Investigating Welfare State Change: The Dependant Variable Problem in Comparative Analysis

Investigating Welfare State Change Investigating Welfare State Change The ‘Dependent Variable Problem’ in Comparative ...

Author: Jochen Clausen | Nico A. Siegel

16 downloads 466 Views 1MB 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

Investigating Welfare State Change

Investigating Welfare State Change The ‘Dependent Variable Problem’ in Comparative Analysis

Edited by

Jochen Clasen Professor of Comparative Social Policy, University of Edinburgh, UK

Nico A. Siegel Senior Research Manager, TNS Infratest Sozialforschung, Munich, Germany

Edward Elgar Cheltenham, UK • Northampton, MA, USA

© Jochen Clasen and Nico A. Siegel 2007 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical or photocopying, recording, or otherwise without the prior permission of the publisher. Published by Edward Elgar Publishing Limited Glensanda House Montpellier Parade Cheltenham Glos GL50 1UA UK Edward Elgar Publishing, Inc. William Pratt House 9 Dewey Court Northampton Massachusetts 01060 USA A catalogue record for this book is available from the British Library Library of Congress Cataloguing in Publication Data Investigating welfare state change : the ‘dependent variable problem’ in comparative analysis / edited by Jochen Clasen, Nico A Siegel. p. cm. Studies from ‘thematic seminar’ held in May 2005 at the University of Stirling, and then from a panel organized at the 2005 ESPAnet [European Social Policy Analysis network] conference in Fribourg, Switzerland, with the addition of results and subsequent discussions on the subject Includes bibliographical references and index. 1. Welfare state—OECD countries—Congresses. 2. Social policy—OECD countries—Evaluation—Congresses. 3. Public welfare—Evaluation—OECD countries—Congresses. I. Clasen, Jochen. II. Siegel, Nico A. HN18.3.I58 2007 330.126—dc22 2006102954 ISBN 978 1 84542 739 9 (cased) Printed and bound in Great Britain by MPG Books Ltd, Bodmin, Cornwall

Contents List of figures List of tables Contributors Acknowledgements

vii viii xi xii

PART I THE ‘DEPENDENT VARIABLE PROBLEM’ IN COMPARATIVE WELFARE STATE RESEARCH 1. Comparative welfare state analysis and the ‘dependent variable problem’ Jochen Clasen and Nico A. Siegel 2. More than data questions and methodological issues: theoretical conceptualization and the dependent variable ‘problem’ in the study of welfare reform Christoffer Green-Pedersen 3. Too narrow and too wide at once: the ‘welfare state’ as dependent variable in policy analysis Giuliano Bonoli

3

13

24

PART II MEASURING AND ANALYSING ‘WELFARE EFFORTS’: SOCIAL EXPENDITURE REVISITED 4. When (only) money matters: the pros and cons of expenditure analysis Nico A. Siegel 5. Social expenditure under scrutiny: the problems of using aggregate spending data for assessing welfare state dynamics Johan De Deken and Bernhard Kittel 6. Social rights, structural needs and social expenditure: a comparative study of 18 OECD countries 1960–2000 Olli Kangas and Joakim Palme

v

43

72

106

vi

Contents

PART III BEYOND SPENDING: WELFARE STATE GENEROSITY, SOCIAL RIGHTS AND OBLIGATIONS 7. Welfare state generosity across space and time Lyle Scruggs 8. Levels and levers of conditionality: measuring change within welfare states Jochen Clasen and Daniel Clegg 9. Exploring diversity: measuring welfare state change with fuzzy-set methodology Jon Kvist

133

166

198

PART IV CAPTURING THE NATURE OF WELFARE STATE CHANGE 10. Convergence in European welfare state analysis: convergence of what? Julia S. O’Connor 11. (In)Dependence as dependent variable: conceptualizing and measuring ‘de-familization’ Sigrid Leitner and Stephan Lessenich 12. Pension reform: beyond path dependency? Sven Jochem References Index

217

244 261

281 313

List of figures 3.1 Spending on old (OSR) and on new (NSR) social risk policies in OECD countries as a percentage of GDP, five-year averages, 1997–2001 4.1 Public social expenditure in 21 OECD countries, 1980, 1993 and 2001 4.2 First order differences, annual total social expenditure, 1980–2001 6.1 The determinants of public social expenditure 6.2 Social rights and social spending in 18 OECD countries, 1960–2000 6.3 Politics, social rights and social spending in 18 OECD countries, 1960–2000 7.1 Comparative welfare state entitlements decommodification indices (Esping-Andersen scoring method) 7.2 Revised (and original) decommodification scores for 18 OECD countries 7.3 Revised decommodification scores: mean values, 1970–2001 8.1 Levels and levers of conditionality in social programmes 8.2 Conditionality shifts in unemployment benefit reforms in four countries 9.1 The dichotomy of residual and institutional welfare states 9.2 Accessibility and generosity of social rights 9.3 Analytical property space for social citizenship 12.1 A classification of European pension systems 12.2 Pension characteristics in the early 1980s 12.3 Pension dynamics – programmatic profiles

vii

38 45 65 109 114 125 156 159 161 175 183 200 202 203 266 269 279

List of tables 3.1 3.2

Forms of social protection in Western Europe, early 1990s Indicators of pension generosity at the end of the ‘trente glorieuses’ 4.1 Social rights and welfare efforts 4.2 First order differences, annual total social expenditure, 1980–2001: year and country specific averages 5.1 Expenditure on mandatory private and voluntary private sickness benefits in Germany and the Netherlands as a percentage of GDP 5.2 Expenditure on sickness benefits as a percentage of GDP according to SOCX and ESSPROS 5.3 Old-age and survivors’ pension benefits as a percentage of GDP according to SOCX and ESSPROS 5.4 Receipts and expenditure on benefits by Dutch pension funds in million € (and as a percentage of GDP) 5.5 Public, mandatory private and voluntary private social expenditure on old-age cash benefits and survivors’ benefits in four countries as a percentage of GDP 5.6 Coverage of types of social expenditure by the SOCX database 5.7 Overview of the various dependent variables 5.8 Correlations between various social expenditure measures 5.9 Social expenditure dynamics, 1993–2001: dependent variables compared 5.10 Social expenditure dynamics, 1994–98: dependent variables compared 5.11 Pension expenditure dynamics, 1994–98: dependent variables compared 5A.1 A comparison of SOCX and ESSPROS categories 6.1 Regressions of social spending (OECD data) on aspects of social rights and structural factors, 1960–2000, unstandardized coefficients 6.2 Regressions of spending on unemployment on aspects of social rights and structural factors, 1960–2000, unstandardized coefficients viii

29 37 60 64

77 77 78 83

85 87 89 90 95 97 98 105

117

117

Tables

6.3

6.4

6.5

7.1 7.2 7.3 7.4 7.5 7.6 7.7 7.8 8A.1 8A.2 8A.3 8A.4 9.1 9.2 9.3

9.4

9.5 10.1 10.2 10.3

Regressions of spending on sickness benefits on aspects of social rights and structural factors, 1960–2000, unstandardized coefficients Regressions of spending on pensions on aspects of social rights and structural factors, 1960–2000, unstandardized coefficients Regressions of spending on child allowances on aspects of social rights and structural factors, 1960–2000, unstandardized coefficients Dimensions of the decommodification/generosity index Unemployment benefit replacement rates Sickness benefit replacement rates Evolution of social pension replacement rates Evolution of standard pension replacement rates Evolution of unemployment and sickness insurance coverage Pension take-up rates Decommodification and generosity scores, 1980 Major legislative changes in the conditionality of unemployment support, 1980–2005 (United Kingdom) Major legislative changes in the conditionality of unemployment support, 1980–2005 (Germany) Major legislative changes in the conditionality of unemployment support, 1980–2005 (France) Major legislative changes in the conditionality of unemployment support, 1979–2005 (Denmark) Truth table of social citizenship Eight-cell table of social citizenship Specification of empirical indicators and the translation of raw data into fuzzy membership scores and verbal labels Fuzzy membership scores for Danish unemployment insurance benefits in ,  and , 1990–98 Fuzzy membership scores for seven European countries in unemployment insurance ideal types, 1990–99 Social protection expenditure (as a percentage of GDP) Social protection expenditure in Purchasing Power Standards GDP per capita as percentage of EU 15 100 selected years, 1970–2004

ix

119

121

123 141 144 145 146 148 150 153 158 187 189 192 195 204 204

207

209 212 228 229 233

x

Tables

10.4 Social protection expenditure as a percentage of GDP, relative to EU 15 100 and PPS for selected years, 1970–2003 10.5 Cohesion Countries: catch-up relative to EU 15 accession to 2003 11.1 Conceptual reductionism in the debate on ‘de-familization’ 11.2 Intuitive and counterintuitive relations between social/ economic (de-)familization and economic/social (in)dependence from the care giver’s perspective 12.1 Pension dynamics – spending and politics

234 236 253

253 278

Contributors Giuliano Bonoli is Professor of Social Policies at IDHEAP in Lausanne, Switzerland. Jochen Clasen is Professor of Comparative Social Policy in the School of Social and Political Studies at the University of Edinburgh, UK. Daniel Clegg is Lecturer in Comparative Social Policy at the University of Stirling, UK. Johan De Deken is Lecturer in Sociology in the Department of Sociology and Anthropology, University of Amsterdam, the Netherlands. Christoffer Green-Pedersen is Research Professor in the Department of Political Science at the University of Aarhus, Denmark. Sven Jochem is Assistant Professor of Politics at the University of Konstanz, Germany, and the University of Lucerne, Switzerland. Olli Kangas is Research Professor at the SFI, the Danish National Institute of Social Research, Copenhagen, Denmark. Bernhard Kittel is Professor of Methods of Empirical Research, Department of Sociology, University of Oldenburg, Germany. Jon Kvist is Senior Research Fellow at the SFI, the Danish National Institute of Social Research, Copenhagen, Denmark. Sigrid Leitner is Lecturer in Sociology in the Institute of Sociology, University of Göttingen, Germany. Stephan Lessenich is Professor of Sociology in the Institute of Sociology, University of Jena, Germany. Julia S. O’Connor is Professor of Social Policy in the School of Policy Studies, University of Ulster, Belfast, UK. Joakim Palme is the Director of the Institute for Futures Studies, Stockhom, Sweden. Lyle Scruggs is Associate Professor of Comparative Politics, West European Politics and Comparative Political Economy in the Department of Politics at the University of Connecticut, USA. Nico A. Siegel is Senior Research Manager, TNS Infratest Sozialforschung, Munich, Germany.

xi

Acknowledgements The idea for this volume originated during the second annual conference of ESPAnet (the European Social Policy Analysis network) in Oxford in 2004. Each ESPAnet conference involves at least one stream of papers which deal with comparative methods in cross-national social policy analysis. In this context the editors of this volume discussed the current international debate about welfare reform and welfare state change, and the notion that the strong interest in this topic within the scientific community has not been matched by considerable academic progress. There are plenty of leitmotifs aimed at capturing the nature of welfare state change (or the lack of it), as well as competing accounts of how to explain policy reform. At the same time, there has been a noticeable paucity of how to empirically and systematically conceptualize, operationalize and measure change within and across welfare states. It is this ‘dependent variable problem’ and its various aspects which led the editors to invite international experts to a ‘thematic seminar’ in May 2005 at the University of Stirling, and subsequently to organize a panel on this topic at the 2005 ESPAnet conference in Fribourg, Switzerland. This book provides the results of these and subsequent discussions on the subject. It aims to reflect on a number of dimensions of the ‘dependent variable problem’ within comparative welfare state analysis, and to provide suggestions as to how to address and tackle it empirically. We are grateful to ESPAnet for providing the necessary fora for preliminary discussions, and to the Department of Applied Social Science, University of Stirling for hosting the above mentioned seminar in May 2005. We would like to thank Jacqueline Davidson for helping with its organization and all participants, and especially Frank Castles and Jane Lewis, for useful comments and advice. We owe thanks to our previous employers, i.e. the Universities of Stirling and Kent respectively, for enabling us to put the necessary time into editing this volume. We are indebted to all contributors for their efforts and patience with us and our repeated requests for revision and refinement. Finally we would like to thank Catherine Elgar for her interest in the original idea for this project and her continuous support throughout. Jochen Clasen Nico A. Siegel xii

PART I

The ‘dependent variable problem’ in comparative welfare state research

1. Comparative welfare state analysis and the ‘dependent variable problem’ Jochen Clasen and Nico A. Siegel INTRODUCTION Reforms of public pension schemes, health care systems and labour market programmes have been amongst the most salient political issues for some years. The reasons for governments to engage in welfare reform vary across countries, but generally include budgetary pressures and projected increases in spending on health care, social services and public pension systems. Economic internationalization and associated shifts in production and employment patterns have jointly contributed to new labour market risks, problems of long-term unemployment or labour market inactivity. Changing household formations as well as political forces, either of a domestic or a supranational nature (e.g., Europeanization), put additional pressures on policy makers to adapt existing forms of welfare state provision, and reinforce perceptions of welfare reform as a political topic which is likely to remain high on the public policy agenda for some time to come. Moreover, in terms of the political discourse, traditional left concepts favouring big welfare statism have become rather scarce. Instead of the public provision of social protection perceived as market restricting and correcting, current debates and new policies appear to present a shift towards market enabling principles. Within social science the notion of changing welfare states as a topic for comparative research is not new. In fact, it was the emergence and subsequent expansion of social protection until the 1970s which stimulated a large number of investigations into the causes for welfare state growth (e.g. Flora, 1987a) as well as processes of cross-national variation (e.g. Esping-Andersen, 1990). However, mirroring the more recent challenges to and current reform initiatives within advanced welfare states, the focus of research has shifted profoundly over the last decade. Whereas in the 1970s and 1980s most social scientists were studying the economic context, social 3

4


configurations and political coalitions and institutions which had enabled the massive expansion of the welfare state in the 20th century, nowadays they are investigating how governments are able to ‘impose losses’ without losing political power. An impressive number of books and articles have dealt with this ‘new politics of the welfare state’ (Pierson, 2001). Analysing the political logic of ‘welfare retrenchment’ and ‘blame avoidance’ has become a major research field for political scientists. Many sociological accounts of welfare state change have shifted their focus too. The new risk configurations which seem to have emerged in the transition to postindustrial societies and which challenge welfare state arrangements that had been established in the context of old, traditional risk contexts of industrial societies have been at the centre of a number of recent research projects and books (Taylor-Gooby, 2004; Bonoli, 2005; Armingeon and Bonoli, 2006). Despite somewhat subject specific perspectives, authors across different disciplines seem to agree that we are witnessing an era of considerable challenges to existing arrangements for social protection. Yet whereas some diagnose far reaching transformations, others identify incremental adjustments and stepwise recalibration. More generally, there is little consensus on either the nature of welfare state change or its scope. Indeed, despite a growing availability and comparability of relevant data, the body of comparative welfare state research cannot be described as resting on a widely shared empirical basis or common understanding about how much change there is, what drives change, or how the nature of change should be understood or conceptualized. This lack of progress in the description and explanation of welfare state change across countries and time is, we argue, at least to a considerable extent due to a ‘dependent variable’ problem, i.e. a noticeable absence of reflection on how to conceptualize, operationalize and measure change within welfare states. Of course, the kinds of problems we identify and discuss in this volume may be similar to those in many other areas of comparative studies in social sciences. Cross-national inquiries into processes of social dynamics, political regime change, social revolutions, or of structural changes in the economy, to name just a few examples, all face problems of conceptualization and operationalization. Comparative studies of welfare state change may thus not be a special case. Yet, as this volume aims to demonstrate, the ‘dependent variable problem’ deserves to be taken more seriously within comparative welfare state analysis than it has been to date. More than from a purely methodological perspective it raises important questions of theoretically guided concept building, and it has important implications for policy discourses too. With this volume we are hence taking a step back in order to focus on some of the most salient dependent variable problems in comparative

Comparative analysis and the ‘dependent variable problem’

5

welfare state research – and on ways in which they might be addressed or even overcome. Ultimately such a step, we hope, will help to enable a more cumulative build-up of empirical evidence and contribute to constructive theoretical debates about the causes for welfare state change. Two points of clarification should be made. First, we acknowledge that a concentration on the welfare state as ‘dependent variable’ cannot overlook the point that welfare states can figure as important independent variables, too. For example, the size and institutional structure of national pension systems or health care programmes may influence the speed and direction of welfare reform. Equally, mature welfare states tend to be rather large employers in many countries. As Esping-Andersen has demonstrated in his ground breaking study of welfare regimes (Esping-Andersen, 1990), the welfare state itself is a major force stratifying societies and personal life contexts. Accounts of welfare state dynamics have thus to consider preferences and interests not only of relevant policy makers who are perceived as ‘exogenous political actors’ on the one hand and welfare state clientele on the other, but also possible vested interests of welfare state employees and the complex interplay of actors shaping the development of contemporary welfare arrangements. However, this volume does not intend to provide an all-encompassing manual on the challenges and pitfalls within the study of change and reform in modern welfare states. Adopting a less ambitious and more parsimonious approach, we deliberately restrict ourselves to discussions about different aspects of the ‘dependent variable problem’. Put differently, our intention is to discuss why and in which ways the ‘dependent variable problem’ is relevant for comparative accounts of welfare state change, not only in terms of basic methodological or technical problems but also in terms of substantive and theoretical aspects within this field of study. Second, we do not use the term ‘dependent variable’ in a narrow, technical or even exclusively statistical sense. Instead the term has been chosen in order to underline the general task of all chapters, i.e. to engage with the challenge of conceptualizing and measuring welfare state change – rather than contributing (further) to the debate about causes for change. Of course, it can be argued that the former cannot be discussed in isolation from the latter. Indeed, as Green-Pedersen points out in Chapter 2 of this volume, different theoretical perspectives tend to favour different types of dependent variables. For example, a focus on what Bonoli (2006; and Chapter 3 this volume) has referred to as ‘new social risks’ (such as single parenthood, having relatives in need of long-term care, possessing obsolete skills, insufficient social insurance coverage), and the extent to which welfare states have responded to their emergence, would suggest certain types of concepts and indicators of change. Social spending as a central

6


parameter for assessing change might not be suitable for approaches which focus on gender related aspects of welfare states, or those which aim to test the impact of left parties on welfare state generosity. Instead, as it was not a goal of welfare state advocates to struggle for higher social spending per se (cf. Esping-Andersen, 1990) but to improve the degree of social protection in the context of market economies, composite indices of ‘de-commodification’ (operationalized as a form of benefit generosity) seem to offer a more appropriate means for assessing the influence of prowelfare state advocates on policy making. In other words, as GreenPedersen argues in Chapter 2, the choice of a dependent variable should follow theoretical considerations.

