynamics 155 systemic, and network stability should not be too high, offering a certain volatility of network membership, for the sake of novel combinations. This is where the thesis of the strength of weak ties comes into its own. At several points, the analysis has implications for local embedding. In particular, while reputation mechanisms, instability and volatility of network structure, wide scope of communication, frequency of interaction and trust could all occur at a geographical distance, they are greatly facilitated by the proximity and local embedding found in and characteristic of clusters. By hypothesis, in a network for exploitation, conditions are more or less the reverse of those that apply to a network for exploration. First we turn to network structure. Dominant designs have emerged, and technological and market uncertainty have decreased. Knowledge becomes more codified and stable, absorptive capacity increases, and knowledge gets widely diffused. With new entry into the emerging market, competition shifts to competition on price and marketing, so that considerations of efficiency become crucial. These pressures on cost yield a need to: •

•

•

utilize economies of scale, achievable because, owing to decreased uncertainty on the part of customers, the market has enlarged. As a result, there is increase of scale, a shakeout of producers and resulting concentration of production. search more widely for the cheapest sources of supply, made possible by reduction of uncertainty and emerging standards. Thus there are both needs and opportunities to loosen activities from their local embedding and to extend the network beyond the cluster. eliminate redundant ties, which is now possible owing to increased certainty about network structure, the location and relevance of knowledge, and the ability to absorb it.

Thus there is a requirement for a less dense structure, which is made possible by the fact that now one can identify what competencies are and will remain relevant, who has those competencies and who is likely to survive in the industry. Owing to the extension of the network, reduction of cognitive distance, codification of knowledge and slow-down of knowledge change, spillover risk increases. Owing to diffusion of knowledge and stabilization of the network, routinization of established practice and the emergence of standards, interaction becomes less intensive and shifts from developmental to transactional. The increased codification of knowledge furthers diffusion without the need for relation-specific investments in mutual understanding. Investments shift to large-scale production, distribution systems and brand name, which are all longterm. In view of such large and often sunk investments, with a long economic life, and to maintain efficient division of labour, network structure is likely to be stable. Under systemic conditions, exploitation may require considerable centrality. Concerning the strength of ties, implications of these investments for the duration of ties depend on the extent to which they are relation-specific

156 Bart Nooteboom investments, which in turn depends on the flexibility of technology: more generic or flexible technology entails that investments are less relation-specific. With an increasing division of labour for the sake of efficiency, there is an increase in specialization, so that ties entail more specific knowledge, on a narrower scope of issues. There is less need for relation-specific trust, and a basis arises for institutionbased trust. Reduced uncertainty and codified, diffused knowledge on a more narrow range of issues, made possible the specification of contracts and the monitoring of compliance, entailing a shift from trust and incentive control to opportunity control. Increased specialization, reduced scope and reduced need for trust reduce the frequency of interaction, that is, interaction in the exchange or joint production of new knowledge (purely in terms of transactions, there may be very frequent ‘just-in-time’ deliveries from suppliers). In exploitation, the extended reach of the network, into new markets of outputs and inputs, more formality of control, and less dependence on reputation mechanisms and trust, entail both opportunities and pressures for local disembedding, or ‘declustering’. The hypotheses for different conditions for networks for exploration and exploitation are summarized in Table 7.2. Within these categories of exploration and exploitation, there is still considerable variability of parameters, particularly in exploitation, as a function of contingencies of technology, market, and institutions: systemic or stand-alone, tacitness of knowledge, speed of knowledge change, flexibility of technology, economic life of investments, economies of scale, entry barriers, legal institutions, institutions for trust etc. These vary with both industry and location. Table 7.2 Networks for exploration and exploitation Network features

Exploration

Exploitation

Network structure density stability centrality

high low low

low high often high

wide high

narrow low

limited1 high low medium to high high

often long low high low to medium generally low

Strength of ties scope investment in mutual understanding duration frequency of interaction opportunity control incentive control trust/openness

1 especially when technology is systemic Source Nooteboom and Gilsing (2003)

Innovation, learning and cluster dynamics 157

Empirical illustrations An example of separation in place, with an ongoing transfer of activities from exploration networks to exploitation networks, is the Dutch pharmaceutical or biotechnology industry, where Gilsing (2005) found the following structure. Biotech firms take up an intermediary position, straddling exploration networks with universities and networks with pharmaceutical companies for the transfer from exploration to exploitation. On the whole, their ties with universities conformed to the characteristics of exploration networks, and their ties with pharmaceutical companies conformed to the characteristics of exploitation networks. In the exploration network of universities-biotech, Gilsing found high network density, high frequency of interaction, and high specific investment in mutual understanding. However, counter to expectations, he found that ties were fairly strong in control, quite strong in duration and weak in scope. This deviated from expectations, but could be explained as follows. While knowledge in the process of conducting scientific research is highly tacit, the knowledge output that was exchanged between players was highly codified, which opened opportunities for contracting. In contrast with the development and commercial use of technology, scientific research is much less concerned with issues of organization, production, commercialization and distribution, so that scope could be low. Long duration of ties was understandable from the fact that mutual understanding required high specific investment, but still appeared to yield a problem of insufficient flexibility and variability, for the sake of novel combinations. However, it turned out that a core network of durable relations in the Netherlands biotech sector was complemented with a periphery of more variable ties to universities abroad. Note that, in the theoretical analysis, durable ties were seen to be problematic for exploration only if they are also exclusive. Durable ties need not limit the diversity needed for learning if the nodes involved tap into non-overlapping outside sources, in shorter and more variable ties. In other words, here we find a hybrid network, with a core and a periphery that have compensating strengths and weaknesses. Perhaps this finding can be generalized, as follows. If a core network of stable relations is needed, for exploitation, or to recoup large specific investments, or to build and maintain trust, the potentially negative effects for exploration may be eliminated by tacking on a peripheral network that feeds the core network with diversity. There may be a lesson here for clusters, to maintain local embedding while tapping into outside sources of diversity. As an example of separation in time, with transformations from exploration to exploitation networks, and back again, Gilsing (2005) found a clear case in the emergence of the Netherlands multi-media industry. However, there was an intermediate stage, which did indeed show decreased density, increased stability, ties beyond the original local setting, increased centrality, decreased scope, and some increase of contractual control, as hypothesized. But Gilsing still found considerable trust and frequency of interaction. This could be explained by the fact that the provision of new Web-based information services

158 Bart Nooteboom still needed considerable fine-tuning of mutual fit, in a systemic coherence of elements of the overall system (hardware, software and content).

Back to the cycle of discovery Now we can give a more detailed account of transitions in the cycle of discovery. In the consolidation stage, one can expect a transition from exploration to exploitation networks, summed up as follows. In network structure, a reduction of density takes place, with the elimination of redundant ties, the emergence of centrality for co-ordinating specialized production activities, and increasing stability, to maintain systemic structures of production. Concerning strength of ties, scope decreases; duration in the division of labour and in specialization increases, to ensure stability and the recouping of increased and durable specific investments; frequency of knowledge exchange decreases, as a result of standardization; contractual control increases, owing to reduced uncertainty; and trust declines, owing to more arms-length contacts and extension of the network beyond local boundaries, in search of a widened market and cheapest sources of inputs. Next, in the generalization and differentiation stage, in new contexts of application, the expectation is that the core exploitation network in the home niche is complemented with a peripheral network that taps into novel contexts. This entails a reduction of centrality, reduced stability and increased size of the total network. We would expect emerging features of exploration in the peripheral network, and increased frequency of interaction throughout the network, to cope with growing problems of coordination between centre and periphery. In the reciprocation stage, the expectation is that the overall network begins to loosen up, and may break up altogether. A crucial contingency here is the extent to which the old exploitation system needs to be maintained along with exploration, and the extent to which it is systemic. In the multi-media case discussed above, technology was systemic, and development occurred outside the old exploitation system. At the periphery, structure becomes denser, and new investments are required for local understanding. Local commitment, at the periphery, may exceed loyalty to the centre. Stability decreases further, ties become weaker in control, and need to become stronger in trust, locally and between centre and periphery. If that is not feasible, in the maintenance of the old exploitation structure, the network is likely to break up, in a separation between core and periphery, to allow for experimentation. Finally, in the accommodation stage, novel opportunities begin to prove their worth, new networks are formed, and move towards consolidation. Existing networks in the old centre come under pressure, and may break up in the face of emerging novel dominant designs. Note that in both separation in place and separation in time we may meet hybrid networks, with a relatively stable and tight core and a looser peripheral network. In fact, as described, the process of transformation from exploitation networks to exploration networks may entail a break-up and separation in place.

Innovation, learning and cluster dynamics 159 It is very difficult, though not inconceivable, to accomplish separation in time while maintaining the overall network, owing to contradictions in the combination of a narrow focus with a wide focus and the corresponding mix of incongruent organizational cultures, stability and instability, formal control with informal trust, short and durable ties, in local embedding and outside linkages. In sum, the cycle of discovery, thus specified in terms of networks for exploitation and exploration, with attendant features in terms of structure and strength of ties, may serve as a model of cluster development.

Conclusions for clusters and cluster policy This chapter started from the claim that in order to understand ‘learning regions’ the causal effects of regional features on innovative performance should be elaborated in terms of the relations between firms and organizations that intervene between regional features and regional performance. For the analysis of such causality, it elaborated the notion of embeddedness, in three dimensions: • • •

relational embedding: strength of ties between firms, with seven dimensions structural embedding: structure of ties, with four dimensions institutional embedding: regional variables that affect conditions of competence and governance in relations between firms, partly through the dimensions of structure and strength of ties.

Innovative performance and development of clusters were analysed in terms of the combinations and transitions between exploration and exploitation. According to a ‘cycle of discovery’, exploitation emerges from exploration, on the basis of consolidation, and exploration arises from exploitation, in generalization, differentiation, reciprocation and accommodation. The combination of exploitation and exploration, in the same place and at the same time, is problematic to the extent that exploitation is systemic, i.e. entails a dense, tightly coupled structure of many elements. Under such conditions, there are two basic structural forms for combining exploration and exploitation: separation in time and separation in place. Exploitation and exploration networks were analysed and contrasted in terms of the dimensions of structure and strength of ties. Some of the analysis contradicts the thesis of the ‘strength of weak ties’. In exploration there are good arguments for density of ties, and for strength of ties in four or five of the seven dimensions of tie strength. For exploitation, density is less and ties tend to be strong, depending on further contingencies, in dimensions where they were weak in exploration. Thus, cluster development can require fairly drastic changes in the features of embedding, to allow for transitions between exploitation and exploration. The emergence and early consolidation of novelty may require protection from premature competition. Exploration requires local embedding mainly for the

160 Bart Nooteboom sake of governance by reputation and trust based on institutionally embedded morality and close interaction. On the other hand, disembedding and break-up of structure, and an escape from established interests, may be needed to proceed from exploitation to exploration. Inertia of a cluster, in the maintenance of established structure and closure to the outside, for the sake of exploitation, may seriously inhibit exploration. In the combination of exploitation and exploration, a trade-off needs to be made between sufficient durability of relations to call forth specific investments in mutual understanding and in the building of trust, and sufficient flexibility to make possible new variety for innovation. In the search for optimal cognitive distance, a trade-off must be made between distance for the sake of novelty and proximity for the sake of understanding; and between durability to encourage specific investments in mutual understanding and trust, and flexibility or non-exclusiveness, with external linkages, to maintain variety and novelty. In this view, therefore, public cluster policy is problematic. Overall, the case for such policy seems dubious. The purpose, structure and performance of clusters are connected with cluster-specific, local conditions, issuing from a history of development, under complex constellations of circumstances, as indicated in this chapter. This can yield problems for the transplantation of a successful cluster form from one local institutional context to another. A cluster may arise as a compensation for local weaknesses that do not arise elsewhere. The much-lauded development of networks or clusters (districts) in Italy can be attributed, at least in part, to a lack of reliable legal institutions, and a climate of corruption. Similar conditions currently apply in China. Then there is no opportunity for generalized institutions-based trust, and one has to fall back on personalized trust in specific relations. According to Pagden (1988), in southern Italy this goes back to a systematic breakdown of institutions in the eighteenth century, as a deliberate policy of the Spanish Habsburg emperor, ruler of the kingdoms of Naples and Sicily, to prevent coherent opposition and sedition. Thus, one should beware of ambitions for a generic blueprint for clusters that can be applied anywhere (Martin and Sunley, 2003). Clusters yield solutions to specific problems or opportunities in specific contexts. As a result, public policy, if such policy is viable at all, should probably retreat from the design of cluster structure to the facilitation of processes of cluster development, as a function of local conditions, in ‘giving a nudge here and there’, to yield triggers or remove obstacles. This requires an understanding of how clusters may develop and adapt to changing internal and external conditions. Even that, however, requires caution. The policies appropriate for one stage of cluster development may be inappropriate to another stage. For example, in an early stage of exploration one may need to support escape from established structures, and to shield embryonic novel ventures from the competitive power of established dominant designs and interests, allowing for and indeed stimulating locally embedded support. For consolidation, on the other hand, one may need to stimulate standards, efficient division of labour, for the purpose of exploitation. After consolidation one may need to shift policy towards the stimulation

Innovation, learning and cluster dynamics 161 and facilitation of local disembedding and outside reach. When new radical innovation arises from outside, one may need to encourage the break-up of localized clusters, and to eliminate entry barriers, to allow for novel architectures of old and new components. It is not easy for public bodies to identify which policies are appropriate at what time in which clusters. And when they are able to do this, the delay between design and implementation of policy may yield counter productive effects, since by the time that policies are fully implemented actual cluster requirements may have changed.

References Ashheim, B.T. and A. Isaksen (2002). ‘Regional innovation systems: The integration of local “sticky” and global “ubiquitous” knowledge’, Journal of Technology Transfer, 27 (1), pp. 77–86. Bettis, R.A. and C.K. Prahalad. (1995). ‘The dominant logic: Retrospective and extension’, Strategic Management Journal 16/1, pp. 5–14. Bogenrieder, I. and B. Nooteboom (2004). ‘Learning groups: What types are there?’, Organization Studies 25 (2), pp. 287–314. Boschma, R.A. and J.G. Lambooy (2002). Knowledge, Market Structure and Economic Co-ordination: The Dynamics of Italian Industrial Districts, University of Utrecht, Faculty of Spatial Sciences, Department of International Economics and Economic Geography. Burt, R. (1992). ‘Structural holes: The social structure of competition’, in N. Nohria and R. Eccles (eds), Networks and Organizations: Structure, Form and Action. Boston: Harvard Business School Press, pp. 57–91. Cohen, M.D. and D.A. Levinthal (1990). ‘Absorptive capacity: A new perspective on learning and innovation’, Administrative Science Quarterly 35, pp. 128–52. Coleman J.S. (1988). ‘Social capital in the creation of human capital’, American Journal of Sociology 94, pp. 95–120. Cooke, P. and R. Huggins (2001). ‘High-technology clustering in Cambridge (UK)’, in A. Amin, S. Goglio and F. Sforzi (eds), The Institutions of Local Development, Deutsch, M. (1973). The Resolution of Conflict: Constructive and Destructive Processes. New Haven: Yale University Press. Dimaggio, P. and W. Powell (1983). ‘The iron cage revisited: Institutional isomorphism and collective rationality in organizational fields’, American Sociological Review 48, pp. 147–60. Gilsing, V.A. (2005). The Dynamics of Interfirm Networks and Strategic Alliances, Exploration, Exploitation and Co-evolution, Cheltenham: Edward Elgar. Granovetter, M.S. (1973). ‘The strength of weak ties’, American Journal of Sociology 78, pp. 1360–80. Gulati, R. (1995). ‘Familiarity breeds trust? The implications of repeated ties on contractual choice in alliances’, Academy of Management Journal 38, pp. 85–112. Hansen, M.T. (1999). ‘The search-transfer problem: The role of weak ties in sharing knowledge across organization subunits’, Administrative Science Quarterly 44, pp. 82–111. Henderson, R.M. and K.B. Clark (1990). ‘Architectural innovation: The reconstruction of existing product technologies and the failure of established firms’, Administrative Science Quarterly, 35, pp. 9–30.

162 Bart Nooteboom Krackhardt, D. (1999). ‘The ties that torture: Simmelian tie analysis in organizations’, Research in the Sociology of Organizations 16, pp. 183–210. Lewicki, R.J. and B.B. Bunker (1996). ‘Developing and maintaining trust in work relationships’, in R.M. Kramer and T.R. Tyler (eds), Trust in Organizations: Frontiers of Theory and Research, Thousand Oaks: Sage Publications, pp. 114–39. McAllister, D.J. (1995). ‘Affect- and cognition based trust as foundations for interpersonal cooperation in organizations’, Academy of Management Journal 38, pp. 24–59. March, J. (1991). ‘Exploration and exploitation in organizational learning’, Organization Science 2, pp. 101–23. Nelson, R.R. and S. Winter (1982). An Evolutionary Theory of Economic Change. Cambridge: Cambridge University Press. Nooteboom, B. (1992). ‘Towards a dynamic theory of transactions’, Journal of Evolutionary Economics 2, pp. 281–99. Nooteboom, B. (1999). Inter-firm Alliances: Analysis and Design, London: Routledge. Nooteboom, B. (2000). Learning and Innovation in Organizations and Economies. Oxford: Oxford University Press. Nooteboom, B. (2002). Trust: Forms, Foundations, Functions, Failures and Figures. Cheltenham: Edward Elgar. Nooteboom, B. (2003). ‘Problems and solutions in knowledge transfer’, in D. Fornahl and T. Brenner (eds), Cooperation, Networks and Institutions in Regional Innovation Systems Cheltenham: Edward Elgar, pp. 105–25. Nooteboom, B. (2004a). Inter-firm Collaboration, Learning and Networks: An Integrated Approach. London: Routledge. Nooteboom, B. (2004b). ‘Competence and governance: How can they be combined?’, Cambridge Journal of Economics 28 (4), pp. 505–26. Nooteboom, B. and V.A. Gilsing (2003). Density and Strength of Ties in Innovation Networks: A Competence and Governance View. Draft paper, Erasmus University Rotterdam (obtainable from www.bartnooteboom.nl). Nooteboom, B., W.P.M. van Haverbeke, G.M. Duysters, V.A. Gilsing and A. van den Oord (2005) ‘Optimal cognitive distance and absorptive capacity’. Research paper, Eindhoven Center for Innovation Studies (ECIS), Eindhoven (obtainable from www.bartnooteboom.nl). Oinas, P. and E.J. Malecki (2002). ‘The evolution of technologies in time and space: From national to regional to spatial innovation systems’, International Regional Science Review 25, pp. 102–31. Ouchi, W.G. (1980). ‘Markets, bureaucracies, clans’, Administrative Science Quarterly 25, pp. 129–43. Pagden, A. (1988). ‘The destruction of trust and its economic consequences in the case of eighteenth-century Naples’, in D. Gambetta (ed.), Trust, the Making and Breaking of Cooperative Relations. Oxford: Blackwell, pp. 127–41. Polanyi, M. (1962). Personal Knowledge, London: Routledge. Shapiro, S.P. (1987). ‘The social control of impersonal trust’, American Journal of Sociology 93, pp. 623–58. Spender, J.C. (1989). Industry Recipes, Oxford: Basil Blackwell. Williamson, O.E. (1993). ‘Calculativeness, trust and economic organization’, Journal of Law and Economics 36, pp. 453–86. Wuyts, S., M.G. Colombo, S. Dutta and B. Nooteboom (2005). ‘Empirical tests of optimal cognitive distance’, Journal of Economic Behaviour and Organization, 58/2, pp. 277–302.

Innovation, learning and cluster dynamics 163 Zand, D.E. (1972). ‘Trust and managerial problem solving’, Administrative Science Quarterly 17, pp. 229–39. Zuchella, A. (2003). Geographic Co-location and Global Value Chains: Cluster Dynamics and Strategic Innovations in Cluster-Based Firms. University of Pavia.

8

Do clusters or innovation systems drive competitiveness? James Simmie

Introduction The concept of clusters has risen to prominence and general acceptance by policy makers at almost all levels of government in a remarkably short space of time. The main attraction seems to be an apparent ability to link the microeconomic requirements of individual firms with both macro-economic policy and the different spatial levels where these economic activities take place. As a result many policy makers at both central and regional or local levels of government have been seduced by the apparent intellectual coherence that provides them with practical levers to enable and facilitate the activities of private firms thereby enhancing the average competitiveness of their particular localities. What tends to be pushed to one side in all these arguments and policies is the fact that there is nothing new in certain kinds of industries locating together in particular places. A cursory examination of the histories of traditional manufacturing sectors in England easily shows the concentrations of woollen, cotton and pottery manufacturing in certain places owing to their endowments of early sources of energy or relevant raw materials. Such concentrations or clustering did not stop their eventual decline and even extinction in the face of external competition and structural economic change. The economic landscape of Britain is littered with all too many uncompetitive and declining concentrations of industries. Being highly clustered on its own has not prevented such decline taking place. It is also open to question whether clustering per se was even a cause of their initial success. Thus the main question addressed in this chapter is whether clustering is a cause or an effect of competitiveness. The arguments are pursued in the light of the fact that there are a number of other factors that can be argued to drive up the competitiveness of industries whether or not they are clustered in any meaningful way. The issue of causality is crucial to the case for or against clusters. A much stronger argument can be made that innovation and not clustering is the key to competitiveness and exports. Thus it can be argued that the production of new products or processes that are new to markets provide first-mover advantages and temporary monopoly profits to the firms that first introduce them. This is an important incentive to innovate. Following this argument, in order to be

Do clusters drive competitiveness? 165 competitive localities would be more successful if they encouraged local systems of innovation rather than clusters. Even though there could be some evidence of clustering present in fully functioning innovation systems they could be regarded as an effect of successful innovation rather than as a cause. It is argued here, therefore, that Porter’s early cluster theory is really an extension of traditional agglomeration theory. It also offers no analytical criteria for establishing the geography of clusters or an explanation of why some clusters grow and are successful while others decline and are not competitive. It can also be argued that, in his later theory, trading clusters are by definition competitive but that functioning innovation systems rather than clustering dynamics drives this success.

Porter: innovation-productivity-competitiveness In his seminal work The Competitive Advantage of Nations Porter (1990) sought to explain the competitive advantages of national economies. This focus on competitiveness led him to analyse the underlying causes of observed differences in economies. He argued that the first of these is productivity. Thus he says, ‘A region’s or nation’s standard of living (wealth) is determined by the productivity with which it uses its human, capital and natural resources. The appropriate definition of competitiveness is productivity’ (Porter 2002). Productivity, however, is not a characteristic that can be developed in a vacuum. In its turn it is also highly dependent on innovation. Porter argues that innovative capacity is the key to productivity and that competitiveness can be equated with productivity. In the context of globalization First World economies need to concentrate on high-value-added products and services and to be innovative in doing so (Porter 2003). In these economies it is ‘Productivity and innovation – not low wages, low taxes, or a devalued currency – [that] are the definition of competitiveness’ (Porter 2000, p. 30). So the key link for Porter between innovation and competitiveness is that innovation is a significant driver of productivity. Porter defines innovation as an attempt ‘to create competitive advantage by perceiving or discovering new and better ways of competing in an industry and bringing them to market’ (Porter 1990, p. 45). More broadly the concept can be defined as the introduction of a new or changed product, process, service or new form of organization into the marketplace. In short, innovation is the commercialization of new ideas. These can include new manufactured products, new ways of producing products or, more frequently, but much more difficult to measure, new services. In addition it should also be noted that innovation is not just a technological and economic process. It is also a complex social and geographic process. It is highly dependent on new knowledge and the ways in which individuals and groups exchange that knowledge. Interest in innovation has increased enormously since the recessions of the early and late 1980s. It is increasingly seen as the main economic objective of the developed economies as they are confronted by international competition

166 James Simmie from the newly industrializing countries (NICs) and less developed countries (LDCs) economies based on price and low labour costs. There is plenty of evidence to show that innovation does play an important role in driving competitiveness and hence economic growth. The OECD, for example, estimates that between 1970 and 1995 about half of the total growth in output of the developed world resulted from innovation (OECD 2000) and the proportion is increasing, as the economy becomes more knowledgeintensive. They go on to argue that between 25 and 50 per cent of economic growth comes from technological progress (OECD 2000). The Enterprise Directorate of the European Commission also estimated that, in 2002, 40 per cent of the variation in per capita income between the regions of Europe can be explained by differences in innovative performance (European Commission Enterprise D-G 2002, p. 12). The links between innovation and productivity and thence to competitiveness are complex and not particularly well understood. In principle it seems fairly plausible that process innovations can lead to more efficient forms of production and therefore productivity gains. Product and service innovations can also lead to higher sales, increasing returns to scale and therefore to productivity gains. Beyond this there is still much to unpack in these relationships. With these reservations in mind it is also possible to accept for the time being that there are further relationships between productivity and competitiveness. These tie innovation into the complex of forces that underlie the competitiveness of national economies. Measuring competitiveness is also not an easy task. The most conventional indicator used is the share of world markets taken by the products and services of a particular industry. This idea is closely related to the argument that the demand for a region’s exports drives growth. The demand for a region’s exports is determined by their competitiveness that results, in part, from regions specializing in goods and services where they have a comparative advantage. The usual operational definition of competitiveness is the ‘market share’ taken by a particular industry or sector. In this sense ‘An industry can be considered competitive if it is successful in gaining market share’ (DTI 2001). This is particularly the case if it is successful in gaining market share in world markets. Three benchmarks have been used to define what constitutes competitiveness in international markets (Trends Business Research 2001 Annex 3, p. 22). These are: • • •

Proportion of world exports that originate from national industry (UK standard 5.08%) Proportion of world values added that is created by national industry (UK standard 2.93%) Excess of national industry exports over imports by national industry.

Thus in the UK an industry could be considered to be relatively competitive if it is able to take something over 5 per cent of world exports or about 3 per cent of world values added or it runs a balance of payments surplus.

Do clusters drive competitiveness? 167 Up to this point in the argument it is therefore quite possible to agree with Porter that innovation can drive productivity in various ways that lead to new processes or sales. It also contributes to competitiveness both through its contributions to productivity and in its own right. Thus, in its own right, innovative new products and services are exportable and therefore push up competitiveness as measured by the conventional standards. So the key question that emerges from this argument is that, if innovation and productivity are linked and are the main forces underlying competitiveness, what is it that drives innovation and its effects on productivity and competitiveness?

Clusters In general terms Porter’s answer to this question is that supply-side ‘localized’ micro-economic dynamics and environments drive competitiveness and the complex of forces underlying it. It is these dynamics that he refers to as clusters. Initially he defines clusters as ‘Geographic concentrations of interconnected companies, specialised suppliers, service providers, firms in related industries, and associated institutions (for example universities, standards agencies, and trade associations) in particular fields that compete but also co-operate’ (Porter 1998, p. 197). Porter-type clusters must therefore have two key characteristics. The first is firms in the cluster must be linked in some way. Clusters are constituted by interconnected companies and associated institutions linked by commonalities and complementarities. The links are both vertical (buying and selling chains), and horizontal (complementary products and services, the use of similar specialized inputs, technologies or institutions and other linkages). Most of these linkages involve social relationships or networks that produce benefits for the firms involved. Porter’s emphasis on inter-firm linkages is highly influenced by the work of Piore and Sabel (1984). Like them he argues that ‘Businesses have been deverticalising and developing systems of flexible production’ (Porter 2003). In the examples cited in Emilia-Romagna and Baden-Württemberg this has entailed developing networks of firms that both collaborate and compete. In these instances space matters because physical proximity reduces the transaction costs of such a system of production. This closely parallels the network paradigm of Granovetter (1973). But, to the extent that the sharpness and pervasiveness of the ‘New Industrial Divide’ is exaggerated then the significance of local space attributes in the use and upgrading of factors of production is also overstated. Multinational, multilocational companies play a major role in the globalized economy. Their complex spatial divisions of labour use different factors of production in different locations, all of which contribute to eventual outputs. In such cases space matters but not so much because of local uses of factors of production so much as the interconnections between different places involved in internationally distributed value chains (Castells 1987, Amin and Thrift 1992, Simmie 2002a and 2002b).

168 James Simmie The second key characteristic is that clusters are geographically proximate groups of interlinked companies. Co-location encourages the formation of and enhances the value creating benefits arising from networks of interaction between firms. Quite how proximate firms need to be is never made clear. It appears that clusters can be found at almost any level of spatial aggregation. ‘They are present in large and small economies, in rural and urban areas, and several geographic levels (for example nations, states, metropolitan regions and cities)’ (Porter 1998, p. 204). Their geographical scope can extend as far as ‘a network of neighbouring countries’ (Porter 1998, p. 199). The vagueness on the geographic scale of clusters is repeated in Porter (2002, p. 7). The practical examples of clusters provided by Porter usually seem to operate at the geographical scale of large sub-regions within American states. The Californian wine industry is one such example. There is also strong presumption in the new supply-side models of endogenous economic growth that the geographic extent of clusters must be limited because of the importance attached to transaction costs and untraded interdependencies in providing local competitive advantages. It is argued that only fairly close geographic proximity can develop the confidence and trust needed to both compete but also collaborate. Porter’s analysis of these ‘localized’ forms of interaction is depicted in the form of his well-known diamond model of the effects of localities on competition. This has four significant interrelated influences. These are factor inputs ranging from tangible assets such as physical infrastructure to information, the legal system and university research institutes that all firms draw on in competition. The context for firm strategy and rivalry refers to the rules, incentives and norms governing the type and intensity of local rivalry. Demand conditions in a locality have much to do with whether firms can and will move from imitative, low-quality products to competing on differentiation. Advancement means the development of more demanding markets. Finally, related and supporting industries and the connections across industries are regarded as fundamental to competition, productivity and to the direction and pace of new business formation and innovation (Porter 2000, pp. 18–20). Martin and Sunley (2003) correctly argue that both the geographic and the industrial definitions of clusters in Porter’s work are ridiculously elastic. They go on to say that there is nothing inherent in the concept itself that provides a way of defining the spatial range of a cluster or which are the key dynamic processes at different geographical scales. This imprecision helps to account for the enormous popularity of the concept in policy circles. It can mean all things to all policy makers. It does mean, however, that it is very difficult to investigate the dynamics of clustering starting from a spatial definition.

Clusters and competitiveness Regardless of the geographic imprecision of the concept of clusters, it is clear that the key dynamic within them is the interlinkages between different sectors, institutions and actors as portrayed in the diamond analysis. It is therefore

Do clusters drive competitiveness? 169 interesting to investigate whether, as a general principle, competitive industries are usually characterized by strong and significant interlinkages with others. Empirical evidence on the nature and extent of inter-firm linkages is thin on the ground mainly because the only really satisfactory way of identifying such linkages is by expensive, detailed original survey work. As a result studies such as those by Bennett et al. (1999) and Trends Business Research (2001) assume that significant linkages exist between different sectors either because they are co-located or because they are simply concentrated in a particular locality. Even in Porter’s (2003) latest analysis using County Business Patterns data at the four-digit SIC level across the whole of the United States, the only indicator used to show sectoral interlinkages is the geographic correlation of different sectors in particular localities. Apart from original survey work, one other method that has been used to identify real inter-sectoral linkages is input/output analysis. Input/output tables define the spatial level of interlinkages at the geographic scale of the national economy. There are no similar data sources on interlinkages at the level of the urban region. As part of their study of clusters for the DTI, Trends Business Research (2001) defined which UK industries could be considered competitive according to the criteria outlined above. They also analysed the input/output tables for 123 sectors of the UK economy to define statistically which industrial sectors were significantly interlinked with each other irrespective of the geographic extent of those linkages. From these analyses it is possible to identify which UK sectors could be defined as competitive in 2001 and which were interlinked with each other. Table 8.1 shows that some 20 sectors in the UK economy could be considered competitive in conventional terms. That is they were either running a balance of payments surplus or they exceeded the UK’s average share of world markets or its average contribution to world values added. In terms of the ratio of exports to imports the top eight competitive industries are services. They include banking and finance, insurance and pension funds as well as producer services such as legal, market and accountancy services. The most competitive manufacturing sector on this criterion is weapons and ammunition. Other manufacturing sectors that take a significant share of world markets are pesticides, agricultural machinery, aircraft and spacecraft, and metal forging. The results of the Trends Business (2001) input/output analysis are shown in Table 8.2. The first and most significant finding is that they could only identify some 12 major interlinked clusters of industries in the whole of the UK economy. In his latest analysis that emphasizes the significance of traded clusters even Porter (2003) identifies only 41 key traded clusters in the whole of the US economy each containing about 29 sectors. Thus the all too frequent analyses that identify a handful of firms in a single sector as a cluster are a complete nonsense. The second key finding shown in Table 8.2 is that out of the twelve clusters only five contained sectors defined as competitive on the basis of their export performance as shown in Table 8.1. Not surprisingly, the most competitive

170 James Simmie Table 8.1 UK competitive industries Industry

Ratio of UK Share in world exports/imports exports (%)

Share in world value added (%)

Banking and finance Public administration Insurance and pension funds Legal activities Auxillary financial services Market research Education Accountancy services Weapons and ammunition Architectural etc. activities Pesticides Agricultural machinery Other business services Soap and toilet preparations Printing and publishing Pharmaceuticals Computing services Aircraft and spacecraft Metal forging, pressing etc Owning and dealing in real estate

16.70 9.08 5.79 5.33 4.51 2.95 2.61 2.60 2.48 2.44 2.29 2.01 1.91 1.67 1.63 1.48 1.41 1.18 1.00

8.4 3.9 9.5 6.9 6.2 5.8 3.2 5.1 4.2 4.6 8.0 5.1 5.1 5.6 7.1 5.9 4.4 6.5 7.4

UK average

7.6 16.1 11.3 10.5 10.1 13.2 10.9 11.7 11.9 11.3 5.8 5.08

2.93

Source Trends Business Research (2001) Annex 3: The UK Economy, its Competitive Performance and Linkages, p. 23

cluster is composed of services concerned with knowledge, organization and management. On the export criteria used here this is by far the most competitive cluster. It is followed by the industrial chemicals and plastics part of the chemicals cluster embracing two competitive sectors, metals and mechanical engineering, motor vehicles, and cultural and leisure industries each with one competitive sector. On this evidence we may conclude that the mere existence of interlinkages within clusters, said to be one of their main defining characteristics, does not always lead to competitive performance by those clusters. In fact the contrary case can also be made. This is that there are competitive industries that are not associated with clusters. Table 8.3 shows that there are some sectors that are competitive but not particularly interlinked with other sectors. These include some of the best export performers such as weapons and ammunition, soap and toilet preparations, and aircraft and spacecraft. This raises questions over what role the most significant attribute of clustering, namely interlinkages, may play in driving competitiveness and hence economic prosperity. Admittedly this top-down approach is limited by the lack of more detailed sectoral divisions. Spatial analysis below the level of the national economy is also dependent on subjective calculations derived

Do clusters drive competitiveness? 171 Table 8.2 Interlinked trading clusters in England and Wales Main cluster

Sector linkages

Buildings and materials

Construction industry Home ownership and rentals Bricks

Textiles and clothing

Artificial fibres Textile fibres Knitted goods Weaving Carpets and rugs Made-up textiles Other textiles Apparel Textile finishing

Wood and woodworking

Forestry Timber and wood products Furniture Small tools

Knowledge, organization and management

Education Research and development Accounting Law Architecture and design Estate agents Business services Consultancy and market research Computing services

Logistics

Railways Transport services Shipping Air transport Couriers Postal service Wholesale distribution

Electronics

Electronic components Office machinery TV and radio transmitters TV and radio receivers Precision instruments Electric motors

Metals and mechanical engineering

Miscellaneous manufacturing (including recycling) Ore extraction Non-ferrous metals Jewellery Iron and steel Miscellaneous metal goods Structural metal products continued

172 James Simmie Table 8.2 continued Main cluster

Sector linkages Forging Casting Engines Machine tools General purpose machines Special machines Containers and other transport equipment

Motor vehicles

Motor vehicle distribution Agricultural machinery Rubber and tyres Rental Domestic appliances

Food and agriculture

Fertilizers Feedstuffs Agriculture Meat processing Leather tanning and footwear Dairy products Oils and fats Sugar Confectionary Miscellaneous foods Milling Bread Fishing and fish processing

Chemicals Industrial chemicals and plastics

Consumer chemicals

Organic chemicals Pesticides Inorganic chemicals Synthetic resins Plastic products Industrial gases Soft drinks Paints and dies

Energy

Crude oil and natural gas Gas distribution Processed fuels Coal mining Electricity

Cultural and leisure industries

Publishing including music Recreational services including broadcasting Advertising Paper and board and their products Membership organisations

Note Internationally competitive sectors in italics Source Trends Business Research (2001) London, DTI, Annex 3: The UK Economy, its Competitive Performance and Linkages, pp. 26–7

Do clusters drive competitiveness? 173 Table 8.3 UK competitive sectors not interlinked with other industries Banking and finance Public administration Insurance and pension funds Auxilliary financial services Weapons and ammunition Soap and toilet preparations Aircraft and spacecraft Note Internationally competitive sectors in italics Source Trends Business Research (2001) London, DTI. Annex 3: The UK Economy, its Competitive Performance and Linkages, pp. 23 and 26–7

from the original 123 sector tables. There may be smaller clusters that exist below the level of the national economy. It should also be pointed out that locally delimited clusters could and have had exactly the opposite effects to those suggested by Porter. Martin and Sunley (2003) point out that ‘Economic landscapes are littered with local areas of industrial specialization that were once prosperous and dynamic but have since gone into relative or even absolute decline’. Too much intellectual inbreeding can lead to ‘lock-in’. This is a process that arises if too much reliance is placed upon local knowledge and face-to-face relationships. This may work within certain parameters but makes the cluster very vulnerable to major shifts in technology such as the development of electronic as compared with mechanical watches and digital publishing compared with hot lead type setting. Cluster advocates offer no explanation of how the dynamics of clusters can go into reverse in this way. They are also quiet on the issue of how cluster policies should respond to these dangers. So far we can conclude that the available evidence is mixed on the contribution of clustering to competitiveness. First genuine clustering is a much more limited phenomenon than many policy analyists have suggested. There are at most only a relatively small number of interlinked sectors in the UK economy as shown by input/output analysis. Even Porter can detect only 41 important trading clusters in the whole of the US economy. Second, clustering per se is no guarantee of competitiveness. There are a limited number of internationally competitive sectors in the UK economy. Some of these are clustered and some are not. There is also a majority of clustered sectors that are not internationally competitive and quite large numbers of previously competitive and clustered industries that have also gone into decline. In these mixed circumstances we should not expect to find that innovative industries that are usually competitive would also normally regard local clustering as one of the keys to their success.

174 James Simmie

Porter analysis of relationships between competitiveness, innovation and clustering Porter arrived at the notion of clusters as a result of following his arguments on the nature and characteristics of competition. Accordingly he argues that ‘Competition is dynamic and rests on innovation and the search for strategic differences. Close linkages with buyers and suppliers and other institutions are important, not only to efficiency, but also to the rate of improvement and innovation. Location affects competitive advantage through its influence on productivity and especially on productivity growth’ (Porter 2000, p. 19). In this respect the key to successful competition is based on the ability first to produce continuous streams of innovation and secondly to position a company strategically in the marketplace in such a way as to produce products that are both different from and superior to those of rivals. Given Porter’s concern with national competitiveness the major function of clusters is to contribute to national prosperity. The interlinked elements of this contribution include the building of innovative capacity leading to improvements in productivity. Porter claims that there are a number of advantages to be gained with respect to the key activity of innovation by operating in a cluster. These advantages include the ability to perceive and react to new buyer needs more quickly owing to the proximity of demanding and sophisticated customers. In addition firms can see the evolution of new technologies and understand their implications and possibilities more quickly. Local relationships, including those with universities, are said to facilitate this process (Porter 2000, p. 23). From this perspective the cluster concept has become increasingly associated with the ‘new’ or ‘knowledge’ economy. The argument here is that the processes that drive the development of new economic knowledge and its application and commercialization in innovation are facilitated by localisation (Martin and Sunley 2003). Norton (2001), who argues that the success of the US in the ‘new’ economy derives directly from the growth of large and dynamic clusters of innovation and entrepreneurialism, supports this idea. Baptista (1996) has also argued that ‘geographical concentration is of foremost importance for organisational improvement and technological innovation’ (Baptista 1996, p. 60). In summary, Porter argues that localized clusters deliver innovation because: • • • • • •

They allow rapid perception of new buyer needs. They concentrate knowledge and information. They allow the rapid assimilation of new technological possibilities. They provide richer insights into new management practices. They facilitate ongoing relationships with other institutions including universities. The knowledge-based economy is most successful when knowledge resources are localized.

Do clusters drive competitiveness? 175

The geography of innovation There is a prima facie case for the idea that innovation processes benefit from local clustering dynamics. This is because as a result of the ways in which new economic knowledge is created and exchanged there is a strong regional geography of innovation. In the United States Audretsch and Feldman (1996) used a 1982 Small Business Administration census of innovation citations taken from over a hundred scientific and trade journals to identify the geography of US innovations. The census included a total of 4,200 new product announcements that contained information on the location of the enterprise that introduced the innovation. Their first finding was that the spatial concentration of innovative activity in particular industries was much greater than for all manufacturing. For example, 41.7 per cent of all recorded innovations in the computer industry were in California. A further 12 per cent were listed in Massachusetts. As a result these two states alone account for more than half of all the innovations in the computer industry. Altogether the most innovative sectors provide 80 per cent of all innovations. Beyond this, 11 states account for 81 per cent of all innovations. California is the state in which the greatest numbers of innovations were listed. New York, New Jersey and Massachusetts followed this. Audretsch and Feldman (1996) comment that ‘A particularly striking feature . . . is that the bulk of innovative activity in the United States occurs on the coasts, and especially California and New England’. In Europe, a study by Hilpert (1992) of the location of scientific funding from the European Community, national governments and the regions found that up to three-quarters was concentrated in ten ‘Islands of Innovation’. These were identified according to the following criteria: • • • •

islands which specialize in more than one of the three studied technoscientific fields islands which are covering more than 20 per cent of public R&D expenditures in the country strong presence in the islands of both research institutions and enterprises islands which are European ‘knots’ in the web of co-operation links (Hilpert 1992, p. iv).