THREE ASPECTS OF THE ‘DEPENDENT VARIABLE PROBLEM’ Collectively, subsequent chapters tackle three separate aspects of the ‘dependent variable problem’: questions of conceptualization, operationalization and measurement. As for the former, depending on analytical interest, the welfare state (or welfare regimes) might be regarded as an inappropriate unit of analysis since its individual components (transfer programmes, services) might differ substantially in terms of the type of risk management, dominant actor constellations, and interdependencies with other policy fields, e.g. as integral parts of political economies (Hall and Soskice, 2001; Ebbinghaus and Manow, 2001). Over the last decade a growing number of authors have argued that the comparative analysis of the welfare state should be disaggregated to investigations into the dynamics of particular policy domains in order to be able to better capture variation both in terms of policy outcome (Huber and Stephens, 2001) as well as processes of reform (e.g. Ferrera and Rhodes, 2000b; Pierson, 2001). Of course, whether changes of or within welfare states are of interest reflects different research questions and approaches. Conventionally and narrowly defined, welfare state programmes consist of statutory benefits and service provision (generally leaving aside education). More broadly defined, the ‘welfare state’ can be conceptualized as all mechanisms which provide social protection against and redistribution of market mechanisms and outcomes. Hence the welfare state not only comprises transfers and services, but also tax expenditures, minimum wages, state regulation of labour and product markets, state recognition of collective bargaining and other interventions, all of which ‘disconnect or buffer income streams from market outcomes’ (Schwartz, 2003). As Giuliano Bonoli discusses in Chapter 3, the decision in favour of a broader or a more narrow boundary


7

of the term ‘welfare state’ is not merely a matter of perspective. It also determines the level of analytical abstraction and thus influences whether or not the explanatory power of different theories of the welfare state can be assessed against each other. To put it differently: both the choice of analytical perspectives and of empirical indicators are shaped by (meta)theoretical considerations and they affect central issues of descriptive and analytical inferences in comparative welfare state research. Following on from the more general aspects covered in this introduction as well as in Chapters 2 and 3, subsequent sections deal with indicators, measurement and concepts. Since it has figured so prominently within comparative welfare state analysis, each of the three chapters of Part II is devoted to the value of ‘social spending’ as a yardstick for measuring the size of the welfare state and to assessing its scope of change. Part III comprises three chapters which present alternative parameters of change which could be used instead of, or to complement, social spending, such as welfare state generosity and benefit conditionality. Finally, Part IV addresses the usefulness and problems of some ‘big concepts’ which have been employed as leitmotifs in analyses aimed at capturing the nature of change in contemporary welfare states: ‘convergence’ (Julia O’Connor in Chapter 10), de-familization (Sigrid Leitner and Stephan Lessenich in Chapter 11), and finally, path dependence (Sven Jochem in Chapter 12).

SOCIAL SPENDING AND OTHER INDICATORS OF WELFARE STATE CHANGE The amount of money spent on social protection programmes has figured as a major indicator particularly within ‘variable oriented’ quantitative comparative studies of welfare state development. Improved and more comprehensive data sources (Castles and Obinger, 2006), disaggregated data at the level of individual social policy programmes (Castles, 2004), as well as more sophisticated statistical techniques (e.g. Kittel and Obinger, 2003) have contributed to the continuous popularity of social spending as a central parameter of studies of change or stability within, and convergence or divergence across, welfare states. Nevertheless, for some time now questions about the nature and appropriateness of social expenditure as the (only) ‘dependent variable’ for quantitative welfare state comparisons have been asked even by prominent scholars within this field (e.g. Castles, 1994). More recently, the debate about the pitfalls of expenditure based welfare state accounts has been revitalized because analyses for the post-1980s era have produced more inconclusive results than studies which concentrated on the expansion of social expenditure in the 1960s and 1970s. Sensitivity

8


towards the choice of statistical techniques, model specifications, and spending indicators contributed to this inconsistency. At a time when the OECD started to publish regular updates of probably the best source for social expenditure figures ever available for comparative welfare state research (the so-called ‘SOCX’, OECD Social Expenditure Database, see OECD, 2005a), critical voices were raised concerning the comparability of data across countries and time. Yet in most comparative studies, such issues tend to be ignored or neglected. Three chapters in this volume critically reassess the use of expenditure analysis in comparative welfare state research. In Chapter 4, Nico A. Siegel presents a general discussion of the use of expenditure data, illustrating the strengths and importance as well as problems of comparative inquiries which mainly (and sometimes exclusively) describe and analyse welfare state change in the light of levels of social spending. In Chapter 5, Johan De Deken and Bernhard Kittel focus on more specific problems. They compare the quality of spending data provided by major international agencies such as the OECD or EUROSTAT. Since definitions of national social policy programmes differ across countries and time, researchers face severe problems when comparing expenditure dynamics. Whereas methodological innovations such as pooled time series analysis seem to offer ever more sophisticated techniques of macroquantitative data analysis, the chapters by Siegel and De Deken and Kittel both demonstrate that research practice is plagued by basic (and some seemingly trivial) problems of data quality, comparability, and non-random measurement error of the dependent variable. Together, these issues generate knotty problems for cross-national studies. Consequently, comparative welfare state researchers run the risk of building their analytical inferences on rather shaky empirical foundations, i.e. on descriptive inferences which may inherently suffer from measurement bias and related issues of the dependent variable problem. The final chapter of Part II deals with the comparative analysis of social expenditure and particularly its relationship to other indicators, such as welfare state generosity as an expression of ‘social rights’. As Olli Kangas and Joakim Palme demonstrate in Chapter 6, levels and changes of social expenditure are affected by two major factors: the generosity of social rights (which is basically a direct consequence of political decisions affecting benefit levels and conditions of eligibility and entitlement) and levels of ‘structural need’ which reflect demographic and other socioeconomic factors. The authors argue that social expenditure data, particularly at the most aggregated level (‘total public social expenditure’), should not be used to infer changes in the generosity of social rights. In short, the categorical imperative of comparative research aimed at investigating


9

the extent of welfare state expansion or retrenchment, rather than analysing merely the change in ‘welfare efforts’, is to move beyond a purely expenditure based and macroscopic perspective of the ‘welfare state’ and engage with more detailed accounts of changes and policies at the level of individual social policy programmes. Indeed, the complexity of advanced welfare states seems to make it tremendously difficult to reduce major dynamics of change to just one or a few social policy indicators, and to explain the dynamics of change with a ‘catch all master theory’ (Siegel, 2002). The suggestion of complementing (or substituting) social expenditure with other indicators of change is the topic spanning the three chapters in Part III. A prominent alternative, also propagated by Kangas and Palme, are ‘social rights’, which are expressed as entitlement conditions (benefit level and duration) and, often ignored, eligibility conditions. Lyle Scruggs has developed a new method and data source for quantitative measures of welfare state generosity. His innovative data set combines measures of several dimensions of welfare state ‘generosity’, covering core social policy programmes such as unemployment benefits, public pensions, and child allowances for a time span of more than 30 years (1970–2001) for 18 OECD democracies. In Chapter 7, Scruggs thus illustrates how the empirical infrastructure of comparative welfare state research can be improved. His fine grained instruments for measuring benefit generosity provide surprising results and challenges, contesting the idea of clearly distinct and conventionally conceived welfare regimes. Moreover, his programme specific analysis suggests a certain extent of convergence since the 1980s while questioning the notion of welfare state resilience against cutbacks. In Chapter 8, Jochen Clasen and Daniel Clegg point out that many current reform initiatives are about the management and reallocation of risks across different social groups and between the public and the private sphere. Even a ‘social rights’ based approach to studying welfare state change may thus miss a central aspect of change. All social rights are conditional and involve obligations on the part of benefit claimants. In principle, changes in (various dimensions of) conditionality could thus be employed for measuring the scope of change within welfare state programmes which are often assumed, but rarely systematically investigated, as signalling paradigmatic shifts towards ‘enabling’ or ‘social investment’ states. Linking up with these two chapters of Part III, in Chapter 9, Jon Kvist employs empirical indicators of ‘generosity’, as well as two dimensions of conditionality, i.e. the ‘accessibility’ of transfer payments and the obligations imposed on benefit claimants. However, he operationalizes indicators somewhat differently and makes use of one of the more recent methods in

10


comparative analysis, fuzzy-set methodology. Applying his analysis to unemployment insurance in seven countries, Kvist demonstrates that fuzzyset methodology provides a powerful tool for the identification of social policy programmes shifting between ideal typical welfare state categories. In contrast to social expenditure based accounts of change, fuzzy sets, used for a comparative exploration of changes over time, not only offer a more encompassing method of capturing multidimensional changes within individual welfare state programmes but also present a more ‘concept driven’, and therefore theory guided comparative method than the statistical analysis of spending accounts.

CONCEPTUALIZING CHANGE The final part of this volume deals with macro conceptualizations of the nature of change within modern welfare states. As already referred to above, ‘retrenchment’ has become perhaps the most prominent concept within the comparative social policy analysis in recent years. Hinrichs and Kangas (2003: 574) pointed out that it was Paul Pierson (1994) who initiated what they call the ‘retrenchment business’ in comparative welfare state research, i.e. cross-national investigations into the causes for and political management of imposing losses within modern welfare states. However, Pierson’s theoretical interest in ways in which decision makers are able to ‘avoid blame’ (Weaver, 1986) has rarely been matched by reflections as to how to conceptualize and measure ‘retrenchment’ within or across welfare state programmes (Alber, 1996). However, similar criticisms can be made of many other concepts aimed at capturing the nature of at times different types of welfare state change which have mushroomed lately. These include ‘re-commodification’, ‘cost containment’, ‘recalibration’ (Pierson, 2001), ‘individualization’ and ‘de-familialization’ (Ostner, 2003; Leitner et al., 2004), or ‘residualisation’ (Powell, 2004). It is beyond the remit of Part IV of this volume to review the range of concepts and leitmotifs on change within contemporary welfare state research (Hobson et al., 2002, cover quite a few). Instead, as a way of illustrating the complexity, problems involved and usefulness of concepts which have been employed frequently in recent analyses, the three chapters of Part IV focus on ‘convergence’, ‘(de)-familization’, and ‘path dependence’ respectively. The question whether advanced welfare states are undergoing major convergent trends, or whether persistent diversity can be observed, has been at the heart of welfare state analysis for a long time. However, whereas the question of convergence or divergence in the 1960s figured most prominently in comparisons of economic systems (capitalism vs state socialism,


11

e.g., Pryor, 1968) and was linked to the so-called ‘end of ideology debate’, in the 1980s and 1990s the convergence-divergence question was raised in a different context and closely coupled to the study of globalization, or more generally, economic and political denationalization. As Julia O’Connor’s contribution in Chapter 10 shows, the concept of convergence is related to general issues of globalization and Europeanization. Whereas new concepts and measures of convergence have been suggested by several authors during the last two decades, the question of developing and selecting appropriate indicators for the assessment of the various types of convergence remains a major challenge. The concept of ‘de-familization’ has gained a prominent position in contemporary analyses of welfare state change not least as a response to the concept of ‘de-commodification’ (Esping-Andersen, 1990) which has been criticized as unduly focusing on the (expansion or retrenchment) of social rights arising from (male) paid work in the labour market, thereby ignoring the ‘familization’ of care work, which is largely done by women. ‘De-familization’ has been seen, as Sigrid Leitner and Stephan Lessenich argue, as a process of ‘unburdening the family’ (de facto, women) from care responsibilities by providing rights to payments for care which secure economic independence of the carer and her dependants, as well as access to paid work via, for example, the expansion of public services. However, the authors show that the concept is ‘multi faceted and complex’, involving ‘economic’ as well as ‘social’ independence, and must be perceived from two perspectives, the carer and the person who is cared for. By deconstructing the concept of ‘de-familization’ Leitner and Lessenich demonstrate that it can be employed as a ‘dependent variable’ in empirical comparative analysis, although its multidimensionality needs to be recognized. Finally, one of the most prominent concepts in the recent study of the political reform processes of welfare state changes is that of path dependence. The term has become somewhat of a catch-all theorem, often without clear conceptualization, let alone operationalization or measurement. It characterizes the logic of change as much as the nature (or process) of change. In Chapter 12, Sven Jochem not only provides a critical discussion of several dimensions of path dependence but also operationalizes the concept, which allows researchers to generate criteria for the categorization of welfare state reforms as path stabilizing or path deviating. In order to demonstrate that there are ways of solving the ‘dependent variable problem’ of specifying critical thresholds, Jochem discusses one of the areas of welfare state reform where path dependence is to be expected, i.e. pension reform. Similarly to Scruggs (Chapter 7), Clasen and Clegg (Chapter 8), and Kvist (Chapter 9), his analysis provides evidence against the notion that contemporary welfare states are locked in distinct

12


trajectories which tend to reinforce themselves. Given that even the ‘elephants of the welfare state’ are on the move (Hinrichs, 2001), it is hardly plausible to assume that the more lightweight welfare state programmes suffer from inertia. So, what are the lessons of these investigations into welfare state change and particular aspects of the dependent variable problem? For us, perhaps the most valuable point is the importance of investing in the theoretical and empirical infrastructure for comparative research. The development, upgrading and updating of this infrastructure might not be regarded as the most attractive academic endeavour given that it is time consuming and does not promise short-term gains or spectacular breakthroughs in scientific debates. However, we regard it as an essential long-term investment, building robust foundations for systematic empirical cross-national analysis. As the subsequent chapters demonstrate, real welfare state change is complex and difficult to capture, especially in comparative analysis. This however, is rarely acknowledged within overly parsimonious or elegant models or ‘generalized’ theories. It thus seems worthwhile to go back to the roots of comparative research: conceptualizing, operationalizing and measuring the kind of changes which may transform democratic welfare capitalism around the globe.

2. More than data questions and methodological issues: theoretical conceptualization and the dependent variable ‘problem’ in the study of welfare reform Christoffer Green-Pedersen INTRODUCTION1 Studies of welfare state reform2 have been booming in recent years. Today the literature in the area offers a large number of theoretical arguments on what, for instance, causes cross-national variation in the scope or degree of welfare reforms implemented in OECD countries. In the literature claims are made about the importance of party politics in different versions (Ross, 2000a; Kitschelt, 2001; Green-Pedersen, 2002; Korpi and Palme, 2003; Allan and Scruggs, 2004), the role of political institutions (Bonoli, 2000; Swank, 2002a), political discourse (Cox, 2001; Schmidt, 2002) and economic pressures (Castles, 2001). However, the literature actually offers very few established facts as to what factors or what combination of factors matter for cross-national variation in welfare state reform. One clear example is the continuation of the classical ‘does politics matter?’ debate. Some studies (Korpi and Palme, 2003; Allan and Scruggs, 2004) claim that politics still matters, or more precisely that social democratic parties in government matter, implementing less welfare state reform, especially retrenchment reforms, than centre-right governments. Other studies (Ross, 2000a; Green-Pedersen, 2002) argue the opposite, i.e. that social democratic governments introduce more rather than less retrenchment than rightwing governments, thereby reversing the classic ‘politics matters’ claim. Yet other studies (Castles, 2001; Huber and Stephens, 2001; Siegel, 2001; Kittel and Obinger, 2003) imply that politics does not matter any longer. In one way, the continuing disagreement is helpful since it keeps the scientific debate alive, but in another way it is also deplorable. After all, the role of party politics in welfare retrenchment and reforms is an empirical 13

14


question to which welfare state research should be able to offer fairly clear answers. So why are there conflicting answers to basic empirical questions such as ‘does politics matter?’ (and others, see below) in the central literature on welfare state reform? The argument proposed here is that this is partly caused by what I label the ‘dependent variable problem’ (GreenPedersen, 2004), i.e. the problem of what is meant by welfare state reform and how reforms can be measured. The chapter argues, firstly, that more attention needs to be paid to these questions in order to advance comparative welfare state research; secondly, it will outline some of the issues which the debate needs to address. The chapter firstly argues that the ‘dependent variable problem’ is as much a question of theoretical conceptualization as it is a question of empirical indicators. A clear theoretical definition of what one wants to measure is a necessary first step before considering indicators. Secondly, a clear distinction has to be made between welfare state retrenchment and welfare state restructuring, with the former concept needing further refinement from both an output and an outcome perspective. Finally, the chapter argues that the often emphasized distinction between quantitative and qualitative studies of welfare state reform is misplaced in the sense that both types of study should aim to make use of systematic empirical indicators, albeit not necessarily expenditure data. Only the combination of different types of indicators will advance the study of welfare state reform and retrenchment in terms of answering basic empirical questions. Continuing Disagreement – the Literature on Welfare State Retrenchment As indicated, the literature on welfare state reform, especially retrenchment, is divided on the classical question ‘does politics matter?’. What lies behind this disagreement? It is noticeable that studies which challenge and reverse the traditional ‘politics matters’ argument, i.e. argue that social democratic governments tend to retrench welfare programmes more than centre-right governments, are based on studies of a few countries (Ross, 2002a; Green-Pedersen, 2002). Thus one explanation for the disagreement may simply be found in the number of countries studied which tended to be much higher in previous research which favoured the classic, i.e. socialdemocratic ‘politics matters’ proposition. However, recent research covering 18 OECD countries is divided too, with some studies concluding that politics does not matter anymore (Castles, 2001; Huber and Stephens, 2001; Kittel and Obinger, 2003) while others continue to subscribe to the classic ‘politics matters’ argument (Korpi and Palme, 2003; Allan and Scruggs, 2004). Further, in terms of statistical methods, apart from Castles (2002) all studies use a pooled time cross-sectional approach and still reach

Theoretical conceptualization and the dependent variable ‘problem’

15

different conclusions (for a more detailed discussion, see Siegel, Chapter 4 in this volume). However, studies which argue that politics matters focus on trends in welfare generosity, using mainly benefit replacement rates in selected welfare programmes. By contrast, investigations which claim that politics does not matter anymore focus on the development of social expenditure in relation to GDP. In other words, the answer to the question whether ‘politics matters’ depends on the dependent variable and how it is operationalized. There are other examples in the literature on recent welfare state reform. Based on a qualitative assessment, Anderson (2001) argues that the Swedish pension system was retrenched more than the unemployment scheme. Making use of the economic model applied by the Swedish Ministry of Finance, Lindbom (2005) concludes the opposite. Hacker (2002; 2004) argues that Pierson’s (1994) conclusion about continuity in the American welfare state is based on an exaggerated focus on active changes in formal welfare programmes, thereby ignoring changes in the societal context of welfare programmes. Thus, also in these cases the operational definition of the dependent variable seems to determine the conclusions. The examples above indicate that much more attention needs to be paid to the dependent variable within studies of welfare state reform. Following Kitschelt (2001) with regard to the disagreement on the dependent variable, Pierson (2001: 420) argues that ‘it is difficult to exaggerate the obstacle this dissensus creates for comparative research . . ., it is impossible to seriously evaluate competing explanations when there is no agreement about the pattern of outcomes to be explained.’ Nevertheless, to date the issue has not been sufficiently addressed in the literature, despite some notable exceptions. For example, Castles (2002) discusses how new opportunities for disaggregating social expenditure can ease some of the problems involved in using social spending as an indicator of welfare state development, while Allan and Scruggs (2004) also pay attention to the same issue. There is a lively debate, for example, on the usefulness of the pooled time series cross-sectional statistical approach (Kittel and Winner, 2005; Plümper et al., 2005). Such a discussion of statistical techniques is of course highly appropriate, but it is striking that this aspect receives more attention than the question of the appropriateness of the data that are analysed. Theoretical and Operational Definitions As a starting point for discussing these matters, a clear distinction between theoretical and operational definitions is crucial, i.e. between what one wants to measure and how one measures it. While this point is straightforward, it

16


receives surprisingly little attention in the welfare state retrenchment literature. For instance, in his well known article on the new politics of the welfare state, Pierson (1996) argues that welfare state retrenchment is defined as (1) significant increases in the reliance on means-testing; (2) major transfers of responsibility to the private sector; (3) dramatic changes in benefit and eligibility rules that signal a qualitative reform of a particular programme. This is a theoretical definition of retrenchment which leaves open the question about how to measure, for instance, ‘significant increases in the reliance on means-testing’. Without such an operational definition, it becomes impossible for others to reproduce the findings. Exactly this presents a major weakness of many studies based on ‘qualitative assessment’, often lacking an operational definition that spells out how, and on the basis of which criteria, assessments were made, thereby diminishing their reliability significantly. It is simply often difficult for other researchers to arrive at the same conclusion when making similar assessments. At the same time, many other studies use social expenditure data without much discussion on the usefulness and appropriateness of the data. Reliability is rarely a problem here, but judging the validity of expenditure as a measure is impossible without a reflection of what one wants to measure.

WHAT TO MEASURE? THEORETICAL DEFINITIONS OF WELFARE STATE RETRENCHMENT Regarding the question of what one wants to measure, a couple of distinctions and delimitations are important. First, the welfare state may be viewed either from an output or from an outcome perspective. The output perspective regards the welfare state as a number of government programmes or policies, and welfare state reforms as changes to these outputs or programmes. This raises questions concerning the underlying assumption about what the welfare state is (see also Chapter 3 by Bonoli). The literature generally agrees on defining the welfare state as social policy in a broad sense, i.e. including most cash transfers to households such as pensions, unemployment benefits, child allowances and housing benefits, and social services such as health care, child care, elder care and social housing. However, should educational policy be regarded as part of the welfare state too? What about macroeconomic policy? Adopting an outcome perspective the welfare state is viewed more from recipients’ perspective, as a government’s commitment to, for instance, minimizing inequality or maintaining full employment. In other words,


17

when looking at welfare state reform, the starting point would be to examine how outcomes such as inequality or replacement rates have developed. Developments in these outcomes are of course influenced by outputs or policies, but other factors are relevant too, such as changes in the extent of social need. In short, it is imperative to reflect on the concept of ‘welfare state’ since there is a difference in adopting an outcome as distinct from an output perspective. Second, it is important to distinguish between two types of welfare state reform, namely retrenchment on the one hand and restructuring or institutional change on the other. The former refers to changes which cut back or reduce social entitlements by, for example, reducing benefit levels, tightening eligibility rules or shortening entitlement periods. The latter refers to changes to the institutional rules surrounding a scheme. These can be changes in the administration or the funding of benefits, but also shifts in the principles which govern the calculation of benefits, for instance changes in the scope or type of means-testing (cf. Clasen and van Oorschot, 2002). This distinction is important because the evaluation of certain changes to a programme is highly dependent on the perspective adopted. A reduction of say 5 per cent in benefit level is clearly a form of retrenchment. But it is not necessarily an institutional change. On the other hand, financing or organizational structures may be altered without affecting benefit generosity, just as principles may be changed without implying retrenchment. Imagine, for instance, a flat-rate pension scheme which becomes supplemented by a highly selective pension top-up. Such a change implies an institutional alteration, but clearly not a case of retrenchment. To provide a concrete example, the Swedish pension reform in the first half of the 1990s clearly represented a major form of restructuring. However, the degree of retrenchment is much less clear, partly because the new pension system automatically adjusts pensions downward if life expectancy increases (cf. Anderson, 2001; Lindbom, 2005). It is the future development of life expectancy therefore which will partly determine the retrenchment effects of the reform. When focusing on retrenchment, the distinction between output and outcome perspectives becomes highly important. From the output perspective, retrenchment is ‘negative’ since benefits are reduced or eligibility rules tightened as suggested above. From the outcome perspective, one would focus on for instance rising inequality, which is seen as a retrenchment of the political commitment to minimizing inequality (cf. Clayton and Pontusson, 1998). Rising inequality may be the result of programme retrenchment, but could also be the result of changes to the tax system or wage inequality and requires a much broader perspective on the welfare state than the output perspective. As mentioned, Hacker (2002; 2004) has

18


shown how the development of the US welfare state looks very different once an outcome perspective is applied, in contrast to Pierson’s (1994) output perspective. When discussing different theoretical definitions, it is important to be aware that no definitions are better or worse than others per se. Which theoretical definition of retrenchment should be adopted depends on the overall theoretical question of interest. For example, if it is a government’s ability to implement unpopular policies (as in Pierson, 1994, for example) then an output definition is clearly the most logical choice since the overarching interest are policy measures which governments are politically able to carry through, rather than effects on recipients. However, if the latter is the theoretical focus, an outcome definition seems more useful. The same point applies to institutional aspects of programmes. Following Esping-Andersen (1990), it has become fairly common to focus on the question of selectivity vs. universalism, as Pierson’s (1996) definition above shows. However, it is often ignored that the focus on the institutional aspect of welfare programmes stems from power resources theory and the need of the working class to secure broad political support for the welfare state. If this theoretical perspective is not applied, it is far from obvious why universalism should figure as a focus in welfare state research and not other institutional aspects. In sum, there are several ways to define welfare state reform theoretically and the choice depends on the broader theoretical perspective adopted. There is no such thing as reform per se, and thus no particular definition which would capture the essence of reform. The crucial point is to pay attention to the question of theoretical definitions, how the latter are linked to the broader theoretical perspective underlying a particular study, and how this affects the choice of measurement. Discussing measurement issues without knowing what one wants to measure in the first place is problematic and in this sense the dependent variable ‘problem’ concerns questions of theoretical definition or conceptualization as much as methodological and technical questions.

OPERATIONAL DEFINITIONS Before discussing different types of measurements, it is worth taking up the issue of quantitative vs qualitative research. To some extent the literature on welfare reform reflects a divide between studies of a few countries, which often focus more on processes and politics related to the reform and less on measuring reforms (e.g. Bonoli, 2000: Timonen, 2003), and studies of all OECD countries which apply social


19

expenditure data. The latter tend to omit any discussion of the advantages and limitations of using these data. Such a divide is also evident in several prominent studies of welfare retrenchment such as Huber and Stephens (2001) and Swank (2002a), which combine statistical analysis of expenditure data with case studies of selected countries. Measurement questions seem to relate mainly to expenditure data. One is thus left with the impression of having to choose between two evils: either using expenditure data, which, as will be argued below (see also De Deken and Kittel’s contribution to this volume), incorporates several problems, or conducting case studies without much reflection on the measurement of the scope of reforms implemented (see Jochem, Chapter 12, this volume). The latter often makes case studies vulnerable to ‘fuzzy judgments’, as Alber (1996) critically noted. However, as argued by King et al. (1994) or Brady and Collier (2004), there is no fundamental difference between quantitative and qualitative research. Both share the same scientific ambition and the same basic standards for research. Thus, discussing operational definitions and different types of indicators should essentially be a reflection of the pros and cons of different ways of measuring theoretical variables. In other words, the issue at stake here is the validity and reliability of different measures or indicators, not one of different research aspirations or one of numbers vs judgments. For good reasons, social expenditure is by far the most common measurement in studies of welfare state retrenchment and of welfare state expansion. Data are easily available for all relevant countries and cover almost all relevant welfare state programmes, although data on social service expenditure remain somewhat of a problem (see Huber and Stephens, 2001). Furthermore, social expenditure provides a summary measure of different aspects of programmes. How else would it be possible, for example, to add together changes in benefit levels, eligibility rules and benefit duration? However, there are at least four sets of problems attached to using expenditure data. Firstly, in order to make figures cross-nationally comparable, calculating social expenditure as a percentage of GDP is most common. The problem is, of course, that the measure is then also affected by the development of GDP (see also Scruggs, Chapter 7, this volume; Clayton and Pontusson, 1998). Secondly, expenditures on certain transfer programmes, especially unemployment benefits, are strongly influenced by the state of the economy. A government can easily limit unemployment benefit levels and at the same time experience rising expenditure due to higher unemployment. This problem is well recognized, and Castles (2002) has suggested leaving unemployment benefits out of analyses based on expenditure data, whereas Siegel (2001) has adjusted data by keeping the number of recipients constant. Still, the problem with such strategies is the loss of information on the

20


development of the programmes. Changes in the number of unemployed, for example, are influenced not simply by the level of unemployment, but also by eligibility and entitlement (benefit period) rules. Thirdly, there is a ‘time-lag’ problem (Pierson, 1994: 14). This refers to the fact that many retrenchment reforms are designed to have a gradual rather than immediate effect. Consequently, many legislative changes are not immediately reflected in expenditure data. The relevance of this problem varies, but with regard to pension systems it is a very serious caveat since pension reforms are typically designed to take effect in the long run. The expenditure effects of many pension reforms which were implemented in OECD countries in recent years will become visible only gradually over the next decades. Fourthly, an investigation into reforms in welfare services, rather than transfers, generates questions about the usefulness of expenditure data from the perspective of people receiving such services. One clear example of this is health care expenditure. Does the fact that a country spends more on health care imply a better health care system or just one which is more inefficient? Public health care spending in the US is at the same level as in many European countries, and the US spends twice as much if private health care spending is included. However, about 15 per cent of the US population have no health care insurance (Wilkerson, 2003; Green-Pedersen and Wilkerson, 2006). In other words, there are significant problems attached to expenditure data as a measurement of the extent of retrenchment in relation to core welfare state programmes such as pensions, health care and unemployment benefit. To a large degree, the use of social expenditure data is problematic because they are effectively outcome data. All four types of problem mentioned above are to some extent caused by other factors than changes in welfare programmes affecting social expenditure. Returning to the discussion above, the use of social expenditure as outcome measure is not a problem if the theoretical definition of retrenchment is based on outcome. However, most studies which use social expenditure data in reality have an output focus. They aim to measure changes in social policy programmes in order to assess the importance of government partisanship, for example. Other studies, such as Hacker (2002; 2004) adopt an explicit outcome focus, but the outcomes they focus on, such as inequality, are poorly captured by social expenditure data. The discussion above should not be read as an argument against the use of expenditure data, but as underlining that there are strong reasons to look for other data sources which might supplement social spending. Recent studies (Korpi and Palme, 2003; Allan and Scruggs, 2004) have used net replacement rates in selected welfare programmes to study welfare state retrenchment (see Kangas and Palme, Chapter 6, and Scruggs, Chapter 7, this volume). This


21

avoids some of the problems of using expenditure data, e.g. rising number of recipients, but raises other problems, which also partly stem from using an outcome measure to evaluate output related questions. Firstly, replacement rates are only one aspect of welfare programmes. Moreover, changes in replacement rates are influenced by other factors such as wage levels and taxation rules. In other words, the link back to government policy can be quite indirect. Governments may, for instance, improve benefits and at the same time replacement rates might decline due to wage increases. Although relatively underdeveloped, an alternative to outcome measures such as expenditure data or replacement rates for the assessment of retrenchment are output measures. Kitschelt (2001) suggests an index measuring changes made in social security in relation to the level of benefits, eligibility criteria, etc. Thus, retrenchment could be measured by ‘microdata’, that is data providing a quantitative measure of the degree of retrenchment implied by individual changes in social security schemes. Elsewhere I have applied such ‘output data’, measuring the degree of retrenchment following from a specific change in a welfare state programme by its likely budgetary effect (the percentage of the total amount of cash spent on the programme) (Green-Pedersen, 2002; cf. also 2000). This measurement is calculated using statistical information about the scheme and material from the parliamentary reading of legislative proposals. Another example of an output measure has been developed by Lindbom (2005). Using the economic model of the Swedish Ministry of Finance, he calculated what the current expenditure on different programme would had been had the rules which applied in 1991 not been changed, and then compared this with actual expenditure. The latter method is clearly more reliable, but of course requires access to an economic model. Compared with outcome measures the strength of output measures, such as the two examples presented above, is the much clearer connection with political decisions, which is, of course, particularly important if the underlying theoretical definition is output related, e.g. in Green-Pedersen (2002), who provides retrenchment measures for each programme change. Furthermore, it allows an assessment of long-term effects thus avoiding time-lag problems associated with expenditure data, while being able to take all aspects of programmes into account. However, output measures have major drawbacks too. Even with access to statistical models, various assumptions about recipient numbers, etc. have to be made in order to forecast effects arising from legislative changes. In some cases, such as the Swedish pension reform, changes in the programmes were deliberately made in ways which make their future effects uncertain. Finally, it is also a very time consuming exercise requiring language skills, etc. in order to develop output measures.

22


When considering operational definitions of welfare state restructuring, output perspectives focusing on institutional aspects might be justified in line with Esping-Andersen’s argument against spending, i.e. his claim that ‘it is difficult to imagine anyone struggled for spending per se’ (1990: 21). With regard to the question of operational definitions, it is crucial to remember that Esping-Andersen did not replace expenditure data with qualitative assessments, but with indices measuring institutional traits of welfare state programmes. Thus the need for operational definitions and indicators also relates to studies focusing on welfare state restructuring. One way is to construct indices of change. Of course, this will always imply an element of arbitrary judgement. However, this should not lead to dispensing with empirical indices. Instead, it requires an explicit discussion of the criteria which guided their construction and the robustness of indices against changes in such criteria. Chapter 8 by Clasen and Clegg in this volume provides a fine example of such an index approach, while Kvist (1999 and Chapter 9 in this volume), as well as Vis (forthcoming), also show how the fuzzy-set methodology can be used to measure the restructuring aspects of welfare state change.