The ten major European islands identified in this way are Greater London, Rotterdam/Amsterdam, Ile-de-France, the Ruhr area, Frankfurt, Munich, Lyon, Grenoble, Turin, and Milan. See Simmie (2001) for a study of five of these city regions. Much of the explanation of the geography of innovation has focused on versions of a supply-side cluster-type regional endogenous growth model. These argue that innovation is concentrated in certain regions because of the internal supply characteristics of those regions (see for example Camagni 1991, Todtling 1992, Aydalot and Keeble 1988, Grabher 1993, Storper 1995, Cooke 1997,

176 James Simmie Malecki 1997, Jaffe et al. 1993, Feldman 1994, Swann et al. 1998, Audretsch and Feldman 1996, Scott 1988). Nevertheless, while, on the one hand the mechanisms through which innovations are conceived and brought to market are increasingly international, on the other hand they are all conceived in particular localities and so subnational ‘hot-spots’ are formed in particular fields (Metcalf et al. 2002). These ‘hot spots’ are often concentrated in city regions giving rise, for example, to the distinctive urban European geography of innovation as shown by Hilpert (1992), Simmie (2001), Huggins (2001) and IAURIF (2002). Thus there is strong evidence to show that innovation is concentrated in a limited number of city regions. This provides the prima facie case for clustering making an important contribution to innovation. But innovation is an internationally distributed phenomenon with international linkages between different ‘hot’ city regions, which suggests that international connections may be at least as significant as local clustering dynamics.

Some evidence on why innovative firms are concentrated in a few ‘hot spots’ In a recent piece of research (Simmie et al. 2002) 160 innovative firms were interviewed in five major European city regions by teams of locally based researchers. Among other things, the firms were asked to rate on a scale of 1 to 5 the importance to them of a list of reasons for locating in their particular cities. Principal component analysis was then used to group their replies into a series of related sets of reasons. The rankings given to these groupings are shown in Table 8.4. The highest ratings were given to fairly traditional factors associated with agglomeration economies. These included professional and skilled labour and business services, and transport and communications. The two most significant reasons for firm location were good access to a major airport (mean score 3.39) and the availability of professional and technical labour (mean score 3.86). In contrast the types of reason that might be expected to indicate the significance of clustering such as production and consumption linkages and networks, and social networks tended to score lower than agglomeration advantages. Proximity of collaborators (mean score 2.85) followed by proximity of suppliers at 2.58 were the top rated cluster linkage type of reasons for innovative firms to be located where they were. At the geographic level of city regions the characteristics of traditional agglomeration economies can easily be mistaken for evidence of firms interlinked in clusters. The original analyses of why geographically proximate groups of firms appear and persist can be traced back to the work of Marshall (1919), Hoover (1937, 1948), Perroux (1950), Isard (1951), Chinitz (1961, 1964) and Mills (1980). The general analyses of pure agglomeration usually start with the classification proposed by Hoover (1937, 1948). He grouped the sources of agglomeration

Do clusters drive competitiveness? 177 Table 8.4 Reasons for location of innovative firms in five European cities Main components

City regions mean scores

1. Labour, premises and capital Cost of labour Availability of premises Cost of premises Access to finance capital Low levels of traffic congestion

2.23 2.44 2.47 2.05 2.18

2. Professional and skilled labour and business services Availability of professional experts to recruit Availability of skilled manual labour Access to private general business services Access to private specialised business services Proximity of business services

3.86 2.89 2.11 2.10 2.03

3. Transport and communications Good access to central city Good rail connections Good access to national road network Good access to major airport

2.47 2.69 3.19 3.39

4. Production and consumption linkages and networks Proximity of customers Proximity of suppliers Proximity of competitors Proximity of collborators Proximity of sources of information

2.26 2.58 1.34 2.85 2.19

5. Social networks Local public business support services Presence of ex-colleagues Presence of friends

1.82 1.57 1.34

6. Public knowledge, information, training and research provision Contributions from TECs Contributions from Business LINKS Contributions from universities

1.60 1.47 2.38

Notes Scores 1 = not important to 5 = very important Source Simmie (2002)

advantages into three categories. These were internal returns to scale, localization economies and urbanization economies. Contrary to much of the more recent analyses of clusters, the pure model of agglomeration presumes no form of co-operation between actors beyond what is in their individual interests in an atomized and competitive environment. The key variable is the size of the

178 James Simmie agglomeration. Greater size increases the chances of profitable local interactions through chance, the law of large numbers and natural selection of the businesses that can benefit from the multiple opportunities on offer. Since Hoover’s work (1937, 1948) major economic changes have taken place. One of the most important of these, particularly since the 1970s and accelerating during the 1990s (Veltz 1993, Gordon 1996) is the globalization of the world economy. This has involved, among other phenomena, internationalization, growing instability in product markets, more intense competition, and greater emphasis on competition based on quality and variety rather than price. These changes place a competitive premium on economies that may accrue locally but are initially, at least, external to both the firm and the urban region. Some evidence of the significance of international trading linkages for innovative firms can be gleaned from the third Community Innovation Survey (CIS 3). This is becoming the most comprehensive Europe-wide survey of innovation. Local agents for each of the fifteen member states conduct it on a four-yearly cycle. The methodology is based on the recommendations of the Oslo manual (OECD/Eurostat 1997). In the UK, the Office of National Statistics (ONS) conducted the latest survey in 2001 for the Department of Trade and Industry (DTI). It involved a two-stage sample of all firms in the UK. In the first stage 13,315 firms were sent a postal questionnaire in April 2001. A top-up survey of 6,287 was conducted in November of the same year. This produced a total sample of 8,172 firms. The results were weighted to represent all firms in the production and construction industries, wholesale trade (excluding motor vehicles), financial intermediation and business services. The weighted results constitute the best estimates of the innovation activities of firms across the entire UK for the period 1998–2000. Table 8.5 shows an analysis taken from CIS-3 of the locations of collaborators for innovative and non-innovative firms. It may be seen that, in general, innovative firms tend to have higher rates of collaboration and therefore linkages than non-innovative firms do. This could suggest some contribution to innovation by clustering. On the other hand the highest rates of collaboration are recorded with national rather than local firms and institutions. Furthermore, higher rates of collaboration with suppliers, customers and competitors are recorded for Europe and the US than with their local equivalents. These data indicate the complex nature of the kinds of linkages that contribute to innovation within firms. They show that local agglomeration economies are still important and suggest that city size plays a more significant role in providing the kinds of assets required by innovating firms than does any form of clustering. They also show that while linkages at numerous geographical scales are important, purely local linkages of the kinds associated in some of the literature with clusters are often less significant than national and even international connections and collaborations. This supports the view of innovation as a set of internationally distributed systems located in city regional ‘hot spots’ in the more advanced national economies.

0.1

Specialized Private research institutes

Source Community Innovation Survey 3

3776

0.4 0.3

Institutional Universities or other higher education institutes Government research organizations

Total N=100%

1.3

0.1

3154

0.7

4.5 0.7

3.8 4.5 1.1 2.6

5.1

0.5 0.6 0.3 0.5

0.5

Market Suppliers of equipment, materials, comps. software Clients or customers Competitors Consultants Commercial laboratories or R&D enterprises

Internal Other enterprises within group

3776

0.3

0.5 0.3

0.4

1.2 0.9 0.3 0.6

0.7

National

Local

Local

Noninnovators

NonLeading innovators innovators

Table 8.5 Local co-operation arrangements by type of collaborator

3154

3.0

7.7 3.1

4.0

13.4 14.0 4.0 7.6

5.9

National

Leading innovators

3776

0.0

0.1 0.1

0.1

0.4 0.3 0.1 0.1

0.4

European

Noninnovators

3154

0.8

3.1 0.6

1.3

7.1 7.3 2.5 1.2

5.2

European

Leading innovators

Leading innovators

3776

0.1

0.1 0.1

0.1

0.2 0.2 0.0 0.1

0.5

3154

0.4

0.5 0.3

1.1

3.2 4.8 1.5 1.4

5.0

United States United States

Noninnovators

180 James Simmie

Trading clusters In his earlier work, despite the fact that globalized economic interactions are increasing in importance, Porter argued that such linkages mitigate disadvantages rather than create advantages. He said, ‘distant sourcing is a second-best solution compared to accessing a competitive local cluster in terms of productivity and innovation’ (Porter 2000, p. 32). As a result he emphasized the significance of micro-economic conditions and the ability to improve them in order to improve the competitiveness of the macro economy in general. Nevertheless, globalization appears to reduce the incentives for firms to invest time and resources in purely local clusters. Instead they clearly need to be competitive in international markets. This requires capabilities for fast changing business strategies, flexibility and constant re-combinations of specialized suppliers and other business partners. Globalization and changing products have also reduced the importance of traditional localized factors of production. All these factors seem to emphasize the importance of ‘weak ties’ (Granovetter 1973) which are multiple, open-ended and changing and link both national producers and international customers. More recently Porter (2003) has taken these arguments on board. He now argues that it is primarily export-oriented clusters that drive regional prosperity. Exporting clusters tend to pay higher wages than those serving purely local markets do and so they help to pull up other wages in the regional economy. Export clusters are, however, much more likely to have national and international linkages than to be based on purely local connections. The critical importance of these extended linkages in the context of a globalized international economy calls into question the relative significance of the kinds of limited and local connections so often stressed by local policy makers supposedly following Porter’s analysis. While the earlier version of the cluster hypothesis had much in common with traditional agglomeration economy theory, this latest version of the cluster hypothesis has much in common with traditional export-base theory. Export-base models were founded on the theory that demand for a region’s exports drives growth. They were developed originally by Ohlin (1933), North (1955), Tiebout (1956) and Richardson (1969). They argued that a region’s growth is determined by the exploitation of natural advantages and the growth of the regional export base, which are in turn largely influenced by the level of external demand from other regions and countries. The demand for a region’s exports is determined by their competitiveness that results, in part, from regions specializing in goods and services where they have a comparative advantage. Further development of export-based models also emphasized the impacts of cumulative causation and agglomeration. Kaldor (1970) and Dixon and Thirwall (1975) developed the idea that regions are able to exploit the benefits of economies of scale and specialization. This improves their export performance and in turn raises output growth. Later developments incorporated the effects of external economies of scale. Here it is argued that geographical concentrations of economic activity improve productivity and thereby raise output.

Do clusters drive competitiveness? 181 This theoretical approach has seen something of a revival over the past decade. One leading economist, Paul Krugman, has labelled this the ‘new economic geography’. Among other things, this revival now recognizes the key importance of cities and regions in shaping a nation’s competitive performance (Krugman 1991a; Fujita, Krugman and Venables 1999). At the heart of this recognition is the argument that the competitiveness of a nation’s industries in the global marketplace is shaped in large part by the extent to which those industries are able to benefit from the increasing returns that flow from localized specialized agglomeration. One of the more recent expositions of this theory was propounded in the Kalecki Memorial Lecture, by Rowthorn, who argued that ‘The prosperity of a region is determined primarily by the strength of its export base’ (Rowthorn 1999, p. 23). In this case the export base is defined as ‘all those activities which bring income into the region by providing a good or service to the outside world, or provide locals with a good or a service which they would otherwise have to import. The alternative term “tradables” is also used to denote such activities’ (Rowthorn 1999, p. 22). The export base of a city region is important not just for its local supply-side characteristics but also because of its international demand-side linkages. From this perspective exports and trade bring in external knowledge into the innovation processes of cities. While some export-based growth models were designed, in the first instance, to explain the development of NICs and LDCs, innovation and trade are vehicles for technological spillovers in numerous directions. They can provide the knowledge and experience needed for these less advanced economies to catch up with the more advanced. There is also an empirical relationship between accumulated R&D expenditures and total factor productivity. The benefits of R&D can spill across both the less and more advanced countries through trade. This effect is larger the more open an economy is to foreign trade. The significance of innovation to competitiveness in general and to exports in particular is shown in Table 8.6. This shows a comparison between innovative and non-innovative firms in the UK taken from the CIS 3. It may be seen that between 1998 and 2000 the mean growth in turnover among leading innovators was around three times that for non-innovating firms. The same applied to exports with the mean growth in exports among leading innovators being more than three times that for non-innovators. There is therefore a strong correlation between innovation and export growth. In so far as exports are a good indicator of competitiveness then it may be argued that innovation is a key driver of competitiveness. Thus far it is possible to agree with Porter that exports are the key to economic growth in particular localities. This is primarily because they bring in new capital, revenues and ideas to city regions. The recognition of the significance of exports to the economic growth of localities is not new and represents the rediscovery of traditional export base theory. It is also clear that innovation is one of the key drivers of exports. The development of market leading goods,

182 James Simmie Table 8.6 Growth in key indicators of competitiveness of non- and leading innovators 1998–2000 (mean scores %)

Growth Turnover

Non-innovators

Leading innovators

???

???

6.4

19.5

Exports

29.1

100.2

Capital expenditure

29.5

46.3

7.2

9.8

Employees

Source Community Innovation Survey 3

processes and services provides the firms that accomplish this with comparative advantages over their rivals and sometimes early-mover monopoly profits in the early years of an innovation’s product life-cycle. What is not clear is whether clustering drives exports either directly or through innovation.

Innovation = competitiveness = exports So far it has been argued that the kinds of local supply-side assets found in innovation ‘hot spots’ have more in common with traditional agglomeration economies than with clustering. The sheer size of capital cities such as London and Paris provides multiple possibilities for inter-firm collaborations, the circulation of new ideas and informal and ad hoc relationships during the development of innovations. Second, given Porter’s increasing stress on trading clusters rather than purely local clusters, it is also argued that this has more in common with demand side export base theories than clusters per se. The supply and demand sides for innovative firms both involve complex multi-layered local, national and international linkages. While many of these could well be similar to the clustering dynamics of the Porter diamond, they are not confined within a single geography. With respect to understanding the contribution that innovation makes to competitiveness, the concept of clusters is therefore of limited value. It is possible to conceive of an alternative argument that starts with innovation and offers some explanation of competitiveness and export success without recourse to the concept of clusters. This could run something along the following lines. Given that economic growth is unlikely to be based on selling the same products and services on a permanent basis, competitiveness requires the continual commercialization of new ideas in the form of innovations. It is therefore the systems that facilitate and enable this continual process to take place that are the key to economic competitiveness. Innovation systems are complex and internationally distributed. The different elements of these internationally distributed systems are also highly concentrated in a limited number of city regions. Collectively these city regions make up

Do clusters drive competitiveness? 183 their respective national economies. At the same time this concentration enables particular city regions to gain comparative advantages and to export competitively into world markets. The export base of cities is therefore highly reliant on the presence and functioning of local innovation systems. The main elements of a local innovation system include: • • •

financing by risk and venture capital investment in firms investment in intellectual capital and the creation of new ideas investment in human capital, particularly in the quality of the workforce.

In addition to adequate investment in these forms of capital, local innovation systems also require: • • • • • • •

an entrepreneurial business culture that is tolerant of failure commercialization and marketing a social culture that is tolerant of diversity and new ideas and ways of doing things high levels of specialization supplying the best in national and international markets knowledge-generating firms and institutions knowledge exploitation firms high levels of technical sophistication among producers or users of technology.

The combination of these characteristics forms the basis of competitive exports. These also bring benefits to the local innovation system. These benefits include: • •

economies of scale international knowledge spillovers from sophisticated customers providing the local innovation system with information on leading-edge knowledge, products and services.

This system can create a virtuous circle running from capital investment through to innovation and competitive exports. The elements of this system by themselves are necessary but not sufficient to create such a virtuous circle. In addition they need to interact as a system. This requires connectivity in terms of transport, communications and social networks at regional, national and international levels. Connectivity is the crucial dynamic of innovation systems that turns them from a series of qualities and activities in their own individual right to something greater than the sum of the individual parts. In these favourable circumstances it is possible for some evidence of clustering to emerge. But in this argument competitive clusters are an effect of the interactions of a functioning innovation system and not a cause. Unless a local economy develops the capacity to innovate, change and adapt it is likely to stagnate. It is the lack of these capacities that also begins to explain the large numbers of co-located firms that constitute uncompetitive clusters.

184 James Simmie

Conclusions The cluster idea has taken many academics and policy makers by storm. It has become the accepted wisdom more quickly than any other major idea in the field in recent years. This accelerated progress has been at the expense of previous explanations and lacking in relevant empirical evidence. On the one hand explanations of why firms co-locate in particular localities that rely on the idea of agglomeration economies seem to have been forgotten. These provide very plausible explanations of why firms locate in large city regions that do not rely on the interlinkages hypothesized by the Porter cluster theory. On the other hand when the focus shifts to trading clusters traditional exportbase theory also does not receive the recognition that it deserves. Again it provides a perfectly acceptable explanation of why local economic growth requires exports as a base that does not rely on any elements of cluster theory. Another feature that is thin on the ground with respect to cluster theory is adequate empirical evidence on a number of issues. First the interlinkages of various kinds that are said to be a key element in clustering are seldom revealed by new survey work. Instead they are hypothesized to exist on the basis of the geographic co-location of industrial sectors. This is not good enough because, as has already been pointed out, agglomeration theory provides an explanation of why concentrations of industrial sectors may be found together in the same locality without any need for interlinkages between them. Second, the theory provides no a priori criteria for the identification of even approximate cluster boundaries. In the absence of a clearly defined object of study it has been very difficult to conduct comparable studies that could establish what they do on a consistent basis. Thirdly, there is also the vexed question of causality. Porter is quite clear that productivity gains lead to competitiveness and innovation leads to improved productivity. The dynamic interactions within a given cluster drive all three and, in particular lead to competitiveness. This chapter, however, raises the issue of whether cluster dynamics should be seen as a cause or an effect of these other phenomenon. With respect to innovation, empirical evidence has been presented that suggests clustering is not one of its major causes. In fact the reverse is the case. Functioning local systems of innovation rely on a combination of different forms of investment together with an entrepreneurial culture, and sophisticated technical and managerial knowledge to produce competitive exports. Sometimes the activities of such a system come to resemble that of a cluster as described by Porter. But the direction of causality is from innovation to clusters rather that the reverse. These arguments have some implications for policy makers. The first is that the rush to accelerate the development of clusters is misplaced. It has been shown that in a majority of cases in the UK clustered sectors are not competitive. Therefore a simple cluster building strategy is just as likely to promote the growth of uncompetitive and non-exporting firms as the reverse. Second, if exports, along with FDI and tourism, are one of the keys to local economic growth, then it is essential to re-focus policies on the export base of

Do clusters drive competitiveness? 185 cities. One of the most effective ways to do this is to establish functioning local innovation systems. These have been shown to improve export performance. Some elements of such systems are susceptible to public policy interventions. Paradoxically one of the key elements of local innovation systems and place competitiveness is connectivity. By this is meant transport, ICT and network linkages. Some of these phenomena are also associated with cluster dynamics. Thus the promotion of local systems of innovation could also have the effect of generating some of the kinds of linkages that are said to be one of the defining characteristics of clusters. This should not be allowed to obscure the fact that they are an effect of innovation rather than a cause and that they also serve different purposes from the more descriptive linkages hypothesized in much of the literature on clusters.

References Amin, A. and Thrift, N. (1992) ‘Neo-Marshallian nodes in global networks’, International Journal of Urban and Regional Research, 16 (4), pp. 571–87. Audretsch, D. B. and Feldman, M. P. (1996) ‘R&D spillovers and the geography of innovation and production’, The American Economic Review, 86, pp. 630–40. Aydalot, P. and Keeble, D. (eds) (1988) High Technology Industry and Innovative Environments: The European Experience. London, Routledge. Baptista, R. (1996) ‘Research round up: Industrial clusters and technological innovation’, Business Strategy Review, 7 (2), pp. 59–64. Bennett, R. J., Graham, D. J. and Bratton, W. (1999) ‘The location and concentration of business clusters, business services, market coverage and local economic development’, Transactions of the Institute of British Geographers, NS 24, pp. 393–420. Camagni, R. (1991) (ed.) Innovation Networks: Spatial Perspectives. London, BelhavenPinter. Castells, M. (1987) ‘Technological change, economic restructuring and the spatial division of labour’, in Muegge, H. and Stohr, W. B. (eds) International Economic Restructuring and the Regional Economy. Aldershot, Avebury. Chinitz, B. (1961) ‘Contrasts in agglomeration: New York and Pittsburgh’, American Economic Review, 51, pp. 279–89. Chinitz, B. (1964) ‘City and suburb’, in Chinitz (ed.) City and Suburb: The Economics of Metropolitan Growth. Englewood Cliffs, NJ, Prentice-Hall. Cooke, P. (1997) ‘Regional innovation systems – an evolutionary approach’, in Braczyk, H., Cooke, P. and Heidenrich, R. (eds) Regional Innovation Systems. London, UCL Press. Department of Trade and Industry (1999) Biotechnology Clusters: Report of a team led by Lord Sainsbury, Minister for Science. London, DTI. Dixon, R. J. and Thirwall, A. P. (1975) ‘A model of regional growth rate differentials along Kaldorian lines’, Oxford Economic Papers, 27, pp. 201–14. European Commission (2002) European Innovation Scoreboard 2002: Technical Paper No 3: EU Regions. Brussels, EU Enterprise Directorate-General. Feldman, M. P. (1994) The Geography of Innovation. Boston, Kluwer Academic Publishers. Fujita, M., Krugman, P. and Venables, A. (1999) The Spatial Economy: Cities, Regions and Trade. Cambridge, Mass., MIT Press.

186 James Simmie Gordon, I. R. (1996) ‘Territorial competition and locational advantage in the London Region’, paper presented to the American Association of Geographers annual conference, Charlotte, NC. Grabher, G. (ed.) (1993) The Embedded Firm: On the Socio-economics of Industrial Networks. London, Routledge. Granovetter, M. (1973) ‘The strength of weak ties’, American Journal of Sociology, 78, pp. 1360–80. Hilpert, U. (1992) Archipelago Europe – Islands of Innovation, Synthesis Report, Forecasting and Assessment in Science and Technology, Vol. 18, Prospective Dossier No. 1, ‘Science, Technology and Social and Economic Cohesion in the Community’. Brussels, Commission of the European Communities. Hoover, E. M. (1937) Location Theory and the Shoe and Leather Industries, Cambridge, Mass., Harvard University Press. Hoover, E. M. (1948) The Location of Economic Activity. New York, McGraw-Hill. Huggins, R. (2001) Global Index of Regional Knowledge Economies: Benchmarking South East England. Guildford, South East England Development Agency. Institut d’Amanegement et d’Urbanisme de la Region Ile-de-France (IAURIF) (2002) Science and Technology in the Functional Urban Regions in Northwest Europe. Paris, IAURIF. Isard, W. (1951) Location and Space Economy. Cambridge, Mass., MIT Press. Jaffe, A. B., Trajtenberg, M. and Henderson, R. (1993) ‘Geographic localization and knowledge spillovers as evidenced by patent citations’, Quarterly Journal of Economics, 108, pp. 79–104. Kaldor, N. (1970) ‘The case for regional policies’, Scottish Journal of Political Economy, Krugman, P. (1991a) ‘Increasing returns and economic geography’, Journal of Political Economy, 99 (3), pp. 483–99. Krugman, P. (1991b) Geography and Trade. Cambridge, Mass., MIT Press. Krugman, P. (1997) Pop Internationalism. Cambridge, Mass., MIT Press. Malecki, E. J. (1997) Technology and Economic Development. Longman, Harlow, Addison Wesley. Marshall, A. (1919) Industry and Trade. London, Macmillan. Martin, R. and Sunley, P. (2003) ‘Deconstructing clusters: chaotic concept or policy panacea?’, Journal of Economic Geography, 3, pp. 5–35. Metcalf, S., Coombs, R. and Georghiou, L. (2002) ‘Developing technology policy – balancing supply-side with demand’. Paper presented to the Cambridge/MIT Seminar on Technology Policy, London, 19 November 2002. Mills, E. S. (1980) Urban Economics. Glenview Ill., Scott, Forseman and Co. North, D. C. (1955) ‘Location theory and regional economic growth’, Journal of Political Economy, 63, pp. 243–58. Norton, R. D. (2001) Creating the New Economy: The Entrepreneur and US Resurgence. Cheltenham, Edward Elgar. OECD (1997) Oslo Manual: Proposed Guidelines for Collecting and Interpreting Technological Innovation Data Paris, OECD. OECD (2000), A New Economy? The Changing Role of Innovation and Information Technology in Growth Paris, OECD. Ohlin, B. (1933) Inter-regional and International Trade. Cambridge, Mass., Harvard University Press. Perroux, F. (1950) ‘Economic space: theory and applications’, Quarterly Journal of Economics, 64, pp. 89–104.

Do clusters drive competitiveness? 187 Piore, M. and Sabel, C. (1984) Second Industrial Divide. New York, Basic Books. Porter, M. E. (1990) The Competitive Advantage of Nations. London, Macmillan. Porter, M. E. (1998) On Competition Cambridge, Mass., Harvard Business School Press. Porter, M. E. (2000) ‘Location, competition, and economic development: Local clusters in a global economy’, Economic Development Quarterly, 14 (1), February, pp. 15–34. Porter, M. E. (2002) ‘Regional foundations of competitiveness and implications for government policy’, Department of Trade and Industry Workshop, 16 April. Porter, M. E. (2003) ‘The economic performance of regions’, Regional Studies, 37 (6 & 7), August/October, pp. 549–78. Richardson, H. W. (1969) Regional Economics. London, Weidenfeld and Nicolson. Rowthorn, R. (1999) ‘The political economy of full employment in Britain’, Kalecki Memorial Lecture, Department of Economics, University of Oxford, 19 October. Scott, A. J. (1988) New Industrial Spaces. London, Pion. Simmie, J. M. (2001) Innovative Cities. London, Spon. Simmie, J. M. (2002a) ‘Trading places in the global economy’, European Planning Studies, 10 (2), February, pp. 201–14. Simmie, J. M. (2002b) ‘Innovation, international trade and knowledge spillovers’, Italian Journal of Regional Science, 1, pp. 73–91. Simmie, J. M. and Sennett, J. (1999) ‘Innovative clusters: Global or local linkages?’, National Institute Economic Review, 170, October, pp. 87–98. Simmie, J. M., Sennett, J., Wood, P. and Hart, D. (2002) ‘Innovation in Europe: A tale of knowledge and trade in five cities’, Regional Studies, 36 (1), February, pp. 47–64. Storper, M. (1995) ‘The resurgence of regional economies, ten years later: The region as a nexus of untraded interdependencies’, European Urban and Regional Studies, 2 (3), 191–221. Storper, M. (1997) The Regional World: Territorial Development in a Global Economy. New York, Guildford Press. Swann, P., Prevezer, M. and Stout , D. (eds) (1998) The Dynamics of Industrial Clustering. Oxford, Oxford University Press. Tiebout, C. M. (1956) ‘Exports and regional economic growth’, Journal of Political Economy, pp. 160–4. Reply by North D. C. and Rejoinder by Tiebout, C. M. Todtling, E. (1992) ‘Technological change at the regional level – the role of location, firm structure and strategy’, Environment and Planning A, 24 (1), 565–84. Trends Business Research (2001) Business Clusters in the UK – A First Assessment. London, DTI. Veltz, P. (1993) ‘L’économie des villes, entre la montée du global et le retour du local’, Territoires et Sociétés (mimeo).

9

The role of clusters in knowledge creation and diffusion An institutional perspective David B. Audretsch and Erik E. Lehmann

Introduction Ever since Robert Solow (1956) based his model of economic growth on the neoclassical production function with its key factors of production, capital and labor, economists have relied upon the model of the production function as a basis for explaining the determinants of economic growth. Paul M. Romer’s (1986) critique of the Solow approach was not with the basic model of the neoclassical production function, but rather what he perceived to be omitted from that model – knowledge. Not only did Romer (1986), along with Robert E. Lucas (1988) and others argue that knowledge was an important factor of production, along with the traditional factors of labor and capital, but, because it was endogenously determined as a result of externalities and spillovers, it was particularly important. The recognition that knowledge is a key factor determining competitiveness and economic growth was accompanied by two developments that were largely unanticipated. The first was the (re-)emergence of the importance of regions and geographic proximity as important units of economic activity. The second was that much of the innovative activity is less associated with footloose multinational corporations and more associated with small entrepreneurial start-up ventures forming the basis of high-tech innovative regional clusters, such as Silicon Valley, Research Triangle and Route 128. Only a few years ago the conventional wisdom predicted that globalization would render the demise of the region as a meaningful unit of economic analysis. According to The Economist, ‘The death of distance as a determinant of the cost of communications will probably be the single most economic force shaping society in the first half of the next century.’ Yet the obsession of policy-makers around the globe to ‘create the next Silicon Valley’ reveals the increased importance of geographic proximity and regional agglomerations. The rediscovery of the importance of geographic proximity in shaping economic performance has not escaped the attention of scholars. In proposing a new theory of economic geography, Paul Krugman (1991, p. 55) asks, ‘What is the most striking feature of the geography of economic activity? The short answer is surely concentration . . . production is remarkably concentrated in

Clusters in knowledge creation and diffusion 189 space.’ A careful and systematic series of empirical studies provided evidence that what Krugman observed to be true for production was even more pronounced for innovative activity. This finding helped trigger a new literature with the goal of understanding the spatial dimension of innovative activity, specifically the determinants and mechanisms that underlie the propensity of innovative activity to cluster spatially. Knowledge spillovers figure prominently in addressing these issues. Even as scholars assembled the requisite theoretical frameworks and empirical analyses to reach conclusions with a high degree of confidence about the importance of geographic location, agglomerations and clusters for competitiveness and growth, they began to question the role that institutions along with the organization and structure of economic activities play within spatially bounded regions. The purpose of this chapter is to identify what has been learned in this new literature on the role of geographic clusters. The second section draws on the literature that analyzes the re-emergence of the spatial cluster as an important unit of analysis. This literature has focused largely on the impact that globalization has had in shifting the comparative advantage from traditional factors, such as physical capital, to knowledge. However, a key insight of this literature is that a fundamental feature of this new factor involves the propensity for knowledge spillovers to be spatially localized. The third section focuses on institutional factors that may, in fact, impede the spillover of knowledge that is assumed to occur automatically in the Romer model, or what we call the knowledge filter. Mechanisms that can serve to permeate the knowledge filter, and in particular, the organizational form of the entrepreneurial firm are identified. Finally, in the last section a summary and conclusions are provided. In particular, the findings of this chapter suggest globalization has dramatically altered both the spatial and the organizational forms of economic activity, rendering regional clusters once again a key building block for international competitiveness. Public policy has responded with a focus towards fostering the development of knowledge-based regional clusters.

Globalization and the re-emergence of regional clusters That globalization is one of the defining changes at the turn of the century is clear from a reading of the popular press. Like all grand concepts, a definition for globalization is elusive and elicits criticism. That domestic economies are globalizing is a cliché makes it no less true. In fact, the shift in economic activity from a local or national sphere to an international or global orientation ranks among the most significant changes shaping the current economic landscape. Globalization would not have occurred to the degree that it has if the fundamental changes were restricted to the advent of the microprocessor and telecommunications. It took a political revolution in significant parts of the world to reap the benefits from these technological changes. The political counterpart of the technological revolution was the increase in democracy and

190 David B. Audretsch and Erik E. Lehmann concomitant stability in areas of the world that had previously been inaccessible. The Cold War combined with internal political instability rendered potential investments in Eastern Europe and much of the developing world too risky and impractical. During the post-war era most trade and economic investment was confined to Europe and North America, and later a few of the Asian countries, principally Japan and the Asian Tigers. Trade with countries behind the Iron Curtain was restricted and in some cases prohibited. Even trade with Japan and other Asian countries was highly regulated and restricted. Similarly, investments in politically unstable countries in South America and the Middle East resulted in episodes of nationalization and confiscation where the foreign investors lost their entire investments. Such political instability rendered foreign direct investment outside of Europe and North America to be particularly risky and of limited value. The fall of the Berlin Wall and subsequent changes in governments in Eastern Europe and the former Soviet Union were a catalyst for change and accessibility to parts of the world that had previously been inaccessible for decades. As Thurow (2002, pp. 25–6) points out, ‘Much of the world is throwing away its communist or socialist inheritance and moving towards capitalism. Communism has been abandoned as unworkable (China), imploded (the USSR), or has been overthrown (Eastern Europe).’ Within just a few years it has become possible not just to trade with but also to invest in countries such as Hungary, the Czech Republic, Poland, Slovenia, as well as China, Vietnam and Indonesia. For example, India became accessible as a trading and investment partner after opening its economy in the early 1990s. Trade and investment with the developed countries quickly blossomed. Trade and investment with the United States tripled between 1996 and 1997, reflecting the rapid change in two dimensions. First, India was confronted with sudden changes in trade and investment, not to mention a paradigmatic shift in ways of doing business. Second, to the foreign partner, in this case the United States, taking advantage of opportunities in India also meant downward pressure on wages and even plant closings in the originating country. As Thurow (2002, pp. 38–9) concludes, ‘As long as communism was believed to be a viable economic system, there were limits to global capitalism whatever the technological imperatives. Capitalism could not go completely global because much of the globe was beyond its reach. Forty percent of humanity lived under communism.’ With the opening of some of these areas and participating in the world economy for the first time in decades, the post-war equilibrium came to a sudden end. Opportunities associated with the gaping disequilibria were abruptly created. Consider the large differentials in labor costs. As long as the Berlin Wall stood, and countries such as China and Vietnam remained closed, large discrepancies in wage rates could be maintained without eliciting responses in trade and foreign direct investment. The low wage rates in China or parts of the former USSR neither encouraged foreign companies to build plants nor resulted in large-scale trade with the West on the basis of access to low production costs. Investment by foreign companies was either prohibited by local governments

Clusters in knowledge creation and diffusion 191 or considered to be too risky by the companies. Similarly trade and other restrictions limited the capabilities of firms in those countries from being able to produce and trade with the West. Thus, the gaping wage differentials existing while the Iron Curtain stood and much of the communist world was cut off from the West were suddenly exposed in the early 1990s. There were not only unprecedented labor cost differentials but also massive and willing populations craving to join the high levels of consumption that had become the norm in Western Europe and North America. Of course, the productivity of labor is vastly greater in the West, which compensates to a significant degree for such large wage differentials. Still, given the magnitude of these numbers both trade and investment have responded to the opportunities made possible by the events of 1989. While the most salient feature of globalization involves interaction and interfaces among individuals across national boundaries, the more traditional measures of transnational activity reflect an upward trend of global activities. These traditional measures include trade (exports and imports), foreign direct investment (inward and outward), international capital flows, and inter-country labor mobility. The overall trend for all of these measures has been strongly positive. The reason why location has (re-)emerged as an important spatial unit of observation in a rapidly globalizing economy is attributable to the shift in comparative advantage of high-wage countries to knowledge. In particular, despite the claim of ‘The Death of Distance’ (as The Economist put it), to access knowledge, and in particular knowledge spillovers, local proximity to the knowledge source(s) bestows competitive advantage. The spillover of knowledge is a key mechanism in the models of endogenous growth. However, the spatial dimension has been less clear. For example, in disputing the role of knowledge externalities in explaining the geographic concentration of economic activity, Krugman (1991) and others do not question the existence or importance of such knowledge spillovers. In fact, they argue that such knowledge externalities are so important and forceful that there is no compelling reason for a geographic boundary to limit the spatial extent of the spillover. According to this line of thinking, the concern is not that knowledge does not spill over but that it should stop spilling over just because it hits a geographic border, such as a city limit, state line, national boundary or intercontinental ocean. Thus, it took more than theories of knowledge spillover, or knowledge externalities, to explain the (re)emergence of location as a platform for harnessing knowledge and generating innovative activity. The second theoretical leg involves explanations or theories of localization, which explain why the economic value of knowledge tends to decline as it is transmitted across geographic space. As Audretsch and Feldman (1996) explain, the theory of the localization of knowledge spillovers lies in a distinction between knowledge and information. Information has a singular meaning and interpretation. It can be codified at low

192 David B. Audretsch and Erik E. Lehmann cost and the transaction cost is trivial. In contrast, knowledge is vague, difficult to codify and often only serendipitously recognized. While the marginal cost of transmitting information across geographic space has been rendered trivial by the telecommunications revolution, the marginal cost of transmitting knowledge, and especially tacit knowledge, rises significantly with distance. Why is geographic proximity so important for the transmission of knowledge, and especially tacit knowledge? Localization theories suggest that face-to-face interaction and non-verbal communication facilitate the transmission of ideas and intuition that cannot be communicated through codified instructions. While information is often context free, tacit knowledge is often derived from specific contexts. Thus, in order to access knowledge and participate in the generation of new ideas, local proximity is significantly more cost effective than trying to attain the same knowledge across distance. Perhaps it was this insight that led Glaeser, Kallal, Scheinkman and Shleifer (1992, p. 1126) to conclude that ‘intellectual breakthroughs must cross hallways and streets more easily than oceans and continents’. The emergence of knowledge as perhaps the key factor in shaping economic growth, employment creation and competitiveness in globally linked markets has also had an impact on the organization of economic activity. First, it has impacted that spatial organization of economic activity. In particular, globalization has rendered the organization of economic activity for the spatial unit of the region more important. Just as globalization has reduced the marginal cost of transmitting information and physical capital across geographic space to virtually zero, it has also shifted the comparative advantage of a high-cost Standort, or location, in the developed countries from being based on capital to being based on knowledge. This shift in the relative cost of (tacit) knowledge vis-à-vis information has been identified as increasing the value of geographic proximity. To access knowledge, locational proximity is important. Thus, a paradox of globalization is that it has geographic proximity and location as being more important, not in spite of a globalizing economy, but because of it.

The knowledge filter and the missing link in regional clusters In the Romer model of endogenous growth new technological knowledge is assumed to spill over automatically. Investment in new technological knowledge is automatically accessed by third-party firms and economic agents, resulting in the automatic spillover of knowledge. The assumption that knowledge automatically spills over is, of course, consistent with the important insight by Arrow (1962) that knowledge differs from the traditional factors of production – physical capital and (unskilled) labor – in that it is non-excludable and nonexhaustive. When the firm or economic agent uses the knowledge, it is neither exhausted nor can it be, in the absence of legal protection, precluded from use by third-party firms or other economic agents. Thus, in the spirit of the Romer model, drawing on the earlier insights about knowledge from Arrow, a large

Clusters in knowledge creation and diffusion 193 and vigorous literature has emerged obsessed with the links between intellectual property protection and the incentives for firms to invest in the creation of new knowledge through R&D and investments in human capital. However, the preoccupation with the non-excludability and nonexhaustability of knowledge first identified by Arrow and later carried forward and assumed in the Romer model, neglects another key insight in the original Arrow (1962) article. Arrow also identified another dimension by which knowledge differs from the traditional factors of production. This other dimension involves the greater degree of uncertainty, higher extent of asymmetries, and greater cost of transacting new ideas. The expected value of any new idea is highly uncertain, and as Arrow pointed out, has a much greater variance than would be associated with the deployment of traditional factors of production. After all, there is relative certainty about what a standard piece of capital equipment can do, or what an (unskilled) worker can contribute to a mass-production assembly line. By contrast, Arrow emphasized that, when it comes to innovation, there is uncertainty about whether the new product can be produced, how it can be produced and whether sufficient demand for that visualized new product might actually materialize. In addition, new ideas are typically associated with considerable asymmetries. In order to evaluate a proposed new idea concerning a new biotechnology product, the decision-maker might not only need to have a PhD in biotechnology, but also a specialization in the exact scientific area. Such divergences in education, background and experience can result in a divergence in the expected value of a new project or the variance in outcomes anticipated from pursuing that new idea, both of which can lead to divergences in the recognition and evaluation of opportunities across economic agents and decision-making hierarchies. Such divergences in the valuation of new ideas will become greater if the new idea is not consistent with the core competence and technological trajectory of the incumbent firm. Thus, because of the conditions inherent in knowledge – high uncertainty, asymmetries and transactions cost – decision-making hierarchies can reach the decision not to pursue and try to commercialize new ideas that individual economic agents, or groups or teams of economic agents think are potentially valuable and should be pursued. The basic conditions characterizing new knowledge, combined with a broad spectrum of institutions, rules and regulations, impose what Audretsch et al. (2006) term the knowledge filter. The knowledge filter is the gap between new knowledge and what Arrow (1962) referred to as economic knowledge or commercialized knowledge. The greater is the knowledge filter, the more pronounced is this gap between new knowledge and new economic, or commercialized, knowledge. The knowledge filter is a consequence of the basic conditions inherent in new knowledge. Similarly, it is the knowledge filter that creates the opportunity for entrepreneurship in the knowledge spillover theory of entrepreneurship. According to this theory, opportunities for entrepreneurship are the duality of the knowledge filter. The higher is the knowledge filter, the greater are the

194 David B. Audretsch and Erik E. Lehmann divergences in the valuation of new ideas across economic agents and the decision-making hierarchies of incumbent firms. Entrepreneurial opportunities are generated not just by investments in new knowledge and ideas but in the propensity for only a distinct subset of those opportunities to be fully pursued by incumbent firms. Thus, the knowledge theory of entrepreneurship shifts the fundamental decision-making unit of observation in the model of the knowledge production function away from exogenously assumed firms to individuals, such as scientists, engineers or other knowledge workers – agents with endowments of new economic knowledge. When the lens is shifted away from the firm to the individual as the relevant unit of observation, the appropriability issue remains, but the question becomes, How can economic agents with a given endowment of new knowledge best appropriate the returns from that knowledge? If the scientist or engineer can pursue the new idea within the organizational structure of the firm developing the knowledge and appropriate roughly the expected value of that knowledge, she has no reason to leave the firm. On the other hand, if he or she places a greater value on his ideas than do the decision-making bureaucracy of the incumbent firm, he or she may choose to start a new firm to appropriate the value of his or her knowledge. In the knowledge spillover theory of entrepreneurship the knowledge production function is actually reversed. The knowledge is exogenous and embodied in a worker. The firm is created endogenously in the worker’s effort to appropriate the value of his or her knowledge through innovative activity. Typically an employee from an established large corporation, often a scientist or engineer working in a research laboratory, will have an idea for an invention and ultimately for an innovation. Accompanying this potential innovation is an expected net return from the new product. The inventor would expect to be compensated for his or her potential innovation accordingly. If the company has a different, presumably lower, valuation of the potential innovation, it may decide either not to pursue its development or that it merits a lower level of compensation than that expected by the employee. In either case, the employee will weigh the alternative of starting his or her own firm. If the gap in the expected return accruing from the potential innovation between the inventor and the corporate decision maker is sufficiently large, and if the cost of starting a new firm is sufficiently low, the employee may decide to leave the large corporation and establish a new enterprise. Since the knowledge was generated in the established corporation, the new start-up is considered to be a spin-off from the existing firm. Such start-ups typically do not have direct access to a large R&D laboratory. Rather, the entrepreneurial opportunity emanates from the knowledge and experience accrued from the R&D laboratories with their previous employers. Thus the knowledge spillover view of entrepreneurship is actually a theory of endogenous entrepreneurship, where entrepreneurship is an endogenous response to opportunities created by investments in new knowledge that are not commercialized because of the knowledge filter.

Clusters in knowledge creation and diffusion 195 The knowledge spillover theory of entrepreneurship challenges two of the fundamental assumptions implicitly driving the results of the endogenous growth models. The first is that knowledge is automatically equated with economic knowledge. In fact, as Arrow (1962) emphasized, knowledge is inherently different from the traditional factors of production, resulting in a gap between knowledge and what he termed as economic knowledge, or economically valuable knowledge. The second involves the assumed spillover of knowledge. The existence of the factor of knowledge is equated with its automatic spillover, yielding endogenous growth. In the knowledge spillover theory of entrepreneurship the existence of the knowledge filter imposes a gap between new knowledge and new economic knowledge, and results in a lower level of knowledge spillovers. Thus, as a result of the knowledge filter, entrepreneurship becomes central to generating economic growth by serving as a conduit for knowledge spillovers. The process involved in recognizing new opportunities emanating from investments in knowledge and new ideas, and attempting to commercialize those new ideas through the process of starting a new firm is the mechanism by which at least some knowledge spillovers occur. In the counterfactual situation, that is in the absence of such entrepreneurship, the new ideas would not be pursued, and the knowledge would not be commercialized. Thus, entrepreneurs serve an important mechanism in the process of economic growth. An entrepreneur is an agent of change, who recognizes an opportunity, in this case generated by the creation of knowledge not adequately pursued (in the view of the entrepreneur) by incumbent organizations, and ultimately chooses to act on that opportunity by starting a new firm. Recognition of what Arrow (1962) termed the non-excludability of knowledge inherent in spillovers has led to a focus on issues concerning the appropriability of such investments in knowledge and the need for the protection of intellectual property. However, Arrow (1962) also emphasized that knowledge is also characterized by a greater degree of uncertainty and asymmetry that is other types of economic goods. Not only will the mean expected value of any new idea vary across economic agents, but the variance will also differ across economic agents. Thus, if an incumbent firm reaches the decision that the expected economic value of a new idea is not sufficiently high to warrant the development and commercialization of that idea, other economic agents, either within or outside of the firm, may instead assign a higher expected value of the idea. Such divergences in the valuation of new knowledge can lead to the start-up of a new firm in an effort by economic agents to appropriate the value of knowledge. Since the knowledge inducing the decision to start the new firm is generated by investments made by an incumbent organization, such as in R&D by an incumbent firm or research at a university, the start-up serves as the mechanism by which knowledge spills over from the sources producing that knowledge to the (new) organizational form in which that knowledge is actually commercialized. Thus, entrepreneurship serves as a conduit, albeit not the sole conduit, by which knowledge spills over.

196 David B. Audretsch and Erik E. Lehmann As investments in new knowledge increase, entrepreneurial opportunities will also increase. Contexts where new knowledge plays an important role are associated with a greater degree of uncertainty and asymmetries across economic agents evaluating the potential value of new ideas. Thus, a context involving more new knowledge will also impose a greater divergence in the evaluation of that knowledge across economic agents, resulting in a greater variance in the outcome expected from commercializing those ideas. It is this gap in the valuation of new ideas across economic agents, or between economic agents and decision-making hierarchies of incumbent enterprises, that creates the entrepreneurial opportunity. The knowledge spillover theory of entrepreneurship analogously suggests that, ceteris paribus, entrepreneurial activity will tend to be greater in contexts where investments in new knowledge are relatively high, since the new firm will be started from knowledge that has spilled over from the source actually producing that new knowledge. A paucity of new ideas in an impoverished knowledge context will generate only limited entrepreneurial opportunities. By contrast, in a high-knowledge context, new ideas will generate entrepreneurial opportunities by exploiting (potential) spillovers of that knowledge. Thus, the knowledge spillover view of entrepreneurship explains why entrepreneurial activity will tend to result from investments in new knowledge. By serving as a conduit for the spillover of knowledge that otherwise might not have been commercialized, entrepreneurship provides a missing link to economic growth. Because the spillover of knowledge tends to be localized within the spatial context of geographically bounded regions, entrepreneurship becomes an important vehicle in regional clusters by which (regional) knowledge spills over and becomes transmitted into (regional) growth.

Conclusions Globalization has led to a shift in the comparative advantage of developed countries away from the factor of physical capital and towards the factor of knowledge. Because not only does knowledge tend to spill over, but such spillovers are geographically localized, regional clusters have emerged as a key spatial unit of organization of economic activity. However, this chapter has also pointed out that the spillover of investments in new knowledge is by no means automatic. Rather, the knowledge filter can impede the spillover and commercialization of knowledge. By serving as a conduit for knowledge spillovers, entrepreneurship is the missing link between investments in new knowledge and economic growth. Thus, the spillover theory of entrepreneurship provides not just an explanation of why entrepreneurship has become more prevalent as the factor of knowledge has emerged as a crucial source for comparative advantage, but also why entrepreneurship plays a vital role in generating economic growth. Entrepreneurship is an important mechanism permeating the knowledge filter to facilitate the spillover of knowledge and ultimately generate economic growth.

Clusters in knowledge creation and diffusion 197 A generation ago, entrepreneurial firms within the context of regional clusters did not seem to be prominent in the public policy approach to enhance growth and create employment. For example, in advocating a new public policy approach to promote growth and international competitiveness at the European level, Servan-Schreiber warned of the ‘American Challenge’ in the form of the ‘dynamism, organization, innovation, and boldness that characterize the giant American corporations’ (1968, p. 153). Because giant corporations were considered to be the engine of growth and innovation, Servan-Schreiber advocated the ‘creation of large industrial units which are able both in size and management to compete with the American giants’ (1968, p. 159). According to ServanSchreiber (1968, p. 159), ‘The first problem of an industrial policy for Europe consists in choosing 50 to 100 firms which, once they are large enough, would be the most likely to become world leaders of modern technology in their fields. At the moment we are simply letting industry be gradually destroyed by the superior power of American corporations.’ Ironically, the 1988 Cecchini Report identified the gains from European integration as largely accruing from increases in scale economies. However, the more recent insights concerning the role of entrepreneurship and regional clusters have become a focal point in the debate to foster growth and employment. For example, in the Lisbon Accord of 2002, the European Commission made a formal commitment to becoming the entrepreneurship and knowledge leader in the world by 2020 in order to foster economic growth and prosperity throughout the European Union. Similarly, as Bresnahan and Gambardella (2004, p. 1) point out, Clusters of high-tech industry, such as Silicon Valley, have received a great deal of attention from scholars and in the public policy arena. National economic growth can be fueled by development of such clusters. In the United States the long boom of the 1980s and 1990s was largely driven by growth in the information technology industries in a few regional clusters. Innovation and entrepreneurship can be supported by a number of mechanisms operating within a cluster, such as easy access to capital, knowledge about technology and markets, and collaborators. Similarly, Wallsten (2004, p. 229) suggests that ‘Policy makers around the world are anxious to find tools that will help their regions emulate the success of Silicon Valley and create new centers of innovation and high technology’. Whatever those policy instruments may actually prove to be, regional clusters have emerged as a key focal point for organizing and harnessing not just investments in new knowledge, but also the spillover of that knowledge, leading to growth, employment creation and international competitiveness.

198 David B. Audretsch and Erik E. Lehmann

References Arrow, K. (1962). Economic welfare and the allocation of resources for invention. In The Rate and Direction of Inventive Activity. Princeton: Princeton University Press, 609–26. Audretsch, D.B. and Feldman, M. (1996). R&D spillovers and the geography of innovation and production. American Economic Review, 86, 630–40. Audretsch, D.B., Keilbach, M. and E.E. Lehmann (2006). Entrepreneurship and Growth, Oxford University Press, Oxford. Bresnahan, T. and Gambardella, A. (ed.) (2004). Building High-Tech Clusters: Silicon Valley and Beyond. Cambridge: Cambridge University Press. Glaeser, E., Kallal, H., Scheinkman J. and Shleifer A. (1992). Growth in cities. Journal of Political Economy, 100, 1126–52. Krugman, P. (1991). Geography and Trade. Cambridge, MA: MIT Press. Lucas, R. (1988). On the mechanics of economic development. Journal of Monetary Economics, 22, 3–42. Lucas, R. (1993). Making a miracle. Econometrica, 61, 251–72. Romer, P. (1986). Increasing returns and long-run growth. Journal of Political Economy, 94, 1002–37. Servan-Schreiber, J. (1968). The American Challenge. London: Hamish Hamilton. Solow, R. (1956). A contribution to theory of economic growth. Quarterly Journal of Economics, 70, 65–94. Thurow, L. (2002). Fortune Favors the Bold. Cambridge, MA: MIT Press. Wallsten (2004). The role of government in regional technology development: The effects of public venture capital and science parks. In T. Bresnahan and A. Gambardella (eds) Building High-Tech Clusters. Silicon Valey and Beyond. Cambridge: Cambridge University Press, 229–79.

10 Do clusters ‘think’? An institutional perspective on knowledge creation and diffusion in clusters Michael Steiner Changing perspectives in the interpretation of clusters Some years ago, Mary Douglas raised the provocative issue of ‘how institutions think’, embedding the question in a wider framework of the relation between rational choice and collective behaviour, and pointing to the ‘central paradox’ of extending the individuality of thinking to social groups and institutions (Douglas, 1986, p. 36). In this chapter, we seek to ask a parallel question, namely: ‘Do clusters think?’ More specifically, and linking the discussion with Douglas’s focus of enquiry, we are interested in the question of whether and to what extent clusters can be regarded as institutions, above the level of the individual firm, for the creation and diffusion of knowledge. Since Porter’s (1998, 2000) original definition of clusters as ‘geographic concentrations of interconnected companies, specialized suppliers, service providers, firms in related industries and associated institutions in a particular field that compete but also cooperate’, the emphasis of cluster analysis has changed. The initial concentration on the usefulness of the predominantly Porterian cluster concept as a model of ‘regional competitiveness’ has been subjected to several well-argued criticisms that point to many fundamental conceptual, theoretical and empirical questions and that urge a ‘much more cautious and circumspect use of the notion’ (Martin and Sunley, 2003, p. 5). It has also been argued that Porter’s cluster theory embodies ideas arising from quite different conceptual approaches – sometimes complementary, but sometimes contradictory – leading to confusion and ambiguity (Gordon and McCann, 2000; Martin and Sunley, 2003). The recent debate has focussed more on how far and in what ways clusters foster knowledge-creation organizational learning, emphasizing the organicevolutionary dimension of cluster-based industrial agglomerations. Growth of the knowledge base depends on intended and unintended individual processing of experiences, while the interpretation, transfer and use of experiences is influenced by interaction between individuals and between organizations (Cohen and Levinthal, 1989; Anderson, 1995; Hartmann, 2004). Several new strands of theory from diverse disciplines – economics, geography, sociology, psychology – coincide in the exploration of this aspect of the cluster debate:

200 Michael Steiner •

•

•

Knowledge has been recognized as a major source of competitive advantage in an increasingly integrated world economy (Dosi and Malerba, 1996; Grant, 1996; Foss, 1999; Nonaka et al., 2000). The most successful regions are perceived to be those whose firms display innovative capacity, being able to adapt to a rapidly changing marketplace and stay one step ahead of competitors. This is connected with the changing character of knowledge leading to new forms of organizations where the dichotomy between market and hierarchy is challenged by hybrids in the form of networks. Attention has thus focused on innovation as an interactive process involving the sharing and the exchange of different forms of knowledge between actors (Lawson and Lorenz 1999) – knowledge and competence is developed interactively and within subgroups of a (regional) economy (Freeman, 1991; Lundvall, 2002). The critique here has been concerned with the question of whether this interaction is an outcome of (neoclassical) rational behaviour or the result of a more ‘associative-relational’ mode of organization, or what has been termed ‘associative governance’, which leads to the creation of clubs, forums, consortia and other institutional schemes of partnership (Cooke, 1998; Cooke and Morgan, 1998). To what extent this associative approach has specific regional or spatial dimensions has also become a topic of investigation. The focus here is on the necessity, and forms, of proximity for knowledge exchange. The key argument is that the collaborative nature of innovation processes has reinforced tendencies toward geographical clustering because of the advantages of locating in close proximity to other firms in specialized and related industries (Storper, 1995, 1997). Despite the claimed ubiquity of access to information engendered by the rapid growth of telecommunications, access to tacit knowledge based on networks and face-to-face contacts, which offer greater reliability and less risk, tends to be spatially concentrated. The discussion here has centred on the exclusive spatial dimension of ‘proximity’ (Rallet and Torre, 1998). A third element is the growing interest in the role of institutions as a factor shaping economic performance in general, and for knowledge creation in particular. The complexity of co-operation (Axelrod, 1997) is a phenomenon that cannot be explained solely by individual decision-making: strong rationality is not sufficient for relatively effective economic behaviour. The institutional perspective seeks to identify additional factors influencing economic behaviour, factors that lead to co-operation and that emphasize that human behaviour has to be understood as a social and cultural phenomenon influenced by institutions of various kinds (Hodgson, 1998).

These strands of thinking – relatively new in kind and especially their combination – have nevertheless a longer tradition: the influence on institutions on human (economic) behaviour, the role played by knowledge in the creation of wealth, the influence of space and distance for economic decision-making as well as the problem of co-ordination of individual decision-making units are

Do clusters ‘think’? 201 topics which have long attracted discussion. Yet the cluster debate may be used as a device to review these questions from a new perspective. Clusters and their networks may be seen as a phenomenon that combines the three elements of knowledge, spatial proximity and institutional character. In what follows, we will elaborate arguments that favour the specific character of clusters as an institution for knowledge-sharing in a regional context, i.e. as ‘thinking institutions’. We will first outline some general aspects of institutional economics relevant for a knowledge-based interpretation of clusters, leading to arguments regarding clusters as co-ordinating institutions for knowledgesharing, organizational learning and management of new technologies. We then offer a critique of this interpretation, before finally using our arguments to make some comments concerning the organic-evolutionary character of clusters.

In search of guiding institutions The recent renaissance of interest in institutions as a factor shaping economic performance has implications also for the creation and sustained existence of clusters as a tool for knowledge management and as learning organizations within and across regions. Knowledge creation and technology management are not an automatic outcome of individually rational behaviour but need ‘guiding’ institutions. These guiding institutions have to be seen from ‘the perspective that technology and institutions should be understood as co-evolving’ (Nelson, 2001, 19). Development processes do not take place in a vacuum, but rather have profound institutional and cultural roots: ‘The central issue of economic history and of economic development is to account for the evolution of political and economic institutions that create an economic environment that induces increasing productivity’ (North, 1991, p. 98). To what extent can clusters be regarded as part of this co-evolutionary process which is the driving force behind economic growth? Several general ideas of institutional economics (be they ‘old’ or ‘new’) seem to be of relevance and help in understanding the institutional character of clusters in the process of technological development. They also help to answer the ‘why’ question posed by Arrow (1987, p. 734) as the essential perspective of the New Institutional Economics: ‘it does not consist of giving new answers to the traditional question of economics – resource allocation and the degree of utilization. Rather, it consists of answering new questions, why economic institutions emerged the way they did.’ First there is the proposition that ‘institutions do matter’ (Matthews, 1986; Williamson, 2000) and that individual behaviour is moulded by – and in turn constitutive of – social institutions. This was the basic idea developed by Veblen (1899): institutions act upon individuals by changing their habits. This habitformation becomes the more important the more co-operation – instead of competition – is needed. Innovation and productivity gains are based on subtle forms of co-operation, where the creation of new knowledge implies an intense process of interaction which cannot be explained solely in terms of individual

202 Michael Steiner decision making. This is of additional importance if we interpret – as formulated as one of three possible interpretations by Zuckerman (2003, p. 550) – economic networks as ‘economic interactions that are shaped in consequential ways by ascribed or “primordial” relationships where habits are influenced by institutions that pre-exist the market’. A further general implication of institutional economics is that no entity can ultimately be taken as given. It is neither the individual nor the firm which is the sole ‘agent’ in economic and social life: institutions are also potentially useful units. Different levels of social analysis therefore have to be considered. Williamson (2000, pp. 596ff) distinguishes four such levels. The top level – the level of ‘social embeddedness’ – concerns the norms, customs, mores and traditions that are basic conceptual categories for Veblenian institutional economists; the second refers to the ‘institutional environment’ in the sense of formal rules such as constitutions, laws, property rights; at the third level questions of governance are the focus; whereas only the fourth level is the domain of the usual questions of economics concerning the problems of resource allocation and employment. It is the third level that is potentially of most relevance in the context of the cluster model, wherein clusters can be construed as specific governance structures which influence contractual and network relations. Alluding to Commons’s (1932) trilogy of conflict, mutuality and order, Williamson (2000, p. 599) sees governance as ‘an effort to craft order, thereby to mitigate conflict and realize mutual gains’. The prevailing governance structures hence reshape incentives and behaviour. Institutions help to explain stronger forms of co-operation that are needed especially when we deal with learning, and knowledge creation and diffusion. At one level, individuals do not form their preferences in an institution-free ‘state of nature’. Rather, ‘Institutions . . . not [only] constrain options: they establish the very criteria by which people discover preferences’ (Powell and DiMaggio, 1991, pp. 10–11). At another level, this extends to a larger scale where ‘institutions play an essential role in providing a cognitive framework for interpreting sense-data and in providing intellectual habits or routines for transforming information into useful knowledge’ (Hodgson, 1998, p. 171). In short, the generation and diffusion of knowledge are inescapably institutional processes, in that they depend crucially on, and thereby serve to reproduce, complex systems of formal and informal institutions.

Clusters as coordinating institutions for knowledge sharing Thus an institutional perspective might offer tentative answers to our basic question: In what sense do clusters matter for transforming information into useful knowledge necessary for innovation and may hence be regarded as institutions for learning and thinking? Innovation processes in developed economies have essentially been marked by differing forms of innovative milieux and their supporting institutions.

Do clusters ‘think’? 203 Evolutionary economics – as a special interpretation of the institutional perspective – sees these institutions as a moulding device for the technologies used by a society. In this context, drawing on Nelson and Sampat (2001) and Nelson (2001), institutions can be regarded as ‘social technologies’. Whereas ‘physical technologies’ refer to the technical forms of commodities and services, and the ways in which they are produced using particular divisions of labour, ‘social technologies’ are the specific mode of co-ordination once there is a division of labour. Social technologies involving ‘patterned human interaction’ become institutions as soon as they are regarded by the relevant social group as standard and become accepted ways to get things done. In Nelson’s perspective this concept encompasses ways of structuring activity not only within particular organizations but also across organizational borders: institutions are not so much constraints on behaviour but rather an effective support as soon as human co-operation is needed (Nelson, 2001, p. 24). Clusters accordingly can be interpreted as a specific social technology for the co-ordination of the knowledgeintensive use of physical technologies – they are a form of productive pathway co-ordinating human action and combining different factors that are important for growth such as technical advance, physical capital and growth of human capital. Clusters as social technologies are an answer to the problems of achieving agreement and co-ordination between separate decision-making units (such as firms) within a given spatial setting. They combine different additional elements that are important for regional development and economic growth. Clusters as specific social technologies can also be viewed as ‘modes of governance’, as a form of Coasian institution, which serve in effect to integrate the positive external effects of innovation, technological knowledge and development activities. The creation of such institutions may be in response to high transaction costs: because of the specific asymmetric and tacit character of technological knowledge, these transactions have to be mediated by non-market methods. Primarily, these transactions are mediated through networks and other forms of arrangements between organizations and individuals, such as procedures which build trust and work to limit the damaging consequences of asymmetric information. In a sense, clusters can be regarded as economic clubs, which act as institutions for internalizing the problems of effective knowledge transmission. As such, networks provide a substitute both for formal markets and for organizational integration. They therefore fall within the perimeter of non-market devices which firms use to co-ordinate their activities with other firms and with other knowledge-generating institutions. They also guarantee a certain exclusivity and constitute a guild-like privilege of valuable knowledge monopolization. From this evolutionary-institutional viewpoint, cluster activity is based on a different kind of logic, in which we have a paradigmatic change from an optimizing logic to an adaptive form, which stresses the need for political intervention when the system fails. This adaptive logic is based on an Austrian, and in particular a Schumpeterian, evolutionary framework that interprets economic change in the following manner (Metcalfe, 1995, 1998; Steiner, 2002).

204 Michael Steiner Economic change is driven by the variety of economic outcomes between competing and alternative possible ways of fulfilling needs. On the other hand, the variety of economic outcomes depends on the variety of technical and organizational forms. Innovations introduce new varieties; yet imitation (in a certain sense also learning) and competition consume variety, so that economic progress and economic change depend on the balance of these two factors. Variety and diversity are therefore the main forces of economic progress in the context of a competition-oriented market economy. Therefore policy-making has to look not for optimality but for variety and diversity. One principal concern, therefore, is the difference in the behaviour of firms and the resulting variety of experiments. This implies a specific interpretation of firm behaviour which contrasts with traditional theory, in that it is the ‘outlier’ firm, and not the ‘representative’ firm, that is the typical element. There is a substantial diversity between firms – in size, in competence, in knowledge of technological options (Bryant, 2001). Thus the attention of the evolutionary-orientated policy-maker shifts away from notions of efficiency toward notions of creativity, and patterns of adaptation to market stimuli and technological opportunity. The evolutionary policy-maker therefore adapts rather than optimizes, and his or her central concern is with the innovation system and the operation of the set of institutions within which technological capabilities are accumulated. Findings on individual cognition and communication indicate that there is not only the problem of quantitative underproduction of knowledge in markets but also a problem of qualitative underproduction of variety of knowledge (Bünstorf, 2003, p. 92). The canonical policy problem is thus (re)defined in terms of the dynamics of innovation, in a world characterized by immense micro-complexity. Since creativity and the generation of variety are central to this approach, the question of the wider institutional structure which supports the innovative activities of firms is of central concern. This support has primarily to be co-ordinated through non-market-mediated interactions – the unbiased generation and diffusion of knowledge cannot be expected to come about spontaneously (Bünstorf, 2003). A central fact about the modern process of innovation is that it is based on a division of labour, as clearly foreseen by Adam Smith. He early recognized what is now called the social nature of the innovation process. This division-of-labourinduced social process produces efficiency gains from both specialization and professionalization, but also requires a framework to connect the component contributions of the different agents. As far as knowledge and skills are concerned this aspect of connectivity, or technology transfer, cannot be effectively co-ordinated by conventional markets: we are in need of specific institutional arrangements. Yet – as has been outlined recently by Helmstädter (2003) – the idea of connectivity transcends the usual problems of the ‘division of labour’ – there are additional and non-trivial problems of ‘knowledge sharing’ thus far not properly appreciated by the New Institutional Economics. The main line of arguments runs as follows (Brödner, Helmstädter and Widmaier, 1999; Helmstädter, 2003):

Do clusters ‘think’? 205 •

•

•

•

The pure transaction cost approach misses fundamentally the essence of knowledge as an economic resource. ‘The new institutional economics deals with institutions that govern the interactions taking place under the division of labour, but leaves aside the division of knowledge activities that go with it’ (Helmstädter, 2003, p. 14). Once the object of interaction between participating actors is knowledge, the character of interaction changes – the institutional conditions for an efficient division of knowledge are different. Social interaction processes – i.e. networks – have different subject matters of their interactions. Leaving aside the political network, there are networks of economic transactions and the networks of knowledge sharing. The first belongs to the process of division of labour dealing with the exchange of goods and services, the second with knowledge. The main differences reside in the form of interaction and in the impact of interaction. Under the division of labour the transaction of goods and services is paramount, and subject to the rules of competition and to exclusivity of use and consumption. Under knowledge-sharing it is knowledge and skills that are paramount, and these are subject to cooperation and the increase of knowledge for all (inclusivity). Whereas the division of labour involves differentiation and separation of method, mode and product, knowledge sharing involves internalization and recontextualization. The most important ‘institutional’ consequence is that ‘cooperation is the basic institution of the process of the division of knowledge’ (Helmstädter, 2003, p. 32). But the degree of co-operation depends again on the type of knowledge use: application has stronger competitive elements whereas the creation and the transfer are dominated by non-economic competition (status, acceptance) and co-operation. The interest lies here in the institutions that make knowledge sharing efficient.

These strands of evolutionary and institutional thinking in the context of knowledge creation and sharing emphasize that connectivity and the desired efficiency cannot be effectively co-ordinated by conventional markets, but require non-market institutional arrangements for the generation of knowledge. They also emphasize that the growth of knowledge depends on intended and unintended individual processing of experiences, i.e. ‘learning’, while the interpretation, transfer and use of experiences are influenced by interaction between individuals and between organizations, i.e. ‘organizational learning’. Can clusters be interpreted as ‘learning organizations’ generating and diffusing knowledge but also protecting it for cluster members? The concept of learning has changed considerably in recent years. For some time, learning was primarily considered as an adaptive response by an organism to a change in its environment. According to an essentially behaviouristreductionist perspective, learning was a linear process and something that has to start from the level of the individual so that learning in a social context can be understood as the aggregate of individual behaviours.

206 Michael Steiner As Cullen (1998, p. 4) argues, conventional models of organizational learning still retain elements of these positions, taking as a starting point an ‘information processing’ model or ‘black box’ conceptualization of learning, in which information is converted into knowledge and then action. Applied to the concept of organizational learning, it can be understood as a collective and purposive strategy to achieve the goals of the firm; it can furthermore be extended to the notion of clusters as learning organizations with common goals and shared agendas. Yet learning cannot only be regarded as a process leading to changes in capabilities and competencies; it has also to be considered as a social process of ongoing development embedded in a socio-cultural (and regional) context. Learning then becomes essentially a communicative process rather than a cognitive performance, requiring new thinking about the nature and forms of the transmission and dissemination of knowledge within a social and organizational context, such as the firm or, indeed, a cluster (Cullen, 1998, p. 5). This involves more than the concept of ‘locally bounded spill-overs’ (Feldman, 2000). According to some contributions, following Marshall (1890), knowledge (at least locally) ‘is in the air’ and everybody in the cluster benefits (at least in principle) by the existence of such a ‘stock of knowledge and knowhow’, as it is embodied for example in universities and research centres, in other firms, etc. Others argue that knowledge is transferred mainly through faceto-face contacts, through formal and informal conversations. While both mechanisms are certainly important, these representations may fail to capture some fundamental processes and channels through which knowledge is exchanged and created. The concept of ‘organizational learning’ – here extended to clusters – is capable of highlighting some of these processes (Steiner and Hartmann, 1999, 2002). A learning organization is an organization skilled at creating, acquiring and transferring knowledge, and at modifying its behaviour (Garvin, 1993). Organizational learning is the outcome of three overlapping spheres of activity – individual, team and system learning. All three kinds of learning take place simultaneously (Dixon, 1995). Learning systems are therefore the precondition for the transformation from individual learning to organizational and even interorganizational learning (Staehle, 1991). As a medium of communication and exchange that functions independently of the individuals involved, such systems make possible collective learning in and between organizations. Thus, organizational learning takes place when the organization develops systemic processes to acquire, use and communicate organizational knowledge, as learning is conceived as something that should deliberately be pursued by the organization and its members (Argyris and Schon, 1978; Nevis et al., 1995; Pedler et al., 1991; Stankiewicz, 2001). Thus, organizational learning may be recognized by the existence of learning systems that are independent of the individuals (Shrivastava, 1983). In the present context, ‘integrative capabilities’ belong to the most important factors for clusters as prerequisites for regional development. This means that

Do clusters ‘think’? 207 different fragments of knowledge, competencies, etc. have to be not only accessed but also integrated into specific configurations. The prevalence of ongoing learning processes between firms and within clusters indicates the importance of institutional arrangements for the generation of knowledge and learning networks. They serve to fulfil additional functions such as reducing uncertainty about the experimental knowledge of others and increasing the incentives for medium- and long-term investments in diffusion channels (Maskell and Malmberg, 1999). Clusters can help to develop and adapt the research, production, distribution and after-sales strategies of firms to increase the capacity of participants to absorb new information, and can contribute to raising the specificity of knowledge development, processing and diffusion within the cluster to strengthen incentives for the participants to concentrate their investments in the cluster and protect new knowledge against competing clusters.

Critical reflections and open questions The insights gained from the foregoing institutional perspective can be summarized as follows: •

•

•

•

•

New physical technologies are not just there; innovations do not just happen, but need social technologies as pathways co-ordinating human action. Variety and diversity are the main forces for economic progress demanding also a variety of knowledge. Yet besides the quantitative problem of knowledge underproduction, there is also the problem of qualitative underproduction of varieties of knowledge. Knowledge creation and diffusion are in need of specific forms of interaction – knowledge-sharing demands additional mechanisms other than the division of labour leading, to the consequence that co-operation is the basic institution of the process of the division of knowledge. Clusters can be regarded as a special form of governance institution providing a cognitive framework for transforming information into useful knowledge. Clusters then become ‘learning organizations’ and are among the nonmarket devices by which firms seek to co-ordinate their activities with other firms and other knowledge-generating institutions.

So we have a tentative answer to the ‘why’ question concerning the existence of clusters as ‘thinking’ institutions: clusters support processes of knowledge generation, accumulation and diffusion – individuals and firms alone are from an economic point of view not capable of delivering sufficient amounts and varieties of knowledge. Clusters as ‘hybrid institutions’ fulfil this role. It is the specific content, namely knowledge creation and diffusion, and not only the form of networks, that leads to clusters as specific institutions. At the same

208 Michael Steiner time they restrict and protect – more or less efficiently – this knowledge for cluster members generating incentives of belonging to the ‘club’. However, this interpretation itself raises additional questions. Do all clusters ‘think’? Or to put it differently: are all ‘clusters’ institutions for the exchange of knowledge to the same degree? The interpretation of clusters not only as a form of inter-firm interaction but as having a specific content emphasizes a special concept of clusters. If we follow Gordon and McCann (2000), who distinguish three ideal and basic types of clustering – the model of pure agglomeration, the industrial-complex model and the social-network model – not all of them may be regarded in the same degree as ‘thinking’ institutions in the manner outlined above: •

•

•

The model of pure agglomeration – in the tradition of Marshall – based on a local pool of specialized labour, on the increased local provision of nontraded input specific to an industry, and on technological spillovers – may contribute to an ‘evolving localized environment of learning’ (Gordon and McCann, 2000, p. 517). Yet it presumes no form of co-operation, and is without any particular observable organization or inter-agent loyalty leading to conscious efforts of knowledge sharing. The second type aims to minimize spatial transaction costs. These are static and predictable. There are elements of knowledge sharing in the sense that this type of cluster represents a quasi-monopoly for the internalization of the benefits of innovation being created within the (more or less) ‘closed club’. These ties are not formal but contain strong elements of hierarchy and inclusion favouring the exchange of knowledge. The social-network model as the third type – relying on trust and social embeddedness as the dominant links between the cluster firms (and therefore not on deliberate economic decisions based on the minimization of different transaction costs) – also favours the exchange of knowledge, here based on strong interpersonal relationships that transcend firm boundaries and allow for diverse forms of knowledge sharing. These links are then the stronger the more they are based on a variety of elements of this social embeddedness: norms, sets of common assumptions, habits formed by culture, history and of course (but not necessarily) spatial proximity. They form social capital that favours the explicit and implicit sharing of knowledge.

So all three types of clusters contain elements of knowledge sharing but to a different degree and via different mechanisms making them also to a different degree ‘thinking institutions’. Further, this interpretation helps to distinguish between ‘clusters’ as institutions and the (more or less) automatic spillover effects of agglomeration. Are ‘thinking’ clusters a specific institution of the social network model or a hybrid mode of governance of institutional economics? This is a reiteration of Granovetter’s (1994) ‘second Coase’ question: why do firms have costly co-operation with

Do clusters ‘think’? 209 others and get embedded in social networks, and the answer as typically given in the domain of new institutional transaction-cost economics (Williamson, 1996, 2000). Underlying this question are two contrasting models of institutions (Schmid and Maurer, 2003). The sociological approach assumes that coordination mechanisms solely based on decentralized decision-making do not suffice to establish the order needed for continued interaction but that a social process is needed which guides this individual behaviour: human action does not need so much the assumption of rational behaviour but has to be regarded as guided by rules. Individual behaviour is therefore in need of institutions that lead to functioning social relationships. Exchange is only possible once the agents have agreed on institutions that allow such an exchange. The exchange of knowledge, then, is only possible once there is sufficient embeddedness and social capital to enable the firms to share their knowledge. This differs from the economic approach where agents or firms make decisions that form institutions. Interpreted in terms of clusters, firms decide to form spatially proximate and interlinked networks because they regard it as an efficient way to participate in and to contribute to the generation and diffusion of knowledge. Transaction-cost-oriented economics – as outlined by Williamson (2000, 2002) – goes one step further. Social embeddedness is regarded as a higher level of institution and – once well established – influences the choice of contract between market and hierarchy. Once social embeddedness is given, together with adequate property rights, firms can decide about the forms of governance as a third level of institution. Taking these positions for granted, additional insights can be gained by ‘embedding’ the organizational environment into the analysis of transaction problems (Ipsen, 2002, 2003, pp. 206ff) and emphasizing the phenomenon of knowledge sharing (in the above-mentioned sense of Helmstädter, 2003). •

•

•

The organizational environment and the content of the transaction allows a deepened analysis of the choice of governance – the more unstable the environment, the less precise the character of the object to be ‘shared’, the less either market or hierarchy and the more hybrid forms of governance will be chosen. In the context of different degrees of moral hazard risk and availability of social capital, quite different types of hybrids may develop. This opens a research agenda for a deeper analysis of diverse cluster forms. The transaction relation is more than a decision oriented towards costefficient production or exchange. The transactional approach is too narrow to account for long-term aspects such as the adaptation to changing market conditions and also for the development of a knowledge base within and between firms. Special organizational forms, i.e. institutions, are of decisive importance for the solution of such complex decision problems. The isolated perspective of decisions concerning alternative governance forms therefore is not sufficient because it disregards the specific environmental situation and the content of the transactions; yet this applies also to

210 Michael Steiner the opposite perspective – behaviour is not completely determined by social relations, there are still decisions to be made. Are clusters a specific form of changing social technology? If we accept clusters as a useful concept for an analysis of local and regional economic performance, in their role of supporting the creation and diffusion of knowledge, it is of equal importance to emphasize that institutions are not automatically ‘just there’ but are the result of a evolving process shaped both by policy activities and entrepreneurial behaviour responding to new challenges. This implies a changing character of institutions in support of knowledge management. If, drawing again on Nelson and Sampat (2001) and Nelson (2001), we regard institutions as evolving ‘social technologies’, they can be interpreted as a form of productive pathway co-ordinating human action and combining different factors that are important for growth such as technical advance, physical capital, growth of human capital. These social technologies co-evolve with new physical technologies and are therefore in a constant need to change themselves. The rise of mass production – as outlined by Alfred Chandler – implied changes that led to developed industrial economies forming new forms of ‘social technology’. New modes of organizing business were required to take advantage of the new opportunities for ‘scale and scope’, such as professional management, new financial institutions and associated markets, and business schools. Clusters can accordingly also be regarded as an institutional response to a new logic of production where flexible specialization has replaced mass production (Piore and Sabel, 1984). Information-based networks led to the transition from the age of machines to the era of information and knowledge (Antonelli, 1992; Powell and Smith-Doerr, 1994). The innovation process since the 1980s and 1990s in Europe has essentially been marked by differing forms of innovative milieux and their supporting institutions. Here innovation and productivity gains are based on subtle forms of co-operation where the creation of new knowledge implies an intense process of interaction. Clusters then become institutions that support adaptation also by making possible ‘ordered thought, expectation and action, by imposing form and consistency on human activities . . . the hidden and most pervasive feature of institutions is their capacity to mould and change aspirations’ (Hodgson, 2004, p. 656). Do clusters change? If clusters are a certain institutional response to a historically given logic of production then clusters themselves have to undergo change. As long as economic growth is to be understood as an evolutionary process, the nature and dynamics of the organization of production, the role and change of institutions and technology and technological advance have to be specified: ‘The central issue of economic history and of economic development is to account for the evolution of political and economic institutions that create an economic environment that induces increasing productivity’ (North, 1991, p. 98). The question concerning the evolution of clusters, and thus their future, consists itself of several questions. These include the following.

Do clusters ‘think’? 211 In a globalized world of freely moving capital and increasingly freely moving people, it is only social capital that remains tied to specific locations. Thus, the ‘learning economy’ is characterized by the hyper-mobility of information and knowledge and the local character of social capital. What does this mean for the institutional setting of learning networks in an internationalized framework? What is the relative importance of local versus international knowledge exchange (Simmie, 2003)? The relationships between the firms become more complex and risky and require to be redesigned in a long-term perspective. This has compelled firms to devise new organizational forms and contractual arrangements, which may be capable to manage these new and more complex relationships. The question thus concerns the form and content of an internationalization of learning networks and their consequences for regional development (Cappellin and Steiner, 2004). This is closely related to the concept of proximity and its differentiation (Rallet and Torre, 1998; Gallaud and Torre, 2002; Cappellin and Steiner, 2004) – it can be either ‘geographical’ or ‘organizational’. Geographical distance is related to the availability of face-to-face contacts and direct communication, whereas institutional proximity refers to common technological paradigms, or countries which have traditions, norms and institutions in common. Geographical proximity certainly enhances the organizational and institutional proximity between the various local actors. When spatial distances are important, access to knowledge and learning networks depends on the existence of specific skills, of social relationships and of organizations and ‘soft’ infrastructures, which may allow access to tacit knowledge and involvement in processing of new experiences. However, physical distance may represent a sufficient but not necessary condition for the creation of knowledge and innovation networks between firms and organization. As has been shown, new technologies such as the Internet have hardly had the effect on existing innovation networks of a ‘real’ extension to new spatial levels and types of partners (Kaufmann et al., 2003). In fact, the accumulation of tacit knowledge, the building of new skills and knowledge spillovers are enhanced by geographical proximity, but they especially require a common culture, organizational framework, social capital and institutions. Thus, knowledge transfers are not territorially bounded when culture, organizational framework, social capital and institutions are common or harmonized. The two concepts of distance imply a different structure of networks, in particular production, technological and financial networks. After the ‘why’ question certainly comes the ‘how’ question: how do clusters ‘think’. Tentative empirical approaches of case studies are mostly based on theories of (organizational) learning, combining individual, team and system learning. These approaches accordingly differentiate between forms of learning (formal and informal) and categories of learning in order to correct errors (single-loop learning) or to question and modify existing norms and procedures (double-loop learning) (Argyris and Schon, 1978; Nevis et al., 1995). The results confirm that clusters undertake conscious attempts to retain and improve adaptive capabilities by learning. Rabelotti (1995) and Konstadakopulos (1998)

212 Michael Steiner reveal that both informal learning (for example, contacts at trade fairs, membership of ‘old boys’ networks’, and the like) and formal learning (common projects, research groups, regular contacts to regional universities, benchmarking clubs) are important. Raymond et al. (2002) and Wibe and Narula (2002) show that double-loop learning takes place both at the firm as well as at the interfirm level. Further analysis identifies different factors influencing the kind of learning: existing value chains, the competitive structure, working cultures within clusters thus revealing that there is an interaction between the material linkages and the immaterial knowledge flows of clusters (Steiner and Hartmann, 2002). Further research may therefore be guided by the desire to get deeper insights into forms of learning depending on different types of clusters. Can clusters be replicated in the transforming economies of Europe? The enlarged European economy is enriched by a wide diversity of social models and cultural and historical backgrounds. Thus, the same policy framework may have different effects in different regions. Here institutional aspects gain special importance insofar as new institutionalism points to the fact that regulation is basically an answer to uncertainty, in that it is the role of an economic order to reduce this uncertainty and to create a certain standardization and predictability of economic behaviour. Institutionalism (and ‘new’ institutionalism in particular) points out that under conditions of fundamental uncertainty, rather than simply of calculable risk, the activity of the market participants is insufficient to stabilize expectations – the spontaneous order of the market is not sufficient to create endogenously the necessary institutions. As a consequence clusters as an institutionalized form of trust depend on a larger framework of institutionalized credibility of the economic system. As a precondition for an adaptive logic of regional development, clusters themselves require a certain institutional environment to come into existence. Regional policy for new market economies and transition countries consequently means starting a learning process for the establishment of local clusters and networks.