CONCLUSIONS The previous decade has witnessed the emergence of an extensive literature dealing with welfare state reform, and to some extent the knowledge of these reforms is much greater today than it was ten years ago. However, the reform literature is much richer in theoretical hypotheses about the effect of different political and economic factors on welfare state reform than it is on empirical findings about the scope of reforms actually carried through. Notably, the question of which welfare states have been reformed more than others remains in many ways still unanswered. This chapter has argued that the lack of progress relates to the failure of adequately addressing issues connected with the dependent variable, i.e. questions such as ‘What is a reform?’ and ‘How can it be measured?’. The dependent variable ‘problem’ is thus not so much the fact that variables can be difficult to measure and reliable data hard to obtain (which are both common in any type of empirical social science research); the ‘problem’ lies more in the lack of attention paid to the issues and the dominance of studies based on social expenditure data. This chapter has argued that the starting point for such discussions of the dependent variable should be theoretical in nature. Welfare state reform can mean many different things. The suggestion of distinguishing welfare state retrenchment and welfare state restructuring also highlights the importance


23

of the difference between an output and an outcome perspective adopted in research of welfare state reform. In other words, one reason for the lack of progress and the continuing disagreement on the scope and relative extent of reform may simply be found in the varying definitions of reform applied. A country may have retrenched its welfare state significantly without reconstructing it and vice versa. However, another aspect of the dependent variable ‘problem’ lies in the way the literature has typically approached issues relating to measurement of retrenchment, often distinguishing research which relies on expenditure data from studies engaged in ‘qualitative assessment’. As argued above, this is an unhelpful perspective. There are some problems related to the use of social expenditure, but the use of other indicators implies other problems. The use of social expenditure data becomes a serious problem if no other indicators are used allowing for a comparison of findings from different indicators. Thus, an important challenge for research on welfare state reform and retrenchment is to develop empirical indicators, which – together with social expenditure data – can answer questions such as which welfare states were reformed the most.

NOTES 1. This chapter is a revised and updated version of Green-Pedersen (2004). Thanks to the participants in the Stirling seminar, May 2005 on the ‘Dependent variable problem’, especially Jochen Clasen and Nico A. Siegel, for useful comments on an earlier version of this chapter. 2. In the following, welfare state reforms will be used as a general term covering different types of changes to welfare state programmes. We can distinguish between two types of change, namely welfare state retrenchment and welfare state restructuring, see below.

3. Too narrow and too wide at once: the ‘welfare state’ as a dependent variable in policy analysis Giuliano Bonoli INTRODUCTION The ‘welfare state’, the object of a substantial amount of policy research over the last three decades or so, is an ill defined entity. Definitions found in the literature tend to be based either on vague notions such as ‘policies that aim to improve people’s welfare’ or as enumerations of policies that belong to the welfare state. None of these approaches is really satisfactory. In the former case, it is difficult to think of a policy that does not aim to improve people’s welfare. In the latter, by relying on an unspecified listing of policies, we miss the criterion that tells us what belongs to the welfare state and what doesn’t. This confusion has arguably always represented a problem for social policy research, but in the current postindustrial social and economic context, an unclear understanding of what we mean by ‘the welfare state’ is a major obstacle to insightful analysis. Over the last 20 years or so, we have witnessed the development of new policies, such as child care or active labour market policies, which have little in common with the traditional protective and de-commodifying function of postwar welfare states. These new policies do not aim to protect individuals from market forces; instead they have the objective of improving their chances to succeed in the market. They do not promote de-commodification; they improve the quality of life of highly commodified workers. Are these new policies part of the welfare state? In this chapter I argue that a conceptually coherent definition of a dependent variable for social policy research must take into account changes in the fundamental nature of social policy, in particular, those which have occurred as a result of the transition from predominantly industrial social and economic structures to postindustrial ones. On this basis, the ‘welfare state’ is found unsuitable for the analysis of social policies both in industrial societies 24

The ‘welfare state’ as dependent variable in policy analysis

25

and in the postindustrial age. In the former case, the welfare state is too narrow a dependent variable, since countries have used a whole range of instruments to tame markets and improve the living conditions of industrial workers. In the latter, the welfare state is too broad a category, because it conflates the old social policies, inherited from the postwar years and designed to deal with industrial social risks, and the new social policies developed to deal with social problems stemming from postindustrialization. The argument is developed in two stages. First I examine the adequacy of the welfare state as a dependent variable of social policy analysis in the context of industrial societies. On the basis of comparative historical studies, I argue that ‘functional equivalents’ to what is commonly considered as part of the welfare state must be included in the dependent variable. In the second part of the chapter, in contrast, I show how using the welfare state as a dependent variable in the current context conflates policies with different objectives, targeting different groups and developed in response to different social transformations. These should be analysed separately.

SOCIAL POLICY IN INDUSTRIAL SOCIETIES: REDUCING WAGE EARNERS’ EXPOSURE TO MARKET FORCES One of the clearest and most influential definitions of what a welfare state is, is the one developed by Asa Briggs. According to him, ‘A “welfare state” is a state in which organized power is deliberately used [. . . ] to modify the play of market forces.’ Unfortunately, presumably in order to clarify his thought, Asa Briggs continues by listing typical areas of intervention of welfare states ‘first by guaranteeing individuals and families a minimum income irrespective of the market value of their work or their property; second, by narrowing the extent of insecurity by enabling individuals and families to meet certain “social contingencies” (for example sickness, old age and unemployment) . . .; and third, by ensuring that all citizens without distinction of class or status are offered the best standards available in relation to a certain agreed range of social services’ (Briggs, 1961 [2000] p. 228). It is in fact the second part of his definition that proved most influential, being picked up in many social policy textbooks. But in the search for conceptual clarity, it is definitely the first part of Briggs’ definition that helps us the most. The welfare state is understood as actions that make use of political power in order to modify the distribution of goods and services that result from market exchanges. It is a clear definition that, at least on an abstract level, helps us distinguish actions that belong to the welfare state from those that do not. Its main problem, of course, is that

26


it defines the welfare state against a hypothetical state in which all goods and services are exchanged according to market principles. The fact that such a state has never existed in human history makes it difficult to operationalize this definition. All human societies, including the early forms of capitalism, have framed market exchanges so as to modify their outcomes, though with different instruments and to different degrees (Polanyi, 1957). In the United Kingdom, the objective outlined in the first part of Asa Briggs’ definition of the welfare state, the deliberate use of organized power to modify the play of market forces, has been pursued essentially through the instruments that he mentions in the second part of his definition: income support, social insurance and social services. The British welfare arrangement was based on the assumption that the market was to be allowed to operate as freely as possible, while the state would compensate the resulting distribution of resources, but only ex post. This view of the relationship between the state and the market is summed up by T.H Marshall’s definition of welfare-capitalism: The hyphen links two [. . .] different and contrasting elements together to create a new entity whose character is a product of the combination, but not the fusion of the components, whose separate identities are preserved intact and are of equal and contributory status. (Marshall, 1981: 124, emphasis added)

The role of the state here is understood as a form of intervention that takes place only after resources have been allocated according to market principles. But this conceptualization of welfare capitalism does not travel very far. Britain is in fact rather exceptional in this respect. Other industrial countries have relied on a mix of instruments to tame markets, which include the regulation of the labour market and the use of trade policy as an instrument to protect wage earners against the vagaries of modern markets. Social Policy through Labour Market Regulation1 A number of recent studies have pointed to the existence of ‘functional equivalents’ to income replacement programmes and social services, insofar as the provision of economic security to wage earners is concerned. Often these functional equivalents have been found within the realm of labour law or industrial relations (Castles, 1985; Mares, 1996; Whiteside and Salais, 1998; Bonoli, 2003a). Castles, for example, challenges the predominant view of Australia being a typical ‘welfare laggard’ by arguing that for this country a focus on social programmes alone does not accurately reflect the extent to which its citizens have been protected against market risks. A developed system of wage


27

determination plays an important role in enhancing the degree of welfare enjoyed by its citizens. For most of the 20th century, wages have been set by the courts on the basis of social criteria, at a level deemed sufficient for a (male) wage earner supporting a family (Castles, 1985). Other authors have emphasized the role played by labour laws in protecting the incomes of wage earners against the risk of low pay and unemployment. In continental Europe (particularly in the South), minimum wages, collective bargaining systems, and employment protection laws provide wage earners with job security and protection against the risk of being paid poverty wages. These policy instruments are akin to income transfer programmes insofar as their function is concerned: even though they use a different channel, they provide economic security to wage earners (Mares, 1996; Whiteside and Salais, 1998). As argued elsewhere (Bonoli, 2003a), cross-national variation in relation to these labour market based social policy instruments seems to follow a pattern of ‘families of nations’ (Castles, 1993). In general, the legal regulation of employment is stronger in Latin European countries (France, Spain, Portugal, and Italy). Collective bargaining, in contrast, is an important source of economic security for workers in Germanic and in Nordic countries. It is in fact only in the US and in the UK that these instruments did not develop to any significant extent. My claim here is that capitalist societies have developed different instruments for providing economic security to wage earners. The welfare state is only one of the channels that have been used to that effect. In addition to building welfare states, and sometimes as an alternative to them, several market economies have developed sophisticated systems of labour market regulation that either through legislation, or through self-regulation by labour market actors, substantially increase the level of economic security experienced by wage earners. In this section, I focus on three sets of instruments: labour laws, collective bargaining, and redistributive income transfer programmes. Law based regulation of the labour market aiming at improving levels of economic security experienced by wage earners has taken a number of different forms. Perhaps the most obvious example is employment protection legislation. Most industrial countries provide some legal protection against dismissal, generally in order to avoid arbitrary lay-offs, but sometimes also by limiting the freedom of employers to dismiss workers during recessions. Employment protection laws, as a result, can protect wage earners against the risk of unemployment, by securing their place in the labour market. In some countries, governments have also traditionally intervened in the process of wage setting, for example through minimum wages set by law which constitute a floor below which employment

28


relationships are illegal. Wages can also be regulated by imposing collectively negotiated agreements upon employers and workers who are not among the original signatories to the agreement, a process known as extension. These instruments are used to a different extent in different countries (see Table 3.1, columns 1 to 4). Broadly speaking, the use of legislation to regulate labour markets is strongest in Latin Europe and weakest in English speaking countries. Between these two extremes, reliance on labour law as a social policy instrument is relatively weak in the Nordic countries, and somewhat stronger in continental Europe. This clustering of countries could in reality be even clearer than is suggested by Table 3.1. In fact, the information presented refers to formal instruments, which are not necessarily used to the same extent in different countries. For example, even though extension laws of collective agreements are common both in continental and in Latin Europe, they are used much more sparingly in the former (especially in Germany and Switzerland) than in the latter. Collective bargaining is the second labour market based instrument for providing economic security to wage earners examined in this chapter. Typically, countries with all-encompassing collective agreements (sectoral or national) have been able to protect low wages to a much larger extent than countries which lack a tradition of collective bargaining (UK, US), in spite of economic trends that push in the direction of more earnings inequality (Rueda and Pontusson, 2000). In addition, collective agreements generally regulate a host of other aspects of employment contracts, such as working hours, holidays, fringe welfare benefits, and so forth. Table 3.1 (columns 5 and 6) provides information on the importance of collective bargaining cross-nationally. The level at which collective bargaining takes place can be relevant to the effectiveness of this instrument, as usually the bigger the unit the stronger the bargaining power of the unions, and hence the concessions in terms of wages and other aspects of economic security they can obtain. The effectiveness of collective bargaining depends also on the coverage it can achieve. There are important variations, with the highest rates achieved in continental Europe and among Nordic countries. Latin European countries have fair degrees of coverage for collective agreements, but this is often the result of the use of extension laws (Ebbinghaus, 1998). In this respect, high rates of coverage for collective agreements in Latin Europe are not so much the result of selfregulation by labour market actors (as is the case in continental Europe) as a reflection of the tradition of state economic interventionism which is typical of these countries. Third, virtually all European countries have developed income transfer programmes that modify the primary allocation of resources produced by

29

0.35 0.50 0.70 1.00 1.30 1.10 1.10 1.00 1.30 1.10 0.55 1.40 1.5 1.70 1.40

2 1998 37.4 (1999) No No No No No No 50.4 No 51.1 No 55.3 No 49.6 28.8

3

Minimum wage

None Mandatory Voluntary Mandatory Voluntary Voluntary Mandatory Mandatory Mandatory Mandatory Mandatory Mandatory None Mandatory Mandatory

4

Extension

Firm National/Firm Sector National National Sector Sector Nat./Sector Sector Sector Sector Sector/Firm Sector Sector Sector/Firm

5 Level 47 70 80 95 75 83 98 90 90 (West) 81 53 82 70 79 70

6 Coverage

Collective bargaining

14 13 28 25 18 22 – 9 8 12 8 6 9 6 8

7

Tax financed social expenditure as % of GDP

Notes: Columns 1 and 2: Index of protection against dismissal, varies between 0 (no protection) and 2 (maximum protection). Source: Blanchard and Worlfers (2000) quoted in Nickell (2003). Column 3 Minimum wage (if exists) as a percentage of full-time average earnings. Source OECD (1998: 37). For the UK, author’s own calculation. Columns 4–6: Possibility of extending collective agreements; main level of collective bargaining; coverage. Source: Ebbinghaus (1998: 15). Column 7: Eurostat (1995); Nordic Statistical Office, (1995); Flückiger and Cordero, (1995).

0.33 0.50 1.10 1.20 1.55 1.46 0.84 1.55 1.65 1.35 0.55 1.21 2.00 1.59 1.99

1 1973–79

Protection against dismissal

Forms of social protection in Western Europe, early 1990s

United Kingdom Ireland Denmark Finland Norway Sweden Austria Belgium Germany Netherlands Switzerland France Italy Portugal Spain

Country

Table 3.1

30


labour markets. Relative to the previous two instruments, income transfer programmes can be seen as examples of compensatory social policy, or ex post social intervention, in so far as they are based on the acceptance of the market as the primary mechanism of resource allocation, combined with corrections aimed at redressing undesirable social outcomes. This is what is most usually considered under the label ‘welfare state’. Table 3.1 uses the proportion of GDP spent on tax financed social programmes as an indicator that reflects the extent to which states intervene ex post to redress market outcomes. The indicator intentionally excludes spending on contributory social insurance programmes, which are more difficult to classify according the threefold distinction adopted here. In general, social insurance is strictly related to employment, with benefits being understood as a ‘deferred wage’. It also maintains employment based status differentials (Esping-Andersen, 1990; Palier, 2002). In this respect social insurance cannot be seen as a compensatory social policy, as it basically prolongs (labour) market outcomes instead of modifying them. Social insurance programmes essentially mimic the contractual and actuarial features of private insurance schemes. The overall distributional outcome of a pure social insurance scheme is thus often not very different from what would result from a market based system. As a matter of fact, the social insurance based pension systems of France and Germany do not significantly alter the income distribution produced by the labour market, as shown by the fact that in these countries levels of income inequality among working-age individuals and among pensioners are similar (Whiteford and Kennedy, 1995). There are two important exceptions where the role of social insurance as a provider of economic security is substantially different from what would be achieved in a private system. The first one is unemployment insurance, which cannot mimic the actuarial features of private insurance because of the non-random distribution of the risk of unemployment. Social insurance based systems of unemployment compensation provide coverage against the risk of unemployment to low-wage, high-risk individuals who would simply be unable to insure themselves on a private basis. As a result, they contribute to enhancing wage earners’ economic security relative to what they can achieve under pure market conditions. Second, many social insurance programmes do perform some vertical redistribution. This is the case for example of the British or the Swiss basic state pension, in which the link between contributions and benefits is rather loose. In general, however, both unemployment insurance schemes and redistributive social insurance programmes are partly funded through general taxation, so that their redistributive component is captured by the indicator ‘tax financed social expenditure’.


31

The extent to which states rely on redistributive income transfer programmes also varies significantly across countries, being strongest in Scandinavia and weakest in Latin Europe. The English speaking countries occupy an intermediate position, but within the context of overall low levels of social expenditure, which suggests that tax financed programmes play a relatively important role in these countries, in spite of the small proportion of GDP allocated to these programmes. On the basis of the data provided in Table 3.1, we can identify four different ways of combining the three instruments dealt with in this chapter. First, the English speaking countries, where economic security is provided mostly through redistributive income transfer programmes, are weak in relation to the two other forms of social protection. Second, the Nordic countries show rather low levels of legal regulation of the labour market, but they compensate with strong collective bargaining and redistributive income transfer programmes. Third, continental European countries tend to have a fair degree of law based social protection, have strong collective bargaining institutions, but rather weak redistributive income transfer programmes. And fourth, Latin European countries have the highest degree of labour market regulation by law, but make little use of other forms of social protection. Table 3.1 supports the hypothesis that these different forms of social protection are functional equivalents. Countries which developed strong employment protection have welfare states that perform little vertical redistribution and vice versa. Of course, the extent to which these policies are truly functional equivalents depends on the target group under consideration. As far as core workers are concerned, labour market based forms of social protection can provide levels of economic security that are comparable (if not superior) to those found in a fully fledged welfare state. For marginal workers and for non-workers (female and young workers or unemployed people, the long-term unemployed), labour market based social protection is often of little use, hence in these cases the different policy instruments discussed in this chapter are unlikely to be functional equivalents. Overall, although countries combine different forms of social protection, they clearly have preferences for some forms over others. The clustering of countries which emerges from Table 3.1 is reminiscent of Francis G. Castles’ (Castles, 1993) categorization of Western countries into ‘families of nations’, and suggests that a number of historical developments might be relevant in determining the forms of social protection chosen by a society. In a related paper I have argued that these have to do essentially with the shape of state–society relationships and dominant ideological orientations at crucial historical moments in the development of contemporary societies (Bonoli, 2003b).