Conclusions Mary Douglas suggested that social institutions may be regarded as codified information. Yet, according to her view, it is highly improbable that institutions arise continuously as a clustering of congruent ideas and a mixture of force and convention (Douglas, 1986, 179). Also North (1991) has pointed to the possibility of ‘institutional obstruction’ and to the potential failure of economies because of the lack of new institutions capable of adopting available productive technologies. According to Durkheim the elementary social bond emerges only if a model of social order becomes rooted in the thinking of individuals. And Williamson (2000, 595) conceded not only that all forms of institutions are flawed but that we are still very ignorant about institutions and should therefore be accepting of pluralism. Despite their basic institutional character there is a strong diversity of clusters both in form and content. There is evident progress in the conceptualization of

Do clusters ‘think’? 213 the content and form of knowledge-exchange and learning within clusters, as captured by the vast literature extending from the ‘learning region’ to regional ‘interactive/collective learning’ and on different analytical models of regional innovation adding to the description of the variety of institutional arrangements (such as the triple helix, inclusion of intangible assets). Yet there is a need for more empirically oriented approaches to reveal and assess the diversity of knowledge sharing. Governance structures are never deterministic – cluster analysis has to avoid being ‘oversocialized’. Within clusters there is ample room for human agency. One of the basic elements of an evolutionary approach is the creative function of the market as also expressed by innovative behaviour supported by clusters. Yet clusters do have a tendency for exclusivity – part of the goals of networks is to create some kind of knowledge monopolizing market of proximate firms and related support institutions. Institutions are themselves shaped by economic behaviour and hence subject to change. Since there is definitely room for agency there is ongoing interaction between the agents and the clusters, and interaction that itself is a force for the adaptation of clusters to changing conditions of competition, technology and so on. A possible and perhaps necessary change for the future of clusters is their spatial and functional extension. This calls for an appropriate institutional framework at a larger, trans-regional level. Supranational institutions may become an important actor in setting policies which do not merely support particular innovative activities but create a framework by which knowledge processes are harnessed. How endogenous this change will turn out to be is again an open question. In contrast to ‘old’ institutionalism, as proposed by commentators such as Hayek, Menger and von Mises, which emphasized the concept of spontaneous order, ‘new’ institutionalism points to the fact that this spontaneity is not enough – the stabilization of expectations does not result from simple market interactions, but is a feature of institutions themselves. But institutions such as clusters show emergent features in the form of further institutionalization of relationships, interactions and networks of knowledge transfer. So there is in-built endogeneity in the development of clusters: their institutional forms are exogenous in the short run (so setting the framework for economic relationships and development), but become themselves endogenous over the longer run, changing in response to the development of the cluster itself.

References Anderson, J.R. (1995), Learning and Memory: An Integrated Approach, New York: John Wiley. Antonelli, C. (1992), The economic theory of information networks, in: Antonelli, C. (ed.), The Economic Theory of Information Networks, Amsterdam, Elsevier, pp. 5–27. Argyris, C. and Schon, D. (1978), Organizational Learning, London, Addison Wesley. Arrow, K. (1987), Reflections on the essays, in: Feiwel, G. (ed.), Arrow and the Foundations of the Theory of Economic Policy, New York, NYU Press, pp. 727–34.

214 Michael Steiner Axelrod, R. (1997), The Complexity of Cooperation, Princeton, Princeton University Press. Brödner, P., Helmstädter, E. and Widmaier, B. (eds) (1999), Wissensteilung: Zur Dynamik von Innovation und kollektivem Lernen, Munich, Rainer Hampp. Bryant, K. (2001), Promoting innovation: An overview of the application of evolutionary economics and systems approaches to policy issues, in: Foster, J. and Metcalfe, J. S. (eds), Frontiers of Evolutionary Economics, Competition, Self Organization and Innovation Policy, Cheltenham and Northampton, Mass., Edward Elgar, pp. 361–83. Bünstorf, G. (2003), Processes of knowledge sharing: From cognitive psychology to economics, in: Helmstädter, E. (ed.), The Economics of Knowledge Sharing, Cheltenham, Edward Elgar, pp. 74–99. Capellin, R. and Steiner, M. (2004), Enlarging the scale of knowledge and innovation networks: Theoretical perspectives, methodological approaches and policy issues, in: Wink, R. (ed.), Academia-Business Links: European Policy Strategies and Lessons Learnt, Houndmills, Palgrave, pp. 263–86. Cohen, W.M. and Levinthal, D.A. (1989), Innovation and learning: The two faces of R&D, The Economic Journal, 99, 569–96. Commons, J.R. (1932), The problem of correlating law, economics and ethics, Wisconsin Law Review, 8, 3–26. Cooke, P. (1998), Introduction: Origins of the concept, in Braczyk, H. J., Cooke, P. and Heidenreich, M. (eds), Regional Innovation Systems: The Role of Governances in a Globalized World, London, UCL Press. Cooke, P. and Morgan, K. (1998), The Associational Economy: Firms, Regions and Innovation, Oxford, Oxford Economic Press. Cullen, J. (1998), Developing learning organizations in SME clusters: Modelling and recommendations, mimeo, Tavistock Institute, London. Dixon, N. (1995), The Organizational Learning Cycle: How We Can Learn Collectively, Ottawa, McGraw-Hill. Dosi, G. and Malerba, F. (eds) (1996), Organization and Strategy in the Evolution of the Enterprise, London, Macmillan. Douglas, M. (1986/1991), Wie Institutionen denken, Frankfurt, Suhrkamp. Feldman, M.P. (2000), Location and innovation: The new economic geography of innovation, spill-overs and agglomeration, in: Clark, G. L., Feldman, M. and Gertner, M.S. (eds), The Oxford Handbook of Economic Geography, Oxford: Oxford University Press, pp. 371–94. Foss, N.J. (1999), The use of knowledge in firms, JITE – Journal of Institutional and Theoretical Economics, 155, 458–86. Freeman, C. (1991), Networks of innovators: A synthesis of research issues, Research Policy, 20 (5), 499–514. Gallaud, D. and Torre, A. (2002), Are innovation networks localised? Proximity in the diffusion of knowledge (The case of SME’s in ‘agbiotech’), paper presented at the Workshop on Rethinking Regional Innovation and Change: Path Dependency or Regional Breakthrough? at the Center of Technology Assessment, Stuttgart. Garvin, D.A. (1993), Building a learning organization, Harvard Business Review, 71, 78–91. Gordon, I.R. and McCann, P. (2000), Industrial clusters: Complexes, agglomeration and/or social networks?, Urban Studies, 37, 513–32. Granovetter, M. (1994), Business groups, in: Smelser, N.J. and Swedberg, R. (eds), The Handbook of Economic Sociology, Princeton, Princeton University Press, pp. 453–75.

Do clusters ‘think’? 215 Grant, R.M. (1996), Towards a knowledge-based theory of the firm, Strategic Management Journal, 17, 109–22. Hartmann, C. (2004), Die Lernfähigkeit von Clustern: Eine theoretische und empirische Betrachtung, Ph.D. thesis, University of Graz, Department of Economics. Helmstädter, E. (ed.) (2003), The Economics of Knowledge Sharing: A New Institutional Approach, Cheltenham and Northampton, Mass., Edward Elgar. Hodgson, G.M. (1998), The approach of institutional economics, Journal of Economic Literature, XXXVI, 166–92. Hodgson, G.M. (2002), Reconstitutive downward causation: Social structure and the development of individual agency, in: Fullbrook, E. (ed.), Intersubjectivity in Economics: Agents and Structures, London, Routledge, pp. 159–80. Hodgson, G.M. (2004), Reclaiming habit for institutional economics, Journal of Economic Psychology, 25, 651–60. Ipsen, D. (2002), Zum Erklärungspotential von Williamsons Theorie der Wahl der Governance-Formen, in: Ipsen, D. Peukert, H. (eds), Institutionenökonomie: Theoretische Konzeptionen und empirische Studien, Frankfurt, Haag+Herchen, pp. 67–78. Ipsen, D. (2003), Transaktionskostenökonomie und soziologische Institutionen-theorie, in: Schmid, M. and Maurer, A. (eds), Ökonomischer und soziologischer Institutionalismus, Marburg, Metropolis Verlag, pp. 191–209. Kaufmann, A., Lehner, P. and Tödtling, F. (2003), Effects of the Internet on spatial structure of innovation networks, Information Economics and Policy, 15, 402–24. Konstadakopulos, D. (1998), Collective Learning in the West of England, Bristol, University of the West of England. Lawson, C., and Lorenz, E. (1999), ‘Collective learning, tacit knowledge and regional innovative capacity’, Regional Studies, 33, 305–17. Lundvall, B.-Å. (2002), The learning economy: Challenges to economic theory and policy, in: Hodgson, G.M. (ed.), A Modern Reader in Institutional and Evolutionary Economics, Key Concepts, Cheltenham and Northampton, Edward Elgar, pp. 26–47. Marshall, A. (1890), Principles of Economics, London, Macmillan. Martin, R.L. and Sunley, P. (2003), Deconstructing clusters: Chaotic concept or policy panacea?, Journal of Economic Geography, 3 (1), pp. 5–35. Maskell, P. and Malmberg, A. (1999), Localized learning and industrial competitiveness, Cambridge Journal of Economics, 23, pp. 167–85. Matthews, R.C.O. (1986), The economics of institutions and the sources of economic growth, Economic Journal 96 (4), 903–18. Metcalfe, S. (1995), The economic foundations of technology policy: Equilibrium and evolutionary perspectives, in: Stoneman, P. (ed.), Handbook of the Economics of Innovation and Technological Change, Oxford, Blackwell. Metcalfe, S. (1998), Evolutionary Economics and Creative Destruction, Graz Schumpeter Lectures, London and New York, Routledge. Nelson, R. (2001), The coevolution of technology and institutions as the driver of economic growth, in: Foster, J. and Metcalfe, J.S. (eds), Frontiers of Evolutionary Economics, Competition, Self-Organization and Innovation Policy, Cheltenham and Northampton, Mass., Edward Elgar, pp. 19–30. Nelson, R. and Sampat, B. (2001), Making sense of institutions as a factor shaping economic performance, Journal of Economic Behavior & Organization, 44 (1), 31–54. Nevis E., DiBella, A. and Gould, J. (1995), Understanding organizations as learning systems, Sloan Management Review, 37, 123–39.

216 Michael Steiner Nonaka, I., Toyama, R. and Nagata, A. (2000), A firm as a knowledge-creating entity: A new perspective on the theory of the firm, Industrial and Corporate Change, 9, 1–20. North, D.C (1991), Institutions, Journal of Economic Perspectives 5 (1), 97–112. Pedler, M., Burgoyne, J. Boydell, T. (1991), The Learning Company, London, McGrawHill. Piore, M. and Sabel, C. (1984), The Second Industrial Divide: Possibilities for Prosperity, New York, Basic Books. Porter, M.E. (1990), The Competitive Advantage of Nations, London, Macmillan. Porter, M.E. (1998), On Competition, Boston, Harvard Business School. Porter, M.E. (2000), Location, competition and economic development: Local clusters in a world economy, Economic Development Quarterly, 14, 15–34. Powell, W.W. and DiMaggio, P.J. (eds) (1991), The New Institutionalism in Organizational Analysis, Chicago, University of Chicago Press. Powell, W.W. and Smith-Doerr, L. (1994), Networks and economic life, in: Smelser, N.J. and Swedberg, R. (eds), The Handbook of Economic Sociology, Princeton, Princeton University Press, pp. 386–402. Rabelotti, R. (1995), External Economies and Cooperation in Industrial Districts, Ph.D. thesis, IDS, University of Sussex. Rallet, A. and Torre, A. (1998), On geography and technology: Proximity relations in localised innovation networks, in: Steiner, M. (ed.), Clusters and Regional Specialisation: On Geography, Technology and Networks, London, Pion Limited, pp. 41–56. Raymond, L., Blili, S. and Thibault, L. (2002), A model and evaluation of organizational learning in networked SMEs, Paper Presented at the 12 International Electronic Commerce Conference, Bled. Schmid, M. and Maurer, A. (2003), Institutionen und Handeln: Probleme und Perspektiven der Institutionentheorie in Soziologie und Ökonomie, in: Schmid, M. and Maurer, A. (eds), Ökonomischer und soziologischer Institutionalismus, Marburg, Metropolis Verlag, pp. 9–46. Shrivastava, P. (1983), A typology of organizational learning systems, Journal of Management Studies, 20, 7–28. Simmie, J. (2003), Innovation and urban regions as national and international nodes for the transfer and sharing of knowledge, Regional Studies, 37 (6/7), 607–20. Staehle, W. (1991) Management, Eine verhaltenswissenschaftliche Einführung, Munich, Vahlen. Stankiewicz, R. (2001) The cognitive dynamics of technology and its technological system, in Carlsson, B. (ed.), New Technological Systems in the Bio-industry, Dordrecht, Kluwer. Steiner, M. (2002), Clusters and networks: Institutional settings and strategic perspectives, in: McCann, P. (ed.), Industrial Location Economics, Cheltenham, Edward Elgar, pp. 207–21. Steiner, M. and Hartmann, C. (1999), Interfirm cooperation and learning within SMEnetworks, in: Fischer, M., Suarez-Villa, L. and Steiner, M. (eds), Innovation, Networks and Localities, Berlin, Springer, pp. 85–106. Steiner, M. and Hartmann, C. (2002) Material and immaterial dimensions of clusters: Cooperation and learning as infrastructure for innovation. Paper presented at the 42th Congress of the European Regional Science Association, Dortmund. Storper, M. (1995), The resurgence of regional economies ten years later: The region as a nexus of untraded interdependencies, European Urban & Regional Studies, 2 (3), 191–221.

Do clusters ‘think’? 217 Storper, M. (1997), The Regional World: Territorial Development in a Global Economy. London: Guildford Press. Toboso, F. (2001), Institutional individualism and institutional change: The search for a middle way mode of explanation, Cambridge Journal of Economics, 25, 765–83. Veblen, T.B. (1899), The Theory of the Leisure Class: An Economic Study in the Evolution of Institutions, New York, Macmillan. Wibe, M. and Narula, R. (2002), Interactive learning and non-globalisation: Knowledge creation by Norwegian software firms. Paper presented at the DRUID Summer Conference on ‘Industrial Dynamics of the New and the Old Economy – Who is Embracing Whom?’, Kopenhagen. Williamson, O.E. (1996), The Mechanisms of Governance, New York, Oxford University Press. Williamson, O.E (2000), The new institutional economics: Taking stock, looking ahead, Journal of Economic Literature, XXXVIII (3), 595–613. Williamson, O.E. (2002), Empirical microeconomics: Another perspective, in: Augier, M. and March, J. (eds), The Economics of Choice, Change, and Organization, Cheltenham, Edward Elgar. Zuckerman, E.W. (2003), On Networks and Markets by Rauch and Casella, eds, Journal of Economic Literature, XLI, 545–65.

11 Spaces of knowledge flows Clusters in a global context Meric S. Gertler and David A. Wolfe

Is all knowledge local? At the very foundation of the cluster concept is the idea that proximity matters. The literature suggests that this is manifest in a number of important ways. First, the geographical clustering of economic actors facilitates the exchange of knowledge between them, through both traded and untraded means. The interaction that supports this is formal and planned as well as informal or unplanned, with spatial concentration facilitating both forms of contact. Common conventions and norms, and readily available knowledge about the reliability and trustworthiness of individual economic actors, further support the local flow of knowledge – both tacit and codified – within local industry clusters (Storper and Leamer 2001; Storper and Venables 2004). The same conditions enrich close, collaborative vertical interaction with local customers and suppliers, in which learning-through-interacting generates mutual benefits for technology users and producers alike (Lundvall 1988; Gertler 1995). Finally, the geographical clustering of firms in the same industries accentuates competition – and the innovative dynamism arising from it – by enhancing firms’ ability to learn from one another through observation and monitoring (Porter 2000; Malmberg and Maskell 2002). While the above picture represents a coherent and influential consensus view emerging from the cluster literature, it is not free of controversy (Martin and Sunley 2003). Recent conceptual contributions to the literature have begun to propose the unthinkable: to question the overwhelming emphasis on local interaction and knowledge circulation contained within the cluster as the only – or even the primary – source of innovative dynamism for firms in clusters. Instead, these recent commentaries argue that non-local (inter-regional and international) relationships and knowledge flows are crucial sources of vitality, complementing the local ‘buzz’ that has come to be regarded as the hallmark characteristic of the cluster (Bathelt et al. 2004; Owen-Smith and Powell 2004). Appealing as these arguments may be, however, they still rest on a small base of empirical evidence. This chapter aims to address this gap in our knowledge and understanding by synthesizing the results of an ongoing comprehensive national study of cluster

Spaces of knowledge flows 219 development in Canada. The study covers a range of economic sectors – both mature and new – as well as different types of geographical locations: large metropolitan regions as well as mid-size urban regions and rural settings. We provide an analytical overview of the findings from this study, and use it to shed light on the relative importance of local and non-local actors, relationships, and forces in the development of more innovative forms of production. We also build a conceptual typology, based on types of knowledge, institutional frameworks, and the strength of geographical concentration of economic actors. In this way, we aim to make a larger contribution to ongoing debates concerning the importance of the local in the development of economic clusters.

The economic advantages of clusters: agglomeration and knowledge spillovers According to the large literature on the subject, the potential advantages that firms derive from locating within clusters arise from two principal sources: agglomeration benefits and knowledge spillovers. Agglomeration economies arise primarily from the ready access to a collective set of resources available to firms co-locating in the same region or locale. Porter’s (2000) work is consistent with this approach, although he embellishes the benefits attributed to traditional agglomeration economies by setting out the competitive advantages derived from the effects of his ‘diamond’. Porter stresses that the location of a firm within the cluster contributes to enhanced productivity, higher wages, and greater innovativeness by providing easier and/or cheaper access to specialized inputs, including components, machinery, business services, and personnel, whose availability obviates the need for vertical integration or non-local sourcing of necessary inputs. Local sourcing from within the cluster also facilitates communication with key suppliers in the sense that repeated interactions with local supply firms in the value chain create the potential for conducting repeated transactions on the basis of tacit, as well as more codified, forms of knowledge. Clusters offer distinct advantages to firms in terms of the availability of specialized and experienced personnel. The cluster itself can act as a magnet drawing skilled labour to it. Conversely the location of specialized training and educational institutions in the region provides a steady supply of highly qualified labour to the firms in the cluster. While not diminishing the importance of these agglomeration economies, another stream of literature suggests that a key source of competitive advantage for firms located in clusters is their shared access to a distinctive local knowledge base. The central argument is that the joint production and transmission of new knowledge occurs most effectively among economic actors located close to each other. Proximity to critical sources of knowledge, whether they are found in public or private research institutions or embedded in the core competencies of lead or anchor firms, facilitates the process of acquiring new technical knowledge, especially when the relevant knowledge is located at the research frontier or involves a largely tacit dimension. Knowledge of this nature is

220 Meric S. Gertler and David A. Wolfe transmitted most effectively through interpersonal contacts and the inter-firm mobility of skilled workers. However, Breschi and Malerba (2001) argue that this approach overestimates the benefits of physical proximity alone. They argue that sheer proximity is not sufficient to account for local knowledge spillovers. In their view, the body of research on local knowledge spillovers overlooks the broader set of factors and conditions that support the effective transfer of knowledge in clusters: ‘a key feature of successful high-technology clusters is related to the high level of embeddedness of local firms in a very thick network of knowledge sharing, which is supported by close social interactions and by institutions building trust and encouraging informal relations among actors’ (Breschi and Malerba 2001, 819). In other words, the degree to which firms can tap into a common knowledge base at the local level depends on more than just spatial proximity, cultural affinity, or corporate culture. In this sense there is a strong interdependence between the economic structure and social institutions that comprise the cluster. The institutional context of the cluster defines how things are done within it and how learning transpires. As Gertler has argued, it is a function of institutional proximity – the common norms, conventions, values, and routines that arise from commonly experienced frameworks of institutions existing within a regional setting (2003; 2004). It is also critical to differentiate between different kinds of knowledge spillovers. Much of the literature on knowledge spillovers, and in particular the role of tacit knowledge, presumes that the knowledge being shared is highly technical in nature and results largely from the transfer of research results between regionally embedded research institutes and private firms. However, technical research results are only one element of the kinds of knowledge flows that contribute to the competitive dynamics of a successful cluster. One of the most important forms of knowledge flow is the knowledge embodied in highly qualified personnel which flows directly from research institutes to private firms in the form of graduates and also moves between firms in the form of mobile labour (Wolfe 2005). There is a strong suggestion in the literature that the recombination of talent in new configurations through labour mobility and the spinning off of new start-up firms is one of the most important sources of innovative dynamism in clusters (Saxenian 1994; Brown and Duguid 2000). Another form of knowledge flow involves entrepreneurial skills. Feldman et al. place this form of knowledge flow at the centre of their model of cluster formation. Entrepreneurs act as the key agents who build upon the existing base of institutional assets that provide the local antecedents for cluster formation. In their view, an outbreak of entrepreneurial activity is necessary to transform these assets into a cluster (2005). From our perspective, the ongoing transmission of entrepreneurial skills within the cluster is critical for its continued vitality and growth. This form of knowledge can be transmitted through a variety of mechanisms – including the spinning off of new firms from large anchor firms within the cluster, the mobility of key personnel within the cluster, and the transfer of entrepreneurial and managerial skills through angel and venture

Spaces of knowledge flows 221 investing. Closely related to this is knowledge about external market conditions. For small and medium-sized enterprises, an essential piece of knowledge they must acquire to grow and expand concerns the competitive conditions in external markets and which ones constitute the most suitable targets for expansion. Entrepreneurial skills and market information can be transmitted throughout the cluster by a variety of mechanisms – some formal and some informal. One of the most important is the peer-to-peer mentoring and knowledge-sharing that is organized through local industrial and civic associations. The dynamic role played by such associations in facilitating this form of knowledge flow underlines the importance of the local and regional institutional structures. The final dimension of knowledge-sharing crucial for the success of the cluster is achieved through the infrastructural knowledge resources found in the specialized local legal, management, and financial firms that are essential to the success of individual firms in the cluster. These kinds of services often provide vital support to the individual firms in the cluster. In an attempt to elaborate further the role that knowledge plays in sustaining clusters, Maskell (2001) has proposed a knowledge-based theory of the cluster. He suggests that the primary reason for the emergence of clusters is the enhanced knowledge creation that occurs along two complementary dimensions: horizontal and vertical. Along the horizontal dimension, clusters reduce the cost of co-ordinating dispersed sources of knowledge and overcoming the problems of asymmetrical access to information for different firms producing similar goods and competing with one another. The advantages of proximity arise from continuous observation, comparison, and monitoring of what local rival firms are doing, which drives innovation as firms race to keep up with or get ahead of their rivals. The vertical dimension of the cluster consists of those firms that are complementary and interlinked through a network of supplier, service and customer relations. Once a specialized cluster develops, local firms increase their demand for specialized services and supplies. Furthermore, once the cluster has emerged, it acts as a magnet drawing in additional firms whose activities require access to the existing knowledge base or complement it in some significant respect (Maskell 2001, 937). In critical respects, this knowledge-based conception of the cluster takes for granted key aspects of the Porter diamond, by assuming that firms co-located in the cluster tend to be rivals in the same product markets or part of a locally based supply chain, and that close monitoring of competitors or tight buyer–supplier interaction are key elements that tie the firm to the cluster. While these conditions may hold for the most developed clusters in their respective industrial or product segments, there is growing evidence (see following sections of this chapter) to suggest that they do not apply universally to all clusters – especially those in more specialized niches, at an earlier stage of development, or in smaller, more open national economies. If Porter’s conditions do not hold, then this opens up a new line of inquiry about the relationship between the global and the local, and complicates considerably the question we posed at the outset: must local concentrations of

222 Meric S. Gertler and David A. Wolfe firms in the same and related sectors rely exclusively on local sources of knowledge? A knowledge-based theory of the cluster must recognize that relatively few clusters are completely self-sufficient in terms of the knowledge base from which they draw. The development of ever more complex technologies, such as modern passenger aircraft, requires the support of sophisticated organizational networks that provide key elements or components of the overall technology (Kash and Rycroft 2000). While some elements of these complex technologies may be co-located in an individual cluster, increasingly the components of these networks are situated across a wide array of locations (Niosi and Zhegu 2005). This suggests that the knowledge flows that feed innovation in a cluster are often both local and global. Bathelt et al. (2004) maintain that successful clusters are those that are effective at building and managing a variety of channels for accessing relevant knowledge from around the globe. However, the skills required to absorb knowledge from the local environment are substantially different from the ones needed to identify, acquire, and make the best use of knowledge produced elsewhere, and firms in the cluster must be able to manage these different tasks. Bathelt et al. maintain that a robust knowledge-based model of the cluster must account for both dimensions of these knowledge flows. Bathelt et al. refer to these two kinds of knowledge flows as local buzz and global pipelines respectively. Following Storper and Venables (2004), ‘buzz’ arises from the fact of physical co-presence. It incorporates both the broad general conditions that exist when it is possible to glean knowledge from intentional face-to-face contacts, as well as the more diffuse forms of knowledge acquisition that arise from chance or accidental meetings and the mere fact of being in the same location. Buzz is the force that facilitates the circulation of information in a local economy or community and it is also the mechanism that supports the functioning of networks in the community. In this context, it is almost impossible to avoid acquiring information about other firms in the cluster and their activities through the myriad number of contact points that exist. Pipelines, on the other hand, refer to channels of communication used in distant interaction, between firms in clusters and knowledge-producing centres located at a distance. Important knowledge flows are generated through network pipelines. The effectiveness of these pipelines depends on the strength of pre-established social relationships and the quality of trust that exists between the firms in the different nodes involved. The advantages of global pipelines derive from the integration of firms located in multiple selection environments, each of which is open to different technical potentialities. Access to these global pipelines can stimulate new local innovation through the use of knowledge that has contributed to the emergence of successful firms and clusters elsewhere. Firms need access to both local buzz and the knowledge acquired through international pipelines. The ability of firms to access such global pipelines and to identify both the location of external knowledge and its potential value depends very much on the internal organization of the firm, in other words, its ‘absorptive capacity’. The same can be said of local and regional clusters (Bathelt et al. 2004).

Spaces of knowledge flows 223 However, the precise mix of the global and local knowledge flows present in individual clusters must of necessity be indeterminate. There is increasing evidence to suggest that, even in the most advanced clusters, a growing proportion of the knowledge base is not exclusively local. Recent work on Silicon Valley indicates that local production processes are part of a complex production chain that is connected into global production networks. The most dynamic of multinational corporations and a larger proportion of emerging small and medium-sized enterprises have strong linkages to a variety of specialized clusters around the globe. Both types of firms use their presence in these local clusters to access specialized bodies of knowledge created by the local research institutions, or to tap into a specialized skill set or knowledge base developed by cluster-based firms. However, rarely are the local knowledge bases of these clusters, or the production activities of the firms embedded in them, completely self-contained. Rather, according to Sturgeon, ‘what gets worked out in the clusters is exactly the codification schemes that are required to create and manage spatially dispersed but tightly integrated production systems’ (Sturgeon 2003, 200). A greater proportion of the production of complex technologies in sectors ranging from information technology to the aircraft and automotive assembly occurs in these ‘modular production networks’ with activities dispersed across a wide range of global locations. What takes place in the clusters of the more industrialized economies are the core interactions between lead firms and key suppliers that resist easy codification, such as design, development of prototypes, and determining the validity of manufacturing processes. The production of high value-added or low-volume products also takes place in these locations. Sturgeon implies that there is a geographic hierarchy of clusters within specific industrial sectors, with Silicon Valley acting as the key location for standardsetting activity in information technology (Sturgeon 2003, 220). A marked pattern of stronger global (versus local) relations emerges even more clearly in a recent study of opto-electronics clusters in six locations (Hendry et al. 2000). This study found that extra-regional commercial linkages are more important than localized ones, owing to the highly diversified nature of the enduser markets and the complexity of the technologies involved in assembling an end product for the market. The individual clusters in each of the six regions are dominated by a key local actor: either a strong research centre or a lead firm that serves as a catalyst to bring together the firms in the cluster. However, owing to the nature of the technologies involved and the intra- and inter-firm dynamics, there is little local co-operation and few traded relationships among firms within the individual clusters. What the firms in the clusters do share is their common linkage to the leading institution or firm and their common interest in stimulating and maintaining the critical supply of highly skilled labour (Hendry et al. 2000, 140–1). These findings are consistent with a number of objections that have been raised with respect to Porter’s assertions about the critical importance of local factor and demand conditions for the development of clusters. As noted earlier, one of the key sources of knowledge – especially in an era in which nonlinear,

224 Meric S. Gertler and David A. Wolfe interactive innovation processes have become widespread – is the customer or user. While there is a widespread acceptance within the cluster literature that the most demanding customers – Porter’s oft-repeated ‘sophisticated and demanding’ customers (Porter 2000, 258) – must necessarily be local for competitive industries and clusters to emerge, this assertion is contradicted by a growing body of both analytical and empirical research. A number of management scholars question whether the home market is as critical for the emergence of competitive industries and economies as Porter insists. They note that Porter’s original analysis in The Competitive Advantage of Nations, which was conducted at the level of the national economy, drew overwhelmingly on the US case – a large, affluent, and diverse domestic market that is likely to contain a high proportion of linkages to customers within the same country. They draw upon the international trade and other bodies of literature to note that these conditions seldom hold to the same extent in small, open trading economies – such as Canada or the Scandinavian economies, which are tightly linked to larger trading partners (Davies and Ellis 2000, 1202–4). The emphasis on local demand conditions holds up even less well when it is transferred to the local and regional level, where the very definition of the cluster’s boundaries is highly problematic. As Martin and Sunley note, the definition of the local in much of the work on regional clusters is highly elastic (2003, 11). A growing body of evidence suggests the primacy attributed to local demand conditions may be less than universal. Malmberg and Power document numerous cases in which non-local demand dominates industrial markets (2005a).

Cluster dynamics and industrial knowledge bases We suggest that a great deal of the confusion surrounding the importance or unimportance of spatial proximity to the innovation process arises from the failure to differentiate between different types of knowledge that underlie innovative products and processes. For certain industries, reliant on particular forms of knowledge and learning processes, proximity between firms and their customers, suppliers, and competitors will logically be essential; for others, this mutual attraction may indeed be far less powerful. It is therefore useful to have a clearer understanding of how these forms of knowledge might vary systematically between industries. The recent literature in economic geography and evolutionary economics makes the (by now) well-known distinction between tacit and codified forms of knowledge (see Gertler 2003 for a recent review of this literature and related debates). The argument in a nutshell is that, because tacit knowledge is – by definition – more difficult to share in written, symbolic form, and because it is strongly context-specific, it tends to be more commonly transmitted through direct face-to-face (F2F) interaction. Consequently, those firms and industries for which innovation depends heavily on tacit knowledge transmission and application will tend to cluster spatially with their customers, suppliers, and competitors. Conversely, those firms and industries in which codified forms of

Spaces of knowledge flows 225 knowledge are relatively more important will be less constrained spatially to cluster in this way. In other words, local ‘buzz’ dynamics will be less powerful than the reach of global ‘pipelines’. Compelling as this distinction may appear, it has been criticized from a number of angles that are relevant to the current discussion. First, as Nonaka and Takeuchi (1995), Nonaka et al. (2000), and Johnson et al. (2002) point out, the process of producing and using new knowledge involves a dynamic interplay between, and transformation of, tacit and codified forms of knowledge in virtually all sectors of the economy. In other words, these two forms of knowledge are complements to, not substitutes for, one another. Second, Pavitt (1984), Malerba (2005), and others have argued that knowledge bases tend to vary systematically by industry – and so too does the nature of the innovation process. It therefore stands to reason that the geography of knowledge flows – within and between local clusters – will also exhibit characteristic patterns by sector. Recent analyses of this question have found the distinction between ‘analytical’ and ‘synthetic’ knowledge bases (Laestadius 1998) to be helpful in this regard (see Coenen et al. 2004; Asheim and Gertler 2005). A synthetic knowledge base dominates industrial settings where innovation takes place mainly through the application or novel combinations of existing knowledge. Innovation in such industries tends to be driven by the need to solve specific problems arising in the interaction with clients and suppliers. Classic industry examples come from sectors within advanced industrial engineering (such as the development of specialized machinery). In such sectors, research is less important than development. When it occurs, it tends to take the form of applied research, but the most prevalent form of innovative activity is what might be described as incremental product or process development to solve technological or production problems presented by customers. Knowledge tends to be created inductively rather than deductively, through a process of testing, experimentation, and simulation. While the knowledge embodied in technical solutions is at least partially codified, tacit knowledge tends to be more important, since shop floor or office experience, on-the-job training, and learning by doing, using, and interacting are crucial to knowledge generation. Much of this knowledge resides in concrete know-how, craft, and practical skill. In contrast, in those industries where scientific knowledge is highly important, and where knowledge creation is normally based on formal models, codified science, and rational processes, an analytical knowledge base is dominant. Obvious examples of such industries are biotechnology and the ICT sector. Here, the core activity generating new products and processes is systematically organized research and development, both inside the individual firm and in collaboration with universities and other research organizations. Knowledge inputs and outputs in this type of knowledge base are more often codified (or readily codifiable) than in the case of synthetic knowledge, although this does not mean that tacit knowledge is unimportant since – as noted above – both kinds of knowledge are always involved in the innovation process. As

226 Meric S. Gertler and David A. Wolfe Asheim and Gertler (2005) note, the predominance of codification is due to several factors: knowledge inputs are often based on reviews of existing studies, knowledge generation is based on the application of widely shared and understood scientific principles and methods, knowledge processes are more formally organised (e.g. in R&D departments) and outcomes tend to be documented in reports, electronic files or patent descriptions. Knowledge outputs are most frequently in the form of new products or processes, which are more likely to constitute radical rather than incremental innovations. Of course, just as all innovation processes make use of both tacit and codified forms of knowledge, so too do many industries draw significantly upon both synthetic and analytical forms of knowledge. A case in point is the medical devices and technologies sector, in which product development draws upon knowledge from a wide range of fields including bioscience, ICT, software, advanced materials, nanotechnology, and mechanical engineering. Accordingly, it makes more sense to locate individual industries along a spectrum between purely analytical and synthetic knowledge bases, with many – such as the automotive industry – occupying an intermediate position along this continuum. How might this distinction between synthetic and analytical knowledge bases shape our understanding of the geography of knowledge flows and their relationship to clusters? One obvious interpretation would be to argue that localized learning and knowledge flows are more important in synthetic-knowledge industries because of the central role of tacit knowledge and F2F interaction with customers and suppliers. Indeed, many of the examples from the work of Lundvall (1988) and Von Hippel (1988) originate in sectors such as mechanical engineering and specialized industrial machinery, where learning by interacting between users and producers represents the primary mode of innovation. By the same token, in those sectors for which analytical knowledge is more important, the greater prominence of codified and codifiable knowledge in the innovation process leads us to expect that knowledge flows and learning relationships would not be locally bound to the same extent. It should not be surprising, therefore, that the original inspiration for the ‘pipeline’ concept comes from the biotechnology industry (Bathelt et al. 2004; Owen-Smith and Powell 2004). Moreover, in their analysis of the Medicon Valley biotech cluster (in the Öresund region spanning eastern Denmark and southern Sweden), Coenen et al. (2004) provide at least preliminary evidence that non-local knowledge flows, as measured by co-authorship of scientific papers, are strong. Local scientific personnel collaborate actively with colleagues in Germany, the UK, and the US, with roughly one-third of local firms collaborating with partners outside Europe (Coenen et al. 2004, 1013). However, this binary synthetic (tacit) = local, analytical (codified) = global framework is likely to be too simple to capture the complex geography of

Spaces of knowledge flows 227 knowledge flows. Asheim and Gertler (2005) note that, contrary to the above prediction, there is compelling evidence that analytically oriented sectors like biotech in fact exhibit strong clustering tendencies in which at least some forms of knowledge flow are locally bound. They cite the recent literature on knowledge spillovers (see Feldman 2000), in which it has been demonstrated that patent citations exhibit a strongly localized geography. They also review the ‘star scientist’ work of Zucker and Darby (1996), which emphasizes the commercial benefits of close relationships between biotech start-ups and highly productive or highly cited scholars in the same region. Their conclusion is that, despite the codifiability of much scientific knowledge in sectors like biotechnology, there are still some significant advantages to being physically proximate to sources of new knowledge (including knowledge of successful as well as unsuccessful experiments). These dynamics may explain the rather striking geographical concentration of entrepreneurial activity in biotechnology as revealed in Cortright and Mayer’s (2002) landmark study of US metropolitan regions. They also help us make sense of another finding from the Medicon Valley study. Notwithstanding their other findings cited above, Coenen et al. (2004) concede that intra-regional co-authorship activity between local firms and public research organizations (particularly on the Swedish side of the Öresund) remains strong. In sum, the emerging picture seems more or less consistent with the ‘buzz and pipelines’ geography outlined earlier: while global research partnerships and knowledge exchanges are commonplace, these complement (rather than substitute for) strongly localized learning dynamics. Moreover, the two ends of the global pipeline are likely to be active concentrations of research activity, where buzz is rampant. However, it is worth pointing out that customers seem somewhat removed from this picture: the international collaborations described in this literature are largely supply-side in nature, linking individual researchers into non-local epistemic communities (Coenen et al. 2004).

Cluster evolution in Canada To resolve some of these questions empirically, we have directed a large national study of cluster development that documents the emergence and evolution of local clusters in different regions of Canada.1 The goal of our project is to determine the prevalence and success of local industrial clusters across Canada’s diverse regional economies, and to analyse how the formation and growth of these clusters contributes to local economic growth and innovative capacity. Underlying this objective is a set of more substantive questions. How do local assets and relationships between economic actors enable firms – in any industry – to become more innovative? Under what circumstances does ‘the local’ matter, and how important are local sources of knowledge and locally generated institutions (public and private) in strengthening the innovative capabilities of firms and industries? What is the relative importance of non-local actors, relationships, and flows of knowledge in shaping the development trajectories of localized

228 Meric S. Gertler and David A. Wolfe innovation and growth? In contrast to much of the existing international literature on clusters, we do not assume a priori that local relationships and flows trump non-local forms of economic interaction. The project combines both quantitative and qualitative methodologies to study 26 cases across Canada. Each case has been examined using a common research methodology, based primarily on interviews with key cluster participants, though supplemented by statistical analysis at the regional and national level (Gertler and Levitte 2003; Amara et al. 2003). Our methodology allows for systematic comparisons between the case studies, which – again, in contrast to most of the existing work in this field – include metropolitan and nonmetropolitan locations, as well as more and less knowledge-intensive industries. The selection of industries covered reflects the breadth and structure of the Canadian economy, resisting the temptation to focus solely on a narrow list of ‘new economy’ cases. The cases range from highly knowledge-intensive activities such as biotechnology, photonics and wireless equipment, telecommunication equipment and aerospace, to more traditional sectors such as steel, automotive parts, specialty food and beverages, and wood products. Our study also overcomes the tendency in previous policy work to rely on cluster models imported from other countries that may not reflect the Canadian reality. Our intention is to inform cluster-based policy prescriptions that are appropriate for the distinctive circumstances confronting Canada’s regional economies. The major intellectual contributions from this project are now becoming clear. Concerning the relationship between local and global forces in the development of clusters, as noted earlier, some of the most widely cited literature in the field maintains that a strong local market and strong local competition are two essential elements for the development of internationally competitive clusters. In contrast, our findings indicate that, in many successful clusters, the markets served are continental or international, that local customers constitute a relatively small proportion of the firm’s total market, and that firms’ most sophisticated and demanding markets are not local. Perhaps the most vivid examples come from the life sciences, where firms in Canada’s leading biotech clusters (such as Montreal, Toronto, Vancouver, and Saskatoon) have strong non-local backward and forward linkages. Recent analysis of Statistics Canada’s national survey of biotechnology firms (Gertler and Levitte 2003) reveals the complex, dual geography of relationships in which successful firms are embedded. On the one hand, they tap into global knowledge markets by hiring highly qualified personnel from abroad. They also take advantage of other global flows of knowledge, through the use of scientific publications and databases, by licensing their intellectual property to foreign partners, or by licensing the intellectual property of foreign firms for their own use. When they develop collaborative relations with other firms, for both research and marketing purposes, these are both local and global in nature. On the other hand, they rely heavily on local sources of investment capital from private sources (angel investors, family and friends), and are highly likely to have spun off from another local company or research institution at some point in their past.

Spaces of knowledge flows 229 Similarly, there is a strong emphasis within the international cluster literature on the importance of a strong local supply base. Once again, our research has produced novel results. While certain key inputs are predominantly local (see below), relatively few regions can rely exclusively on their local knowledge base to develop, design, and produce innovative products. Conversely, knowledge flows in synthetically oriented sectors like aerospace that increasingly involve the integration of complex technology subsystems draw upon a global network of system integrators to assemble the final product. According to Niosi and Zhegu (2005, p. 22): Four characteristics appear when these knowledge flows are examined. First, they are mostly international. Second, they are mostly constituted of explicit and codified knowledge. Third, they involve several independent companies. And finally, they are closely tied to markets for parts, components and subassemblies. Equally surprisingly, we find comparable results from most of our case studies of ICT clusters as well. What is readily apparent from talking to firms in the ICT clusters, however, is that the amount of inter-firm collaboration in the form of key customer or supplier relationships is relatively low. For the vast majority of firms, the focus of most economic activity – key customers, sources of supply, competitors, and important strategic partnerships – occurs at the global level. Some firms in individual clusters rely upon a local supply base for key inputs, but the vast majority tends to draw components and knowledge inputs from a diverse array of geographic sources. Thus, a core theme that emerges strongly in the study of inter-firm dynamics is the fluid nature of relations between customers, suppliers, and competitors in the cluster. Variation in types of relationships is reflected in many different combinations and permutations of inter-firm dynamics, which are rendered more complex by virtue of the fact that they occur at local or regional, national, global, and ‘virtual’ levels. Consequently, it would seem that explanations of ICT cluster dynamics that privilege inter-firm relationships based on proximity to each other do not capture the whole story. As is true for our biotech case studies, firms in the ICT clusters draw upon a diverse array of sources for their products (Bramwell et al. 2005).