32


Protectionism and Social Policy In addition or as an alternative to standard social polices, industrial countries have also attempted to modify the play of market forces by protecting producers (and their employees) from competition. In most cases this has meant protection from foreign competitors through barriers to trade. This strategy was a key component of the postwar arrangement in countries like Australia and New Zealand. Economic protectionism and a restrictive immigration policy in Australia were the price to pay in order to keep wages high and in line with the expectations of wage earners (Castles, 1985). In New Zealand a protectionist trade policy was used in conjunction with wage restraint as a tool to guarantee full employment throughout the postwar years. Reliance on protectionism stemmed from the need to manage current account deficits in the 1950s, after the ruling out of other options. But in the following years, successive governments kept quantitative restrictions on imports in place, because they permitted the preservation of full employment (Mabbet, 1995: 120–1). It is only in the 1980s, when protectionism begins to undermine the capacity of the New Zealand economy to grow, that this approach was gradually abandoned. A similar use of protectionism as an alternative to social policy is evident in the case of Switzerland. According to OECD expenditure data, this country has traditionally been one of the least developed welfare states, with social expenditure levels similar to those found in the US (except in recent years, when they have risen to Western European levels). But here too, markets have been tamed through other means. Protectionism, for instance, has been a key feature of agricultural policy, and often, through technical regulation, in many other economic sectors (e.g. food manufacturing, construction and related industries; infrastructure). In addition to protecting producers from imports, the Swiss postwar settlement was characterized also by the acceptance of non-competitive practices, such as cartels, among producers (Mach, 1999; Bonoli and Mach, 2000). These were as a result not really subjected to market forces. One consequence of this policy has been the persistence of relatively high levels of employment in the sheltered sectors of the economy in comparison to other countries, until the late 1980s. The other side of the coin has been of course the very high price levels that are common in the country, through which the inefficiency brought about by lack of competition has been financed. Through channels that have little in common with what we usually consider the welfare state, New Zealand and Swiss citizens managed to develop mechanisms against the vagaries of markets that were able to deliver broadly similar social outcomes in terms of access to a decent income and


33

protection from social risk. These peculiar arrangements ran into trouble in the 1980s in New Zealand and in the early 1990s in Switzerland, when costs began to be seen as unacceptable not so much by consumers (who, as workers, where arguably also benefiting from the arrangement through full employment and high wages), but by the export oriented industry that relied on expensive domestic producers for subcontracting and infrastructure (communication, energy, etc.). Taming Markets with Different Tools What is important to stress in the context of this chapter is the fact that for any attempt to map the extent to which industrial societies have managed to control and channel market forces, a focus on income transfer programmes and social services which is adopted in most of the comparative social policy literature, is unlikely to reflect the whole range of interventions that have been developed throughout the industrialized world. If the objective of research is to map differences in the extent to which wage earners are protected against market risks, then these alternative forms of social protection must be taken into account. Broadly similar objectives have been pursued in different countries through completely different means. This finding vindicates the Polanyian view that human societies and free markets are incompatible. Countries like Switzerland and New Zealand, which in welfare state league tables show up at the bottom, simply used other means to contain the socially disruptive character of free markets. This should be taken into account in international comparisons of welfare capitalism, and comparative studies of social policy in the industrial age should use an expanded dependent variable which does justice to the extent of crossnational variation highlighted in the first part of this chapter.

WELFARE CAPITALISM IN THE POSTINDUSTRIAL AGE As suggested by the numerous references to the work of Karl Polanyi, historically the above discussion applies to industrial capitalism only. The welfare state and the other forms of protection outlined above can be seen as responses to the commodification of labour resulting from the industrial revolution. While the aspiration for protection against market forces remains strong in postindustrial societies, over the last few decades we have witnessed the appearance of new demands on the welfare state, essentially in response to emerging new social risks (Bonoli, 2005; 2006). Is it a suitable research strategy to lump these new policies together with those that

34


we have inherited from the industrial days? In the remainder of this chapter, I argue that this is not the case. New Social Risks in Western Welfare States The first set of new demands comes from women. Because of the gender contract that was dominant throughout the 20th century, most women did not suffer exposure to market forces, but experienced a dependency on family ties such as marriage (Lewis, 1992). And women were largely ignored, both by the industrial welfare state and by the alternative forms of protection reviewed above. Income replacement programmes were geared towards preserving the income of the male breadwinner whereas the objective of employment and trade policy was male full employment. But things begin to change in the 1970s, when especially in Northern Europe, we see the emergence of increasingly strong demands from women for new policies that respond to their needs and aspirations. These take different forms. In Sweden, for instance, they are based on claims for free child care while in Germany they emphasize more the recognition of the unpaid domestic work (Naumann, 2005). These new demands have had and are having an impact on the welfare state, as policies that respond to them have been and are being developed throughout Western Europe. These policies, however, have little in common with the market protecting, decommodifying efforts of the old days. They are certainly not aiming at reducing people’s dependence on labour market participation. Quite the contrary, they are likely to facilitate employment, through ‘positive’ commodification (Knijn and Ostner, 2002). The second set of new demands stem from deindustrialization and its consequences, both in terms of mass unemployment and in terms of a labour market where the service sector is predominant. Even though the total volume of employment has not decreased over the last four to five decades, the replacement of industrial jobs by service sector employment has left large numbers of former industrial workers jobless and unable to reintegrate into the labour market in the service sector. The key consequence of this development has been long-term unemployment, a social risk that was virtually unheard of during the trente glorieuses. Industrial longterm unemployment, however, may not be the most important social consequence of de-industrialization, especially since it is a temporary phenomenon, related to the transition from predominantly industrial to predominantly service based economies. As cohorts of dismissed industrial workers reach retirement age, this type of unemployment will slowly fade away. The decline of manufacturing and the expansion of service employment, however, have altered labour markets in a more fundamental way, especially


35

at their lower end, consisting of low-skill, low-value-added jobs. Low-skill individuals have obviously always existed. However, during the postwar years, low-skill workers were predominantly employed in manufacturing. They were able to benefit from productivity increases due to technological advances, so that their wages rose in line with those of the rest of the population. The strong mobilizing capacity of the trade unions among industrial workers further sustained their wages which came to constitute the guarantee of a poverty-free existence. Today, low-skill individuals are mostly employed in the low-value-added service sector or unemployed. Low-value-added services such as retail sale, cleaning, catering, and so forth are known for providing very little scope for productivity increases (Pierson, 1998). In countries where wage determination is essentially based on market mechanisms (US, UK, Switzerland) this means that low-skill individuals are seriously exposed to the risk of being paid a poverty wage. The situation is different in countries where wage determination, especially at the lower end of the distribution, is controlled by governments (through generous minimum wage legislation) or by the social partners (through allencompassing collective agreements). Under these circumstances, the wages of low-skill workers are protected, but job creation in these sectors is limited, so that many low-skill individuals are in fact unemployed (Iversen and Wren, 1998). Overall, the fact of possessing low or obsolete skills today entails a major risk of welfare loss, considerably higher than in the postwar years. Like demands for new policies for women, the plight of low-skill individuals has resulted in new demands put on welfare states. These are to a large extent shaped by the existing labour market – welfare state arrangements. In highly regulated labour markets that produce high levels of lowskill unemployment, these demands translate into policies designed to help the jobless enter the labour market, or active labour market policies. In unregulated labour markets, the effort of policy is directed towards improving the living conditions of the working poor, most notably through tax credit programmes (OECD, 2003). New Policies for New Risks The new social risks that have emerged as consequences of the transition to postindustrial societies have resulted in the development of related social policies. Of course, like the construction of postwar industrial welfare states, the process is uneven and has reached different levels in different countries. But the key point here, from the perspective taken by this chapter, is that the policies adopted in order to respond to new social risks are qualitatively different from those that constituted the cornerstone

36


of postwar welfare states. In the old days, social policy was above all about reducing workers’ dependence on the vagaries of markets (decommodification); today most new policies are about improving the living conditions of those workers and citizens who as a result of postindustrialization are facing new difficulties and dilemmas. These two sets of policies not only have a different objective. They also cater for somewhat different social groups. The key target of the postwar welfare state was the male breadwinner and the protection of his income. Still today, the main beneficiaries of the key programmes that constituted the backbone of the postwar welfare state: old-age pensions, invalidity and sickness benefits, tend to be found among men and older age groups, nationals and long established immigrants. In contrast, the key beneficiaries of the new social policies tend to be found above all among women, young people and new immigrants. These two reasons, a different objective and a different target population, should lead us to consider these two sets of policies as ‘different animals’, which should be analysed separately. What is particularly intriguing is the fact that such an analysis reveals some important differences in terms of degree of development in the two sets of policies. Policies catering for the traditional social risks have reached the highest level of development and generosity in continental and Southern Europe. This view, which obviously contrasts with Esping-Andersen’s ranking of welfare states in terms of de-commodification (Esping-Andersen, 1990), is based on the inclusion in the analysis of two additional elements. The first element is law based regulation of the labour market with social policy objectives, discussed at length in the first part of this chapter. Continental and in particular Southern European countries which score low on de-commodification used a different channel to provide economic security to wage workers, which is not captured in Esping-Andersen’s analysis. The second additional element is the age of retirement. Esping-Andersen focuses on three features of pension systems: the replacement rate, the degree to which entitlement depends on labour market participation, and universality. While these are important aspects in so far as the de-commodification of wage earners is concerned, so is the likely duration of their retirement. And in this respect, continental and Southern European countries provide considerably more generous provision than the social-democratic welfare states of Northern Europe. As shown in Table 3.2, in the early 1980s, the average retirement age was three years lower in continental and Southern European countries than in the Nordic countries. This view is confirmed by an analysis based on social expenditure on each of the two sets of policies. We know from previous research that social expenditure as a proportion of GDP is a very crude and sometimes

37


Table 3.2 Indicators of pension generosity at the end of the ‘trente glorieuses’ Country

Public expenditure on pensions as % of GDP, 1980

Average pension as % of average wage (early 1990s)

Effective age of retirement, men, 1980ca

Denmark Finland Norway Sweden Average

5.79 4.7 4.54 6.65 5.42

36.3 48.7 40.2 – 41.7

64.7 61.1 66.5 63.6 64.0

Germany Italy France Netherlands Spain Average

8.65 7.38 7.59 6.52 4.62 6.95

45.2 – – 40.8 41.8 42.6

62.2 61.1 60.2 59.7 61.5 61.0

Australia Canada UK US Average

3.19 2.83 5.10 4.99 4.03

31.7 44.2 38.0 34.7 37.1

61.2 63.3 62.3 64.1 62.7

Sources: Column 1: OECD (2004); column 2: ILO; column 3, Scherer (2001).

inadequate indicator to reflect the effort made by a country in a given policy field (see also Chapter 4 by Siegel, Chapter 5 by De Deken and Kittel and Chapter 7 by Scruggs, this volume). However, it is a useful and convenient way to provide a first approximation. Figure 3.1 presents aggregate expenditure data on both old and new social risk (NSR) policies. The picture emerging from Figure 3.1 confirms what has been argued above. The most generous (or costly) postwar welfare states are found in continental and Southern Europe (especially Italy, Austria, and for the time being in Switzerland), while the most comprehensive systems of new social risk policies are located in the Nordic countries (but also in France). This finding raises a number of issues, the most important of which is probably the question of what the factors are which explain the development of postwar welfare states and NSR policies respectively. The fact that these did not develop in parallel suggests that the independent variables of each set of policies are different, or exert their effect in a different way. Elsewhere (Bonoli, 2006) I argue that the timing of the emergence of new social risks in a country in relation to the process of population ageing is

38

The ‘dependent variable problem’ in comparative welfare state research ♦ GR

%

Spending on OSR policies

16.00

♦I

♦ CH ♦ A ♦ B ♦F ♦ FIN

12.00

♦P ♦E

♦ LUX ♦ UK ♦ NL

♦D ♦S

♦N

8.00

♦ US ♦ JAP

♦ NZ

♦ DK

♦ ICE ♦♦ AUS CAN

♦ IRL

4.00 2.00

4.00

6.00

8.00

%

Spending on NSR policies Note: Old social risk policies includes old age cash benefits, survivors and incapacity cash benefits and unemployment cash benefits; New social risk policies include spending for families (cash and services) and active labour market policies, old age services, invalidity services and social assistance (cash and services).

Figure 3.1 Spending on old (OSR) and on new (NSR) social risk policies in OECD countries as a percentage of GDP, five-year averages, 1997–2001 crucial in determining the development of the corresponding social policies. If NSR develop relatively early, when the financial requirements of pensions and health care systems are still modest, then we are more likely to see the development of a comprehensive system of NSR policies (Nordic countries). In contrast, if NSR emerge when pension and health care systems have to face additional expenditure due to population ageing, then the development of such a new welfare state will be considerably more difficult. But the relevant point in relation to the question addressed by this chapter is that there is little reason to conflate old and new sets of social risk policies into a construct called ‘the welfare state’. These two sets have different objectives, use different instruments, target different groups of people and have developed in response to different social transformations. It is difficult to see why they should be considered together.


39

CONCLUSION: THE APPROPRIATE DEPENDENT VARIABLE This chapter has argued that more attention should be paid to the choice of an appropriate dependent variable in social policy analyses. In comparative studies the dependent variable has often been chosen and defined rather uncritically, on the basis of institutional traditions, in particular because of the UK’s historical lead in social policy research on the British institutional tradition, rather than on the basis of sound intellectual arguments. The discussion above suggests that it is possible to identify two sets of criteria that should be of help in choosing the appropriate dependent variable. First, the best dependent variable depends on the period under consideration. Studies of industrial social policies should define their dependent variable broadly, so as to include functional equivalents developed in different countries. Otherwise, they run the risk of misrepresenting the extent to which wage earners are protected from the vagaries of markets. In contrast, studies of postindustrial social policy should not conflate those policies that were inherited from the postwar years with those developed in the current postindustrial socioeconomic context. Second, dependent variables should be adapted to the research question that is being investigated. If the objective is to assess the extent to which citizens are protected from social risks, whether historically or today, then the broader understanding seems more appropriate. If the objective of research is to study the determinants of policy making, in particular within the political arena, then a narrower dependent variable may be more suitable. In more general terms, we can conclude by arguing that dependent variables in comparative policy research should be selected accurately, and that they should be coherent and comprehensive. They must be coherent, because a high degree of internal unity is needed, if one wants to use them in comparative analysis. But they should also be comprehensive; otherwise we risk missing relevant instruments that just happen to be categorized under different policy areas in different countries.

NOTE 1. This subsection provides a summary of an argument developed in Bonoli (2003a).

PART II

Measuring and analysing ‘welfare efforts’: social expenditure revisited

4. When (only) money matters: the pros and cons of expenditure analysis Nico A. Siegel Expenditures are epiphenomenal to the theoretical substance of welfare states. (Esping-Andersen, 1990) Money is not all there is to policy, but there is precious little policy without it. (Klingemann, Hofferbert and Budge, 1994) The whole notion of a ‘race to the bottom’ is premised on dog-eat-dog cuts in expenditure and taxation. (Castles, 2004)

INTRODUCTION1 This chapter focuses on a particular dependent variable problem in comparative welfare state research. It will assess the strengths and limits of comparative inquiries analysing welfare states mainly or even exclusively on the basis of social expenditure data. Although the major aim of this chapter is to discuss the specific problems of expenditure based analyses, it will also address more general methodological issues which are related to the dependent variable problem in macroquantitative comparative welfare state research. Over the last decade the comparative analysis of welfare reform has become a booming research field. More than a decade ago, Paul Pierson’s seminal work on the political logic of and limits to welfare state retrenchment sparked off a lively and ongoing debate about the so-called ‘new politics of the welfare state’ (Pierson, 1994). This discussion about the old and new politics of the welfare state is far from being settled (Pierson, 2001; Castles, 2004). Scholars from various disciplines and from distinct analytical angles have investigated welfare state change in advanced societies and presented strikingly divergent results concerning both the scope of change 43

44

Measuring and analysing ‘welfare efforts’

and the factors determining its direction. The core argument of this chapter (as generally of the volume to which it contributes) is that beyond normative and theoretical-analytical reasons, basic methodological problems – and in particular a virulent dependent variable problem – have contributed to the strikingly divergent conclusions characterizing the current state of the art in the analysis of welfare state change in advanced societies. This chapter provides both an overview of the strengths and weaknesses of expenditure analysis in general, and a more detailed account of research problems that result from the specific nature of expenditure data when welfare state change is analysed. In what follows, I will first outline why and how the dependent variable problem is related to expenditure based welfare state comparisons before providing a discussion of the potential strengths and problems of using social expenditure data for a comparative assessment of welfare state change. The final section concludes.