Towards a knowledge-based typology of clusters Can we use any of the conceptual arguments presented earlier to classify our findings into some form of typology? The principal challenge in developing any typology is to select the key categories for organizing the typology along both the horizontal and vertical axes. The proposed typology organizes cases along two key dimensions, knowledge dynamics on the one hand, and the geography of knowledge flows on the other. The categories arranged along the knowledge-base dimension build on the typology of knowledge bases introduced earlier (based on Asheim and Gertler

230 Meric S. Gertler and David A. Wolfe 2005). To the two original categories of knowledge base – analytical and synthetic – we add a third ‘hybrid’ to reflect those industries that draw significantly from both synthetic and analytical knowledge bases. Within the Canadian economy, these hybrid sectors might include industries, such as winemaking, specialty food products, or wood products, which have a strong link to an agricultural or natural resource base. While the generation of new product and process innovations may depend to a large extent on analytical, lab-based scientific methods that draw from a codified international body of science, their successful production also relies on craft-based know-how and tacit knowledge. While production may not be driven by the solution of customers’ specific problems, there is nevertheless a strong element of niche-based differentiation. Also belonging to this category would be sectors such as medical technologies and aerospace. In both cases, while codified science and engineering knowledge make major contributions to product innovation, so too do the solution of particular customer problems and demands. The categories for the geography of knowledge flows reflect three situations – one in which sources of knowledge are primarily local, another where global sources dominate, and a third in which firms draw significantly from both global and local sources of knowledge. Combining these two dimensions together produces the three-by-three typology in Table 11.1, which shows the classification for a selection (one-half ) of our 26 cases. Table 11.1 Typology of knowledge bases and flows Knowledge base

Geography of knowledge flows Strong global sources

Synthetic

Global and local sources

Strong local sources

Ontario steel

Sudbury mining S&S1 Windsor auto parts/TDM2

Hybrid

Montreal aerospace

Okanagan (BC) wine Niagara (Ont) wine Toronto specialty food

Analytical

Saskatoon agri-biotech

Montreal, Toronto, Vancouver biotech Ottawa telecom/ photonics

Notes 1 Supply and services 2 Tool, die and mould

Toronto medical technologies

Spaces of knowledge flows 231 From the distribution of entries in this matrix, several important insights emerge. First, while there is a tendency for synthetic-knowledge industries to source their knowledge locally, this is not universally true. In the case of Ontario steel, firms such as Dofasco in Hamilton are embedded in both local and international knowledge networks. Second, while cases such as agricultural biotech in Saskatoon support the predicted correspondence between analytical knowledge and global sourcing, other analytical-knowledge cases such as biotech in Montreal, Toronto, and Vancouver, or telecom equipment and photonics in Ottawa, depend on a mix of strong local and global knowledge sources and flows. Third, hybrid sectors show no clear tendency toward one scale or the other. Cases such as Toronto’s medical technologies industry – for which analytical knowledge is a strong complement to synthetic forms of knowledge – show strong dependence on local knowledge sources. Food and wine clusters in Toronto, Niagara, and Okanagan rely significantly on both scales of knowledge flow, while Montreal aerospace draws very little unique knowledge from local sources. While these results defy easy generalization, they certainly demonstrate the limits to claims by Porter and others concerning the predominance of local knowledge flows.

What remains of the local? Notwithstanding the complex, multi-scalar geography of knowledge described above, certain characteristics and properties of local industrial clusters remain critically important to the competitive success of firms in a wide range of industries. Despite the importance of non-local markets, knowledge flows, and (in some cases) supply bases, our research confirms that the local dynamics of social interaction between members of the cluster are crucial. These intra-cluster relationships promote the local circulation of knowledge, underpinning the learning processes that enable firms to succeed at innovation (a finding that would appear to support the arguments in Malmberg and Power 2005b). Our work documents the nature and significance of these knowledge flows, and the various forms they take. The local participants in these social learning systems include firms, institutions of education and research, venture capitalists, producer associations and specialized government research labs. In this way, the case studies document a balance between local and non-local relationships and knowledge flows – in other words the dynamic tension between the local ‘buzz’ described above and global ‘pipelines’ that circulate knowledge between clusters. Furthermore, the case studies suggest that the most successful clusters have profited from the development of strong social networks at the community level and the emergence of dedicated, community-based organizations. These entities link leaders in the individual clusters to a broader cross-section of the community. They appear to be supported by new institutions of civic governance that identify problems impeding the growth of the cluster and help mobilize support across the community for proposed solutions. We have found some

232 Meric S. Gertler and David A. Wolfe evidence to suggest that size is a critical variable in the success of civic engagement, with some of the larger urban centres encountering greater difficulty in achieving effective degrees of mobilization. Another finding of fundamental importance, relating to the role of local assets in the innovation process, concerns the relationship between knowledge infrastructure and cluster emergence and evolution. The international literature on the most celebrated clusters identifies research infrastructure, especially post-secondary educational institutions, as the essential ingredient for cluster formation (Gibbons 2000; Kenney and Patton forthcoming). Significantly, and to the contrary, our research indicates that, with a few notable exceptions, local research infrastructure (as a key part of the regional innovation system) plays a supporting, not a causal, role in the growth of clusters in Canada. In some significant instances, the local development of advanced educational and research programs clearly follows the emergence of a dynamic local cluster, rather than preceding it. In most cases, the presence of a strong research infrastructure constitutes a local antecedent that lays the groundwork for the emergence of a cluster. This research infrastructure also contributes to the presence of a ‘thick’ labour market in the local economy, which serves as a magnet for firms in search of highly skilled labour. It may also attract firms to a city-region in the expectation of tapping into the knowledge base that exists. However, strong research infrastructure and a thick labour market are underlying conditions that extend beyond the boundaries of individual clusters. One of the most consistent findings from our work concerns the role of local labour markets and talent. If there is one type of input that is overwhelmingly local, it is highly skilled labour. It is clear that the depth and breadth of the local labour market is the key ingredient defining a cluster’s ability to support knowledge-intensive production. It is also the factor that is most amenable to public policy influence. However, our work suggests that the creation of a talented labour pool in turn depends on many different factors, including not only the strength of local post-secondary education and specialized training institutions but also a set of ‘quality of place’ characteristics that determine a region’s ability to retain well-educated labour and attract it from elsewhere (Gertler and Vinodrai 2005). However, this finding has also revealed a potential downside to the talent factor: not all locations in the country will be equally successful in the pursuit of this objective. Some of our cases have encountered significant obstacles in developing a deep labour market, despite persistent efforts. The results presented in this chapter, based on our national study of cluster development in Canada’s regions, suggest that many of the taken-for-granted qualities of clusters – and especially the geography of knowledge flows supporting innovation – may not obtain in reality. Clearly, this does not mean necessarily that ‘the local’ is unimportant. Rather, it becomes important for different reasons, relating to social interaction, leadership dynamics, and labour markets. Nevertheless, the risk of implementing ill-conceived policy initiatives is great so long as public agencies labour under misguided notions concerning the local self-sufficiency of ‘successful’ clusters.

Spaces of knowledge flows 233

Notes 1

See http://www.utoronto.ca/isrn for a full project description and publications arising from this work. This project, funded by the Social Science and Humanities Research Council of Canada, has run from 2001 to 2005. For the most recent collection of case studies emerging from this project, see Wolfe and Lucas (2005).

References Amara, N., Landry, R. and Ouimet, M. 2003. Milieux innovateurs: determinants and policy implications. Paper presented at the DRUID Summer Conference, Elsinore, Denmark, June 12–14. (available at http://www.druid.dk/conferences/summer 2003/Papers/AMARA_LANDRY_OUIMET.pdf). Asheim, B.T. and Gertler, M.S. 2005. The geography of innovation: Regional innovation systems, in J. Fagerberg, D.C. Mowery and R.R. Nelson (eds) The Oxford Handbook of Innovation. Oxford: Oxford University Press, pp. 291–317. Bathelt, H., Malmberg, A. and Maskell, P. 2004. Clusters and knowledge: Local buzz, global pipelines and the process of knowledge creation, Progress in Human Geography, 28:1, 31–56. Bramwell, A., Nelles, J. and Wolfe, D. 2005. Knowledge, innovation and institutions: Global and local dimensions of the ICT cluster in Waterloo, Canada. Paper presented at the DRUID Academy Winter Conference, Jan. 27–9 (available at http://www. druid.dk/ocs/viewpaper.php?id=347&cf=2). Breschi, S. and Malerba, F. 2001. The geography of innovation and economic clustering: Some introductory notes, Industrial and Corporate Change, 10:4, 817–33. Brown, J.S. and Duguid, P. 2000. Mysteries of the region: Knowledge dynamics in Silicon Valley, in C.-M. Lee, W.F. Miller, M.G. Hancock and H.S. Rowen (eds) The Silicon Valley Edge. Stanford: Stanford University Press, pp. 16–39. Coenen, L, Moodysson, J. and Asheim, B.T. 2004. Nodes, networks and proximities: On the knowledge dynamics of the Medicon Valley biotech cluster, European Planning Studies, 12:7, 1003–18. Cortright, J. and Mayer, H. (2002). Signs of Life: The Growth of Biotechnology Centers in the U.S. Washington, DC: Centre on Urban and Metropolitan Policy, The Brookings Institution. Davies, H. and Ellis, P. 2000. Porter’s Competitive Advantage of Nations: time for the final judgement? Journal of Management Studies, 37:8, 1189–213. Feldman, M.P. 2000. Location and innovation: The new economic geography of innovation, spillovers, and agglomeration, in G.L. Clark, M.P. Feldman, and M.S. Gertler (eds) The Oxford Handbook of Economic Geography, Oxford: Oxford University Press, pp. 373–94. Feldman, M.P., Francis, J. and Bercovitz, J. 2005. Creating a cluster while building a firm: Entrepreneurs and the formation of industrial clusters, Regional Studies, 39:1, 129–41. Gertler, M.S. 1995. ‘Being there’: Proximity, organization, and culture in the development and adoption of advanced manufacturing technologies, Economic Geography, 71, 1–26. Gertler, M.S. 2003. The undefinable tacitness of being (there): Tacit knowledge and the economic geography of context, Journal of Economic Geography, 3, 75–99. Gertler, M.S. 2004. Manufacturing Culture: The Institutional Geography of Industrial Practice. Oxford: Oxford University Press.

234 Meric S. Gertler and David A. Wolfe Gertler, M.S. and Levitte, Y.M. 2003. Local nodes in global networks: The geography of knowledge flows in biotechnology innovation. Paper presented at the DRUID Summer Conference, Elsinore, Denmark, June 12–14 (available at http://www. druid.dk/conferences/summer2003/Papers/GERTLER_LEVITTE.pdf). Gertler, M.S. and Vinodrai, T. 2005. Anchors of creativity: How do public universities create competitive and cohesive communities?, in F. Iacobucci and C. Tuohy (eds) Taking Public Universities Seriously. Toronto: University of Toronto Press, pp. 293–315. Gibbons, J.F. (2000) The role of Stanford University: A dean’s reflections, in C.-M. Lee, W.F. Miller, M.G. Hancock, and H.S. Rowen (eds) The Silicon Valley Edge. Stanford: Stanford University Press, pp. 200–17. Hendry, C., Brown, J. and Defillippi, R. 2000. Regional clustering of high-technology based firms: Opto-electronics in three countries, Regional Studies, 34:2, 129–44. Johnson, B., Lorenz, E. and Lundvall, B-Å. 2002. Why all this fuss about codified and tacit knowledge? Industrial and Corporate Change, 11, 245–62. Kash, D.E. and Rycroft, R.W. 2000. Patterns of innovating complex technologies: A framework for adaptive network strategies, Research Policy, 29, 819–31. Kenney, M. and Patton, D. forthcoming. The co-evolution of technologies and institutions: Silicon Valley as the ideal-typical high technology cluster, in P. Braunerhjelm and M.P. Feldman (eds) Cluster Genesis: The Emergence of Technology Clusters and the Implication for Government Policy. Oxford: Oxford University Press. Laestadius, S. 1998. Technology level, knowledge formation and industrial competence in paper manufacturing, in G. Eliasson et al. (eds) Microfoundations of Economic Growth: A Schumpeterian Perspective. Ann Arbor: University of Michigan Press, pp. 212–26. Lundvall, B-Å. 1988. Innovation as an interactive process: From user-producer interaction to the national system of innovation, in G. Dosi, C. Freeman, G. Silverberg and L. Soete (eds) Technical Change and Economic Theory. London: Pinter, pp. 349–69. Malerba, F. 2005. Sectoral systems: How and why innovation differs across sectors, in J. Fagerberg, D. Mowery and R. Nelson (eds) The Oxford Handbook of Innovation. Oxford: Oxford University Press. Malmberg, A. and Maskell, P. 2002. The elusive concept of localization economies: Towards a knowledge-based theory of spatial clustering, Environment & Planning A, 34, 429–49. Malmberg, A. and Power, D. 2003. (How) do (firms in) clusters create knowledge? Paper presented at the DRUID Summer Conference, Elsinore, Denmark, June 12–14. (available at http://www.druid.dk/conferences/summer2003/Papers/ MALMBERG_POWER.pdf). Malmberg, A. and Power, D. 2005a. On the role of global demand in local innovation processes, in P. Shapira and G. Fuchs (eds) Rethinking Regional Innovation and Change: Path Dependency or Regional Breakthrough? Amsterdam: Kluwer Academic Publishers. Malmberg, A. and Power, D. 2005b. True clusters: A severe case of conceptual headache, in B.T. Asheim, P. Cooke, and R.L. Martin (eds) Clusters and Regional Development. London: Routledge (forthcoming). Martin, R. and Sunley, P. (2003) Deconstructing clusters: Chaotic concept or policy panacea, Journal of Economic Geography, 3:1, 5–35. Maskell, P. 2001. Towards a knowledge-based theory of the geographic cluster, Industrial and Corporate Change, 10:4, 921–43.

Spaces of knowledge flows 235 Niosi, J. and Zhegu, M. 2005. Aerospace clusters: Local or global knowledge spillovers, Industry and Innovation, 12:1, 5–29. Nonaka, I. and Takeuchi, H. 1995. The Knowledge Creating Company. Oxford: Oxford University Press. Nonaka, I., Toyama, R. and Nagata, A. 2000. A firm as a knowledge-creating entity: A new perspective on the theory of the firm, Industrial and Corporate Change, 9:1, 1–20. Owen-Smith, J. and Powell, W.W. 2004. Knowledge networks as channels and conduits: The effects of spillovers in the Boston biotechnology community, Organization Science, 15, 5–21. Pavitt, K. 1984. Sectoral patterns of technical change: Towards a taxonomy and a theory, Research Policy, 13, 343–73. Porter, M.E. 2000. Locations, clusters, and company strategy, in G.L. Clark, M.P. Feldman and M.S. Gertler (eds) The Oxford Handbook of Economic Geography. Oxford: Oxford University Press, pp. 253–74. Saxenian, A. 1994. Regional Advantage: Culture and Competition in Silicon Valley and Route 128. Cambridge, MA: Harvard University Press. Storper, M. and Leamer, E.E. 2001. The economic geography of the internet age, Journal of International Business Studies, 32, 641–65. Storper, M. and Venables, A.J. 2004. Buzz: Face-to-face contact and the urban economy, Journal of Economic Geography 4:4, 351–70. Sturgeon, T.J. 2003. What really goes on in Silicon Valley? Spatial clustering and dispersal in modular production networks, Journal of Economic Geography, 3, 199–225. Von Hippel, E. 1988. The Sources of Innovation. Oxford: Oxford University Press. Wolfe, D.A. 2005. Innovation and research funding: The role of government support, in F. Iacobucci and C. Tuohy (eds) Taking Public Universities Seriously. Toronto: University of Toronto Press, pp. 316–40. Wolfe, D.A. and Gertler, M.S. 2004. Clusters from the inside and out: Insights from the Canadian study of cluster development, Urban Studies, 41:5/6, 1071–93. Wolfe, D.A. and Lucas, M. (eds) 2005. Global Networks and Local Linkages: The Paradox of Cluster Development in an Open Economy. Montreal and Kingston: McGill-Queen’s University Press and Queen’s School of Policy Studies. Zucker, L.G. and Darby, M.R. 1996. Star scientists and institutional transformation: patterns of invention and innovation in the formation of the biotechnology industry, Proceedings of the National Academy of Science, 93, 12709–16.

Acknowledgement The authors gratefully acknowledge the Social Sciences and Humanities Research Council of Canada for supporting this research through its Major Collaborative Research Initiatives Program

12 Spatial and organizational patterns of labor markets in industrial clusters The case of Hollywood Allen J. Scott Setting the scene: clusters as labor markets It is nowadays more or less routine in the literature on industrial clusters to affirm that their form and functions depend on the differential articulation of three fundamental sets of ‘Marshallian’ variables, i.e., structures of inter-firm transacting, learning and innovation processes, and local labor market relationships. The first two of these sets of variables have been the objects of enormous and widespread research attention over the last couple of decades. The third, by contrast, has suffered from relative neglect. There is, to be sure, a scattered literature on the local labor market dimensions of industrial clusters (as the references at the end of the chapter attest), but nothing to compare with the massive attention that has been paid to the other two. In particular, the specific ways in which the operation of local labor markets can enhance the overall competitiveness of industrial clusters have been given rather short shrift in the literature, which is all the more to be regretted in view of what I take to be their absolutely critical role in sustaining high levels of performance in local economic systems. In the present chapter, I seek to redress some of his imbalance by means of an empirical examination of the local labor market dynamics of the Hollywood motion picture complex, with special emphasis on the notion of the ‘labor relation’ as an object of socio-economic regulation in geographic space, and specifically within clusters. For our purposes, the history of the Hollywood labor market is divisible into two main episodes. The first corresponds to the classical studio system of production that developed over the 1920s and 1930s, when workers functioned for the most part as permanent company employees with regular remuneration (Ross, 1941, 1947). This state of affairs obtained for workers of all gradations, from the blue-collar manual workers at the bottom of the job ladder to the stars at the pinnacle. The second episode coincides with the new Hollywood that gradually came into existence after the 1950s as the old studio system broke down, and when the employment relation was largely externalized for all but selected groups of workers providing managerial and administrative continuity within the firm. In this new order of things, the majority of workers now assumed freelance status, operating on a commission basis or being taken on by

Spatial and organizational patterns 237 production companies as temporary employees, and moving irregularly from job to job depending on the fluctuations of productive activity.

Historical emergence of the employment relation in Hollywood The classical studio system The core of the Hollywood production system in the golden age of the 1920s and 1930s was constituted by the large vertically integrated studios or majors that dominated the industry at that time. Over the 1920s the bases of the system were put into place through a succession of company mergers and buy-outs, and by 1930, Columbia Pictures, Fox, Metro-Goldwyn-Mayer, Paramount, RKO, Universal, and Warner Brothers led the field ahead of a number of smaller independent production companies. The main studios were vertically integrated over the whole chain of business activities from production through distribution to exhibition. In addition, production itself was organized on a vast in-house basis in which each studio brought together under a single roof the tasks of writers, directors, producers, actors, composers, musicians, costume designers, cinematographers, etc., and a supporting army of manual, crafts, and technical workers (Lovell and Carter, 1955). Several analysts have erroneously referred to this system as a form of mass production (cf. Bordwell et al., 1985; Storper and Christopherson, 1987), whereas it is more accurately describable as a combination of bureaucratic and craft production models, in which the constant efforts of studio managers to streamline production persistently ran up against the commercial exigencies of product differentiation. On the one hand, the system was overlain by large internal bureaucracies seeking wherever possible to control the labor process through divisions of labor and to standardize the final product. On the other hand, integrated studio production was centrally focussed on craft labor entailing personalized skills and rather freely flowing team work. I have argued elsewhere that the classical Hollywood studios can best be described as ‘systems houses’, meaning that they occupy a position somewhere between the extremes of mass production and flexible specialization, but never achieving the advanced level of routinization essential to the former or the degree of vertical disintegration and narrowness of core competency characteristic of the latter (Scott, 2002). The classical studios possessed the advantage of always having full complements of essential skills and talents on hand to keep production going at a rapid pace. Moreover, the stars they created served under long-term salary agreements (usually of seven years’ duration). The stars could thus not appropriate the full value of the rents that they generated as their celebrity expanded (Kindem, 1982). It would be an error, however, to presume that there was no temporary or short-term employment in Hollywood in the pre-war years. Even at the height of the classical era, the studios as well as independent production companies sought to fine-tune their use of labor by frequent hiring and lay-off

238 Allen J. Scott at the margins, especially of low-skill workers. Great throngs of extras added to the pool of floating short-term labor in Hollywood at this time. The problems created by the expansion of this body of day-workers became so pressing that the studios combined together as early as 1926 to set up the Central Casting Corporation to serve as a clearing-house for casual acting jobs (Ross, 1947). Numerous attempts were made during the classical studio era to unionize the workers, a daunting task under any circumstances, but exceptionally difficult in Los Angeles at this period, when the local elite was resolutely committed to the maintenance of the open shop (Davis, 1997; Ross, 1941). Over the 1920s and 1930s, political clashes between the studios and the nascent unions were rampant. In an attempt to forestall full-blown labor organizing, the studios went so far as to create the Academy of Motion Picture Arts and Sciences in 1927, its prime mission being to function as a company union (Nielson and Mailes, 1995). The Academy, as initially constituted, had five branches representing producers, writers, directors, actors, and technicians. It failed, however, to win legitimacy as a labor organization, and in the aftermath of the depression, when motion-picture workers across the board were asked by the studios to take deep pay cuts, it steadily gave ground to independent unionization movements. The Academy, which still exists, has long since abandoned its original goals, and is now mainly known for its patronage of the annual Oscar awards. The rise of an independent union movement in Hollywood in the pre-war years is a complicated story that has been treated at length elsewhere (e.g. Amman, 1996; Horne, 2001; Prindle, 1988; Ross, 1941). We need only note here that the urge to unionization of the labor force was energized by the passage of the Wagner Act in 1935, and that by the time of the outbreak of World War II, most of the workers in the large studios had been incorporated into either blue-collar unions or professional guilds. The new Hollywood Immediately after the war, two dramatic sets of changes completely disrupted the classical Hollywood system. One set stemmed from the Paramount Decree issued by the Federal Department of Justice in 1948, forcing the large studios to divest themselves of their theater chains. The result was greatly intensified competitiveness in the motion-picture industry as market forces came into play at the point of intersection between distribution and exhibition. The other set of changes turned on the rise of television and the shrinkage of motion-picture audiences over the immediate post-war decades. These two developments gave rise to greatly augmented uncertainty and risk throughout the motion-picture industry, and the destabilization of production to the point where the studios could no longer maintain the high fixed costs and relatively steady output schedules that had characterized their operations before the war. Industrial sectors subject to stresses and strains of these sorts commonly respond by some degree of vertical disintegration. The push to vertical disintegration was all the more pressing in Hollywood because of the large numbers

Spatial and organizational patterns 239 of workers that the studios carried on their permanent payrolls, and that could not always be kept in full-time employment under the new competitive conditions. The studios thus went through a difficult period of downsizing and outsourcing from the late 1940s to the early 1960s, leading eventually to an entirely different pattern of production in comparison with the classical studio era (Storper and Christopherson, 1987). On the one side, this pattern was shaped by a great expansion in the number of independent production companies and specialized service firms. Many of these independents began to work in shifting network arrangements with the restructured studios. On the other side, there was also a shift away from permanent employment in favor of limited-term contracts covering the duration of particular projects. An illustration of this point is provided by the observation that there were 804 professional actors under contract to the Hollywood studios in 1945, whereas the number had dwindled to 139 by 1960 (Cantor and Peters, 1980). Today, the majority of workers in the industry are freelancers who for the most part circulate through the job system in spasmodic bouts of employment and unemployment. As the 1960s came and went, then, radical adjustments were made in the local labor market, and these are still in many ways working themselves out. One conspicuous outcome was that the stars at the top of the job ladder, freed from long-term contractual obligations to individual studios, were now fully able to appropriate the rents due to their celebrity. But the mass of workers in the industry, no matter what their standing in terms of skill or talent, were faced with the greatly increased risks of a project-oriented employment system. The human resource management functions of the firm now gave way to the selfmanagement of workers. Equally, traditional forms of advancement based on the building up of firm-specific human capital and seniority were supplanted (for all but a cadre of managers and administrative workers) by a process based on reputation as the main currency of worker evaluation. Workers adopted as far as possible the strategic imperative of planning their credits and experiences across many different jobs in the quest to mold their reputations (cf. Menger, 1991). Along with these changes came a distinctive manner of classifying workers according to their labor-market power. Workers were either ‘above the line’ (signifying individuals whose salaries are individually negotiated and who are named explicitly as line item entries in any project budget), or ‘below the line’ (with remuneration being set according to wage schedules defined in collective-bargaining agreements). As a general, though by no means absolute, rule, professional guilds became the collective representatives of above-the-line workers, and manual, crafts, and technical employees’ unions of below-the-line workers. These local labor market developments consequent upon the advent of the new Hollywood represent a dramatic vanguard case of the flexibilization of work arrangements that was later to spread much more widely. In the 1950s and 1960s, Hollywood was already pioneering on a large scale the contingent model of employment that is nowadays taken for granted in many if not most sectors of the American economy.

240 Allen J. Scott

Employment, local labor market dynamics, and the metropolis Hollywood is still dominated by a handful of majors, most of them direct descendants of the old studios that had surged to the fore in the 1920s. Over the last decade or so, seven majors have accounted consistently for over 80 per cent of the US motion-picture box office, namely, Metro-Goldwyn-Mayer, Paramount Pictures, Sony Pictures Entertainment, Twentieth-Century Fox, Universal Studios, Walt Disney Co., and Warner Brothers. These seven majors are allied together in the MPAA (Motion-Picture Association of America), a trade association that serves as their collective mouthpiece. Contemporary Hollywood also comprises masses of smaller independent production companies and a huge network of specialized input suppliers. Product formats, moreover, have become steadily more diversified with the passage of time, and the main outputs of Hollywood, in addition to motion pictures, now include television programs, videos, and commercials. The unions and guilds are active in all these segments of the industry. Cluster-based agglomeration economies and the local labor market Dense industrial districts, like Hollywood, typically thrive on the basis of the agglomeration economies (or localized increasing returns effects) that come into being when many different specialized but complementary producers locate in close proximity to one another. Agglomeration economies have various sources, but local labor markets are certainly one of the more important of them, not least because of their character as concentrated pools of workers living in neighborhoods accessible to places of employment (Kim, 1987; Scott, 2000). This geographic relationship ensures that commuting costs will be relatively low and that upward pressures on wages will be to that degree contained. Local labor markets are also repositories of diverse skills and aptitudes more or less corresponding to the array of specialized tasks that make up the entire production system. In addition, the mutual propinquity of home places and employment places in dense centers of production and work facilitates effective matching of individuals and jobs. The thick local tissue of jobs and residences makes it possible, moreover, to provide collective educational and training infrastructures on a significant scale. Equally, local labor markets are sites of socialization and habituation such that individuals are in various subtle ways – even in their social lives – pre-adapted to the experience of work in the production system. Local labor markets, too, are almost always shot through with networks of inter-personal relationships through which large quantities of useful information (about employers, job opportunities, conditions of work, and so on) constantly circulate. In this fashion, Hollywood and its surrounding communities constitute mutually reinforcing elements of a territorial system whose viability over time depends, in part, on the place-specific agglomeration economies that it generates.

Spatial and organizational patterns 241 The local labor market in time and space Aggregate employment in the motion-picture industry of Hollywood has expanded rapidly since the early 1980s. Figure 12.1 shows the number of workers in SIC 78 (Motion Pictures) in Los Angeles County on a month-bymonth basis from January 1983 to March 2002. At the start of this period, employment stood at about 65,000. It subsequently grew to a peak of 158,300 in October 1998. Since then employment has declined somewhat, and in March 2002 it was just above 130,000. The reasons for this fall over the last few years are probably related to a combination of employment decentralization (runaway production) and deteriorating conditions in the larger US economy. Employment patterns over time also exhibit strong short-term fluctuations, reflecting both random instabilities in the labor market and a persistent seasonal effect. Employment has a tendency to turn downward in spring and upward again in autumn as work expands in preparation for the winter season. The prominent decline in employment in SIC 78 over the summer of 1988, as indicated in Figure 12.1, reflects the strike by members of the Writers’ Guild between March and August of that year. Employment and residential locations are linked together by the daily journey to and from work. This relation is spatially expressed in a relatively centralized system of job locations, and a relatively dispersed pattern of residences, but where the outward extension of the latter is constrained by rising commuting costs as a function of increasing distance from workplaces. Something of this state of affairs is captured in Figures 12.2 and 12.3 where the residential locations of 165,000

Employment

145,000 125,000 105,000 85,000 65,000

02

01

00

99

98

97

96

94

94

93

92

91

90

89

88

87

86

85

84

Jan 83

45,000

Figure 12.1 Monthly employment in SIC 78 (Motion Pictures), Los Angeles County, January 1983 to March 2002 Source California Employment Development Department, http://www.edd.ca.gov/

242 Allen J. Scott

Pa

cif

ic

O

ce

an

Figure 12.2 Residential locations of members of IATSE Local 80 (Studio Grips and Crafts Service) in Southern California. The center of the Hollywood labor market (as defined in union and guild contracts) is represented by a cross marking the intersection of Beverly and La Cienega Boulevards, and the thirty-mile circle around this point is shown. The map is based on zip code data provided by Local 80

members of two representative Hollywood labor organizations are mapped out. Figure 12.2 shows the locations of the 2,934 members of IATSE (International Alliance of Theatrical Stage Employees) Local 80, or Studio Grips and Craft Service. This is a largely low-skilled group of workers whose jobs are mostly concerned with arranging production sets, props, and lighting. Figure 12.3 exhibits residential locations of members of the Writers’ Guild West. The Guild’s total membership is 7,727, though in contrast to IATSE Local 80 a significant proportion (16.5 per cent) lives outside Southern California. Screen writers are, of course, highly skilled professionals, but with widely varying incomes depending on ability and experience. Figures 12.2 and 12.3 are representative of the spatial distribution of the residences of workers in the Hollywood motion-picture industry generally. In both cases, workers are densely concentrated in the West Side of Los Angeles, the San Fernando Valley, and the Pasadena/San Gabriel Valley, though members of Local 80 are more prone to fan out into areas of suburban tract housing, just as they avoid the upscale Hollywood Hills where writers’ residences are more

Spatial and organizational patterns 243

Pa

cif

ic

O

ce

an

Figure 12.3 Residential locations of members of the Writers’ Guild in Southern California. The map is based on zip code data provided by the Guild

commonly found. Figures 12.2 and 12.3 also clearly bring out the strong geographic symbiosis between residential and employment locations, where the latter are represented schematically by a cross showing the intersection of Beverly and La Cienega Boulevards. This point marks the center of the thirty-mile circle conventionally used in collective-bargaining agreements to differentiate work locations where normal contract provisions prevail from locations where special allowances for travel are made. Unsurprisingly, given the difficulties that AfricanAmericans and other minorities face in obtaining jobs in the motion-picture industry, Figures 12.2 and 12.3, but especially the latter, are largely blank in and around the area of South Central Los Angeles. The entire local labor market is thus deeply but differentially inserted into the urban milieu of Southern California, which serves not only as a reservoir of labor but also as a sort of employment buffer, with its many casual job opportunities helping to keep new entrants to the labor market going until they are able to establish themselves in their preferred careers. The urban milieu also sustains the functions of socialization and habituation by reason of its embedded assets in the form of traditions, memories, cultures, and visual cues that filter into the consciousness of workers, and then feed into the whole creative thrust of Hollywood’s dream factories.

244 Allen J. Scott

Careers and collective order Occupational structures and worker interaction The Hollywood labor market can be described as a pyramid, with a few lavishly paid writers, producers, directors, actors, etc., at the top, and a bloated mass of less well-paid workers at the bottom. In practice, there are really two main overlapping pyramids, one representing manual, crafts, and technical workers, the other – which has many more tiers in the upper ranges – representing creative or talent workers. In the latter case, much of the base is comprised of individuals whose goal is to move into the above-the-line category, and who, in their asyet-undifferentiated labor-market performance, might be said to have equal a priori chances of success. As the competitive sorting process works itself out, with some individuals making their exit from the labor market altogether, and others rising into higher tiers, differentiation in terms of revealed ability and acquired reputation becomes increasingly evident. In the limit, this upward mobility progresses into stardom, with both its extraordinary financial rewards and the fetishization of personality that is one of its essential attributes. The labor market as a whole is structured by an elaborate system of occupational categories, much of it codified in collective bargaining arrangements. These categories reflect the extended divisions of labor that exist within the motion-picture industry, as represented, for example, by job categories such as gaffers, grips, set painters, costume designers, camera technicians, sound engineers, composers, actors, and so on. Different levels of skill within the same trade are also subject to conventionalized definitions. Thus, the Directors’ Guild, specifically identifies the following occupational distinctions within its membership: director, unit production manager, first assistant director, second assistant director (graded into key, second, and additional second assistant director), and technical co-ordinator. IATSE 700 (the Editors’ Guild) recognizes 26 different job categories in its contract. This formalization of occupational categories, and the rigid demarcation of job assignments that is one of its effects, helps to impose some order on a production process whose extremely fluid character might otherwise induce it to veer frequently out of control (Sydow and Staber, 2002). But it can also contribute to frustration among many types of workers as they contrast the tasks they are actually called upon to perform with their ultimate creative aspirations. The experience of certain kinds of writers on television shows, and their reduction to the status of, say, dialogue technicians, is a case in point (Pasquier and Chalvon-Demersay, 1993). Occupations on the manual, crafts, and technical side of the industry are differentiated from one another across an enormous range of attributes and skills. Most of these occupations are hourly rated, but some, such as costume designers, cinematographers, or film editors, are remunerated on a salary or commission basis. Even for the least skilled occupations, wages are high compared to wage rates generally in Southern California, and the sweatshop conditions that prevail in other industries, such as clothing and furniture, are virtually nowhere to be

Spatial and organizational patterns 245 found in the motion-picture industry. Cases can nevertheless be observed of firms on the fringes of the system (subcontractors in the set-construction segment, for example) that employ low-wage immigrant labor. The favorable wage conditions in the Hollywood labor market can be ascribed both to its tradition of craft skill and to its relatively high level of unionization. The longrun success of Hollywood entertainment products on international markets and the absence hitherto of serious competition from other production locales have also certainly played a part. The rising tide of runaway production to Canada and elsewhere, however, may yet begin to exert serious negative pressure on labor market conditions in Hollywood. Creative or talent occupations in the motion-picture industry are primarily (but not exclusively) represented by actors, directors, and writers, with producers forming an ambiguously situated group somewhere between creative and management workers. It is above all in the creative segment that persistent labor over-supply at the base of the job pyramid occurs. This phenomenon stems from the status of Hollywood as a magnet for migrants from all over the world with burning ambitions to work in the motion-picture industry, together with the numerous aspirants willing to make sacrifices in the short term for the sake of potentially massive monetary and psychological gains in the long term. In these respects, Los Angeles functions much like other centers of creative talent, such as London, Paris, or New York, all of which exert virtually irresistible attractions on neophyte artists seeking fulfillment in their chosen field (Kraft, 1994; Menger, 1993; Montgomery and Robinson, 1993). Individuals then filter up into higher ranges of the pyramid as they accumulate job credits, as their reputation expands, and as they become enculturated into the norms of the industry, including acquisition of the inter-personal skills that sustain effective team work (Batt et al., 2001; Faulkner, 1983; Jones, 1996). Success at one moment in time, however, is no guarantee of continued upward mobility. Weiss and Faulkner (1983) have demonstrated for the cases of directors, screenwriters, and producers that one screen credit is more often followed by anonymity than by more credits. Accordingly, the top echelons of the industry are dominated by a remarkably small number of super-productive individuals who account for the vast majority of successful projects (Jones, 1996). Each of these two main fractions of the labor force is pre-eminently given to freelance work. Workers at every level of skill and experience move between employers as projects come and go, and this creates vicissitudes that workers seek in part to manage by means of informal social organizations (cf. Ekinsmyth, 2002). These in turn rely heavily on what Ursell (2000) has called an ‘economy of favors’, i.e. a process in which workers build relationships of reciprocity with one another. The most obvious instance of this phenomenon can be found in the tangled inter-personal networks through which information in the local labor market circulates (Batt et al., 2001; Blair et al., 2001). In many segments of the industry, too, workers come together in small teams that migrate from project to project, gaining and losing members as they do so (Blair, 2001). Certain combinations of skilled workers (e.g. producer/composer or

246 Allen J. Scott director/actor pairs) evidently recur with some frequency. Faulkner and Anderson (1987) have observed that the individuals in any team are almost always equally matched in terms of experience and accomplishments. Relations of reciprocity are consolidated by the occasional sharing out of assignments when individuals have more work than they can personally handle at any given time (Faulkner, 1983). The local labor market in Hollywood, then, is a field of turbulent but structured social activity in which large numbers of individuals constantly confront the need for strategic planning of their careers. Once committed to a particular project, workers are mobilized in co-operative and often rather freely flowing teamwork efforts. In this manner, each individual’s creative labor makes an essential contribution to the final output (Elliott, 1972). In the industrialized world of Hollywood teamwork is the very foundation of the system of production, and, even in a personalized cinéma d’auteur, it is a critical element of the labor process as a whole (cf. Darré, 1986; Scott, 2000). Co-ordinating mechanism In addition to the informal networks that represent one kind of collective response to uncertainty and latent disorder in the local labor market, three complementary instruments of co-ordination need to be described. First, the proclivity for communication gaps to appear in local labor markets encourages the rise of intermediaries who trade on the consequent demand for useful information. Hollywood is rich in such intermediaries not only because there are many discontinuities throughout the system but also because the shortterm nature of many jobs means that both firms and workers constantly have to search out new relationships with one another. Agents, casting directors (who specialize in finding actors for roles), and talent managers are of major importance in this respect. Figure 12.4 displays their geographic locations and reveals their tendency to cluster at the core of Hollywood – a locational trait that enhances their accessibility to prospective employees and employers, thereby facilitating their role as network brokers. Size and successful past performance are the main measures of achievement for firms in this segment, and only a small elite of large intermediaries participates in the top deals executed in Hollywood (Bielby and Bielby, 1999). Many talent agencies are now seeking to extend their traditional functions by assembling composite packages (say, a concept, a writer, a production team, and a cast), which are then put out on offer to production companies. Second, above and beyond collective bargaining organizations, there is a proliferation of formal and informal associations in Hollywood providing workers with diverse types of backup support. The Academy of Motion Picture Arts and Sciences occupies a central place in this associational environment. The Academy’s annual Oscar awards ceremony is the most important occasion in North America for recognizing the achievements of elite workers in the industry, and is a public relations event with worldwide impact. There are scores of other

Spatial and organizational patterns 247

Agents Casting directors Managers

Figure 12.4 Locations of agents, casting directors, and managers in Los Angeles. Observe the clustering of firms in the central core of Hollywood, together with the alignment of firms along Ventura and Wilshire Boulevards Source of address data Hollywood Creative Directory, Agents and Managers (Los Angeles, 2002)

associations serving the needs of workers in the industry, including, for example, the Production Assistants Association, the Society of Motion Picture and TV Art Directors, Women in Film, the Stuntmen’s Association, the Visual Effects Society, and so on. Associations like these constitute forums in which problems of common interest can be discussed and acted upon, while providing useful information, contacts, mutual support, and training programs. Workers on the creative side of the industry are especially prone to participate in workrelated associations, and this again is probably an instance of the deployment of risk-reducing and opportunity-enhancing strategies as a way of countering the insecurities of the employment relation. Third, many different institutions and agencies throughout Southern California provide educational and job-training opportunities for workers (and would-be workers) in the motion-picture industry. The region is home to numerous colleges and universities, which in their turn are an important element of the agglomeration economies of Hollywood. Several local branches of the California State University system have professional programs focused on film, television, and media (at Northridge, Long Beach, Fullerton, and Los Angeles).

248 Allen J. Scott Other colleges and universities, such as Los Angeles City College, Chapman University, and Loyola Marymount University offer diplomas in film-related activities, as do the more specialized Los Angeles Film School and the American Film Institute. At the head of the list of local educational institutions providing industry-oriented programs is UCLA’s School of Theatre, Film, and Television and the University of Southern California’s School of Cinema-Television, both of which have alumni who have attained to the highest levels of accomplishment in acting, producing, directing, writing, and so forth. Also, a profusion of part-time training programs, workshops, and courses is constantly available throughout the region, on issues ranging from studio management to costume design or lighting techniques.

Unions and guilds The political upheavals that attended the formation of the unions and guilds before World War II continued unabated once the war was over. A bitter rivalry existed at this time between IATSE and the CSU (Conference of Studio Unions), the former having had a history of mob control, the latter being a leftleaning amalgam of skilled trades. The studios were strongly inclined in favor of IATSE, and the CSU rapidly faded away after a mass lockout in 1946 that greatly undermined its influence. Thus, from the late 1940s on, IATSE locals came to function as the principal unions for the manual, crafts, and technical workers in Hollywood (Horne, 2001). The immediate post-war years in Hollywood were also marked by red scares and black lists, and – from 1947 to 1954 – by the scrutiny of the House of UnAmerican Activities Committee, all of which had profound effects in disciplining the labor force as well as in regulating the political content of films (Ceplair and Englund, 1983; Prindle, 1988). In spite of (or perhaps because of) these clashes in the post-war period a remarkably stable and resilient system of unions and guilds emerged over the 1950s and 1960s, and it has played a critical role in maintaining a smoothly operating system of labor relations and collective bargaining in Hollywood down to the present. A list of unions and guilds in Hollywood today is presented in Table 12.1. The organizations identified are all collective bargaining units certified by the National Labor Relations Board. Table 12.1 provides information on total membership (as of March 2002) in the three guilds, twenty-three IATSE locals, and two other union locals most closely tied to the motion-picture industry. A number of other organizations such as the American Federation of Television and Radio Artists, the National Association of Broadcast Employees and Technicians, and the American Federation of Musicians are also bona fide collective bargaining agencies, but their membership extends far beyond workers in the motion-picture industry, and for this reason no reference is made to them in Table 12.1. The organizations listed in the table have an aggregate membership of 162,955, a suspiciously high number compared to the 131,800 workers actually employed in SIC 78 in Los Angeles County according to the California Employment Development in March 2002. The discrepancy between the two

Spatial and organizational patterns 249 Table 12.1 A comprehensive list of union locals and guilds with collective-bargaining authority in contemporary Hollywood. Membership numbers refer to the year 2002 Union or guild Professional guilds: Directors’ Guild Screen Actors’ Guild Writers’ Guild West IATSE locals: 44 Affiliated Property Craftspersons 78 Plumbers 80 Motion-Picture Studio Grips and Crafts Service 600 International Cinematographers 683 Film Technicians 695 Production Sound Technicians, TV Engineers, and Video Assistant Technicians 700 Motion-Picture Editors 705 Motion-Picture Costumers 706 Makeup Artists and Hair Stylists 724 Studio Utility Employees 728 Studio Electrical Lighting Technicians 729 Motion-Picture Set Painters 755 Plasterers, Modelers, and Sculptors 767 First Aid Employees 790 Studio Art Craftsmen 816 Scenic, Title and Graphic Artists 818 Publicists 839 Motion-Picture Screen Cartoonists and Affiliated Optical Electronic and Graphic Arts 847 Set Designers and Model Makers 871 Script Supervisors and Continuity Coordinators 876 Motion-Picture and Television Art Directors 884 Studio Teachers and Welfare Workers 892 Costume Designers Other locals: International Brotherhood of Electrical Workers, Local 40 Teamsters Local 399, Studio Transportation Drivers Total union and guild membership:

Total membership 12,420 98,000 7,727 6,755 109 2,934 4,643 1,725 2,224 1,983 1,625 1,454 2,553 1,357 345 359 186 227 310 1,800 233 1,738 723 173 352 672 5,469 162,995

figures can be reconciled in part by the fact that some of the organizations designated in Table 12.1 represent workers from all over the United States, and in part by the circumstance that only a fraction of union or guild members is employed at any one time. The largest organization designated in the table, the Screen Actors’ Guild, has 98,000 members, but only about two-thirds of them live in Southern California, and only about 25 per cent of these actually have jobs at any one time. Members of IATSE Local 80 occupy a relatively stable

250 Allen J. Scott niche in the local labor market, but even they face an average unemployment rate of 30 per cent. While the majority of talent workers in the industry are members of a guild, only about half of all manual, crafts, and technical workers belong to a union (cf. Paul and Kleingartner, 1996). Indeed, union (but not guild) density is widely thought to be decreasing at the present time, owing in part to the great expansion of small independent production companies able to sidestep labor-organizing activities in lower-skilled occupations, as well as to rising levels of runaway production. The main business of the unions and guilds is collective bargaining, especially with the AMPTP (Alliance of Motion Picture and Television Producers). The latter acts as the MPAA’s instrument in all collective bargaining situations (Counter, 1992; Gray and Seeber, 1996; Wasko, 1998). Contracts are negotiated by each union or guild separately with the AMPTP on a three-year cycle, with pattern bargaining commonly occurring. As a rule, contracts signed between the guilds and the AMPTP are co-signed by large numbers of independents; but IATSE contracts tend to attract fewer independent signatories, especially in cases involving the less skilled locals. Contracts negotiated between the unions and guilds with the AMPTP contain three critical clauses in regard to remuneration. These provide for (1) minimum pay scales for specified occupational categories and levels of experience, (2) personal services contracts, which is industry jargon for individually negotiated above-scale wage or salary payments, and (3) the administration of residuals, a form of compensation based on secondary runs or releases of any given entertainment product (Kleingartner, 2001; Paul and Kleingartner, 1996). In the case of IATSE locals, residuals are paid into pension funds; in the case of the talent guilds, they are appropriated by designated individuals. Union and guild contracts also make stipulations about health and pension payments and vacation benefits, and specify the rules governing conditions of work performed beyond the thirty-mile circle. The IATSE locals pool their welfare plans in the centrally administered Motion-Picture Industry Pension and Health Plans, while the guilds maintain individual plans. In either case, the system that has emerged in practice has the advantage that individuals’ benefits packages are not tied to any single employer but are fully portable from firm to firm. Thus, the Hollywood unions and guilds play a role somewhat analogous to that played in France by the government-sponsored Intermittence du Spectacle, which provides unemployment compensation and other benefits to workers in the entertainment industry (Rannou and Vari, 1996; Scott, 2000). Further functions of the unions and guilds are (1) the codification and regulation of professional categories, (2) accreditation of members’ work experiences, and (3) the provision of educational, labortraining, and other qualification-enhancing services. The roster system that operated in at least some IATSE locals in the 1950s according to Christopherson and Storper (1989) no longer exists, though the qualifications lists maintained by some organizations have a somewhat similar sorting effect. The unions and guilds, then, are a source of significant benefits, above all by correcting market failures that would otherwise occur in the local labor market,

Spatial and organizational patterns 251 and by reducing the potential for exacerbated inter-worker competition over jobs and pay. They lower the labor-market risks and raise the living standards of their members, which no doubt also helps to maintain the overall supply of workers to the industry. The continuing success of the talent guilds in holding and expanding their membership reflects the real advantages that they offer, above all through their capacity to negotiate attractive wage and salary scales and residuals payments. The manual, crafts, and technical workers’ unions also continue to be a significant force in Hollywood, though the fact that union pay scales (unlike those of the guilds) do not seem to differ much from market wage rates1 is perhaps a further sign that their power may be waning. The strength and influence of the guilds is especially evident in the circumstance that they are more apt to engage in industrial action than the politically weaker IATSE locals. The Screen Actors’ Guild and the Writers’ Guild lead the way in this respect. The Directors’ Guild is more circumspect, having gone on strike only once in its existence (for a few hours in 1987). The latter state of affairs is probably ascribable to the close association that is usually to be found between directors and production managers on film projects. The most recent strike of the Actors’ Guild was from May to October 2000 as a result of a breakdown in negotiations over the guild’s commercials contract. The Writers’ Guild engaged in a lengthy strike from March to August 1988, and threatened to strike again in spring 2001 over pay structures. This threatened strike was averted only by a last-minute agreement with the studios. The last major work stoppage by an IATSE local was in the late summer of 1982 when IATSE 839 called an unsuccessful strike over the subcontracting of animated-film work to firms in Asia (cf. Scott, 1988; Lent, 1998).

A concluding comment The local labor market in Hollywood has evolved through many mutations over the last several decades. Despite a number of built-in degrees of inertia, this system has proved itself to be remarkably capable of adaptation to changing circumstances. Today, a further wave of change is in progress as various new competitive pressures make themselves felt. The destiny of Hollywood as a viable industrial district hinges in particular on continued fine-tuning of the local institutional framework so as to sustain socio-spatial order of the labor market and the continued inflow of new talent in the face of the exigencies and vagaries of the working environment. At the same time, the flexible labor market arrangements that emerged on a large scale in Hollywood at an unusually early stage offer a useful lesson as similar arrangements diffuse ever more insistently to other types of industrial clusters around the world. Thus, flexibilization does not always result in an endless downward spiral for workers, provided that appropriate social safeguards can be constructed to combat its most deleterious effects. Contrariwise, strong labor organizations do not necessarily mean that business is destined to continual disruption. From all that has gone before, it is evident that the unions and guilds of Hollywood perform critical functions in

252 Allen J. Scott bringing forms of order to the labor market that are ultimately beneficial to firms as well as workers. Not least among these functions is the part that unions and guilds play in regulating labor market insecurities, in promoting the long-term commitment of their members to a vocational or quasi-vocational sense of their work, and hence also to the maintenance of the creativity and innovation that have always been one of the hallmarks of Hollywood. For several decades, and notwithstanding numerous vicissitudes, this delicate political balance has served Hollywood very well indeed. The implications for other high-performance industrial clusters reside not so much in direct imitation of the institutional arrangements of Hollywood as in concerted efforts – in the context of each individual cluster’s own peculiar history and culture – to ensure that the naked forces of inter-firm competition do not begin to undermine complex local ecologies of production, work, and social life. This proposition implies, among other things, that local labor market operations need to be underpinned by forms of collective order and personal reward sufficient to maintain the highest possible levels of human and social capital. According to Monitor (1999), as many as 20,000 jobs were lost in Hollywood in the late 1990s as a result of runaway production, and the hemorrhage evidently continues apace (EIDC, 2001). This state of affairs represents one of the main challenges facing Hollywood at the present time, and it will undoubtedly lead to painful local readjustments over the next few years (Kleingartner and Raymond, 1988). Conversely, the very agglomeration economies built into the production system and its associated local labor market suggest that there are probably limits as to how much further this process can go, certainly in regard to more skilled types of workers. A measured prediction, then, is that continued decentralization of jobs will probably not undermine the core functions of Hollywood in any irrevocable fashion, even if it does generate increasing employment at a number of satellite production locations. Loss of influence among the IATSE locals in Hollywood will doubtless persist, but the professional guilds will assuredly continue to play a large role as Hollywood moves into the next main phase of its development.

Note 1

This judgement is based on extensive interviews with officials and individual members of different unions and guilds.

References Amman, J. (1996). The transformation of industrial relations in the motion-picture and television industries: Craft and production. In L. S. Gray and R. L. Seeber (eds), Under the Stars (pp. 113–55). Ithaca: Cornell University Press. Batt, R., Christopherson, S., Rightor, N., and Jaarsveld, D. V. (2001). Net Working: Work Patterns and Workforce Policies for the New Media Industry. Washington, DC.: Economic Policy Institute. Bielby, W. T., and Bielby, D. D. (1999). Organizational mediation of project-based

Spatial and organizational patterns 253 labor markets: Talent agencies and the careers of screenwriters. American Sociological Review, 64, 64–85. Blair, H. (2001). You’re only as good as your last job: The labor process and labor market in the British film industry. Work, Employment, Society, 15, 149–69. Blair, H., Grey, S., and Randle, K. (2001). Working in film: Employment in a projectbased industry. Personnel Review, 30, 170–85. Bordwell, D., Staiger, J., and Thompson, K. (1985). The Classical Hollywood Cinema. New York: Columbia University Press. Cantor, M. G., and Peters, A. K. (1980). The employment and unemployment of screen actors in the United States. In W. S. Hendon, J. L. Shanahan, and A. J. MacDonald (eds), Economic Policy for the Arts (pp. 210–18). Cambridge, MA: Abt Books. Ceplair, L., and Englund, S. (1983). The Inquisition in Hollywood. Berkeley and Los Angeles: University of California Press. Christopherson, S., and Storper, M. (1989). The effects of flexible specialization on industrial politics and the labor market: The motion-picture industry. Industrial and Labor Relations Review, 42, 331–47. Counter, N. (1992). New collective bargaining strategies for the 1990s: Lessons from the motion-picture industry. Industrial Relations Research Association: Proceedings of the 44th Annual Meeting, 32–8. Darré, Y. (1986). Les créateurs dans la division du travail: le cas du cinéma d’auteur. In R. Moulin (ed.), Sociologie de l’art. Paris: La Documentation Française. Davis, M. (1997). Sunshine and the open shop: Ford and Darwin in 1920s Los Angeles. Antipode, 29, 356–82. EIDC (2001). MOWs – A Three-Year Study. Los Angeles: Entertainment Industry Development Corporation. Ekinsmyth, C. (2002). Project organization, embeddedness and risk in magazine publishing. Regional Studies, 36, 229–43. Elliott, P. (1972). The Making of a Television Series. London: Constable. Faulkner, R. R. (1983). Music on Demand. New Brunswick, NJ: Transaction Books. Faulkner, R. R., and Anderson, A. B. (1987). Short-term projects and emergent careers: Evidence from Hollywood. American Journal of Sociology, 92, 879–909. Gray, L. S., and Seeber, R. L. (1996). The industry and the unions: An overview. In L. S. Gray and R. L. Seeber (eds), Under the Stars (pp. 15–49). Ithaca: Cornell University Press. Horne, G. (2001). Class Struggle in Hollywood, 1930–1950. Austin: University of Texas Press. Jones, C. (1996). Careers in project networks: The case of the film industry. In M. B. Arthur and D. M. Rousseau (eds), The Boundaryless Career (pp. 58–75). Oxford: Oxford University Press. Kim, S. (1987). Diversity in urban labor markets and agglomeration economies. Papers of the Regional Science Association, 62, 57–70. Kindem, G. (1982). Hollywood’s movie star system: An historical overview. In G. Kindem (ed.), The American Movie Industry (pp. 79–93). Carbondale: University of Southern Illinois Press. Kleingartner, A. (2001). Collective bargaining: Hollywood style. New Labor Forum, Fall/Winter, 113–21. Kleingartner, A., and Raymond, A. (1988). Hollywood goes international: Implications for labor relations. Los Angeles: University of California, Institute of Industrial Relations, Working Paper No. 156.

254 Allen J. Scott Kraft, J. P. (1994). Musicians in Hollywood: Work and technological change in entertainment industries. Technology and Culture, 35, 289–314. Lent, J. A. (1998). The animation industry and its offshore factories. In G. Sussman and J. A. Lent (eds), Global Productions (pp. 239–54). Cresskill, NJ: Hampton Press. Lovell, H., and Carter, T. (1955). Collective Bargaining in the Motion-Picture Industry. University of California, Berkeley: Institute of Industrial Relations. Menger, P. M. (1991). Marché du travail artistique et socialisation du risque: le cas du spectacle. Revue Française de Sociologie, 32, 61–74. Menger, P. M. (1993). L’hégémonie parisienne: économie et politique de la gravitation artistique. Annales: Economies, Sociétés, Civilisations, No. 6, 1565–600. Monitor (1999). US Runaway Film and Television Production Study Report. Santa Monica: Monitor Company. Montgomery, S. S., and Robinson, M. D. (1993). Visual artists in New York: What’s special about person and place? Journal of Cultural Economics, 17, 17–39. Nielson, M., and Mailes, G. (1995). Hollywood’s Other Blacklist. London: British Film Institute. Pasquier, D., and Chalvon-Demersay, S. (1993). Les mines de sel: auteurs et scénaristes de télévision. Sociologie du Travail, 4, 409–30. Paul, A., and Kleingartner, A. (1996). The transformation of industrial relations in the motion-picture and television industries. In L. S. Gray and R. L. Seeber (eds), Under the Stars (pp. 156–80). Ithaca: Cornell University Press. Prindle, D. F. (1988). The Politics of Glamour. Madison: University of Wisconsin Press. Rannou, J., and Vari, S. (1996). Les Itinéraires d’emploi des cadres, techniciens, et ouvriers intermittents de l’audiovisuel et des spectacles. Paris: Ministère de la Culture, Département des Etudes et de la Prospective. Ross, M. (1941). Stars and Strikes. New York: Columbia University Press. Ross, M. (1947). Labor relations in Hollywood. Annals of the American Academy of Political and Social Science, 254, 58–64. Scott, A. J. (1988) Metropolis. Berkeley: University of California Press. Scott, A. J. (2000). The Cultural Economy of Cities. London: Sage. Scott, A. J. (2002). A new map of Hollywood: The production and distribution of American motion pictures. Regional Studies, 36, 957–75. Storper, M., and Christopherson, S. (1987). Flexible specialization and regional industrial agglomerations: The case of the US motion-picture industry. Annals of the Association of American Geographers, 77, 260–82. Sydow, J., and Staber, U. (2002). The institutional embeddedness of project networks: The case of content production in German television. Regional Studies, 36, 215–27. Ursell, G. (2000). Television production: Issues of exploitation, commodification and subjectivity in UK television markets. Media, Culture and Society, 22, 805–25. Wasko, J. (1998). Challenges to Hollywood’s labor force in the 1980s. In G. Sussman and J. A. Lent (eds), Global Productions (pp. 173–90). Cresskill, NJ: Hampton Press. Weiss, P. R., and Faulkner, R. R. (1983). Credits and craft production: Freelance social organization in the Hollywood film industry. Symbolic Interaction, 6, 111–23.

13 Cluster and hinterland When is a proactive cluster policy appropriate? G. M. Peter Swann

Introduction While research on agglomeration has a long and distinguished history, Porter’s (1990) discussion of clusters has been very influential in shaping policy for competitiveness and innovation.1 One important reason for this is that research has shown that clustering can promote growth, performance and innovation in companies located in those clusters. Despite this, there are several reasons why a cluster policy erected on this observation may be problematic. First, these beneficial effects on growth and innovation are not automatic – hence our emphasis above on the word can. Second, where these beneficial effects exist, entrants will be attracted anyway, and it is not clear that governments need do more to encourage entry. Third, these benign effects can be offset in large clusters by several costs of clustering (such as congestion). Fourth, this focus on the benefits enjoyed within the cluster pays no attention to the (often adverse) implications for the hinterland.2 The aim of this chapter is to explore these issues, and hence to ask when it is appropriate for governments to have a proactive cluster policy. The structure of the chapter is as follows. The next section notes that the term ‘cluster’ has taken on several different meanings, and that some cluster policy is flawed because it does not adequately recognize these distinctions. Then we draw an interesting parallel between the economics of clusters and the economics of networks. The latter has identified three distinct theories (or ‘laws’) that indicate how the value of a network may relate to its size and composition. We then address the question of whether clusters will tend to be too small or too large, and find that the answer depends on these distinctions. We can build on this to explore the dynamics of entry into clusters, and why the ultimate entry may be excessive from the perspective of social welfare. So far, the only costs of clustering considered are those (e.g. congestion costs) found within the cluster. The next two sections then describe two other types of costs that can arise from clustering. One arises because a trajectory of specialization and clustering leads to an increased (derived) demand for transportation, which in turn creates congestion on transportation networks. The other arises because any activity that seeks further to develop the strongest

256 G. M. Peter Swann clusters will do so at the cost of the hinterland. In the light of this we take a fresh look at cluster policies, and tries to identify which of these find a clear economic rationale and which do not.

Clustering: a spectrum of interpretations As Martin and Sunley (2003) have shown, use of the term ‘cluster’ is not standardized, and different authors mean different things. The reader might respond, ‘When was it ever otherwise?’ In any complex and emerging field of social and economic research, non-standardization in research terms is always an issue, and it is common to find that one term is used to cover a wide variety of different phenomena. To some degree, this is inevitable and I doubt scholars studying clusters would meet with much sympathy in business and policy circles if we were to say, ‘Don’t do anything about clusters until we have worked out our etymology!’ Nevertheless, this non-standardization does matter because some cluster policy seems to rest on the (implicit) assumption that all clusters entail the richest forms of cluster activity, when in fact many of them may show only much weaker forms of cluster activity. If people think a cluster-supportive policy will create another mythical ‘Silicon Valley’, then they may be pretty supportive of all kinds of policy directed at such a goal. But if in reality cluster-supportive policies just create more congestion in the South East of England and on our transport networks, with none of the flavour of Silicon Valley, then I suspect that far fewer will be supportive. We shall see that these different attitudes are well placed because the economic rationale for a proactive cluster policy is much stronger in the first example than in the second. In Figure 13.1 I try to summarize in a very simple way some of the different interpretations of the cluster concept. It is a simple one-dimensional spectrum of interpretations, from rich to shallow. The first items may be a necessary condition for the last items in the list, but they are not sufficient. Now some will say that the things at the top of my list are not sufficient evidence of clusters: a cluster is only a cluster when it has the things at the bottom of the list. Actually, I am not too uncomfortable with the diversity of interpretations as such. But I am concerned that some people speak of clusters as if one is going to get the whole spectrum all at the same time, when in reality one may only get the top end. Many quantitative (economic or econometric) studies tend to use a fairly shallow definition of cluster (towards the top of my spectrum) and find evidence that there are such cluster effects. Many detailed qualitative studies tend to work with a fairly rich definition of clusters and find little or no evidence of such rich cluster effects. At first sight this may appear to be a puzzle, but it is not necessarily a puzzle. The econometric studies tend to find indirect evidence: companies perform better in clusters (using several measures of performance) but we don’t quite know how this effect works. Swann et al. (1998) and Beaudry and Swann (2001)

When is a proactive cluster policy appropriate? 257 Co-location

Shallow

Easy to measure

Rich

Hard to measure

Co-location and technological proximity Input/output table complementarities Co-location and superior performance Marshallian externalities Network firms Labour mobility Explicit collaboration Informal knowledge spillovers

Figure 13.1 Different interpretations of the cluster concept

summarize some of this evidence.3 The qualitative studies find that it is usually not the mythical Silicon Valley phenomena of informal knowledge spillovers. Then the explanation must lie elsewhere. For example, McCann and Arita (2002) have argued that, in the Japanese model, clusters are effectively ‘sealed’. They argue that the interactions between agencies and research institutes are very limited, except between an institute and its associated department at the university (many people have joint positions). It is just a really nice place to live, so firms are happy to invest there to get a good quality labour supply. They argue that information exchanges depend primarily on these labour hysteresis issues rather than informal inter-firm spillovers as such. Moreover, on the basis of numerous visits to Silicon Valley, McCann and Arita (2000) argue that this is now the norm in Silicon Valley too.4 Of course, this one-dimensional spectrum is itself too simplistic. We really need to classify the use of the term ‘clusters’ on several other dimensions. In addition to this ‘shallow–rich’ spectrum, some use the term even when they have only indirect evidence (e.g. firms in clusters have superior performance’) while some reserve the term for cases where they have direct evidence (e.g. evidence of informal knowledge-sharing interactions). There are several other useful dimensions, moreover: economies of agglomeration versus economies of urbanisation, demand-side clustering benefits versus supply-side clustering benefits, and so on. I shall not however attempt a comprehensive list here, because it is not necessary for my immediate purpose.

258 G. M. Peter Swann

Cluster effects and network effects At the start of this chapter we said that our aim was to ask whether governments need to do anything to foster the growth of clusters. I shall argue that our answer must depend on the form of the cluster effects operating within the cluster: that is, where on the above spectrum from shallow to rich. To see this, it is helpful to develop an interesting linkage between the economics of clusters and the economics of networks. This brings out very clearly why different interpretations of cluster will have different economic characteristics. From this perspective the cluster is a special form of network that requires geographical proximity. The economics of networks has provided three theories or ‘laws’ which summarize the benefit to be gained from a network as a function of the size of that network. I have put the word ‘laws’ in quotation marks here, because recent discussion has cast much doubt on the empirical validity of these. But nevertheless they are interesting limiting cases, even if not really robust enough to earn the rank of ‘law’. Each of these three different laws captures some different characteristic of a network. Sarnoff ’s ‘Law’ argues that the aggregate value enjoyed by a broadcast network (or other one-way network) is proportional to the number of viewers (n). This is valid if all viewers attach equal value to the broadcast, so that total value is proportional to n. It would also be valid on an expected basis if we add viewers to our audience in a random order, so that the expected aggregate value is proportional to n. This is probably the most reliable of the three ‘laws’ if perhaps the least spectacular. Metcalfe’s ‘Law’ (named after the founder of the Ethernet, Robert Metcalfe) argues that the aggregate value enjoyed by a telephone network (or other twoway communication network) is proportional to the square of the number of viewers (n2). The rationale for this is that aggregate value depends on the number of pair-wise conversations, and this increases roughly with the square of network size. The number of possible pair-wise conversations in a network (neglecting the issue of whether A calls B or vice versa) is

n

C2 =

(

)

n n +1 2

For large n, that is roughly proportional to n 2. Of course, it is unlikely that each network member attaches equal value to each possible call, but this squarelaw relationship can be retrieved in some special cases (Swann, 2002). Third comes Reed’s ‘Law’, which argues that the aggregate value of a groupforming network increases exponentially with the size of the network. The total number of sets of size r that can be formed from a population of size n is nCr. We can show that the sum of this for r = 0 to n is equal to 2n. Excluding the empty set and trivial (one-member) sets, the number of non-trivial sets is 2n – n – 1. For large n, this is exponentially increasing in n. This is the most spectacular of the ‘laws’ and probably the least common, because it is most unlikely that each individual will value all possible groups.

When is a proactive cluster policy appropriate? 259 Table 13.1 Approximate relationship between network value and network size, n Value to individual network member as a function of n

Aggregate value to all network members as a function of n

Sarnoff’s Law

constant

n

Metcalfe’s Law

n

n2

Reed’s Law

2n n

2n

These three ‘laws’ are summarized in Table 13.1. If we apply these network results to clusters, we see why the nature of cluster effects has such an impact on the relationship between cluster size and economic performance. If cluster effects are mostly of an infrastructure kind then it seems that Sarnoff ’s ‘Law’ is the most relevant. Individual members of the network do not benefit as the network grows. Only the new entrants benefit. The aggregate (gross) benefits from the cluster are at best proportional to cluster size. Moreover if the order of entry into the cluster is not random, but early pioneers derive greatest benefit while later entrants derive least benefit, then aggregate benefit increases less than proportionately with cluster size. On the other hand, if the cluster effects are of the richest ‘Silicon Valley’ type, including informal knowledge exchange and the easy formation of creative groups, then Metcalfe’s ‘Law’ (or possibly even Reed’s ‘Law’) could be relevant. In this case, established members of the network do benefit as the network grows, so that aggregate benefit increases more than proportionately with cluster size. In terms of our simple spectrum above, it seems clear that Metcalfe’s ‘Law’ relationships would apply only to the ‘rich’ effects at the bottom end of our list. For the items at the top of the list, Sarnoff ’s ‘Law’ is more relevant and the aggregate (gross) benefits from the cluster are at best proportional to cluster size. Such simple models only give the beginning of the story, moreover. There is nothing in these three ‘laws’ about diminishing returns to cluster size, and in practice diminishing returns are a strong possibility. Nevertheless, we shall see that this distinction is useful in assessing whether we find too little clustering, or perhaps too much.

Clusters too small? Or too large? Clusters may be a ‘good thing’ but they are not unambiguously a ‘good thing’. The individual firm faces advantages and disadvantages from locating in a cluster. The advantages are what we might call cluster or network effects, while the disadvantages include congestion costs from locating in densely populated clusters. Moreover, even if clusters are on balance a good thing for a particular entrant, an increase in the size of a cluster is not necessarily a good thing for existing incumbents, or indeed for third parties. More generally, the net benefits

260 G. M. Peter Swann from increasing the size of a cluster will depend on whose perspective we are taking. If companies benefit from location in clusters, then they are liable to want to locate in clusters. That seems a reasonable conclusion. But do we jump from this to a policy prescription that companies should be organized into clusters or that governments should aim to ‘grow clusters’? I would say no, there can be no such general presumption. A standard economic perspective on such questions is to ask: what are the externalities from cluster entry? If entrants convey a positive externality to other cluster stakeholders, then it is possible that there may be too little entry – because entry stops when the benefits to the entrant fall to zero, even if further entry would still be of benefit to other stakeholders. In that sense incentives for entry would be too small for the public good. But if entrants convey a negative externality to other cluster stakeholders, then it is possible that there may be too much entry – because entry continues so long as the entrant benefits from entry, and even if further entry imposes costs on other stakeholders. In that sense incentives for entry would be too large for the public good. In practice, are the externalities from entry more likely to be positive or negative? This depends on the form of the cluster effects – Sarnoff ’s ‘Law’ or Metcalfe’s ‘Law’. If these effects follow Sarnoff ’s ‘Law’, then new entry does not benefit the incumbent or third party. And indeed to the extent that entry increases congestion costs, then entry probably imposes negative externalities on these incumbents and third parties. But if these effects follow Metcalfe’s ‘Law’ (or Reed’s ‘Law’), then new entry does benefit the incumbent or third party. And unless these benefits are completely offset by increased congestion costs, then entry probably delivers positive externalities. In answer to our question, ‘too much or too little clustering?’, we can conclude that it depends on whether the cluster effects follow Sarnoff ’s ‘Law’ or Metcalfe’s ‘Law’. If Sarnoff ’s ‘Law’, then there may be too much clustering; if Metcalfe’s law, then there may be too little clustering. Governments may be called on to have a proactive policy if the incentives for firms to enter clusters are too small for the public good, so that too little clustering takes place. But if the incentives for firms to enter clusters are too high for the public good, so that too much clustering takes place, then it is not clear why governments should have such a proactive policy. We return to this below.

The dynamics of excess entry The discussion in the last section took a very simple static perspective. In fact the dynamics of entry and its desirability is more complex. Table 13.2 shows a typical pattern of benefits and costs to different cluster stakeholders at different stages of cluster development. This is typical but not universal. A general observation about the growth of clusters which applies in many contexts is that entrants gain more than the social net benefit they contribute. Or, to put it another way, the necessity to cluster (for an individual entrant) is a private

When is a proactive cluster policy appropriate? 261 Table 13.2 Effect of new entry on cluster stakeholders Effect on: Effect of:

Next entrant

All incumbents

Third parties

Consumers

Pioneers

+

+

+

+

Third-party-optimum entrant

+

+

0

+

Incumbent-optimum entrant

+

0

–

+

Post-incumbent-optimum entrant

+

–

–

+

Final entrant

0/–

–

–

+

necessity rather than a social necessity. Entrants may soak up positive externalities but can create negative externalities. Pioneer entrants may benefit everyone. For third parties in the cluster (i.e. those in other businesses who do not benefit directly from the growth of the cluster), there is probably an optimum rate of entry beyond which the congestion costs imposed exceed any benefits. The entrant who optimizes the size of the cluster from the point of view of third parties makes a zero marginal contribution to third parties. But that entrant still generates positive externalities to incumbents and to subsequent entrants, so these latter stakeholders both wish to see further entry. For incumbents in the cluster (i.e. those in the industry around which the cluster is focused) the optimum rate of entry may be somewhat higher if there is any manifestation of Metcalfe’s ‘Law’. The entrant who optimizes the size of the cluster from the point of view of incumbents makes a zero marginal contribution to incumbents, but a negative contribution to third parties. However, this entrant continues to create positive externalities to subsequent entrants. So while, from the perspective of incumbents, there is no desire to grow the cluster further, there is still a desire on the part of further entrants to join the cluster. This process continues until we meet the ‘final entrant’. This is the one that finds just enough net positive externality in the cluster to find entry attractive, but whose entry drives the net positive externality down to zero – thus making further entry unattractive. It is worth reflecting on the state of the cluster when this ‘final entrant’ has entered. This final entrant (and several others before) have imposed a negative net externality on incumbents, and an even greater one on third parties. The cluster has grown too big for most stakeholders – excluding the final entrant and consumers.5 In this example, the outcome is that with free and unfettered entry, clusters will attract a degree of entry that is beyond the wishes of most stakeholders in the cluster – except the last entrant and consumers. In some countries, of course, entry to clusters is not unlimited. The interests of incumbents and third parties demand that entry is stopped before the cluster

262 G. M. Peter Swann reaches the ‘final entrant’. Some science and business parks may apply such constraints. Some episodes of cluster policy – for example the Science City approach in Japan (Grayson, 1993) – have been directed at ‘cooling down’ those clusters that have become ‘overheated’. By contrast, some aspects of UK cluster policy – or that collection of policy measures that impinge on clusters – seem to be directed at removing barriers to further entry. We shall return below to ask if such a proactive policy is appropriate.

Clustering and transportation So far the only costs of clustering we have considered are the costs that accumulate within the congested cluster. However, we shall also see that the emergence of specialized clusters can impose other costs. In this section and the next, we consider two such: congestion costs (this section) and hinterland costs (next section). The growth of specialized clusters can lead to an equally rapid (or possibly faster) growth in demand for transportation between clusters. That either means congestion, or ever increasing investment in transportation infrastructure to overcome congestion. Why should this be? The easiest way to understand this is to open up a personal computer and look at the origins of the various components. Table 13.3 below gives an indication of what we find.6 The components are highly standardized and come from a wide variety of countries. Moreover, PC manufacture has become highly reducible and is split into many steps. But the process is not infinitely reducible, and there are strong economies of scale and agglomeration in the production of each component. As a result, what we find is clusters specialising in the production of particular components, and these are dispersed across many countries. In short, the transportation involved in assembling a computer as above consists of a very large number of component miles. If we were to look at the earlier PCs, in the 1980s, we would find that the count of component miles was much smaller since far more of the components were produced in the USA. This is a good example of a more general phenomenon. The continuing process of specialization, standardization and intra-industry trade increases the count of component miles. In short, the emergence of specialized clusters and inter-cluster trade can lead to an equally rapid (or possibly faster) growth in demand for transportation between clusters. That means either congestion, or ever increasing investment in transportation infrastructure to overcome congestion. The following very simple model helps us to analyse these effects. Imagine that an economy is made up of dispersed towns or clusters at equal distance from each other. For simplicity imagine that each cluster is the same size. Then we can relate the total amount of transportation required to the degree of specialization in the economy. Suppose that there are a large number of industrial sectors (N), and for simplicity let these be all of the same size. Suppose also that a collection of

When is a proactive cluster policy appropriate? 263 Table 13.3 Origin of components for typical personal computer (2003) Brand: Final assembly and dispatch: Main box: Chips on motherboard: Battery on motherboard: Power supply: CD ROM drive: CD-R (consumables): Hard disk drive: 3.5 in. floppy disk drive: Modem card: Graphics card: Specialist video card: Monitor: Keyboard: Mouse: Child’s mouse: Loudspeakers: Microphone: Inkjet printer: Zip drive: Scanner: Webcam: Power supplies (peripherals): Manuals: Environmental certification:

USA Ireland Ireland USA, Korea, Taiwan, Philippines Philippines China China (assembled from Japanese Parts) Germany Singapore Philippines Netherlands (chips from USA, Korea, Taiwan) China (chips from USA, Korea, Taiwan) USA UK (origin of components?) Mexico Mexico Taiwan Malaysia Mexico Spain Malaysia Taiwan China Taiwan, China, Malaysia, Mexico Scotland, Ireland, Wales, Germany Sweden

Source: Swann (2006)

clusters (n) is group-sufficient, in the sense that those n clusters cover between them all N industrial sectors. Suppose moreover, that the N industrial sectors are distributed equally amongst the n clusters. Then these n clusters trade amongst themselves. For any one cluster, a small fraction (1/n) of its production is kept for ‘domestic’ consumption, while the rest (1 – 1/n) is ‘exported’ to the n–1 other clusters in the group. With this basic model structure, we can calculate the relationship between total transportation and n. It depends on the detailed structure of the transportation network. Here we work with two examples: the lattice/grid network and an orbital network.7 Suppose that a group-sufficient group of size n self-select so that they are in a contiguous block in the lattice. In that way, total transportation is minimized for any given n. Then we can calculate two relationships between T (the total amount of transportation) and n (the size of the group-sufficient group). If n is 1, then each cluster is self-sufficient, and so transportation requirements are limited

264 G. M. Peter Swann

Figure 13.2 The lattice/grid transport network

Figure 13.3 The orbital transport network

to those inside the cluster. If n is small, then a small group of clusters cover all industrial sectors, so we can say that they are not very specialized. When n is large (and in particular when n → N), each cluster is very highly specialized, and has to trade with a large number of other clusters.

When is a proactive cluster policy appropriate? 265 Here, total consignment miles means the number of consignments that are sent from one cluster to another. As each cluster becomes more and more specialized, it trades with a larger group of other clusters (n), and sends out a larger number of consignments. Moreover, since it is trading with ever more distant clusters, the total consignment miles (number of consignments × average miles transported) increases steadily. For the M25/M60 circular transportation network, we can show that total consignment miles increase with the square of the number of clusters that trade with each other (n2).8 For the grid or lattice transportation network, total consignment miles increases with the number of clusters that trade with each other raised to the power 3/2 (i.e. n3/2). However, each of these consignments gets smaller as any one cluster specializes in supplying a smaller number of items to a larger number of trading partners. Accordingly, the total ton-miles (number of consignments × average weight per consignment × average miles transported) does not rise so fast. For the M25/ M60 circular transportation network, total ton-miles increase in proportion to the number of clusters that trade with each other (n). For the grid or lattice transportation network, total consignment miles increase with the square root of the number of clusters that trade with each other (i.e. n1/2). These different possibilities are summarised in Table 13.4. Which of these is most relevant? It is hard to say definitively. The optimist will perhaps like to choose the bottom right corner while the pessimist may fear that we end up in the top left corner. If transportation logistics are perfect, then the total transportation load would just depend on the ton mileage. Many small consignments can be transported together and spare capacity is not wasted. The pessimist however will point to the amount of under-utilization of capacity on transportation networks, and might argue that the market in transportation capacity is not so perfect, so that total transportation loads will more likely be a function of consignment mileage. The average is probably somewhere in the middle. Equally, the optimist will probably argue that the lattice/grid network is more common than an M25/M60 orbital network. While the lattice/grid assumed in our model is more perfect than any actual road/rail infrastructure, the mean probably lies nearer to the lattice than the orbital. So, to a first approximation, a reasonable guess is that total transportation loads are roughly proportional to n. How does this amount of traffic translate into a congestion cost? A wide class of systems can use the mathematics of computer Table 13.4 Relationships between degree of specialization in clusters (n) and total transportation (T) Total consignment miles

Total ton miles

M25/M60 transport network

T ∝ n2

T∝n

Lattice/grid transport network

T ∝ n3/2

T ∝ n1/2

266 G. M. Peter Swann system design to map from loads to delays. One of the simplest relationships in that field is that delays are proportional to the inverse of unused system capacity. When a system is very quiet, so that there is lots of spare capacity, delays are small and a slight addition to load makes little difference to delay. When however, a system is near to saturation, then delays are long and a slight addition to load can easily lead to gridlock. The latter seems an accurate description of the character of many transportation systems in this country – at least during times of peak load. Figure 13.4 describes these properties. The left-hand curve shows the throughput-delay curve for one system with saturation capacity given by Cap0. The right-hand curve shows the same for an enhanced system with higher saturation capacity (Cap1). Take a given system, with capacity Cap0. If clustering proceeds unchecked along with increased specialization and increased intra-industry trade, then transportation demand will grow towards saturation (Cap0) and Figure 13.4 shows that we suffer rapidly growing congestion on transportation networks. Another possibility is that government, in the face of ever increasing demands to invest in the transportation infrastructure, increases the capacity of the transportation system to Cap1. So long as demand stays in the region of Cap0, this will markedly reduce congestion costs (or delays). However, this stability of demand may be a forlorn hope, as the trend towards greater specialisation, clustering and intra-industry trade is liable to continue until the capacity of transportation systems is exhausted. In short, transport demand expands to exhaust the additional capacity; investments in additional capacity only bring a temporary relief from congestion. Delay

Cap0

Figure 13.4 Throughput-delay curves

Cap1

Load

When is a proactive cluster policy appropriate? 267

The hinterland The second additional cost from clustering is felt in the hinterland. To assess the appropriateness of any proactive cluster policy, we need to look beyond the cluster alone and assess the balance between the benign effects on clusters and the less benign effects on the hinterland. We shall argue that each policy measure to promote the cluster has a shadow effect on the hinterland, and to assess any cluster policy we need at least to look at the net effects on the aggregate economy. Imagine a distribution of economy activity (say local GDP) between the cluster and the hinterland, described by Xc (for c = 1, . . ., C) and Xh (for h = 1, . . ., H) – where the suffix c indicates one of the clusters of economic activity (C in number) and the suffix h indicates one of the hinterland regions (H in number). Now consider some proactive policy which seeks to strengthen one (or more) of the clusters. This policy will have some influence on location decisions, and, once it has worked through, we can expect a change in the distribution of economic activity. Define this change in activity as ΔXc (for c = 1, . . ., C) and ΔXh (for h = 1, . . ., H). We can identify three (or more) stereotypical attitudes to cluster policy. The qualified enthusiast for cluster policy would argue that proactive cluster policy is justified because: C

H

c=1

h =1

∑ ΔX c + ∑ ΔX h > 0 That is, the aggregate effect across the economy as a whole is positive. This is so because (in the enthusiast’s view) C

∑ ΔX c >> 0 c=1

even if H

∑ ΔX h < 0

h =1

and even if ΔXc < 0, for some c. That is, the positive effects in strong clusters more than offset negative effects elsewhere. The extreme enthusiast might hope that ΔXc ≥ 0, for all c, and that ΔXh = 0 for all h, so that cluster policies are purely additive. That is, they increase activity in strong clusters, but without cost to the hinterland or any weaker clusters. That seems rather implausible. An extreme pessimist, by contrast, might argue that the distribution of activity is a zero sum game, so that any policy which strengthens the strong clusters does so at the direct cost of other weaker clusters and/or the hinterland:

268 G. M. Peter Swann C

H

c=1

h =1

∑ ΔX c + ∑ ΔX h = 0 That also seems implausible. Some policy pronouncements justify clustering in the following way: ‘The alternative to strengthening our strongest clusters is not a revival of the hinterland; rather, it is a decline of total national economic activity.’ This is consistent with the views of the qualified enthusiast above. Moreover, there is something in common between this argument and the lenient approach to competition policy, which can be paraphrased as follows: ‘If we disallow this merger, we don’t get two vigorous national competitors, we get two declining companies unable to compete against international giants.’ The latter argument tries to justify national monopoly as a lesser evil than the loss of a national champion. The former argument is similar: it justifies growing inequality in the geographical distribution of activity on the grounds that the alternative (decline in total national activity) is even worse. If this sort of argument is to be accepted as a justification for an impoverished hinterland, then it seems essential that we can be sure of the beneficial effects of clustering. We need to be sure that the qualified enthusiast is right. We would be happier to tolerate this inequality if the total effect is large, and that requires that the cluster effects are of the rich sort described above. But if they are of the weak sort, if the benefits from clustering accrue mainly to the entrants and not to others in the cluster, and if those entrants relocate from other clusters and/or the hinterland, then it must be in doubt whether the net contribution to national economic activity is all that great. Above all, we must recognize that any policy intervention to strengthen the cluster is liable to have negative effects on the hinterland. A cursory look at economic development in the UK suggests that economic decline in the hinterland carries huge social costs. These are quite different from the costs felt in the congested cluster. A proper analysis of cluster policy must take account of these negatives as well as the positives. Analysis which looks only at the positives is surely incomplete.

Policies to reduce the barriers to clustering? What does all this tell us about cluster policy? It seems that proactive cluster policy is based on the assumption that there are excessive barriers to clustering and hence insufficient clustering. If that were not so, why would governments have to be proactive? Above we argued that this assumption of insufficient clustering would be justified if new entrants to a cluster create positive externalities to incumbents in the cluster. We argued that this could obtain when the cluster effects were of the rich sort in the spectrum of Figure 13.1. In that case cluster effects follow Metcalfe’s ‘Law’ (or even Reed’s ‘Law’) where incumbents benefit from new entry.