THE DEPENDENT VARIABLE AND EXPENDITURE ANALYSIS One set of questions that is of central importance in comparative research focusing on welfare reform in advanced societies may be summarized as follows: what happened to advanced welfare states during the last two and a half decades of the 20th century? Did welfare states undergo a process of retrenchment or of continued expansion? As welfare states represent entities, their change might not be measured adequately if one-dimensional (quantitative) scales are used, hence: do we have to conceptualize welfare state change in more complex multidimensional ways than terms such as ‘retrenchment’ or ‘expansion’ suggest? The relevant literature provides several answers to these questions. Most country studies based on qualitative in-depth analyses of social policy programmes and comparative inquiries into legislative changes have presented evidence showing that the generosity of social rights was adjusted downwards in OECD countries during the 1980s and 1990s. According to these studies, the majority of OECD countries implemented some sort of cost containment measures during the ‘silver age of welfare capitalism’ (TaylorGooby, 2004). While most authors agree that in most countries and contexts cutbacks were rarely of a radical nature, they also agree that cutbacks have contributed to reduced benefit generosity in core welfare state programmes such as pensions and health care (Clayton and Pontusson, 1998; Siegel, 2002). In contrast to these findings suggesting considerable change, macroquantitative studies that have mainly analysed trends in (total gross) social

The pros and cons of expenditure analysis

Public social expenditure % GDP

40.0

45

1980 1993 2001

30.0

20.0

10.0

0.0

es at m St o d gd ite Kin Un d d ite rlan Un itze n Swede Sw n ai al Sp tug r y d Po wa lan r ea No Z nds w la g Ne her our t b Ne em x Lu an p Ja ly Ita and l Ire ece y re n G ma er G nce a Fr and nl rk Fi ma n a De ad n Ca gium l Be ria st lia Au tra s Au Source: OECD (2005a).

Figure 4.1 Public social expenditure in 21 OECD countries, 1980, 1993 and 2001 expenditure in the post-1980 era have found little or almost no evidence for successful cost containment policies. Whereas the remarkable growth of social expenditure (as a percentage of GDP) was a common post-World War II feature across the OECD world up to the 1970s, spending increases slowed down from 1983/84 onwards and particularly during the second half of the 1980s (see Figure 4.1 and Table 4.2). For the 1990s, one may differentiate two major sub-periods. During the first few years most countries experienced a sharp increase in total public social expenditure. Even without an in-depth investigation of the legislative changes which may have contributed to this rise in social expenditure, it is possible to demonstrate by regression analysis that in most countries the economic recession at the end of the 1980s and early 1990s contributed decisively to the expansion of welfare efforts in that period.2 However, the economic recession of the early 1990s was soon superseded by an economic upswing. Thus, in the majority of OECD countries welfare efforts started to decrease again after they had peaked in 1993 or 1994. However, a comparison of social expenditure levels in 1980 (again measured as a percentage of GDP) with those in 2001 does not suggest a massive rollback of the welfare state. Some countries with

46


initially relatively low levels of social expenditure (for example Australia, Portugal and particularly remarkably, Switzerland), went through a notable catch-up process. In contrast, for some (but not all) of the ‘top spenders’ in the OECD league, ‘catch-down’ was characteristic after spending had peaked in 1993 (e.g. in Denmark and Sweden) or 1994 (the Netherlands). A comparison of detailed country studies and comparative social rights based analyses (see also Chapter 7 by Scruggs, and Chapter 6 by Kangas and Palme in this volume) with the results of most quantitative inquiries based on aggregate spending trends reveals divergent findings. This highlights why and how the choice of the key indicator for the dependent variable affects the results of descriptive accounts of welfare state change. It also exemplifies a specific dependent variable problem of comparative welfare state research that has been identified by several authors: the lack of attention paid to a discussion about appropriate (macro and micro) measures of welfare state change and about the operationalization of concepts such as ‘retrenchment’ or ‘restructuring’ (Alber, 1996; Clayton and Pontusson, 1998; Clasen et al., 2001; Siegel, 2002; Green-Pedersen, 2004). In this respect, comparative welfare state research seems to be a showcase for a problem which Francis G. Castles has identified as a common one in social science research in general, i.e. that ‘social sciences systematically under-invest in measurement and hypothesis testing’ (Castles, 2004: 10). The robustness of conclusions in macrocomparative welfare state research suffers from a general lack of more sophisticated measurement tools.3 One might disagree with the second part of Castles’ diagnosis however, which states that social sciences also tend to under-invest in hypothesis testing. One of the most striking trends over the last two or three decades suggests exactly the opposite: abundant studies that have taken hypothesis testing and retesting very seriously – and according to some experts, at times perhaps too seriously if one takes account of the obvious substantial methodological problems these exercises regularly face, but which are generally not discussed in great detail (Kittel, 2004; also De Deken and Kittel, Chapter 5, this volume). Due to technological innovations such as the development of highly efficient statistical software and new developments in regression analyses, a fast growing body of studies has tested and retested a rather small set of standard hypotheses on the basis of more or less the same case selection using expenditure data. Yet this expansion of the replication business in comparative welfare state research has not triggered impressive progress in terms of refining existing theories or developing new ones. Instead, it has failed to produce robust empirical findings and delivered ambiguous conclusions which often seem to be built on rather thin empirical evidence.


47

Compared to the growing number of quantitative studies in comparative welfare state research that keep the ‘replication business’ of quantitative hypothesis re-testing booming and suggest far reaching generalizations about complex processes of change, only a small number of studies have explicitly tried to systematically address basic questions of measurement and concept specification – or tried to build bridges between theory and methodology research (however, see Green-Pedersen in Chapter 2 and Scruggs in Chapter 7 of this volume). As a consequence of the under-investment into the ‘empirical infrastructure’ of macroscopic welfare state research, fascinating concepts such as retrenchment, reform, recalibration, adjustment or restructuring have been introduced but not been followed by more rigid and thus testable operational definitions. The lack of attention to measurement issues causes serious problems for descriptive inferences in macroqualitative and macroquantitative studies. Things can only be more delicate at the level of analytical inferences. A fuzzy conceptualization of a descriptive dimension of welfare state change hardly provides a fruitful ground for robust analytical inferences (Alber, 1996). This leads us to a second set of key questions which is directly related to the dependent variable problem in the use of expenditure data in comparative welfare state research. It concerns the analytical dimension of investigating welfare state change. A fast growing body of literature has dealt with the social, economic and political factors that shape welfare state change in Western democracies. Some of the classical questions about the relevance of impact of national polities and politics have repeatedly been asked: do parties and power resources of collective social actors still matter for the politics of welfare state reform (Korpi and Palme, 2003)? Do we observe the ‘new politics’ of welfare state recalibration, shaped by path dependent trajectories of incremental adjustment and blame avoidance strategies of politicians whose strategic actions are dominated by (re-)election rationales (Pierson, 1994; 2001)? Given the enormous dissent concerning the basic problems of describing welfare state change, the dissent on the crucial factors shaping welfare state reforms is hardly surprising (Kitschelt, 2001: 299). Hence, the current literature on welfare state change is characterized by at times stunning overlaps and numerous contradictions. One striking example may be used to illustrate this point. Based on rather distinct research design both quantitative and qualitative studies have reported strong evidence for Pierson’s claim that retrenchment is not the ‘the mirror image of expansion’ (Pierson, 1994: 1) and for ‘the new politics of the welfare state’ (Ross, 1997 and 2000b; Huber and Stephens, 2001; Pierson, 2001). However, several other studies have

48


contested the new-politics paradigm and suggested that the ‘old politics of the welfare state’ is indispensable for understanding the political logic of contemporary welfare state reform (Clayton and Pontusson, 1998; Korpi and Palme, 2003). Again, other scholars claim to have found robust evidence for the seemingly paradoxical conclusion according to which the partisan composition of governments, a major ‘old politics’ factor, still shapes welfare state outcomes in an era of permanent cost containment – but in the opposite direction if compared with the ‘golden age’ of welfare state expansion. According to this view, left parties are in a better position to mobilize support for unpopular welfare state reform (such as benefit cutbacks) as they are usually perceived as the most ‘natural’ pro-welfare state party. Consequently, at least in most national party system contexts, left parties may convincingly argue that any alternative political party to their right would implement even more severe cutbacks (Armingeon et al., 2001). According to this ‘Nixon-goes-to-China logic’ social-democratic governments should be in a better position to implement cutbacks than any parties to their right. However, based on a sophisticated pooled time series analysis of social expenditure trends in the post-1980s, Kittel and Obinger (2003) have challenged this ‘reversed partisan theory’. In contrast to previous social expenditure developments, their findings suggest that the 1990s did not witness a significant partisan impact on welfare state spending. Finally, some authors have even dismissed the notion of a clear demarcation line between ‘old’ and ‘new’ politics of welfare state change. According to this notion, the two standard paradigms may not be perceived as strictly disjunctive ones. Rather, they may be viewed as at least partly overlapping analytical approaches which do not completely rule each other out. Due to the complex interplay of several factors shaping welfare state outcomes, the internal complexity of processes of welfare state change, and variable time lags between causes and effects, it is very difficult, at times perhaps indeed impossible, to differentiate between new and old politics factors in any real categorical sense (Siegel, 2002).4 In sum, what this chapter addresses might be labelled as the problem of ‘artless sophistication’ in macroquantitative comparative welfare state research. Artless sophistication is the consequence of innovations in ever more sophisticated techniques of data analysis which have been introduced into comparative social sciences without a systematic or critical assessment of their potential and their limits when it comes to analytical inferences. As a result, basic but substantial problems of macroquantitative (welfare state) research are often ‘assumed away’ for purely pragmatic reasons.5 As problems of data quality and consistency and, more substantially, of causal


49

inference are not discussed explicitly, but implicitly built into model specifications of regression analyses, research tends to produce rather unrobust findings. In what follows, this chapter provides a summary of major advantages and disadvantages of using expenditure data for exploring the dynamics of welfare state change. It will address the following questions: what are the major strengths of an expenditure sensitive approach to welfare state analysis? Why is expenditure analysis even more important in the era of ‘permanent austerity’ (Pierson, 2001) than during the golden age of the welfare state? What are the major pitfalls of expenditure based analysis in (a) describing and (b) analysing welfare state change? Why does the analysis of welfare state change over time raise more sensitive methodological issues than a purely cross-sectional perspective?

WHY AND WHEN MONEY REALLY MATTERS: EXPENDITURE ANALYSIS IN AN ERA OF PERMANENT AUSTERITY It may seem paradoxical but for a more extensive discussion of the advantages of expenditure analysis in comparative welfare state research it is worth revisiting one of the most frequently cited criticisms. It was Gøsta Esping-Andersen who argued that ‘[E]expenditures are epiphenomenal to the theoretical substance of welfare state’ (Esping-Andersen, 1990: 19) and that ‘[I]t is difficult to imagine that anyone struggles for spending per se’ (Esping-Andersen, 1990: 21). There are indeed good reasons to subscribe to Esping-Andersen’s critique of expenditure based welfare state studies which apply readily available sets expenditure data sets for ‘testing’ theories of the welfare state. Two major caveats against a sweeping critique of expenditure based welfare state analysis deserve to be mentioned however. First, as I will argue below, both the history of the welfare state and the development of welfare state research provide us with numerous and important examples demonstrating that expenditure based approaches can offer most valuable insights into welfare state change.6 The second point is related to the ‘new’ political economy of the welfare state which emerged in the aftermath of the first oil price shock and manifested itself in the 1980s and 1990s. It is particularly characterized by the increasing salience of cost containment budgetary efforts and is framed more generally by a political discourse that puts more emphasis on the costs of social provision than on the social and political achievements of welfare states.

50


Expenditure Analysis and its Pundits: Myths and Practice in Comparative Research It is worth putting Esping-Andersen’s frequently cited critique of expenditure analysis into the context of his own work. At several points in Chapter 1 of The Three Worlds of Welfare Capitalism, Esping-Andersen bluntly dismisses the value of comparative studies which carry out ‘theory tests’ by picking up easily available expenditure figures as ‘data fast food’ for the production of statistical outputs. He disapproves of the practice of using expenditure figures as the main dependent variable without critically discussing what expenditure actually indicates. His interest in structural and qualitative aspects of welfare statism reflects a much more ambitious approach to macrosociological research. However, a closer inspection of the empirical foundations on which Esping-Andersen’s regime typology is built shows that he himself was not reluctant to utilize spending data. The most striking example is the use of data for four out of seven indicators, of the public and private social expenditure for the construction of the stratification index presented in Chapter 3 of The Three Worlds of Welfare Capitalism (Esping-Andersen, 1990: 69–78). Choosing a quite pragmatic approach to expenditure data EspingAndersen thus hardly qualifies as a principal witness for a fundamental critique of any kind of expenditure based welfare state analysis. The main target of his criticisms was the apparently context-blind (mis)use of expenditure data for hypothesis testing which (implicitly) assumes that expenditure figures may be used as proxies for welfare state generosity. Thus the valuable criticism made by Esping-Andersen and others does not mean that sensitive expenditure approaches cannot provide instructive insights into welfare state profiles and developments. A disaggregated analysis of the structure of social expenditure, for example, can offer important insights into the inherent spending asymmetry between different branches of the welfare state. A closer inspection of the composition of the overall social budget can provide valuable estimates of the political salience of different welfare state programmes. Although old-age cash benefits generally consume the largest share of total social expenditure in national accounts, the profiles of OECD countries differ significantly in their structure of public expenditure (Siegel and Jochem, 2004). These findings can be immensely important for researchers (as well as politicians) involved in the kind of ‘cost containment’ debates or controversies which discuss the financial sustainability of existing welfare state arrangements. There are other important substantial and pragmatic arguments against a general critique of expenditure analysis. Studies which have mainly or exclusively applied expenditure data have been pivotal to comparative


51

welfare state research since the late 1950s (e.g. Wilensky and Lebeaux, 1958). From the second half of the 1970s onwards and particularly in the 1980s some of the most influential studies in comparative social policy have used expenditure data as one of their prime measure of welfare state size and effort (Castles, 1982b; Schmidt, 1982). Because social expenditure has become the largest component of total government outlays in all OECD democracies it is also the main predictor for the size of the tax state. Thus, the analysis of social expenditure trends is of the utmost importance for a broad range of questions concerning the history and the future of public policies in democratic countries (Castles, 1998; Schmidt, 1998; Siegel and Jochem, 2004). As long as spending figures are not misread as proxies for welfare state generosity (see below) a cautious analysis of social expenditure can certainly be an important and in some cases indispensable instrument of comparing welfare states and their change over time. However, there are important problems involved in using spending data for comparative research and about the way social expenditure is used in the current literature on welfare reform. One major pitfall is the use of increasing (or decreasing) levels of social expenditure as indicators for more generous (or restrictive) welfare provision. As will be discussed in more detail below, social expenditure should not be used as a proxy for the generosity of social rights (see also the Chapter 7 by Scruggs and Chapter 6 by Kangas and Palme in this volume). The generosity of social rights is directly affected by political decisions, and is a major predictor of social expenditure. However, as social need and economic output also affect levels and changes of social expenditure it is hazardous to use changes in social expenditure to infer (proportional) changes in the generosity of benefits and/or services. There are ways to adjust social expenditure in order to make it more sensitive to questions of welfare state generosity. Several authors (Castles, 1982b; Clayton and Pontusson, 1998; Huber and Stephens, 2001) have analysed social expenditure in accordance with levels of social need, thereby calculating welfare-to-need ratios or ‘standardized welfare efforts’ (Siegel, 2002). Again, it is worth bearing in mind that an ideal indicator might exist in research theory but certainly not in research practice. Thus, estimates of social need such as the level of unemployment or the size of the pensioner-age population should only be interpreted as rough proxies of more complex constellations of need. In other words, the assessment of social need deserves to be explored in much greater detail in the context of more thorough analyses of programme focused studies. But indicators such as the percentage of the population 65, of the unemployed, or the share of ‘income poor’ households may be used as reasonable need estimates for