When is a proactive cluster policy appropriate? 269 However, we have suggested that in most empirical studies of clustering it seems that the observed clustering effects are more those at the weak end of our spectrum, and for these Sarnoff ’s ‘Law’ is more appropriate. In this case, incumbents do not benefit from new entry and may indeed suffer increased congestion costs. In this case, therefore, there are negative externalities and that would suggest that free entry into the cluster leads to too much rather than too little clustering. In this case, therefore, the above justification for proactive cluster policy does not work. From this perspective, indeed, increasing congestion is ‘nature’s way’ of providing negative feedback, so as to limit the size of the cluster, and in general it should be allowed to operate. But some recent policy trends seem preoccupied with adding infrastructure to overcome any such congestion effects,9 and because of this the negative feedback is not allowed to bite. If policy is directed at trying to remove those barriers that deter excess entry, then there is a danger that further entry will impose even greater congestion costs on incumbents and third parties. In short, there must be some doubt about proactive policy which is uncritically directed at encouraging clusters to become bigger. For example, some policy is directed at trying to reduce the costs of congestion within big clusters (e.g. housing policy, transportation infrastructure within large conurbations). The rationale for this is said to be that congestion costs are stopping further entry into the cluster, and so policy is directed at reducing these barriers. But if the above argument is right, then it is more likely that the observed level of clustering is excessive rather than insufficient. In that case, a policy to reduce barriers to clustering does not look so sensible. However, some policy is directed at trying to foster the sorts of rich cluster interaction described at the top end of our spectrum in Figure 13.1. In short, it aims to go beyond the shallow aspects of clustering in our spectrum, and create something much richer. This makes much more sense. The rationale here is not so much that there will be insufficient clustering because of the negative externalities, but that the creation of such rich interactions involves increasing returns. Economists recognize that increasing returns are one of the conditions under which market failure can occur, and hence proactive policy may be justified.

Conclusion We have looked at the rationale for proactive cluster policy, designed to increase entry into clusters. This rationale looks much stronger when the cluster effects observed are from the ‘rich’ end of our spectrum in Figure 13.1 – or if cluster policy directly attempts to create such rich effects. In that case it is reasonable to expect that entrants to clusters create positive net externalities to incumbents. This means that free entry is below the socially desirable level of entry, and hence proactive cluster policy is justified in trying to enhance the (private) incentives to enter. The rationale looks weaker when cluster effects observed are from the ‘shallow’ end of our spectrum in Figure 13.1. In that case it is reasonable to

270 G. M. Peter Swann expect that entrants to clusters create negative net externalities to incumbents. This means that free entry is already above the socially desirable level of entry, and proactive cluster policy to enhance entry is unnecessary and undesirable. Moreover, we have seen that the above arguments neglect two other important potential costs arising from clustering. A complete assessment of cluster policy needs to take account of these two additional costs. First, we have argued that any trend towards specialized clusters and intra-industry trade must lead to an increased (derived) demand for transportation. Arguably this trend will continue until available transportation capacity is exhausted. The growth of clusters creates congestion not only within the clusters, but also on the transportation networks between clusters. Second, we have argued that any proactive cluster policy designed to benefit the strongest clusters will have shadow effects on the hinterland. In assessing any cluster policy, it is not enough to look at benefits to the chosen cluster on its own. Even if the net effect on the whole economy is positive, these negative effects on the hinterland need to be taken into account.

Notes 1 2 3 4 5

6 7 8 9

Krugman (1991) was also influential in rekindling the interest of mainstream economists in these questions. In German, this means ‘back country’ or ‘behind country’ – a good term for those regions that suffer when economic activity is sucked into the centre. Some important contributions to this tradition include Audretsch and Feldman (1996), Feldman (1994), Jaffe et al. (1993), Quadrio Curzio and Fortis (2002). Some other important contributions to this tradition include Cooke and Morgan (1994), Keeble (1988), Saxenian (1994), Segal Quince Wicksteed (1985). We have not space here to discuss how consumers benefit from increased entry to the cluster. But if, for example, competition in the cluster follows an n-player Cournot model, then all entry drives down prices, and that is of benefit to consumers – even if the effect of the final (nth) entrant is trivially small. These are computers used in the UK. The pattern for computers used in the USA would be somewhat different – notably the location of final assembly, and the origins of CD-Rs, monitors and manuals. Like the M60 around Manchester or the M25 around London. Detailed workings are not shown here, but are available in an unpublished note from the author. Notable examples include the planning reforms described in Department for Transport Local Government and the Regions (2001a, 2001b).

References Audretsch, D. and M. Feldman (1996) ‘Knowledge spillovers and the geography of innovation and production’, American Economic Review, 86, 630–40 Beaudry C. and G. M. P. Swann (2001) ‘Growth in industrial clusters: A bird’s eye view of the UK’, SIEPR Working Paper 00–38, Stanford University Cooke, P. and K. Morgan (1994) ‘The creative milieu: A regional perspective on innovation’, in M. Dodgson and R. Rothwell, eds, The Handbook of Industrial Innovation, London: Edward Elgar Publishers

When is a proactive cluster policy appropriate? 271 Department for Transport, Local Government and the Regions (2001a) Planning: Delivering a Fundamental Change, Green Paper, available at: http://www.planning.dtlr.gov.uk/consult/greenpap/index.htm Department for Transport, Local Government and the Regions (2001b) New Parliamentary Procedures for Processing Major Infrastructure Projects, available at: http://www.planning.dtlr.gov.uk/consult/majinfra/index.htm Feldman, M. P. (1994) The Geography of Innovation, Boston: Kluwer Academic Publishers Grayson, L. (1993) Science Parks: An Experiment in High Technology Transfer, London: British Library Jaffe, A. B., M. Trajtenberg, and R. Henderson (1993) ‘Geographic localisation of knowledge spillovers as evidenced by patent citations’, Quarterly Journal of Economics, 108, 577–98 Keeble, D. (1988) ‘High-technology industry and local environments in the United Kingdom’, in P. Aydalot and D. Keeble, eds, High-Technology Industry and Innovative Environments: The European Experience, London: Routledge Krugman P. (1991) Geography and Trade, Cambridge, Mass.: MIT Press McCann, P. and T. Arita (2000) ‘Industrial alliances and firm location behaviour: Some evidence from the US semiconductor industry’, Applied Economics, 32, 1391–403 McCann, P. and T. Arita (2002) ‘The spatial and hierarchical organization of Japanese and US multinational semiconductor firms’, Journal of International Management, 8: 121–39 Martin R. and P. Sunley (2003) ‘Deconstructing clusters: Chaotic concept or policy panacea?’, Journal of Economic Geography, 3: 5–35 Porter M. (1990) The Competitive Advantage of Nations, London: Macmillan Quadrio Curzio, A. and M. Fortis (eds) (2002) Complexity and Industrial Clusters: Dynamics and Models in Theory and Practice, Heidelberg: Physica-Verlag Saxenian A. (1994) Regional Advantage: Culture and Competition in Silicon Valley and Route 128, Cambridge, Mass.: Harvard University Press Segal Quince Wicksteed (1985) The Cambridge Phenomenon, Cambridge: Segal Quince Wicksteed Swann G. M. P. (2002) ‘The functional form of network effects’, Information Economics and Policy, 14 (3), 417–29 Swann, G. M. P. (2006) ‘Place is what we think with: Spatial history, intellectual capital and competitive distinction’, in G. Vertova (ed.), The Changing Economic Geography of Globalisation, London: Routledge (in press) Swann, G. M. P., M. Prevezer and D. Stout (eds) (1998) The Dynamics of Industrial Clustering, Oxford: Oxford University Press

14 Putting clusters in their place Nick Henry, Jane Pollard and Paul Benneworth

Introduction A central concern of this book is to put forward a theoretically informed assessment of the cluster notion – its strengths, attractions, weaknesses and limitations. Furthermore, the book seeks to apply this assessment across the variety of domains over which the concept has galloped – namely, from driving force within theories of regional innovation, competitiveness and development to multi-scalar empirical construct to the focus of a growing range of policy instruments delivered in the name of regional (cluster) development activity. Within this chapter our aim is similarly to provide an assessment by asking: ‘What is the place of clusters?’ We ask this question particularly of the concept’s place within the conceptual armoury of regional development and we attempt to answer it by drawing on a variety of example cases. Our starting point is to respond to the cogent critique of clusters written by Martin and Sunley (2003), which ends by asking just what is the ‘value added’ by the cluster concept to the range of literature on geographical agglomeration and regional development. We begin by arguing that ‘clusters’ should be understood as a set of multiple perspectives each of which encompasses a conceptual underpinning and methodological approach. Multiple perspectives explain both some of the confusion that surrounds the concept yet, to our minds, also create the potential for value added from theoretical, empirical and policy cross-fertilization (see, also, Benneworth and Henry, 2004). To support our argument in this section we draw on two exemplars – business services in London and the Dutch ICT cluster. In both cases we outline briefly how a range of analyses grouped under the clusters rubric – yet drawing on different theoretical traditions, techniques and evidence bases – can be usefully combined (if done so carefully) to provide a more confident, holistic and deeper understanding of a particular geography of economic activity. Whilst we use these examples to argue for the potential value added of the cluster concept we do, nevertheless, recognize a number of areas of theoretical development required to bring the concept to full maturity. The second section of the chapter discusses two particular areas of development – namely the role of finance in cluster development and the relationship of clusters to firm (as against regional) performance. As geographies of money and finance emerge as

Putting clusters in their place 273 a distinctive arena of research, so it seems apposite to ask what role clusters play, if any, in the provision of firm finance. We use original research on the Jewellery Quarter of Birmingham, UK, to suggest some answers. In turn, better access to finance may represent only one of the supposed advantages accrued by firms within clusters but evidence for the relationship between cluster location and the competitive performance of individual member firms remains imprecise. The chapter provides a short review of a range of studies that have attempted to unravel this critical relationship and, drawing on the discussion in section one, begins to sketch out what a robust research design for identifying the posited performance differential of clusters might look like within a multiperspectival approach. Finally, that clusters (and the synergies of spatial proximity they represent) matter does not imply that all significant economic activity takes place within some bounded area. Nor does it imply that clusters are the inevitable and dominant form of contemporary economic development. Furthermore, as cluster policy continues to be rolled out across the world, a question that remains implicit all too often is what outcomes and impacts are expected from public policy support. In particular, if clusters are supported generally under the rubric of wealth creation and regional competitiveness, questions rarely asked of clusters are those pertaining to social equity and (ironically) uneven regional development. Using the final example of ‘clusters for the inner city’, a discussion of understandings of economic practice and economic activity provides the backdrop to ask exactly what clusters can, and cannot, be expected to deliver for regional policy makers as the chapter concludes by seeking to put clusters firmly ‘in their place’.

Placing the value of clusters: clusters as a multiperspectival approach to understanding geographies of economic activity In 2003, Martin and Sunley published a powerful review of clusters. Crystallizing growing concerns and dissatisfaction, they sought to deconstruct a concept that, for them, has become ‘chaotic’ in its universalistic reification of a diversity of economic processes of localization. Identifying, amongst other things, theoretical eclecticism, methodological inconsistency and the ‘power of the brand’, ultimately, they questioned the ‘value added’ of cluster theory over and above existing theoretical explanatory strands. In a recent response (Benneworth and Henry, 2004), the authors of this chapter agree with much of the argument laid forth by Martin and Sunley but, nevertheless, seek to identify the value added that clusters can bring to the theoretical table. Taking as a starting point that clusters are an (emergent) set of multiple perspectives (and not merely a singular approach propounded by Porter, 1998), we argue that an epistemological position of ‘hermeneutic theorising’ (Barnes, 2001) allows for the potential of theoretical eclecticism to drive forward processes of theoretical, empirical and policy cross-fertilization.

274 Nick Henry, Jane Pollard and Paul Benneworth Hermeneutic theorizing, as against ‘epistemological theorizing’, recognizes that no vocabulary is perfect or final in its (re)description of reality and, in turn, that there is no end of possible sources for vocabularies of theorization (Barnes, 2001). In a context of diverse vocabularies, theorizing is a creative and openended process of interpretation performed by community members. Critical scrutiny by communities of practice will establish the ‘usefulness’ of any one theoretical competitor but the broader philosophical point to be drawn is that theorizing is circular, reflexive, indeterminate and perspectival (Bohman, 1993, p. 116, quoted in Barnes, 2001). Knowledge is situated and partial – even if certain vocabularies may hold sway as useful for long periods of time. Indeed, one might argue that the aims and contents of this book reflect a community of practice engaged in just such a process of theory development. And, certainly, many of the chapters highlight, and review, the diversity of theoretical traditions that comprise contemporary debates on clusters from geographical econometric models to normative recipes for generalized economic success (see also Benneworth et al., 2003; Gordon and McCann, 2000; Martin and Sunley, 2003; Newlands, 2003). For us, it is this range of theoretical perspectives that allows for the possibility of theoretical debate and crossfertilization. We do not dispute the inconsistencies, difficulties and problems that Martin and Sunley highlight within this eclectic literature on clusters. In particular, lack of a suitable marriage of theory and method within each perspective in use (cf. Massey and Meegan, 1985) ranks highly as just such an inconsistency yet, carefully treated as a portmanteau concept, cluster analyses drawing on different empirical studies and their related theories can make sense if the component ideas are each recognized to evolve along their own internal and logically coherent pathways. Different theoretical perspectives seek different forms of evidence collected through different sets of techniques (McKendrick, 1999; Swann, 2002; Martin and Sunley, 2003), and the vitality of the clusters concept can arise from the intersection of these different perspectives within the concept’s explanatory framework. From our perspective, then, ‘clusters’ are more akin to a series of proximate debates and the clusters approach partly the act of holding together these dissonant threads in conversation. Both combining and disentangling the threads produces the possibility of new academic knowledges and, even, the possibility of a broader analysis in which the total knowledge of the ‘cluster situation’ may be greater than that of the component parts. In (anti)essence, multiple strands need not imply incoherence. Coherence is an emergent outcome of how effectively academics perform the theorization process and is contingent on theorization being done satisfactorily and convincingly in the eyes of the community of practice. To exemplify our argument we can, for instance, use the range of work on the London business services cluster (for expansion, see Benneworth and Henry, 2004). Early work on the agglomeration by Allen (1992) and Coe and Townsend (1998) was developed by Gordon and McCann (2000), who used three

Putting clusters in their place 275 disciplinary views of clustering – regional economics, business/management, and geography/sociology – and their associated methodological tools and requisite evidence bases to reconcile earlier findings on the nature of this agglomeration. Recently, Keeble and Nachum (2002) produced the most comprehensive analysis of the cluster to date through careful operationalization of an even broader set of theoretically informed explanations of clustering activity in the region. Within their approach, conceptual work drawn from regional science, economics, economic geography, strategic management, economic sociology and science and technology studies (STS) is utilized to develop robust conclusions. The result, to our minds, is a more confident identification of the existence of the cluster, a more holistic and deeper understanding of what is, or is not, driving its existence and, critically, a highly geographical (scaled) model of the spatial extent and constitution of this particular form of regional development (Benneworth and Henry, 2004, pp. 1018–19). Similarly, the less well-known case of the ICT-sector in the Netherlands provides an example of how analytical threads drawn from different theoretical perspectives can come together under the rubric of clusters to provide insight into emergent economic geographies. The value of the cluster conceptual framework lies in helping to clarify some of the complexity in the situation, and to indicate that the sector’s economic and territorial development is not easily explained by a single set of factors. Moreover, in this case, once revealed, this complexity and multi-dimensionality is seen to be an integral part of the claims for the economic significance of the cluster. Initially, using a standard quantitative SIC mapping process, Brouwer and Den Hertog (2000) demonstrate that the ICT industry is economically significant for the Dutch economy, in terms of employment, value added and research activities. Den Hertog et al. (2000) have used an input/output analysis to reveal that there are four main sub-sectors in the cluster – electronics, services, traditional media and new media – with strong interactions within those subsectors, and between them in the cluster as a whole. Boumans and Bouwman (2000) used an eclectic methodology to explore these four sectors – arguing that they have very different but interdependent geographies, and understanding each is necessary to understanding the cluster as a whole. In particular, they distinguish the media cluster in and around Amsterdam and Hilversum from the much more technology-driven cluster around the Philips central research laboratories in the south of the country, in Eindhoven. Quah (2001) has argued, in turn, that these local concentrations have very different roles in international ‘clusters’; arguing, in particular, that the Amsterdam cluster is part of what he terms a macro-cluster, a central crescent of activity stretching from London to North Italy. Further, Van Oort (2004) has used a spatial econometric approach to demonstrate that the ICT sub-clusters based around Philips and the universities in the east and south are much more dependent on those large agents driving the clusters than in the Randstad area to the west, where there are strong dissociated knowledge pools derived from diffuse inter-firm networks.

276 Nick Henry, Jane Pollard and Paul Benneworth Taken in isolation, it would be easy to dismiss the Amsterdam multi-media agglomeration or the Eindhoven electronics concentration as engaging anecdotes. A cluster framework allows these regional geographies to be understood within a more significant ‘regional system’ at the national level, and assists in building a better understanding of the role and nature of the ICT industry in the development of the Dutch economy. With the growing complexity of production systems, appreciating and explicating the linkages between places and scales that influence comparative economic success is of growing concern (Henderson et al., 2002) and, arguably, a carefully applied and validated cluster framework aids this explanation (Bathelt et al., 2004; Wolfe and Gertler, 2004). In summary, our two examples above suggest how different knowledges can be brought together to create insight and understanding. Nevertheless, in synthesizing appealing cluster stories, we wholeheartedly agree with Martin and Sunley (2003) that there needs to be methodological rigour and consistency. The value added of the clusters approach (drawing on hermeneutic theorizing) lies in, first, allowing for and explicitly promoting theoretical conversations and, secondly, the potential this affords for multiple explanations that can interact conceptually to provide a richer understanding of the situation than permitted by theoretically monistic approaches. For us, the portmanteau concept of clusters provides a potential uniting thread to bring multiple perspectives to bear on industrial agglomeration. The inconsistencies and ambiguities of the cluster concept identified by Martin and Sunley (2003), we would suggest, are part of the ‘work in progress’ around this immature, yet politically powerful, concept.

Clusters as work in progress: finance and firm performance Just as Martin and Sunley (2003) identify, for example, a lack of theoretical development of the concept of social capital within Porter’s work so, in the spirit of ‘work in progress’, we wish to highlight two themes for further development within the clusters literature. Financing clusters In 1990, Becattini (1990, p. 37) outlined an attempt to conceive ‘a framework for a theory of the industrial district’. His discussion ranged over the values and views of a local community, a population of firms and human resources, markets, the balance of competition and co-operation, technological change, adaptability, class and consciousness and, notably, a local credit system. The credit system was described as ‘crucial’ (p. 47) in facilitating continuous development, most especially in the context of districts dominated by large numbers of small firms. Yet the financial systems element of Becattini’s formulation has been largely ignored in subsequent literatures dissecting the organizational, technological and knowledge architectures of agglomerated production complexes.1 Moreover, this neglect continues in spite of many of the insights from economic sociology

Putting clusters in their place 277 that have made their way into economic-geographical research on agglomerated production systems, particularly the concern with the social and cultural embeddedness of economic behaviours and institutions. Although access to finance, and especially venture capital finance, is acknowledged as an important element in industrial development (see Florida and Kenney, 1988; Malecki, 1997; Gompers and Lerner, 1999), there have been very few explorations of the spatiality of firms assets and liabilities and, more broadly, the geographic anatomy of financial infrastructure, institutions, knowledges, regulations and practices that constitute their financial reproduction (see Zook, 2004). With few exceptions,2 questions concerning how and where firms obtain and use external finance, how firms’ capital structures are managed and how individual entrepreneurs, firms and institutions reach decisions about financing are largely the domain of economists and finance theorists in business schools. Here, we wish to argue that it is important to consider whether debates about the significance of spatial proximity are relevant to the production and reproduction of firms’ financial networks. Firms are embedded in national and regionally varying constellations of financial infrastructures, regulations and practices and their assets, customers and suppliers are also grounded in real spacetime. It is also important to investigate the social and cultural dynamics of firms’ financial networks and how they are shaped by gender, class and religious adherence (to name just three axes of differentiation); not simply to counter the under-socialized accounts of agency that predominate in economics and finance literatures, but also to acknowledge that entrepreneurs understand and navigate their relationships with money and financial institutions in different ways that can have profound material effects on the running of their businesses (Pollard, forthcoming). In turn, while clusters have become an important unit of analysis in regional economic theorizing, little is known about their financial architectures and whether there is a relationship between financial infrastructure and cluster development and performance. Just as innovation, learning, problemsolving and other facets of economic co-ordination are now understood as geographically rooted, socially constructed achievements, there is a pressing need to examine a firm’s (and a cluster’s) financial reproduction. The relative neglect of the role of finance in clusters arguably stems from geographers’ (and others’) use of bodies of theory that downplay the significance of the circulation of money and credit in understandings of firm behaviour and production systems. Neoclassical theory, for example, conceives money in neutral terms and describes an economy in which there is perfect competition in financial markets. In addition to the standard neoclassical precepts about utility maximization and the frictionless, costless flow of information, it is assumed that information is symmetric between borrowers and lenders, that there is perfect mobility of investment capital (moving to highest rates of return) and non-existent or insignificant geographic differentials in costs of capital (see Gertler, 1984). If these assumptions are accepted and capital can flow freely to the highest rate of return to facilitate adjustment to equilibrium, then financial flows are of minor importance in regional growth. Indeed, Modigliani and

278 Nick Henry, Jane Pollard and Paul Benneworth Miller’s (1958) influential work on the capital structure of firms argued that, under these neoclassical conditions, the financial structure of a firm is essentially irrelevant because internal and external funds should be substitutable and a firm’s choice of equity or debt finance should have no bearing on its market value or profitability. Neoclassical theory is not alone in holding to theoretical commitments that detract attention from examining the finance-production nexus. Economicgeographical research since the 1970s has been heavily influenced by Marxian political economy that placed production, and not circulation or consumption, at the heart of economic geography. Marx followed the Physiocrats and Adam Smith in arguing that production, and not exchange and money handling, was the source of wealth creation. This stance slowed engagement with issues of circulation (and also consumption, see Miller, 1995) and generated debate amongst radical political economists about the significance of money and credit (Dymski, 1990). This is not to say, however, that political economic approaches have had nothing to say about money and finance in geography. David Harvey (1973; 1982), for example, has long encouraged geographers and others to use Marxian categories to examine how the financial system orchestrates flows of credit to (re)produce exploitation, inequality and uneven development (Harvey and Chaterjee, 1973; Harloe and Lebas, 1981; Dymski and Veitch, 1992; Squires, 1992; Leyshon and Thrift, 1995); although there has been a clearly discernible trend to concentrate on ‘the big picture’ of the global financial order (Leyshon, 2000).3 It is now widely accepted that financial systems have a role to play in economic development, and that there exist significant national differences in financial systems, and their regulation (Thompson, 1977; Zysman, 1983); differences that generate non-trivial variations in how, and on what terms, capital is channelled into industries. Rather less evident, however, has been interrogation of the micro-foundations of financial markets, behaviours and practices and, most especially, issues of allocation and distribution (Sayer, 1995; Storper, 2001). For clusters, the underresearched question remains: do different geographical configurations of financial institutions affect the supply and price of credit available for firms? Further, is firm behaviour affected by the nature, source and terms of their finance? There is a body of research, rooted in new-Keynesian (and later) readings of financial markets, that works from the premise that markets, as social institutions, always involve informational differences between buyers and sellers and that such asymmetries are especially acute in credit markets where borrowers know more about their collateral and industriousness than potential lenders (Leland and Pyle, 1977). These literatures – developed and applied predominantly in economics, insurance and finance literatures – have generated a series of important contributions, explaining the rationing of credit, small firms’ tendency to rely on bank debt as opposed to other forms of finance, the role of financial intermediaries in ameliorating information uncertainties through screening and monitoring, and the costs involved in operating in markets in which informa-

Putting clusters in their place 279 tional asymmetries, uncertainty and opportunism are pervasive (Spence, 1974; Leland and Pyle, 1977). Together with emergent work by geographers that interrogates the social-cultural shaping of financial markets, knowledges and institutions (O’Neill, 2001; Pollard, 2003; Tickell, 2003), a clear agenda has opened up for further research into the role of clusters within the financial infrastructure of firm finance. For example, Russo and Rossi (2001), in a study of 1,700 small and medium firms between 1989 and 1995, found that for some time firms located in industrial districts in central and northern regions of Italy enjoyed lower interest rates and greater access to bank loans than their counterparts outside industrial districts. Theoretically, the few studies available might explain this outcome through agglomeration facilitating fewer information asymmetries for contractmaking, a smoother relationship between the supply and demand for finance through shared understanding (‘industry is in the air’ including for financial intermediaries), or reduced opportunistic behaviour due to peer networks (Becattini, 1990; Dow, 1990; Stiglitz, 1994; Carnevali, 1996; Alessandrini and Zazzaro, 1999; Li et al., 2001; Russo and Rossi, 2001). A further avenue for research concerns the significance (or not) of clusters as financial spaces, as understood by entrepreneurs. One of the hallmarks of much contemporary economic geographical research is an insistence on the social nature of the economic (Thrift and Olds, 1996), that economic contracts are understood and framed in a non-contractual background (Hodgson, 1988), and that the accumulation of knowledge and rules and norms are fundamental in shaping expectations, beliefs and decision making in conditions of uncertainty (Storper and Salais, 1997). In a recent study of designer maker jewellers in Birmingham’s Jewellery Quarter in the UK (Pollard forthcoming), designer makers described the Quarter as an important financial space, home not only to most of their banking transactions and suppliers of credit but also to their peer networks providing intelligence on banks, other financial professionals and sources of funding. Designer makers rely on financial relationships with other designer makers, suppliers, financial institutions and intermediaries in the Quarter that are not automatic or ‘natural’, but have to be produced or reproduced through repeated contact (face to face and sometimes telephone) and occasionally a referral from another designer maker. These relationships are crucial for these businesses because the provision of trade credit, and the ability to negotiate changes in payment terms, allow designer makers to ameliorate the financial uncertainties of co-ordinating turnover times in what is, overall, a relatively low-wage, cash-poor environment. In sum, ‘supporting institutions’ are a commonly attributed advantage of agglomeration and clusters yet whilst there now exists compelling evidence on the relationship between financial infrastructures and the nature and characteristics of production systems, the relationship between financial institutions, clusters and the firm remains both under-theorized and rarely investigated.

280 Nick Henry, Jane Pollard and Paul Benneworth Clusters and firm performance In similar vein, work is still required in building the bridge between micro-scale clustering processes and demonstrating that those cluster processes have positive effects on particular firms’ productivity and competitiveness. Notwithstanding definitional issues concerning how a firm ‘participates’ in a cluster (cf. Gordon and McCann, 2000), this issue arises from how macro-scale cluster analyses have been justified. Lublinski (2003) notes that ‘so far, empirical work on clustering has either been descriptive case studies or econometric studies that measured either the effect of a specific agglomeration advantage or the aggregated affect of all possible centripetal forces’ (p. 454). A range of extensive analyses exist that make claims for a relationship between clustering and economic performance. Porter (1990; Porter and Ketels, 2003), for example, has analysed regional performance and found that where ‘clusters’, broadly defined as specialization or concentration of activities, are present then those regional economies tend to perform well. Established, deep and internationally competitive clusters are argued to boost regional wages and rates of innovation. Similarly, DATAR-OECD, 2001 (quoted in Martin and Sunley, 2003) suggests greater profitability from location within clusters. In contrast, Smith et al. (2002) argue against the importance of urbanization cluster effects for Danish companies on the basis that central location does not increase R&D activity or intensity. Chevassus-Lozza and Galliano (2003) have noted that knowledge spillovers are associated with improved export performance in the French food industry, but note that knowledge spillover effects are different for different sizes of firm. Van Ark et al. (2003) note that productivity has grown fastest in those Dutch service sectors that are dominated by client-led approaches to innovation where proximity to clients is key, but sparse data leaves them unable to draw firm conclusions from their tantalising observations. Relatedly, Dumais et al. (1997) argue that labour mix is the main factor explaining industrial growth and decline, rather than Marshallian spillovers from customers, suppliers and knowledge partners. They conclude, however, by observing that ‘this effect could potentially be occurring because industries with similar labour mixes share ideas as well as workers’ (p. 31) so, in effect, bringing possible cluster benefits back in to the equation. On the whole, however, these analyses remain some distance from spatially nuanced and empirically verified evidence that participation in a cluster produces a particular set of competitive advantage outcomes for (a set of) firms which, in turn, drives a broader process of territorial economic success. As Lublinksi (2003) states, approaches fail ‘to explicitly measure the relative importance of the various agglomerative forces’ (p. 454) and, in particular, he critiques the majority of work for focusing on locational factors in preference to measuring the impacts of cluster location on firms. Baptista and Swann (1998) did attempt to measure whether firms in clusters did innovate more, and found some supporting evidence for this thesis, but the issue of the causal relationship between clusters and innovation performance remains (especially as their method of cluster identification has been disputed, see Martin and Sunley, 2003).

Putting clusters in their place 281 Those approaches that begin from particular clusters themselves have also struggled to deliver a fully convincing case for the existence of demonstrable territorial benefits through participating in ‘clustering’ (Benneworth et al., 2003). A number of cluster studies have argued that clusters and clustering processes have helped particular sets of firms to overcome certain technological problems they have faced. In her work on a local medical/ICT cluster in the east of the Netherlands, for example, Klein Woolthuis (1999) revealed how co-operation and trust built up through common working, and that trust was directly a factor in the development of new products by cluster companies. Similarly, Lorenz (1999) has shown that the build-up of trust through an ongoing learning process was a key part of the successful adjustment of the Lyons machine tools industry in the late 1980s. The issue remains, however, how these findings compare with other similar firms (and the sector) outside of the cluster. Indeed, of particular scarcity are analyses which begin from an assumed cluster, and compare tangible economic aggregates of cluster firms, such as exports, profits or employment growth, with the same industry elsewhere (Ketels, 2003). Lublinksi (2003), for example, is not able to establish that clusters have definite benefits but, instead, is held to the rather less ambitious claim that, in the northwest German aerospace cluster, proximity was more important to firms in the cluster than in a control group of aerospace firms. Overall, however, the continued difficulty (even failure) to link competitive performance benefits with active participation in the cluster by a firm, or firms, represents a significant fissure in the theoretical contentions of cluster advocates. A positive sign is a growing, but still scattered, body of evidence that attempts to rectify this omission (Ketels, 2003). Swann (2002), for example, uses the EU community innovation survey to link co-operation in R&D by firms with improved innovation performance, cost reduction and profitability; however, they are not directly concerned with the spatial location of the collaboration. In 2002, Broersma (2002) applied an innovative methodology for exploring induced innovation effects in clusters, using input/output structures to gauge interaction between sectors and then gauging the flow of innovation based on known levels of R&D in each sector. This concluded that clusters did indeed have higher flows of knowledge between them when this was measured through this imputed input/output analysis. Theoretically, recent work drawing on a knowledge-based view of the firm (Pinch et al., 2003; Tallman et al., 2004) has theorized a set of propositions as to how knowledge flows within clusters can lead to differential firm performance (‘competitive advantage’) across the organizational constructs of industry, cluster and firm but these remain to be empirically tested. Yet, overall, the challenge remains to move away from ‘cluster truisms’ on economic performance and to fully theorize and empirically test cluster processes and their relationship to innovation, productivity and competitiveness within the firm (as well as, and in contrast to, groups of spatially proximate firms or sectors). Indeed, drawing on the above short review, and our earlier discussion on multiple analyses of the Dutch ICT sector, one can begin to tease out what a robust, and multistranded, cluster methodology might comprise:4

282 Nick Henry, Jane Pollard and Paul Benneworth •

•

•

•

initial identification of agglomerations using location quotients and inputoutput analysis to provide possible candidate clusters in any particular industry. More sophisticated technical variants may be applied in this analysis and, for example, might include ‘aggregation’ of certain industrial classifications. If input/output relationships provide one justifiable rationale for aggregation, others might exist such as expert knowledges of emergent production systems (e.g. biotechnology) or systems ‘hidden’ within official classifications (e.g. UK motorsport industry) investigation of a broader range of ‘connectivities or clustering processes’ (labour markets, knowledge systems, institutions, socio-cultural characteristics, and so forth) that suggest a shift in categorization beyond agglomeration to cluster dependent on the research question but, in this case, asking the question ‘do firms perform better inside a cluster than outside’ the selection of matched respondent firms both inside and outside the cluster and collection of data on their comparative performance the testing of propositions for, or evidence of, explanatory relationships for differential performance between cluster and non-cluster firms

Whilst recognizing that the above is necessarily schematic as only one small part of the chapter, we can suggest that the analytical design may be ordered differently according to research questions (or even policy drivers) but that multiple methodologies reflecting a range of approaches can be combined to create more robust identification and deeper understanding of clusters.

Placing clusters in regional development So far in this chapter we have suggested that there is still significant work to be done on the clusters concept but, nevertheless, we believe the approach offers added value in understanding the agglomerative activity evident across the contemporary space economy. Although we have raised concerns about the comprehensiveness of evidence, and theoretical maturity of, the concept of clusters, there is little doubt that there are a range of worldwide ‘paradigmatic’ examples of regional agglomeration that have had a major impact upon models (and policies) of regional economic development. In particular, and to quote Martin and Sunley (2003, p. 6), ‘Clusters, it seems, have become a world-wide fad, a sort of academic and policy fashion item’. Arguably an emblematic example of Thrift’s (2005) ‘soft capitalism’, in which key global institutions and elites rapidly create and live stories of the economy,5 the fad of the cluster concept is losing some of its steam. In turn, and partly contributing to the slowdown, a process of reflection, analysis and review of the concept is under way, almost reclaiming the idea from the policy domain as part of continued theoretical work on agglomeration in economic development. Whilst earlier sections of the chapter have to some extent mirrored this wider process – asking what is new and different, and what work is still required – it

Putting clusters in their place 283 is only now that a more fundamental questioning of the ‘point of clusters’ can begin to be discerned. Highlighting possibly the most powerful aspect of the concept, what has remained an almost constant silence (or given) is exactly what type of regional development outcomes clusters are expected to deliver and the relationship of these outcomes (for example, greater regional competitiveness) to a broader set of regional development objectives such as greater general well-being, quality of life and sustainable development. In this final section of the chapter we wish to ask exactly what form of growth, with what type of economic and social outcomes, can be expected from cluster development activity as a policy instrument. What are the broader regional development objectives of clusters? What has their dominance of regional development policy implied for other developmental avenues for local and regional economies? In asking these questions, our aim is to ‘place clusters’ within the realms of economic activity and policy development and, in particular, to argue for a diverse landscape of economies of which clusters are merely a part. By doing so, we wish to suggest that, as one reading of a particular set of economic relations within the broader economy, clusters should sit alongside a range of approaches to regional development. In this sense, we wish to put clusters firmly ‘in their place’. Clusters as a tool of regional development Fundamentally, clusters have been applied within a framework of the drive to greater regional competitiveness and competitive advantage (Porter, 1990, 1998). Putting aside the issues raised earlier of the current evidence base for this expectation, the question we wish to ask is, if successful, what would the outcomes for a regional economy driven by clusters look like? As Martin and Sunley (2003) put it, the impact of cluster growth on other aspects of the regional economy and regional development remains ‘unresolved’ (p. 27). Within a successful cluster, one might expect rising productivity and profits within member companies. In turn, increased employment opportunities, rising local wages and higher rates of new firm formation might ensue around the core of the cluster. Yet a number of dimensions can be played out around the economic and social outcomes of this growth. First, the nature of the development dynamic within clusters can drive – theoretically – equal or unequal distributional effects (across income and other social axes of differentiation). Previous work on high-technology Cambridge, for example, has outlined the gender inequalities endemic to the growth of this production complex (Henry and Massey, 1995; Massey, 1995; see, also, Allen et al., 1998, on the regionalized growth of the 1980s South East of England). Moreover, growth can trigger ‘spillover’ effects such as tightened labour markets and pressures on housing stock and other regional services (for example, water and transport). This is only likely to increase the pressure on non-core (and/or less skilled and less productive activities) to the point that they might go out of business or move away.6 In effect, then, this is merely to exemplify how Sassen’s (1994) arguments for the

284 Nick Henry, Jane Pollard and Paul Benneworth ‘two sides to growth’ associated with the particular agglomerative growth of global cities could be similarly applied within the realm of clusters. In a rare piece on the distributional effects of clusters, Rosenfeld (2003) asks the question as to what impact this particular form of development policy has on low- and middle-income people, economically distressed urban and rural places and small enterprises. He does so to pinpoint the reasons why certain people, places and firms are ‘excluded’ through the process of cluster development and in an attempt to identify routes to inclusion which might draw on cluster activity, including how examination of systemic relationships may reveal previously unnoticed common or collective competencies, hidden specialized resources, and ways to aggregate strengths that have the potential to take advantage of cluster tools, social capital, and externalities. (Rosenfeld, 2003, p. 376) On the one hand this attempt to engage with, and bend, the model of growth implicit within cluster policy is very welcome; on the other hand it highlights, as Rosenfeld himself suggests, that whilst clusters may be a single development tool more often than not they are interpreted as the framework for understanding and building regional economies. Thus, cluster policy has been applied to revive old industrial regions, stimulate new high-tech and/or knowledge-based agglomerations and, probably most disconcertingly, proposed as the solution to the longstanding problems of the inner city (Porter, 1995). In 1995, Porter argued how cluster policy could find, and/or deliver, competitive advantage to the inner city as a solution to these areas’ ingrained and longstanding economic problems. Whilst he has argued against any attempts to build clusters from scratch, through his ICIC consultancy (Initiative for a Competitive Inner City) he has sought to apply a demand-led and market-based solution to areas in which continued market failure has, arguably, been the root cause of their very definition. On a trip to the USA, ‘clusters for the inner city’ caught the eye of the UK Chancellor Gordon Brown as part of his broader aim to spread enterprise across all areas and communities of the UK. Under what was labelled as the City Growth Strategies programme, in 2003, a total of seven pilot areas were chosen in the UK to implement cluster policy within the inner city under the tutelage of ICIC. Whilst the evaluation report of the pilots argued that it was too early to be able to evaluate the outcomes of the scheme (GHK, 2004a),7 the report also pointed out that no one model of cluster policy could be discerned across the pilots and the lack of any independent evaluation for the effectiveness of the clusters in the inner city initiative within its USA homeland. Nevertheless, a further dozen or so areas have been chosen for City Growth Strategies in the UK. Yet whilst evidence may yet materialize for the effectiveness of the policy, ‘clusters for the inner city’ may also precisely reflect the over-extension of a development tool originally applied at a national level to that of a framework

Putting clusters in their place 285 for understanding and building an economy irrespective of its scale and historical trajectory. For example, at the time of City Growth Strategies in the UK, the UK government was undertaking a national programme of support to Community Development Finance Institutions (CDFIs) that proved to be highly effective at stimulating, and funding, start-up and enterprise expansion across ‘hard to reach’ communities and across areas of much greater breadth than inner cities and that achieved by City Growth Strategies (GHK, 2004b; Freiss, 2004). Indeed, CDFIs are only one institutional form with a broader range of social enterprises, social entrepreneurship activities, community and voluntary organizations, and alternative economic activities around self-provisioning, economies of regard and so forth that may have particular pertinence to ‘inner cities’. Whilst our argument is not to suggest that these examples of diverse economic activities are necessarily ‘the solution’ to the problem of the inner cities or, indeed, are confined to the spaces of the economically disadvantaged, their effectiveness and efficiency as possible policy solutions could, and should, be tested alongside other policy initiatives (including, for example, national support for regional/ inner city champions, regional science policy, inter-regional transfer payments and social security) rather than simply reaching for the cluster tool-kit. In other words other policy solutions do, or could, exist on the basis of different conceptualizations of the economy (and in particular its diversity), conceptualizations which may have much greater resonance than the travelling concept of clusters for some of the problems of local and regional economic development.

Conclusion: putting clusters in their place We began this chapter by defending the potential of the concept of clusters as a theoretical, empirical and policy construct, and its ability to create fruitful dialogue across economic analyses holding to different traditions. Nevertheless, we end it arguing for the rightful place of this potential within the armoury of regional development activity – as only one club within the golf bag of regional development policy. We defend the cluster concept as providing the potential for added value in our theorizing and understanding of regional economic development. As a portmanteau concept, the clusters approach offers considerable potential to combine a variety of perspectives. Furthermore, writing as geographers, we would suggest this potential includes overcoming the economism of cluster analyses delivered within the realms of ‘geographical economics’ (Martin, 1999). In our argument we suggested that much work still remains to deliver this potential and highlighted two particular arenas of activity for further development – the relationship between clusters and financial architectures, and between clusters and firm performance. The variable evidence for the performance of clusters – and the intricate role of finance within firm activity – merely serves, in our view, to highlight the institutional and social practices that ‘make’ the economy, the economic diversity of the economy and the role geography plays in this economic life.

286 Nick Henry, Jane Pollard and Paul Benneworth In the final section of the chapter the context of economic diversity is used to defend the right of clusters as a policy tool but to argue for ‘its place’; to argue against the current over-endowment of this brand within the armoury of regional development policy as the all-encompassing focus of regional development activity. Even if highly successful, the nature of success delivered by cluster policy requires further clarification within the broader objectives of regional development, and reflection on clusters as a development tool, and not a model of the regional economy to the exclusion of all other alternative readings. Such reflection should countenance a number of criteria including the resonance of clusters with a particular set of internationally traded economic activities, the dangers of over-specialization within a regional economy, a recognition that the economic and social outcomes of such policy are unlikely (on their own) to meet broader regional development aspirations and the evidence of the effectiveness of other policy measures based upon different conceptualizations of the economy in delivering economic development. In this regard, Gibson-Graham’s (1995) challenging thesis on economic activity is just one example of alternative conceptualizations of ‘the economy’ that should remind us of both the richness of economic life (and our conceptualisation of this life),8 and the policy initiatives that might ensue. In sum, we strongly suggest a more sparing and nuanced use of a still-to-be-enhanced cluster theory for understanding (elements of) the geographies and territorial performance of regional economies.

Notes 1 2 3

4 5 6 7 8

This section draws on arguments developed in Pollard (2003; forthcoming). Feakins’s (2001) work on commercial bank lending to SMEs in Poland is one such exception. Similarly, there are now well-rehearsed debates about the significance of regional financial networks in an era of ongoing integration of financial markets, agents and regulation. Although ‘the region’ was an important financial unit in the UK in the nineteenth century, when banks were closely tethered to regional industrial structures, its contemporary significance is much less clear (see Pratt, 1998; Klagge and Martin, 2004). Our thanks to Philip Cooke for urging us to at least begin to put a sketch design on paper. See Lagendijk and Cornford (2000) for an early attempt to track the story of clusters. As Martin and Sunley (2003, p. 27) point out, whilst Porter (1998) views this as the ‘correct’ outcome of cluster development it could lead, in aggregate, to negative outcomes within the regional labour market. Positive process outcomes such as business engagement with strategic economic development were identified. See, for example, Henry et al. (2002), Leyshon et al. (2003) and Smith and Stenning (2004) for a range of examples of alternative, but possibly complementary, readings of economic diversity.

Putting clusters in their place 287

References Alessandrini, P. and Zazzaro, A. (1999) ‘A “possibilist” approach to local financial systems and regional development: The Italian experience’, in R. Martin (ed.) Money and the Space Economy. Chichester, Wiley, pp. 71–92. Allen, J. (1992) ‘Services and the UK space economy: Regionalisation and economic dislocation’, Transactions of Institute of British Geographers 7 (2), 292–305. Allen, J., Massey, D., Cochrane, A., Charlesworth, J., Court, G., Henry, N. and Sarre, P. (1998) Rethinking the Region. London, Routledge. Baptista, R. and Swann, P. (1998) ‘Do firms in clusters innovate more?’, Research Policy 27, 525–40. Barnes, T. (2001) ‘Retheorizing economic geography: From the quantitative revolution to the “cultural turn”’, Annals of the Association of American Geographers 91 (3) 546–65. Bathelt, H., Malmberg, A. and Maskell, P. (2004) ‘Clusters and knowledge: Local buzz, global pipelines and the process of knowledge creation’, Progress in Human Geography 28, 31–56. Becattini, G. (1990) ‘The Marshallian industrial district as a socio-economic notion’ in F. Pyke, G. Becattini and W. Sengenberger (eds) Industrial Districts and Inter-firm Co-operation in Italy. Geneva, Institute for Labour Studies, pp. 37–51. Benneworth, P. S. and Charles, D. R. (2001) ‘Bridging cluster theory and practice: Learning from the cluster policy cycle’ in E. M. Bergman, P. den Hertog, D. R. Charles and S. Remoe (eds) Innovative Clusters: Drivers of National Innovation Systems Paris, OECD. Benneworth, P. and Henry, N. (2004) ‘Where is the value-added in the cluster approach? Hermeneutic theorizing, economic geography and clusters as a multiperspectival approach’, Urban Studies 41 (5/6), 1011–24. Benneworth, P. S., Raines, P., Danson, M. and Whittam, G. (2003) ‘Confusing clusters? Making sense of the cluster approach in theory and practice’, European Planning Studies, 11 (5), 511–20. Bergman, E. M., Den Hertog, P., Charles, D. R., and Remoe S. (eds) (2000) Innovative Clusters: Drivers of National Innovation Systems. Paris, OECD. Bohman, J. (1993) New Philosophy of Social Science: Problems of Indeterminacy. Cambridge, MA, MIT Press. Boumans, J. and Bouwman, H. (2000) ‘De catrograife van de multimediasector’ in H. Bouwman and W. Hulisnk (eds) Silicon Valley in de polder: ICT-clusters in de Lage Landen. Utrecht, Lemma. Broersma, L. (2002) ‘The role of services in innovative clusters’. Structural Information Provision on Innovation in Services Working Paper, MINEZ, The Hague. Brouwer, E. and Hertog, P. Den (2000) ‘Key data on the Dutch information and communication cluster’, SIID Project working paper, Utrecht: Dialogic & Centre for Science and Policy University of Utrecht. Carnevali, F. (1996) ‘Between markets and networks: Regional banks in Italy’, Business History 38 (3), 84–100. Chevassus-Lozza, E. and Galliano, D. (2003) ‘Local spill-overs, firm organisation and export behaviour: evidence from the French food industry’, Regional Studies 37 (2), 147–58. Coe, N., and Townsend, A. R. (1998) ‘Debunking the myth of localised agglomerations: The development of a regionalised service economy in South East England’, Transactions of the Institute of British Geographers 23 (3), 385–404.

288 Nick Henry, Jane Pollard and Paul Benneworth Dalsgaard, M.-H. (2001) ‘Danish cluster policy: Improving specific framework conditions’, in E. M. Bergman, P. den Hertog, D. R. Charles and S. Remoe (eds) Innovative Clusters: Drivers of National Innovation Systems. Paris, OECD. Den Hertog, P., Brouwer, E. and Maltha, S. (2000) Innovation in an Adolescent Cluster: The Case of the Dutch Multimedia Cluster. Utrecht, Dialogic. Dow, S. (1990) Financial Markets and Regional Economic Development. London, Avebury Press. Dumais, G., Ellison, G. and Glaeser, E. L. (1997) ‘Geographical concentration as a dynamic process’. National Bureau of Economic Research Working Paper No. 6270, Cambridge, MA, NBER. Dymski, G. (1990). ‘Money and credit in radical political economy: A survey of contemporary perspectives’, Review of Radical Political Economics 22 (2–3), 38–65. Dymski, G. and Veitch, J. (1992) ‘Race and the financial dynamics of urban growth: LA as Fay Wray’, in City of Angels, ed. C. Dersch. Los Angeles, Kendall/Hunt Press, pp. 131–58. Feakins, M. (2001). Commercial Bank Lending to SMEs in Poland: An Empirical Investigation. Oxford, School of Geography, University of Oxford. Florida, R. and Kenney, M. (1988) ‘Venture capital, high technology and regional development’, Regional Studies 22, 33–48. Freiss (2004) Evaluation of the Phoenix Development Fund, http://www.sbs.gov.uk/ content/analytical/evaluation/PhoenixDevFundEvaluation.pdf; accessed 5 December 2004. Gertler, M. S. (1984). ‘Regional capital theory’, Progress in Human Geography 8 (1), 50–81. GHK (2004a) Evaluation of Community Development Finance Initiative, http:// www.sbs.gov.uk/content/analytical/evaluation/CDFI-Evaluation.pdf; accessed 5 December 2004. GHK (2004b) Evaluation of City Growth Strategies, http://www.sbs.gov.uk/content/ analytical/evaluation/CityGrowthStrategies.pdf; accessed 5 December 2004. Gibbs, D. C., Jonas, A. E. G., Reimer, S. and Spooner, D. J. (2001) ‘Governance, institutional capacity and partnerships in local economic development: Theoretical issues and empirical evidence from the Humber Sub-region’, Transactions of the Institute of British Geographers 26 (1), 103–20. Gibson-Graham, J. K. (1995) The End of Capitalism (As We Knew It): A Feminist Critique of Political Economy. Oxford, Blackwell. Gilsing, V. (2001) ‘Towards second-generation cluster policy: The case of the Netherlands’, in E. Bergman, P. den Hertog, D. Charles, and S. Remoe (eds) Innovative Clusters: Drivers of National Innovation Systems. Paris, OECD, pp. 361–76. Gompers, P. and Lerner, J. (1999) The Venture Capital Cycle. Cambridge, MA, MIT Press. Gordon, I. and McCann, P. (2000) ‘Industrial clusters: Complexes, agglomeration and/or social networks’, Urban Studies 37, 512–32. Grabher, G. (ed.) (1993) The Embedded Firm: On the Socio-economics of Industrial Networks. London, Routledge. Harloe, M. and Lebas, E. (eds) (1981) Cities, Class and Capital: New Developments in the Political Economy of Cities and Regions. London, Edward Arnold. Harvey, D. (1973) Social Justice and the City. London, Edward Arnold. Harvey, D. (1982) The Limits to Capital. Oxford, Blackwell. Harvey, D. and Chaterjee, L. (1973) ‘Absolute rent and the restructuring of space by governmental and financial institutions’, Antipode 6, 22–36.

Putting clusters in their place 289 Henderson, J., Dicken, P., Hess, M., Coe, N. and Yeung, W.-C. (2002) ‘Global production networks and the analysis of economic development’, Review of International Political Economy 9, pp. 436–64. Henry, N. and Massey, D. (1995) ‘Competitive time-space in high technology’, Geoforum, 26 (1), 49–64. Henry, N., McEwan, C. and Pollard, J. (2002) ‘Globalisation from below: Birmingham – postcolonial workshop of the world?’, Area 34 (2), 117–27. Hodgson, G. (1988) Economics and Institutions: A Manifesto for a Modern Institutional Economics. Cambridge, Polity Press. Keeble, D. and Nachum, L. (2002) ‘Why do business service firms cluster? Small consultancies, clustering and decentralisation in London and southern England’, Transactions of the Institute of British Geographers 27 (1), 67–90. Ketels, C. H. M. (2003) ‘The development of the cluster concept: Present experiences and further developments.’ Paper presented to Nord-Rhein Westfalen Conference on Clusters, Duisburg, Germany, available online at <www.isc.hbs.edu/pdf/ Frontiers_ of_Cluster_Research_2003.11.23.pdf[more]. Klagge, B. and Martin, R. (2004) ‘Decentralised versus centralised financial systems: Is there a case for local capital markets?’ Paper presented at Financing Regional Enterprise, University of Newcastle, December 2004. Lagendijk, A. and Cornford, J. (2000) ‘Regional institutions and knowledge – tracking new forms of regional development policy’, Geoforum 31, 209–18. Learmonth, D., Monro, A., and Swales, J. K. (2003) ‘Multi-sectoral cluster modelling: The evaluation of Scottish Enterprise cluster policy’, European Planning Studies 11 (5), 567–84. Leland, E. and Pyle, D. (1977) ‘Information asymmetries, financial structure and financial intermediation’, Journal of Finance xxxii (2), 371–87. Leyshon, A. (2000) ‘Money and finance’, in E. Sheppard and T. Barnes (eds) A Companion to Economic Geography. Oxford, Blackwell, pp. 432–49. Leyshon, A., Lee, R. and Williams, C. (2003) Alternative Economic Spaces. London, Sage. Leyshon, A. and Thrift, N. (1995) ‘Geographies of financial exclusion: Financial abandonment in Britain and the United States’, Transactions of the Institute of British Geographers 20, 312–41. Li, W., Zhou, Y., Dymski, G. and Chee, M. (2001) ‘Banking on social capital in the era of globalization: Chinese ethnobanks in Los Angeles’, Environment and Planning A 33, 1923–48. Lorenz, E. (1999) ‘Trust, contract and economic cooperation’, Cambridge Journal of Economics 23, pp. 301–15. Lublinski, A. E. (2003) ‘Does geographical proximity matter? Evidence from clustered and non-clustered aeronautics firms in Germany’, Regional Studies 37 (5), 453–67. McKendrick, J. H. (1999) ‘Multi-method research: An introduction to its application in population geography’, Professional Geographer 51 (1), 40–50. Malecki, E. (1997) ‘Entrepreneurs, networks and economic development: A review of recent research’, Advances in Entrepreneurship, Firm Emergence and Growth 3, 57–188. Martin, R. (1999) ‘The new “Geographical Turn” in economics: Some critical reflections’, Cambridge Journal of Economics 23, 65–91. Martin, R. and Sunley, P. (2003) ‘Deconstructing clusters: Chaotic concept or policy panacea?’, Journal of Economic Geography 3 (1), 5–35. Massey, D. (1995) ‘Masculinity, dualisms and high technology’, Transactions of the Institute of British Geographers 20 (4), 487–99.

290 Nick Henry, Jane Pollard and Paul Benneworth Massey, D. and Meegan, R. (eds) (1985) Politics and Methods: Contrasting Studies in Industrial Geography. London, Methuen. Miller, D. (ed.) (1995) Acknowledging Consumption. London, Routledge. Modigliani, F. and Miller, M. H. (1958) ‘The cost of capital, corporate finance and the theory of investment’, American Economic Review 48 (3), 261–97. Newlands, D. (2003) ‘Competition and cooperation in industrial clusters: The implications for public policy’, European Planning Studies 11 (5), 521–32. O’Neill, P. (2001) ‘Financial narratives of the modern corporation’, Journal of Economic Geography 1 (2), 181–99. Pinch, S., Henry, N., Jenkins, M. and Tallman, S. (2003) ‘From “industrial districts” to “knowledge clusters”: A model of knowledge dissemination and competitive advantage in industrial agglomerations’, Journal of Economic Geography 3 (4), 373–88. Pollard, J. S. (2003). ‘Small firm finance and economic geography.’ Journal of Economic Geography 3 (4), 429–52. Pollard, J. S. (forthcoming). ‘Making money, (re) making firms: Micro-business financial networks in Birmingham’s Jewellery Quarter’, Environment and Planning A. Porter, M. E. (1990) The Competitive Advantage of Nations. New York, Basic Books. Porter, M. (1995) ‘The competitive advantage of the Inner City’, Harvard Business Review 74, 61–78. Porter, M. E. (1998) On Competition. Boston, MA, Harvard Business School. Porter, M. and Ketels, C. (2003) UK Competitiveness: Moving to the Next Stage. London, DTI. Pratt, D. (1998) ‘Re-placing money: The evolution of branch banking in Britain’, Environment and Planning A 30 (12), 2211–26. Quah, D. (2001) ‘ICT clusters in development: theory and practice’, European Investment Bank Papers 6 (1), 85–101. Raines, P. (2002) ‘Cluster policy – does it exist?’, in P. Raines (ed.) Cluster Development and Policy. Burlington, VT Ashgate. Rosenfeld, S. (2003) ‘Expanding opportunities: Cluster strategies that reach more people and places’, European Planning Studies 11 (4) 359–77. Russo, P. and Rossi, P. (2001) ‘Credit constraints in Italian industrial districts’, Applied Economics 33 (11), 1469–77. Sassen, S. (1994) Cities in a World Economy. Thousand Oaks, CA, Pine Forge Press. Sayer, A. (1995) Radical Political Economy: A Critique. Oxford, Blackwell. Smith, A. and Stenning, A. (2004) Social Exclusion, Spaces of Household Economic Practice and Post-Socialism. Mimeo of ESRC Grant Application available from author at [email protected]. Smith, V., Broberg, A. L. and Overgaard, J. (2002) ‘Does location matter for firms’ R&D behaviour? Empirical evidence for Danish firms’, Regional Studies 36 (8), 825–32. Spence, A. (1974) Market Signalling. Cambridge, MA, Harvard University Press. Squires, G. D. (ed.) (1992) From Redlining to Reinvestment: Community Responses to Urban Disinvestment. Philadelphia, Temple University Press. Stiglitz, J. (1994). ‘Peer monitoring in credit markets’, World Bank Economic Review 4, 351–66. Storper, M. (1995) ‘The resurgence of regional economies, ten years later: The region as a nexus of untraded interdependencies’, European Urban and Regional Studies 2 (3), 191–221.

Putting clusters in their place 291 Storper, M. (2001) ‘The poverty of radical theory today: From the false promises of Marxism to the mirage of the cultural turn’, International Journal of Urban and Regional Research 25 (1), 155–79. Storper, M. and Salais, R. (1997) Worlds of Production: The Action Frameworks of the Economy. Cambridge, MA, Harvard University Press. Storper, M. and Walker, R. (1989) The Capitalist Imperative: Territory, Technology and Industrial Growth. Oxford, Blackwell. Swann, G. (2002) The Implications of Clusters: Some Reflections. Paper presented at the Clusters Conference, Manchester Business School, April. Tallman, S., Jenkins, M., Henry, N. and Pinch, S. (2004) ‘Knowledge clusters and competitive advantage’, Academy of Management Review 29 (2), 258–71. Thompson, G. (1977). ‘The relationship between the financial and industrial sector in the United Kingdom economy’, Economy and Society 6, 235–83. Thrift, N. (2005) Knowing Capitalism. London, Sage. Thrift, N. and Olds, K. (1996) ‘Refiguring the economic in economic geography’, Progress in Human Geography 20 (3), 311–37. Tickell, A. (2003) ‘Cultures of money’, in K. Anderson, M. Domosh, S. Pile and N. Thrift (eds) Handbook of Cultural Geography. London, Sage, pp. 116–30. Van Ark, B., Broersma, L. and Den Hertog, P. (2003) ‘Services innovation, performance and policy: A review.’ Structural Information Provision on Innovation in Services Synthesis Paper, MINEZ, The Hague. Van Oort, F. G. (2004) Urban Growth and Innovation: Spatially Bounded Externalities in the Netherlands. Aldershot, Ashgate. Wolfe, D. and Gertler, M. (2004) ‘Clusters from the inside and out: Local dynamics and global linkages’, Urban Studies 41 (5/6), 1071–93. Woolthius, R. K. (1999) Sleeping with the Enemy: Trust Dependence and Contract in Interorganisational Relationships. Enschede, University of Twente Press, Netherlands. Zook, M. (2004) ‘The knowledge brokers: Venture capitalists, tacit knowledge and regional development’, International Journal of Urban and Regional Research 28 (3), 621–41. Zysman, J. (1983) Governments, Markets and Growth. Ithaca, Cornell University Press.

Name index

Abernathy, W. 8 Ablas, L. 4 Acs, Z. 69 Adorno, T. 23 Almeida, P. 62 Amin, A. 81, 94, 167 Anastasia, B. 76 Andersen, E. 117 Anderson, J. 199 Angel, D. 61, 62 Anselin, L. 61 Antonelli, C. 79, 95, 209 Argyris, C. 206 Aron, R. 58 Arrow, K. 192–5, 201 Arthur, B. 69 Asheim, B. 1, 2, 6, 14, 76, 81, 137, 138, 143, 225, 226, 229 Audretsch, D. 61, 65, 175, 176, 193 Aydalot, P. 175 Bagnasco, A. 6 Baldwin, R. 2, 6 Balestri, A. 99 Baptista, R. 61, 69, 174 Bathelt, H. 41, 62, 222, 226 Becattini, G. 2, 4, 6, 12, 69, 90, 99, 276, 279 Belussi, F. 69, 84 Benner, C. 22, 23, 62 Bennett, R. 61, 169 Bertini, S. 99 Bettis, R. 142 Biggeiro, L. 69, 107 Birkinshaw, J. 97 Blanchflower, D. 120, 122 Boari, C. 61 Bogenrieder, I. 150 Borja, J. 97

Boschma, R. 19, 116, 117, 122, 126, 137, 143 Bottazzi, G. 19 Brackman, S. 2 Braczyk, H. 69 Breschi, S. 62, 94, 220 Bresnahan, T. 197 Brown, J. 220 Brown, P. 84 Brusco, S. 2, 6, 21, 30, 76, 93, 95, 96 Buchanan, J. 5 Bunker, B. 148 Burt, R. 138, 141 Bygrave, W. 120 Cairncross, F. 1 Camagni, R. 14, 37, 124, 175 Canïels, M. 19 Cantwell, J. 84 Carroll, C. 120, 126 Casper, S. 19 Castells, M. 97, 167 Chesbrough, H. 17 Chiles, T. 117 Chinitz, B. 176 Christopherson, S. 237, 239 Clark, K. 144 Clifton, N. 19 Coase, R. 16 Coe, N. 61 Coenen, L. 15, 225, 226 Cohen, M. 140, 199 Commons, J. 202 Cooke, G. 78 Cooke, P. 1, 2, 4, 13, 18, 19, 22, 69, 81, 84, 108–10, 138, 175, 200, 269 Cooper, D. 62 Coro, G. 76, 84, 107 Cossentino, F. 69

Name index 293 Crestanello, P. 76 Crevoisier, O. 37 Crouch, C. 2 Czamanski, S. 4, 8 Dahl, M. 62 Dahmén, E. 8–9 De Jong, M. 5 De Man, A. 84 De Propris, L. 19, 61 Dei Ottati, G. 12, 76 Delaney, E. 121 Derudder, B. 97 Deutsch, K. 148 Di Maggio, P. 142, 202 Dicken, P. 53 Dixon, R. 180 Doeringer, P. 3 Dorfmann, N. 69 Dosi, G. 18, 19, 200 Douglas, M. 199 Duguid, P. 220 Dunning, J. 69, 84 Dymski, G. 277 Eliasson, G. 53 Ellison, G. 30 Engelking, S. 118 Engstrom, J. 61 Enright, M. 51, 69, 84, 96 Fagiolo, G. 19 Falzoni, A. 76 Feldman, M. 61, 69, 116–21, 175, 176, 191, 206, 220 Ferrucci, L. 4 Feser, E. 35 Ffowcs-Williams, I. 84 Fine, B. 17,18 Florida, R. 2, 97, 277 Forlai, L. 99 Fosfuri, A. 62 Foss, N. 200 Francis, J. 116, 117 Freeman, C. 200 Fuellhart, K. 61 Fujita, M. 2 Gambardella, A. 197 Garnsey, E. 108 Garretsen, H. 2 Gertler, M. 2, 15, 220, 225–29 Gilsing, V. 138, 143, 147, 149 Gilson, R. 62

Glaeser, E. 30, 61, 126, 192 Gompers, P. 118, 277 Gordon, I. 15, 20, 91, 178, 199, 274, 279 Gottardi, G. 69 Grabher, G. 23, 62, 76, 175 Grandori, A. 91, 95 Granovetter, M. 18, 138, 167 Grant, R. 200 Gray, J. 1 Gray, M. 125 Guerrerri. P, 69 Gulati, R. 141 Håkansson, H. 51 Hallencreutz, D. 61 Hansen, M. 150 Haraldsen, T. 11 Harloe, M. 277 Harrison, B. 69 Hartmann, C. 199 Heidenreich, M. 69 Hench, T. 117 Henderson, R. 144 Hendry, C. 61 Hilpert, U. 175, 176 Hirschman, A. 8 Hodgson, G. 209 Hofer, C. 120 Holtz-Eakin, E. 122 Hood, N. 97 Hoover, E. 33, 176 Howells, J. 61 Hurst, E. 120 Hyry, M. 117–24 Iammarino, S. 69 Isaksen, A. 61, 137, 138, 143 Isard, W. 176 Jacobs, J. 15, 17, 84 Jaffe, A. 61, 69, 176 Jensen, B. 22 Johnson, B. 2 Kaldor, N. 180 Kargon, R. 117–18, 122 Keeble, D. 2, 61, 175 Kenney, M. 117, 122, 277 King, C. 61 Kitson, M. 15 Klein, J. 15 Klepper, S. 20, 126 Kogut, B. 62

294 Name index Krackenhardt, D. 138 Krugman, P. 1, 6, 15, 33, 69, 181, 188 Lambooy, J. 116, 117, 122, 137, 145 Larsson, F. 61 Lawson, C. 62, 200 Lazzeretti, L. 4, 19 Lehrer, M. 19 Lerner, J. 118, 277 Leslie, S. 117–18, 122 Levinthal, D. 140, 199 Lewicki, R. 148 Lewis, T. 62 Leyshon, A. 61, 277, 282 Lissoni, F. 62, 94 Loasby, B. 7, 34 Longhi, C. 105, 108 Lorenz, E. 200, 281 Lösch, A. 8, 13 Lucas, G. 188 Lundan, S. 84 Lundequist, P. 61 Lundmark, M. 61 62 Lundvall, B. 2, 51, 94, 200, 218, 226 Lusardi, A. 120 MacKinnon, D. 61 Madsen, T. 62 Maillat, D. 35–7, 41 Malecki, E. 137, 138, 176, 277 Malerba, F. 200, 220, 225 Malmberg, A. 12, 51–4, 60–1, 206, 218, 224 March, J. 138, 142, 157 Markgren, B. 61 Markusen, A. 15, 45, 115, 117 Marrewijk, C. 2 Marshall, A. 5–7, 24, 30, 31, 34, 69–72, 206 Martin, R. 116 Martin, R.L. 1, 2, 5, 9, 11, 12, 14, 20, 30, 54, 59, 95, 168, 173–4, 199, 272, 280 Maselli, M. 99 Maskell, P. 2, 30, 51–3, 93, 176, 206, 218, 224 Matthews, R. 201 McAllister, D. 148 McCann, P. 91, 199, 257, 274, 279 McNaughton, R. 84 Metcalfe, S. 203 Meyer, A. 117 Mills, E. 176 Morgan, K. 1, 2, 19, 94, 200

Narin, F. 61 Nelson, R. 116, 139 Niosi, G. 222, 228 Nonaka, I. 69, 200, 225 Nooteboom, B. 138, 140–5 North, D. 180, 201 Norton, R. 174 O’Brien, R. 28 O’Malley, E. 15 Ohlin, B. 180 Oinas, P. 137, 138 Onida, F. 76 Orlando, M. 129 Orsenigo, L. 125 Oswald, A. 120, 122 Owen-Smith, J. 61, 218, 226 Pandit, N. 78 Paniccia, I. 2, 6, 69 Parker, E. 127 Patton, D. 117 Pavitt, K. 225 Pedersen, C. 62 Penrose, E. 16, 96 Perroux, F. 4, 11, 176 Perrow, C. 93 Perry, M. 16 Phelps, N. 103 Pietrobelli, C. 69 Piore, M. 1, 4, 6, 69, 76, 93, 167 Pires, A. Da Rosa 4 Pisano, G. 16 Porter, M. xvii, 1–23, 35, 39–40, 50, 52–5, 69–75, 90, 165–75, 199–200, 218–19, 255–7 Poterba, J. 118 Powell, W. 61, 142, 202, 218, 220 Power, D. 53–4, 57, 60–2, 224 Prahalad, C. 142 Provasi, G. 71, 84 Pyke, F. 69 Quévit, M. 35–7 Raines, A. 57 Rallet, A. 75 Rama, R. 108 Refolo, M. 61 Reich, R. 1, 31 Richardson, G. 16, 30 Richardson, H. 180 Roberts, E. 129 Rodrigues-Posé, A. 61

Name index 295 Romanelli, E. 121 Romer, P. 7, 188 Romijn, H. 19 Ronde, T. 62 Rosenfeld, S. 57, 284 Rosenkopf, L. 62 Rowthorn, R. 181 Rullani, E. 69, 81, 84 Russo, M. 11 Sabel, C. 1, 4, 6, 69, 76, 93 Sakakibara, M. 61 Sako, M. 18 Saxenian, A. 4, 69, 108 Schmitz, H. 76 Schmookler, J. 120 Schon, D. 206 Schumpeter, J. 4–18, 203 Scott, A. 1, 2, 4, 127, 176, 237, 246 Sengenberger, W. 69 Senn, L. 35–7 Sforzi, F. 76 Shapiro, S. 148 Signorini, L. 71 Simmie, J. 167, 175, 176 Smith, A. 105–6 Solow, R. 188 Sölvell, Ö. 57 Song, J. 62 Sorensen, O. 121 Soskice, D. 19 Spender, J. 142 Sripaipan, C. 117 Stankiewicz, R. 206 Steiner, M. 16, 203 Storper, M. 1, 51, 97, 175, 200, 218, 222, 237, 278, 279 Stuart, T. 121 Sturgeon, T. 222, 228–9 Stutzer, A. 120 Sunley, P. 2, 7, 11, 13, 14, 20, 30, 54, 59, 75, 95, 199, 218, 224, 272, 280 Swann, P. 61, 78, 176, 257 Teece, D. 16, 96 Terkla, D. 3 Terman, F. 127 Tessieri, N. 75

Teubal, M. 117, 122 Thirlwall, A. 180 Thisse, J. 6 Thrift, N. 61, 81, 167, 277, 282 Thurow, L. 189 Tiebout, C. 180 Toccafondi, D. 99 Tödtling, F. 175 Torre, A. 75 Tyler, P. 15 Utterback, J. 8 Van Egeraat, C. 15 Van Oort, F. 275 Varaldo, R. 4, Veblen, T. 53, 201 Veltz, P. 178 Venables, A. 2, 6, 51, 97, 181, 218, Vernon, R. 8 Viesti, G. 76 Vinodrai, T. 232 Visser, E. 19 Von Burg, U. 122 Von Hippel, E. 51, 226 Wallsten, S. 61, 197 Weber, A. 8, 13 Weber, M. 58 Welz, G. 62 Wernerfeld, B. 96 Whetten, D. 31 Wilkinson, F. 2, 13, 14, 20, 76 Williamson, O. 147, 201 Winter, S. 16, 116, 139 Wymbs, C. 69, 84 Yao, D. 62 Yoon, Y. 5 You, J. 76 Young, A. 16 Zand, D. 149, 154 Zeller, C. 61, 125 Zollo, M. 16 Zook, M. 277 Zuchella, A. 137 Zucker, L. 61, 227 Zysman, J. 277

Subject Index

agglomerations 55, 92 as spatial clusters 1, 54, 69 centripetal/centrifugal, 34 concentrated drivers 20 geographical 8, 92 industrial, 2, 199 and knowledge spillovers, 219–25 local, xvii of SMEs, 90, 92 satellite. 104 technology. 107 theory. 165 urban/metropolitan, 69, 240–50 biotechnology, 3, 8, 16 bioscience, 21, 120 entrepreneur, 120 life sciences, 228–50 Canada, 219, 227–30, 245 China, 12, 24, 190 cities, 1–3, 66–72, 226–30 and clusters, 228 and local labour market dynamics, 240–1 conurbations, 70 resurgence, 1 transportation, 262–4 clusters, 3–8, 20, 30–50, 51–90, 167–80, 218–57 and clustering, 54 and globalization, 162, 189–93 approach, 50, 51–3 categorization, 20, 78, 90–100 composition, 3 concept, xvii, 3, 4, 30, 51 confusion, 51, 59 chaotic, 16, 52 definition, 2, 8, 12, 30, 35, 50–4, 74, 138, 199

dynamics, 78, 90–100 effects, 258–9 embedding, 9, 15, 71, 139–50 first usage, 4 formation, 2, 14, 115 idea, xvii local barriers, 3 local business, 3 model, xvii, 23 Porter’s, xvii, 1–23, 39–40, 50, 52–5, 69–75, 90, 119–21, 165–75, 199–203, 218–22, 255–7 notion, xvii phenomenon, 30 policy, xvii, 3, 20–30, 159–61, 255–71 Smithian, 20 Denmark, 16, 224–6 Öresund, 224–6 Salling, 16 Department of Trade & Industry (DTI), 69, 166 economics, 4–40, 69–90, 122–70, 164–85 capabilities, 69, 96 collaboration, 38, 60, 62 competences, 96, 138–46 competition 14, 60 competitiveness, xvii, 2–9, 31, 35–40, 91, 164–85 diamond, 9–11, 52, 219–21 regional advantage, xvii, 165 cooperation, 2 development, 1 externalities, 4–5, 16, 31, 69 local, 5 lock-in, 145–50

Subject index 297 spillovers, 140, 145–50, 190 technological, 69 returns, 4–8, 62–80 increasing, 4, 8, 69–78 localizaton, 5, 69 neoclassical, 8 evolutionary, 69, 80–90, 123–65, 203–7 cluster dynamics, 14, 123–25, 137–62 complexity, 94 differentiation, 6 integration, 6 mechanisms, 69 nature of capitalism, 120 path, 81–90 systemic nature, 52 institutional, 2–40, 96–116, 138–46, 188–217 agencies, 2 and economic performance, 200 embedding, 138–46 structural, 146 institutional, 146 economics, 201 immaterial, 37 Internet, 116 knowledge-sharing, 202–4 local, 99 network, 40, 96 perspective, 188–217 thinking, 202–5 trade associations, 2 neo-Schumpeterian, 8 entrepreneurship, 85, 115–36, 194–200 entrepreneurial values, 37 high technology, 120 innovation and, 8 in cluster formation, 115 knowledge, 121 knowledge spillover, theory of, 194–6 latent, 120 local milieux, 8, 121 networks, 123 skills, 220 small and medium enterprises (SMEs), 69, 71, 85, 90 start-ups, 194 European Commission, 69, 166 Finland, 17, 127 Oulu, 17, 127 France, 3

geography, 4–10, 30–50, 65–80, 105–9 agglomerations, 8, 92 satellite, 104 concentrated, 106 technology, 107 economic, 4, 51, 69 geographical economics, 6 industrial localization, 6 local interactions, 78 new economic, 6 of innovation, 69 spatial embeddedness, 8 territories, 31 territorial logic, 41 Germany, 3 Baden-Württemberg, 19, 167 Heidelberg, 19 Leipzig, 19 Ruhr valley, 4 Stuttgart, 19 globalization, 1–5, 58–60, 189–233 and regional clusters, 189–97 and technological change, 1 clusters in global context, 218–33 global economy, xvii, 1–5 global mosaic, 2 knowledge, 59 governance, 97–8, 138–50 associative, 200–6 hybrid, 208–13 human capital, 75–7, 120–4, 145–7, 218–21, 240–91 human asset specificity, 146 leadership (Marshallian), 76 managerial, 220 metropolitan, 240–91 occupational structures, 244 university trained, 123 Industrial District, 5, 12, 16, 30–2, 69–77, 90–114 ‘districtualization’, 4, 73 diversification, 93 flexible specialization, 4 high technology, 8 industrial atmosphere, 5, 79 life cycle, 38 Marshallian, 78, 236 neo-Marshallian, 4, 6, 12, 75 specialization, 93 Third Italy, 2, 6, 8, 137, 275 Information and Communication Technology (ICT), 3, 16, 22, 226

298 Subject index clusters, 229 multimedia, 157–8 opto-electronics, 223 software, 226 industry, 1–22, 50–70, 93–7, 115–17, 156–9 anchor, 116 agents, 69 creative, 3 energy, 21 finance, 276–85 firms, 52, 69 horizontal and vertical linkages, 94–6 industrial localization, 1 new industrial spaces, 2 supplier, 52 system, 54 innovation, 5, 78, 91, 135–45, 175–80, 219–20 and knowledge creation, 8 and productivity, 5 cluster-based, 50, 52, 121, 137 disembedded, 137–45 embedded, 2, 8, 78, 137, 145, 208, 219–22 endogenous, 2 geography of, 69, 175–80 innovativeness, 77 milieu, 37, creative, 58 local, 53 networks, 37 network theories, 8 performance, 3 scissors metaphor, 120 spatial dimension, 189 Italy, 6–12, 17, 75, 99 Belluno, 17 Como, 17 Emilia-Romagna, 6, 11, 21, 167 Carpi, 17 Parma, 17 Sassuolo, 11, 21 Florence, 12 Marche, 6 Montebelluna, 17 Prato, 12, 75, 99, 103–5 Santa Croce, 103 Veneto, 6, 75 knowledge, 7, 60, 94–107, 188–217, 218–33 absorptive capacity, 140

accumulation, 26 and competitive advantage, 200–7 asymmetric, 23 codified, 226 communities, 7 creation, 8, 59, 188–217 differentiation, 143, 158 diffusion, 188–217, 199 economy, 3 globalized, 59 entrepreneurial, 121 exchange, 94 exploitation, 138–44 exploration, 138–44 filter, 192–5, 196–8 flows, 218–33 governance, 97–8 institutions, 107 know-how, 37 new economy, 3 pools, 275 sharing, 202–7 spillovers, 60–1, 219–25 tacit, 226 unskilled, 193, 236–53 labour, 36–65, 75–80, 245–52 and capital, 52, 79 crafts, 239 division of, 38, 71 force, 27 highly skilled, 223 local labour market, 76–9 low wage, 137 mobility, 38 organization, 328 shortages, 53 skilled, 62 specialized, 62 technical, 239 learning, 77 collective, 8 co-ordination, 94 dynamic, 36 economies, 2 experimentation, 116 interactive, 51 mechanisms, 69 organizational, 206 Netherlands, The, 3, 157–8 networks, 4–15, 137–58, 257–60 effects, 258–9 exploration and exploitation, 138–56

Subject index 299 GREMI, 14, 31 horizontal, 177 innovation, 37 inter-firm, 7 of entrepreneurs, 123 of firms, 138 relationships, 98 social capital, 4, 208 theory, 14 vertical, 77 Norway, 143 OECD, xvii, 3, 69, 166, 178 policy, 20–33, 157–62, 254–72 cluster, xvii, 20–30, 159–61, 255–71 cluster-based regional innovation, 50 competitiveness, 39 government, 120 institutions, 58 issues (journals), 30 makers, 3, 197 options, 37 partners, 119 towards disembedding, 137 Portugal, 3 proximity, 1–3, 50–9, 138–52 and agglomeration, 59 and efficiency, 91 geographical, 3, 11, 54–5 concentration, 2, 54, 59, 69 death of distance, 188–91 face-to-face interaction, 51, 224 physical, 94 production formations, 16 spatial, 1, 51–2, 94 relational, 35–8, 137–52, 218 capital, 37 cognitive distance, 51, 138–46 embedding139–50 flexibility, 91 ‘matters’, 218 organizational, 94 shared codes, 94 region(al), 34–42, 70, 114–17, 136–42, 280–6 analysis, 1 categorization, 20 clusters, 192–6 cluster-based, 50, 282–5 competitiveness, xvii

competitive advantage, 116 culture, 137 development, xvii, 35–40, 116 development agencies, 3 development policy, 4, 40 economies, 2, 69 industrial transformation, 115 innovation systems, 8, 69, 138–42 knowledge spillovers, 194–6 resurgence, 1 scale, xvii R&D, 17, 35, 79, 125, 140–5, 194–5 science, 28–32, 61–3, 124–5, 227–9 social capital, 14, 16, 17–18 coalitions, 199–201 collaboration, 149 collective behaviour, 199–201 commonalities, 3, 12 complementarities, 3, 12 cooperation, 77 linkage analysis: horizontal and vertical, 3, 34 networking, 37 reciprocity, 37 ‘structural holes’, 141 trust, 18, 37, 77, 146, 154, 208, 232 South Korea, 117 Spain, 21 Basque Country, 21 Taiwan, 119, 128 technology, 52, 69, 78 change, 120, 124 competences, 96 economic, 96 technological, 96, 129 generic, 115 high technology, 79, 84, 129, 220 licenses, 120 patents, 69 path dependent, 124–5 United Kingdom, 3, 24 Scotland, 21 Wales, 105 universities, 60–2, 115–24 and clusters, 2 industry collaboration, 61 pro-technology transfer, 116 research, 119 trained personnel

300 Subject index USA, 1, 23 Boston, 4, 16, 119–22 Cambridge, MA, 4, 118 Harvard, 4 Business School, 11, 69 Route 128, 69, 122–3 Hollywood, 23, 127, 236–53 New Jersey, 117, 126

Silicon Valley, 4, 16, 69, 71, 91, 118–30, 188, 197, 257 Washington DC, 116 Capitol Region, 116–50 Venture Capital, 116, 220–2, 276–9 World Bank, xvii, 3

eBooks – at www.eBookstore.tandf.co.uk

A library at your fingertips!

eBooks are electronic versions of printed books. You can store them on your PC/laptop or browse them online. They have advantages for anyone needing rapid access to a wide variety of published, copyright information. eBooks can help your research by enabling you to bookmark chapters, annotate text and use instant searches to find specific words or phrases. Several eBook files would fit on even a small laptop or PDA. NEW: Save money by eSubscribing: cheap, online access to any eBook for as long as you need it.

Annual subscription packages We now offer special low-cost bulk subscriptions to packages of eBooks in certain subject areas. These are available to libraries or to individuals. For more information please contact [email protected] We’re continually developing the eBook concept, so keep up to date by visiting the website.

www.eBookstore.tandf.co.uk

Clusters and Regional Development (Regions and Cities)

Regional Development in the Knowledge Economy (Regions and Cities)

Knowledge Externalities, Innovation Clusters and Regional Development

Clusters and Globalisation: The Development of Urban and Regional Economies

Local and Regional Development

Sustainable Cities (Regional Development and Public Policy Series)

European Cohesion Policy (Regions and Cities)

Visions of Sustainability: Cities and Regions

Postmetropolis: Critical Studies of Cities and Regions

New Technology and Regional Development

Regional Innovation Strategies: The Challenge for Less-Favoured Regions (Regional Development and Public Policy Series)

Vital Coalitions, Vital Regions: Partnerships for Sustainable Regional Development

Regional Integration and Development (World Bank Trade and Development Series)

Regional Integration and Development (World Bank Trade and Development Series)

Towards Effective Place Brand Management: Branding European Cities and Regions

Endogenous Development: Networking, Innovation, Institutions and Cities

Leadership and Institutions in Regional Endogenous Development

The Spatial Economy: Cities, Regions, and International Trade

Handbook of Regional Growth and Development Theories

Devolution & Regionalism: The UK Experience (Regions and Cities)

Geographies of the New Economy (Regions and Cities)

Innovation Performance and Clusters: A Dynamic Capability Perspective on Regional Technology Clusters

Atomic and molecular clusters

Atomic clusters and nanoparticles

Trade and Development Report 2007: Regional Cooperation for Development

Economics of Agglomeration: Cities, Industrial Location, and Regional Growth

Climate Change and Economic Development: SEA Regional Modelling and Analysis

Japanese Cities in the World Economy (Conflicts in Urban and Regional Development)

Clusters and Regional Development (Regions and Cities)

Regional Development in the Knowledge Economy (Regions and Cities)

Knowledge Externalities, Innovation Clusters and Regional Development

Clusters and Globalisation: The Development of Urban and Regional Economies

Local and Regional Development

Sustainable Cities (Regional Development and Public Policy Series)

European Cohesion Policy (Regions and Cities)

Visions of Sustainability: Cities and Regions

Postmetropolis: Critical Studies of Cities and Regions

New Technology and Regional Development

Regional Innovation Strategies: The Challenge for Less-Favoured Regions (Regional Development and Public Policy Series)

Vital Coalitions, Vital Regions: Partnerships for Sustainable Regional Development

Regional Integration and Development (World Bank Trade and Development Series)

Regional Integration and Development (World Bank Trade and Development Series)

Towards Effective Place Brand Management: Branding European Cities and Regions

Endogenous Development: Networking, Innovation, Institutions and Cities

Leadership and Institutions in Regional Endogenous Development

The Spatial Economy: Cities, Regions, and International Trade

Handbook of Regional Growth and Development Theories

Devolution & Regionalism: The UK Experience (Regions and Cities)

Geographies of the New Economy (Regions and Cities)

Innovation Performance and Clusters: A Dynamic Capability Perspective on Regional Technology Clusters

Atomic and molecular clusters

Atomic clusters and nanoparticles

Trade and Development Report 2007: Regional Cooperation for Development

Economics of Agglomeration: Cities, Industrial Location, and Regional Growth

Climate Change and Economic Development: SEA Regional Modelling and Analysis

Japanese Cities in the World Economy (Conflicts in Urban and Regional Development)

Recommend Documents