52


cross-national comparisons. As comparative studies using needs adjusted expenditure trends have shown, ‘welfare-to-need ratios’ often deviate remarkably from non-adjusted gross social expenditure figures (Clayton and Pontusson, 1998; Siegel, 2002; Castles, 2004). The asymmetry between gross and needs adjusted social expenditure is particularly striking during and shortly after economic crises, e.g. in the aftermath of the first oil price shock in the 1970s, in the early 1980s and during the recession of the early 1990s. While needs adjusted social spending started to decline in several OECD countries, accounts of gross (unadjusted) social spending suggested continuous expansive growth dynamics as a consequence of increasing social need in the context of low or even negative GDP growth. Breaking down total public social expenditure into programme specific trends, thereby climbing down the ladder of aggregation, is another useful tool for comparative welfare state researchers. Disaggregation helps to dismiss generalized judgements about welfare state developments which ignore at times highly variable trends within individual social policy programmes. One example is the huge difference between the level, and growth, of expenditure for old-age pensions on the one hand and for the support for the unemployed on the other. Despite popular beliefs, government spending on unemployment benefits only marginally contributed to the growth of total social expenditure between 1980 and 2001. For the 20 advanced democracies, the OECD SOCX database contains figures for both the years 1980 and 2001 showing that the mean value for spending on unemployment benefits was 0.97 per cent of GDP at the beginning of the period and 1.05 per cent of GDP in 2001. If one bears in mind that the average unemployment rate was more than one percentage point higher in 2001 than in 1980 (6.2 per cent vs 4.9 per cent, based on OECD figures for commonly used national definitions) and taking the stronger expenditure growth for health care and old-age pensions in most countries into account, these figures reveal a rather marginal growth dynamic and reduced welfare-to-need efforts. Expenditure Analysis in the New Political Economy of the Welfare State Concerns about the use of expenditure data as ‘cost indicators’ are particularly salient when it comes to political debates about the cost of welfare in the ‘era of permanent austerity’ (Pierson, 2001). As cost containment policies have made it up to the top of government agendas in most OECD countries, and since the future of national tax states has become a major political battlefield, one may just turn around some of the older assumptions about the (ir)relevance of expenditure figures. Only few governments in Western and Northern Europe have shown a clearly identifiable ideological commitment to rolling back the welfare


53

state, such as in the US and the UK of the 1980s. Nevertheless, as part of a more general shift towards cost containment and saving policies, governments in most OECD countries have implemented at least some entitlement cuts. Cutbacks and saving measures with ‘a bad conscience’ have been the rule rather than the exception in the majority of European welfare states where social-democratic and Christian democratic parties (in various coalition constellations) have often negotiated about cost containments measures within coalition governments. Processes of top-down budget policies which consist of spending caps and similar fiscal policy tools have become increasingly important. Compared to the golden age of Keynesian welfare state expansion, ‘social policy in hard times’ (Huber and Stephens, 2001) often involves redistributive zero-sum policies. Whereas in the 1960s and early 1970s increasing levels of social expenditure were used by centrist and left governments to highlight social achievements, expenditure figures are nowadays often presented as ‘cost indicators’. Both notions, the one focusing on ‘achievements’ only and the other emphasising exclusively the ‘cost’ of welfare, suffer from an ill balanced and somewhat simplistic core assumption – namely that (actual) expenditure equals real costs, which makes neither economic nor political sense. Just as it is misleading to confuse social expenditure levels with measures of social policy generosity, it is equally misleading to use expenditure accounts as measures for the societal cost of the welfare state: an all-encompassing definition of social or economic costs cannot be reduced to the level of outlays within a specified time period as it is reflected, for example, in annual budget figures. The search for an adequate measure that could be used to estimate the real (net) cost of welfare state policies is a knotty problem. One would have to calculate the net effects of complex social, political and economic outcomes which are directly and indirectly affected by welfare state policies. One also needs to specify the time horizon for a proper cost estimate, as many welfare state consequences only materialize over long periods of time or with a significant time lag. As social expenditure should neither be confused with welfare state generosity nor with the complex costs of the welfare state, a rather neutral use of and label for social expenditure (as a percentage of GDP) seems to be most sensible for empirical-analytical welfare state research. ‘Welfare efforts’, a term introduced by Wilensky and Lebeaux (1958: 156) almost half a century ago, offers a reasonable label without emphatic or pejorative connotations. However, in contrast to Wilensky’s simplistic assumption according to which welfare efforts reflect ‘the budget decisions of political elites’ (Wilensky, 1975: 17, footnote 1), they actually depend on a set of intervening and interdependent mechanisms, comprising factors of social

54


need, economic context (impacting on the denominator, GDP) and politically defined generosity of social rights. As soon as one or more of these three dimensions change, one may notify a change in (total) welfare efforts. The crux about welfare efforts is that they (a) present expenditure sums based on data collected for individual social programmes and (b) represent a ratio which is determined by a numerate (social expenditure) and a denominator (usually Gross Domestic or Gross National Product). Hence welfare efforts can be used as a reasonable estimate of the share of national economic output which is channelled through the different pillars of a country’s social security system. Welfare state efforts therefore reflect the elasticity of public budgets for social purposes but they certainly reflect more than the (direct) consequences of political decision making processes.

MAPPING THE LIMITS OF EXPENDITURE ANALYSIS Advanced welfare states of the majority of OECD countries comprise complex sets of diversified institutions and policies, of stratifying regulatory systems that shape societal outcomes in various ways. In most countries, welfare state policies do not only comprise quantifiable features of (direct) state intervention but also private and occupational systems. Not all policies which might be regarded as quantifiable are manifest in spending accounts. ‘Qualitative’ social policies, such as social regulations in health and safety or equal opportunity legislation have become increasingly important, particularly in the ‘spending poor’ context of EU social policy making (Majone, 1994). Until the 1990s, policies affecting the regulation of product and labour markets had often been neglected in comparative welfare state research (but see Samek Lodovici, 2000). Yet regulatory policies represent an essential dimension of market restricting and correcting state intervention (Esping-Andersen and Regini, 2000). In terms of both welfare state size and its internal structure, important features of advanced welfare states are not reflected in public spending accounts. Hence whilst we have argued in the previous section that in many research contexts an expenditure sensitive approach to welfare state analysis is valuable, we also put strong emphasis on the notion that in many (or even most) research contexts approaches that rely exclusively on expenditure indicators provide a too narrow perspective on the welfare state. As all social expenditure counts equally (i.e. regardless of who benefits to what extent from cash benefits or social services), the analysis of aggregated expenditure data particularly does not offer any valuable insights into the


55

underlying structures, the distributional logic or the consequences of different types and changes of welfare state provision. The limits to a purely expenditure based analysis become obvious within in-depth investigations of the process of welfare state change. Expenditure figures are indicators that represent policy outcomes rather than policy outputs (see also Green-Pedersen, Chapter 2 of this volume). If the primary focus of interest is the political process which brings about welfare state change researchers are well advised to focus on policy outputs, as policy outcomes are influenced by a range of social, economic and political factors. In a nutshell, we may be well advised to narrow our focus, shorten causal chains and choose a microscopic approach that analyses programme specific processes and policy outputs. Long, Complex Causal Chains and Narrow Time Perspectives: Social Rights, Welfare Efforts and Pooled Time Series Analysis As already outlined above, a move backwards in the causal chain from expenditure levels or changes therein (measuring a welfare state outcome) to the generosity of social rights (measuring an output) is a risky undertaking due to the many intervening variables that may affect the causal relation between social rights and expenditure (see also Kangas’ and Palme’s contribution in Chapter 6, this volume). One example is the inconsistency of findings which pooled time series analyses (TSCS) of social expenditure trends for the post-1980 period have produced in terms of assumed effects of partisan politics on changes in social expenditure. Relying on pooled analysis most authors have reported overwhelming evidence for a dominant autoregressive ‘level effect’ and the relevance of social and economic variables. GDP growth and levels/changes of unemployment rates are the two social and economic variables which have figured most prominently in studies that have either specified annual changes in total social spending or levels of social expenditure as their dependent variable, controlling for autocorrelation by including a lagged dependent variable (Siegel, 2002; Kittel and Obinger, 2003). However, based on the results of significance tests for partial regression coefficients within multivariate model specifications, political factors should not be dismissed. Political factors may impact on social spending either indirectly or with a considerable and variable time lag.7 Both problems raise serious questions about the standard model specifications that have been used in the majority of studies using pooled time series techniques.8 Political factors may be regarded as important background variables which have a more subtle, cumulative, delayed and therefore obfuscated impact on spending patterns than economic and social need variables.

56


The problem is that the standard (i.e. linear and additive) pooled regression models do not take variable time lags or causal sequences into account. For example, it is important to note that increasing social need is only reflected in expenditure growth if social rights have not been ‘immunized’ against a growing demand effect, i.e. as long as policy makers have not implemented legislative changes limiting the access to social rights or cutting back benefit levels for existing groups of beneficiaries. From a theoretical point of view, statistically significant partial regression coefficients for social need variables in pooled cross-sectional models should be interpreted as evidence that changing social need is a significant short-term source of expenditure changes. But more deeply rooted causes which enable the short-term sources to become visible are to be found in the more deeply rooted policy inheritance (Rose, 1990). This point deserves some clarification. For analytical purposes, levels of social expenditure at the programme level (for social transfers) can be disaggregated into one factor which reflects the number of claimants receiving a benefit and another one which reflects the average benefit payment (OECD, 1985). To assess the consequences of political decisions on expenditure levels the potential relevance of at least two types of policy changes is to be taken into account: (1) policy changes altering the generosity of benefits (entitlements) and (2) changes aiming at reducing the number of beneficiaries (e.g. via changing eligibility criteria). Increasing levels of social need are translated into higher expenditure only if neither the access to nor the level of social rights is significantly affected by policy changes. If benefit conditionality is tightened between two time points (Clasen and Clegg, 2003 and Chapter 8 in this volume), it is likely (and in most cases politically intended) that a lower share of officially recognized unemployed persons qualify for an unemployment benefit payment. Hence the relationship between the level of unemployment and expenditure on unemployment benefits changes, as soon as political interventions distort the causal chain between these two variables. However, in none of the multivariate models estimating the determinants of social expenditure changes do we find a term which would control for changes in the conditionality of benefits, as there is no crossnational time series which would provide comparative researchers with a reasonable quantitative measure of conditionality. Time series analyses of welfare efforts that focus on annual changes face another tricky problem. Particularly during and shortly after economic recessions, when unemployment rates tend to increase sharply, increases in social spending are mainly a consequence of growing social need and reduced economic output, both highly correlated sources of short-term expenditure change. The dramatic increase of total social expenditure in


57

Finland in the early 1990s is a good case in point. According to OECD SOCX data Finnish total social expenditure jumped from just 24.8 per cent of GDP in 1990 to 33.9 per cent in 1991. This dramatic increase mainly reflected a rather sharp decrease of GDP of 5.9 per cent and a doubling of the unemployment rate (from 3.2 per cent to 6.7 per cent). Clearly, the impressive increase of Finnish total social expenditure figures was not the consequence of a unique Finnish experiment lifting the generosity of welfare state provision. Instead, as Saari (2001) has shown, it was the institutional configuration of the Finnish welfare state (the deeply rooted cause) and its interaction with increasing social need, combined with a severe recession of three consecutive years of negative GDP growth which led to a rapidly growing social spending share of GDP in the early 1990s. Time Sensitivity When reconsidering the results of pooled time series analysis on social expenditure trends in the post-1980 period a strong tendency to confirm the (statistical) impact of socioeconomic short term factors plus a level effect indicating catch-up dynamics (as well as ‘catch-down’ effects) can be identified.9 The essential question is what kind of inferences can be drawn from the results of model specifications which are used to analyse annual spending changes. More variable (volatile) short-term factors may for example not decisively impact on the level of social expenditure in the long run. And the level of social expenditure at any time point t1 (the often used autocorrelation term, the ‘lagged dependent variable’) is itself a product of social, economic and political factors that have determined social expenditure at previous time points (t2, t3, . . . tn). What can be demonstrated on the basis of a comparison of the results of pooled time series cross-sectional analysis with OLS cross-sectional analysis is that the relationship between levels and changes of unemployment rates on the one side and levels and changes of social expenditure on the other is extremely time sensitive: in models estimating the determinants of short-term changes of social expenditure, differences in unemployment rate are often statistically one of the most significant predictors of spending changes. In contrast, the level of unemployment often fails to surpass conventional significance thresholds when levels of social expenditure are specified as the dependent variable. This shows that short-term fluctuations of unemployment can cancel each other out in the long run, whereas changes in unemployment contribute to the short-term variation of social expenditure. The comparatively low number of recipients of unemployment benefit – compared to the number of pensioners or health care users – in combination with the comparatively low level of unemployment benefit

58


payments in most countries results in a rather small share of unemployment benefits in total social expenditure. Hence the unemployment levels not usually belong to the most important predictors of social expenditure levels in cross sectional studies. Compared to the short and long-term relationship between the level and/or change in unemployment and levels of total social expenditure the exact opposite time structure is indicative for the relationship between demographic change and social expenditure. Leaving exceptional contexts – such as massive displacements – aside, demographic variables usually do not change dramatically from one year to another. Changes of demographic features are often of a creeping and rather continuous nature. This makes it easy for politicians to play down long-term challenges as the shortterm effects of demographic changes are less visible than in the case of more volatile socioeconomic push factors. Demographic changes represent the kind of slow moving (creeping) process with potentially huge cumulative effects over longer periods of time. The combined effect of higher life expectancy and lower fertility rates represents the most important (domestic) challenge for the financial sustainability of mature welfare states. But it might not be the most obvious short-term challenge for a finance minister in annual budget rounds or in the context of electoral cycles of four to five years. The tricky problem for the analysis of short-term periods is that due to the slow changes in demographic variables they usually do not come up as statistically significant predictors of annual expenditure changes. Thinking about the core set of political factors that have regularly been used as independent variables in most quantitative studies applying expenditure data, such as the partisan composition of governments, indices for the power of trade unions, the extent of corporatist policy making, or measures of the institutional veto density of state structures, it is plausible in most contexts to assume that legislative changes unfold their effects often cumulatively or with a significant time lag between policy changes and significant outcome effects (Huber and Stephens, 2001; Castles, 2004). This further complicates the search for adequate tools to establish robust causal links between expenditure levels/changes, levels and changes of social rights and the key independent variables which plausibly may impact on spending changes. Some of the remedies suggested in the literature, for example ‘cumulative measures’ of partisan incumbency, make sense if they are used for the analysis of expenditure changes over extended time periods. A dynamic approach to analysing spending data can be simulated by regressing on changes between more distant time points, such as changes in spending levels for 20 year intervals for example (Castles, 2004). But this approach does not offer a promising solution to the problem of analysing short-term changes of social expenditure (Huber and Stephens, 2001).


59

Also, extending time intervals causes problems as it ignores information about within-period variation for both descriptive and analytical questions. Variable Impact Chains When drawing conclusions about the relationship between political decisions (and non-decisions!) and levels/changes of welfare efforts it is important to consider variable causal chains between political decisions manifesting themselves in policy outputs – such as the codification of social rights – and their outcome in terms of expenditure effects. In the best of all democratic worlds, politicians do what (the majority of) voters want them to do and achieve what they intend to achieve. In such a utopian democratic context, one may assume a perfect match between policy goals and policy outcomes. However, the relationship between voters’ preferences and government actions, just as between intentions and the achievements of governments, is more complicated. Let us assume a government has chosen to take the most common and frequent policy choice: non-decisions, i.e. the conservation of the policy status quo (Tsebelis, 2002). This choice can result in radically different policy outcomes in terms of social expenditure. It can contribute to an increase, a decrease, or a stabilization of social spending. Only the latter case reflects a congruency between policy output and expenditure outcomes. However, in many circumstances it is extremely difficult to specify the consequences of the hidden dominance of non-decision making in a rigorous way and to assess the consequences of political inertia. Yet for the analysis of social expenditure trends it is of crucial importance to note that political decisions and non-decisions may generate intended but also unintended (and even unforeseen) consequences. At the same time it is exactly the reason why the process of moving backwards in the causal chain, from expenditure changes (as outcome measures) to policy changes (reflecting outputs of the political process) and government intentions, is a highly error prone process of causal inference. The particular theoretical importance of the implication of a legislative status quo preservation of social rights becomes apparent when Tsebelis’ veto player theory (Tsebelis, 2002) is applied. The link between the number of veto players, the ideological policy distance between them and their internal congruence all affect the likelihood of the preservation of the policy status quo on the one hand, and the change in social expenditure on the other hand. Yet this is most difficult to disentangle since status quo preservation in terms of, for example, benefit generosity can have a large, a small or no effect on short-term changes in welfare state budgets. Table 4.1 presents a simplified illustration of the key argument put forward here. It shows that only in a minority of stylised contexts might a

60

Social rights

Table 4.1

Asymmetric ‘Unintended expansion’ as a consequence of conservation of level of social rights + increasing need/ programme maturation Contradictory ‘Loose-loose’ constellation: shrinking generosity, increasing spending Re-commodification without consolidation: unintended spending growth despite cutbacks due to (sharp) increase in social need and/or unfavourable economic context

Decreasing generosity/ recommodification

Congruent Politically induced and Intended retrenchment

Failed cost containment Due to increasing need/low (or negative) economic growth and/or insufficient cutbacks

Contradictory ‘Win-win’ outcome: increasing generosity and less expenditure due to decreasing demand and/or high economic growth Asymmetric Consolidation as a side-product of decreasing need or high economic growth

Decreasing welfare efforts –

Asymmetric

Identical ‘Natural stagnation’. As a consequence of status quo conservation, non-decision making

Asymmetric ‘Cheap decommodification’ due to decreasing need and/or favourable economic context

Congruent Politically intended expansion ‘golden age of the welfare state’ constellation

Stagnation/consolidation at a certain level 0

Expenditure Increasing welfare efforts +

Status quo conservation 0

Increasing generosity/ de-commodification

Generosity/ commodification

Social rights and welfare efforts


61

congruent dynamic between social rights and welfare efforts be expected. While the relationship between social rights and welfare efforts may have been congruent during the golden age of the welfare state, i.e. reflecting a politically induced expansion of social rights resulting in (expected and maybe even appreciated) higher social expenditure levels, the relationship is not that clear-cut in other time periods. Social policy in ‘hard times’ (Huber and Stephens, 2001) is characterized by low economic growth and increases in social need. In such a context what might be observed is expenditure growth despite government initiatives of cutting back social benefits, or preserving the status quo. Germany, in the post-unification period (particularly 1996–98 and 2003 to the present) is a typical example of this unintended expenditure increase, indicating failed cost containment initiatives. As this example illustrates, the potential error probability of any ad hoc inferences from social expenditure dynamics to legislative changes (‘government policies’) affecting welfare state generosity is very high. The Implicit Constant Causality Assumption in Social Expenditure Analysis The problems of drawing inferences from the analysis of total social expenditure using regression analysis are not restricted to the relationship between social rights and welfare efforts. Standard OLS regression in the form of y (Total social expenditure) ab1x1b2x1 . . . bkxme with b1, b2 . . . bk representing partial (unstandardized) regression coefficients of social, economic and political variables are usually built on (rather implicit) assumptions about causal homogeneity (unit homogeneity, see King et al., 1994). The assumptions of causal homogeneity across space (countries) and time have been discussed in the literature on pooled analysis, as one regression coefficient is to be estimated for all countries and time periods, even when fixed unit and or time effects models are specified (Kittel, 1999; Kittel and Winner, 2005; but see also Plümper et al., 2005). However, there is one implicit homogeneity assumption in models that formalize the relationship between aggregate social expenditure and the independent variables which has not been discussed in any detail in comparative welfare state research. As total social expenditure represents the sum of expenditure for different programmes (or functions), and given the complex relationship between a change in a policy output – measures by an indicator such as a change of social rights – and the level of social expenditure at the level of individual programmes, is it plausible to carry out comparative

62


expenditure analysis at the highest level of aggregation? The relationship becomes more complex and sensitive to disturbing intervening variables if both social rights and social expenditure indicators are aggregated. Moreover, it is not only theoretically plausible but, by simple correlation analysis, can be empirically shown that the relationships between political, social and economic variables and expenditure outcomes do diverge substantially across different social policy programmes. In other words, the relationship between variables cannot be assumed to be constant across different social policy programmes. The effect of demographic ageing as a major social need factor, for example, varies between social policy dimensions such as health care, long-term care and old-age and survivors’ pensions, and is not a major predictor of spending on active or passive labour market programmes. In a similar vein, the effect of partisan politics or corporatist policy making may be different across different welfare state domains. This is not the place to discuss all the advantages (and some of the obvious disadvantages) of a more disaggregated approach to expenditure trends in detail. Suffice it to note that, over the last decade, a shift towards more cautious choices in levels of analysis can be observed as an increasing number of authors have focused on disaggregated, programme specific analyses of spending trends (Castles, 1998; 2004; Huber and Stephens, 2001; Siegel, 2002; Kittel and Obinger, 2003). Problems of Data Quality and Measurement Error In textbooks on research methods it is regularly pointed out that all data are in one way or another socially constructed. The data collected and the indicators developed by individual researchers are the end product of a social process of information gathering and standardization which, inevitably, involves critical choices. Statistical data published by national statistical authorities are no exception. Social expenditure figures, such as those published by the OECD, are the product of a complex multistage process. In the case of total social spending aggregation problems are an additional problem. It is worth pointing out some of the issues which illustrate why the quality of data and measurement consistency across countries and time are of crucial importance for comparative inquiries of welfare state change using expenditure data. As De Deken and Kittel (Chapter 5 in this volume) discuss in more detail, the need for data quality and for data consistency are crucial challenges in standardized international data sets provided by OECD or Eurostat. The point can be illustrated with reference to the SOCX dataset compiled by the OECD. One of its specific problems is the absence of spending data on active labour market measures for nine countries for five


63

consecutive years at the beginning of the observation period (1980 to 1984). The inclusion of active labour market expenditure data from 1985 onwards generates a statistical leverage effect on total social expenditure for these OECD countries. Any descriptive, and consequently any analytical, inferences which are drawn from total social expenditure without a correction of this effect are therefore seriously flawed due to a severe statistical artefact. The case of Ireland provides an extreme illustration of uncorrected time series of social expenditure resulting in highly misleading inferences. For the period 1980 to 1984 the OECD has clearly marked the missing data for unemployment cash benefits and expenditure on active labour market policy in Ireland. From the year 1985 onwards, the figures for Ireland are included but the OECD series does not report a statistical break in the tables for total public social expenditure. Active labour market expenditure added up to 1.5 per cent of GDP and unemployment benefits to 3.4 per cent of GDP in Ireland in 1985. Hence, 4.9 per cent of the Irish GDP was spent on active and passive labour market issues in 1985! The dramatic 4.8 percentage increase of total social expenditure from one year to the next is thus not really an intellectual puzzle, but points to a major statistical problem in SOCX. Although not representative, the above example shows that the quality of readily available social expenditure figures in the most aggregated form should not be over-rated. There are good reasons not to praise OECD and ILO expenditure data sets as ‘excellent’ (Amenta, 1993: 752). Rather than interpreting them as exact measures, a more cautious approach would be to view them as proxies for actual spending levels, as ‘point estimates’ with a certain error margin.10 A misleading assumption of much quantitative research is that a national statistical offices can rather easily produce figures for total spending while, at the same time, they are assumed to face problems compiling spending data for individual programmes. Exactly the opposite seems to be realistic in the context of most advanced welfare states. In countries where social expenditure is not entirely financed out of general taxation, the collection of data involves a complex ‘bottom-up process’ which includes several (collective) actors (e.g. social insurance funds) and levels of government. The more fragmented a social security system and its financing structure, the more difficult it is to arrive at a correct overall picture, standardized for national and then for international purposes. The problem of missing data may be corrected in a technical way that enables researchers to carry out reasonable time series analysis without dropping too many observations. But studies which have used the SOCX database often fail to report adjustments which were made to the OECD

64


Table 4.2 First order differences, annual total social expenditure, 1980–2001: year and country specific averages Year interval

Period specific mean values: annual first order differences

1980–81 1981–82 1982–83 1983–84 1984–85 1985–86 1986–87 1987–88 1988–89 1989–90 1990–91 1991–92 1992–93 1993–94 1994–95 1995–96 1996–97 1997–98 1998–99 1999–2000 2000–2001 Mean excluding 1990–93

0.76 0.66 0.76 0.34 0.53* 0.05 0.16 0.12 0.13 1.00 1.06 1.07 0.63 0.29 0.37 0.03 0.50 0.26 0.15 0.39 0.30

Country specific mean values AUSTRALIA AUSTRIA BELGIUM CANADA DENMARK FINLAND FRANCE GERMANY GREECE IRELAND ITALY JAPAN LUXEMBOURG NETHERLANDS NEW ZEALAND NORWAY PORTUGAL SPAIN SWEDEN SWITZERLAND UK USA

0.32 0.17 0.15 0.17 0.00 0.30 0.35 0.21 0.61 0.15 0.29 0.32 0.13 0.24 0.06 0.29 0.49 0.18 0.00 0.58 0.19 0.07

0.19 0.01

Note: * Statistical leverage effect as for nine of the 22 countries data OECD total public expenditure figures contain spending on active labour market measure only from 1985 onwards.

time series. Hence we have to assume that data problems were either not detected, neglected or regarded as a small issue of inconsistency, none of which is satisfactory. Breaks in time series are more than just a minor nuisance. They can cause severe problems for the assessment of changes in social expenditure in more detail. The major reason for this is illustrated in Table 4.2 which reports average changes in total social expenditure for each year for OECD democracies for the period 1980 to 2001. It shows the average annual changes for 22 OECD democracies and 21 annual changes. The average annual first order

65


120

100

Frequency

80 Mean = 0.191; N = 462

60

40

20

0 –2.00

0.00

2.00

4.00

First order differences, total social expenditure per cent of GDP Note: Countries here are the same as in Figure 4.1. Source: OECD (2005a).

Figure 4.2 First order differences, annual total social expenditure, 1980–2001 change of total social spending as a percentage of GDP for 22 countries in the period 1980 to 2001 was 0.19. If we exclude the exceptional crisis years of the early 1990s (1990–93), the average year-by-year change is reduced to a marginal 0.01 percentage points. Figure 4.2 presents a histogram reporting the frequency distribution and range of first order differences. Again, it is easy to see how small annual changes in total social expenditure ‘in normal times’ are. The frequency distribution is highly skewed as in the vast majority of observations, expenditure changes fall within a range of /0.4 per cent of GDP. In real monetary terms, this is of course considerable. However, even if a very small amount of measurement imprecision is assumed, due to the relatively small first order differences minor problems can cause significant problems for any inferences drawn from annual changes in total social spending. If we assume a measurement error of just 1 per cent at an average level of total social spending of 22.5 per cent of GDP (for 22 OECD countries in the year 2001), the estimated error built into spending figures is equal to 0.22 per cent of GDP. Based on well known problems of cross-national time series on social expenditure this estimate of less than a quarter of 1 per cent of

66


GDP appears to be rather conservative and the average measurement error is likely to be significantly larger. The problem for short-term analyses of annual spending changes measured in total welfare efforts is that the estimated measurement error built into total social expenditure figures exceeds the average annual social expenditure change for the majority of observations for the period 1980–2001. Regardless of the fact that small annual spending changes do not cause a technical problem for time series regression analysis, a substantial problem remains: advanced statistical techniques like pooled analysis rely on very optimistic assumptions about the quality of spending data. As De Deken’s and Kittel’s chapter in this volume reveals there are reasons to assume that the difference between measured social expenditure and (the theoretically expected) real levels is not randomly distributed across time and countries. Comparative researchers face time and country specific measurement deviations in total social expenditure data which violate the assumption that the dependent (left hand side) variable is not suffering from systematic measurement error for (time or country specific) subgroups of our sample. If one accepts that the average size of the measurement error is probably larger than the average annual spending change in most years and countries, and if it is acknowledged that the measurement error is not randomly distributed over time and across cases, one might seriously question what kind of descriptive and analytical inferences based on annual total social expenditure changes should be drawn. An alternative option to a short-term perspective focusing on annual spending changes is to extend the time periods for which changes are analysed. In many cases this can reduce the problem of small differences in the dependent variable. But it does not solve the problem of non-random country and time specific measurement bias. It also comes with a price to be paid: the further time intervals are stretched, the higher the risk that important variation of the dependent variable within the time period is lost. Again, an example may help to illustrate this. Comparing total social expenditure for Sweden in 1980 (28.8 per cent of GDP) and 2001 (29.9 per cent of GDP) one might be tempted to conclude that not much has happened to the Swedish folkhemmet in the post-1980s. However, ignoring spending fluctuation between these two points means to ‘aggregate out’ one of the most fascinating periods in the post-World War II history of the Swedish welfare state, both in terms of far reaching legislative changes and in terms of dramatic spending increase (from 30.8 per cent of GDP in 1990 to 36.8 per cent in 1993) and cost containment policies. Similarly, for the same 21 year period, one would miss important developments in German social expenditure trends. (West) Germany was the first country that managed to reduce welfare efforts significantly during the 1980s. However,


67

German unification put an end to this consolidation process. The fast track unification and equalization policy of the centre-right government under Chancellor Kohl triggered off a sharp increase in welfare efforts. A macroscopic perspective which would merely compare the levels of social expenditure at the beginning of the 1980s with those at the beginning of the new millennium would miss one of the most fascinating and dramatic turns in social expenditure trends.

CONCLUSIONS This chapter has summarized some of the major advantages and pitfalls of exploring welfare state change by using historical social expenditure accounts. The discussion was embedded in the wider context of a general debate about the ‘dependent variable problem’ in comparative welfare state research. This is not the place to repeat the major points which could highlight the specific strengths and limits of expenditure based approaches for investigating welfare state change. Instead, two specific comments on expenditure analysis will be made before I move on and re-address some more general issues in comparative macroquantitative welfare state research. First, it is worth pointing out that there is more than one approach to expenditure analysis – just like there is more than one approach to a social rights based comparative welfare state analysis. This chapter has focused on issues related to ‘welfare efforts’, i.e. (total) social expenditure as a percentage of GDP. However, there are alternative expenditure measures which can offer insights in a variety of research contexts (see Schmidt, 1998; Castles, 2004). Decisions about which spending indicators, which levels of analysis, and which techniques of data analysis are to be used for cross-national studies of social expenditure need to be made carefully and within the context of specific research questions and designs. In any case, these choices should reflect the end-result of a theory guided process of decision making. Second, this chapter has argued that in many research contexts expenditure sensitive approaches can indeed offer a valuable perspective for exploring welfare state dynamics. However, this ‘pro-expenditure’ pledge comes with a major caveat. A ‘spending only’ based perspective of analysis can hardly capture a full picture of welfare state change. As all (post)modern welfare states represent complex configurations, it is not reasonable to draw too many generalizations from a rather restricted set of one-dimensional measures of welfare state size. This is a particularly important point since advanced techniques of data analysis might be built on strong assumptions about the underlying structure, consistency and quality of the data to be

68


analysed. As a consequence, causal inferences drawn from the results of highly advanced statistical procedures such as pooled time series analysis should be interpreted with due caution. Within expenditure analysis they have tended to produce unrobust findings as several replication studies have shown (see again De Deken’s and Kittel’s contribution in Chapter 5 of this volume: Kittel and Winner, 2005; Plümper et al., 2005). This lack of robustness reflects a methodological interaction problem between the specific techniques of pooled regression analysis and its use in the context of crossnational expenditure data. So what are the major lessons to be drawn for researchers who wish to explore welfare state change by using advanced techniques of quantitative data analysis? The answer can be illustrated by re-addressing the debate about the ‘new’ or ‘old’ politics of the welfare state. This debate may be misleading insofar as these two theoretical paradigms may not be mutually exclusive and irreconcilable. Rather, they may be understood as, at least partly, overlapping analytical approaches which help to understand specific dynamics of welfare state change. However, what should be avoided is to let the results of macroquantitative correlates decide either in favour or against the one or the other theorem. Reducing the risk of producing methodological artefacts, triangulation may provide a more reasonable and balanced strategy. Rather than ‘topping up’ existing tools for data analysis by ever more advanced analytical techniques, there are good reasons to conclude that comparative welfare state researchers would be well advised to invest more resources in the basic infrastructure of comparative social inquiry and to develop a broader set of measures for analysing welfare state change, in particular diversified sets of indicators in the context of mixed methods approaches. The major argument for such a dependent variable focused triangulation is derived from the assumption that current welfare states represent complex configurations. The most appropriate way to study shifts in these configurations may be found in a multidimensional approach. This disqualifies both too ‘narrow’ research designs and the desperate search for the most parsimonious models maximizing the explained variation in the dependent variable. If one assumes that no single method or measures can offer a full grasp of the dynamics of welfare change, a strong claim for opting out of one-dimensional approaches seems the most plausible lesson to be drawn. More generally, the major conclusion drawn from focusing on key issues related to macroquantitative analyses based on expenditure data is that basic issues in and dilemmas of measuring, describing and analysing welfare state change deserve more attention than at present. Thanks to a rapidly growing body of literature, current comparative welfare state research involves more elaborate theoretical reflections on change and


69

welfare reform than even a decade ago. The lack of an explicit theory or analytical framework for the study of contemporary welfare state development was filled courageously by Pierson (1994, 1996) with his inspiring, politics centred analytical approach to understanding the logic of and limits to retrenchment processes. However, during the last ten years numerous authors have started to develop theories on welfare reform from distinct analytical perspectives. This is, for example, reflected in the intense debate about the explanatory strengths and weaknesses of theories of path dependency in the process of welfare state adjustment (see Jochem, Chapter 12 in this volume). The rise of the ‘new risk’ paradigm that seems to define a major fresh playing field for sociologically inspired welfare state analysis is another example (see Bonoli, Chapter 3, this volume). In addition, new developments in macroquantitative methodology have provided comparative welfare state research with more sophisticated tools for analysing the dynamics of change. Hence one might conclude that innovations in theory and methodology suggest a bright future for comparative welfare state research. However, there are several good and one fundamental reason to be more sceptical. First, the impressive progress in new techniques of data analysis such as pooled time series cannot disguise the problem of basic issues in macrocomparative research deserving more attention than they have attracted so far. Both the development of ambitious theoretical generalizations and the impressively rapid progress in advanced research techniques are vulnerable as long as they are not built on more solid fundamentals for descriptive and analytical inferences. The production of robust, valid and reliable measures of welfare state change represents a key challenge, as the quasi ‘cement’ for any stable foundation on which empirically rooted theories of welfare reform should be built. Both robust descriptive and analytical inferences depend on adequate measures of the dependent variables that are to be investigated. The current debate on welfare state change suffers from a lack of a critical mass of alternative indicators to available measures of social rights and expenditure (see Scruggs, Chapter 7, this volume, for a major innovation in this field). A major investment in the development and collection of a broader range of cross-nationally standardized indicators may not appear to be the intellectually most fascinating undertaking (academic) researchers may like to spend their time (and resources) on. However, as several contributions to this book show, investing in such a foundation of comparative welfare state research, i.e. carefully derived conceptualization and measurements of welfare state change, seems a necessary prerequisite for a promising future for comparative welfare state research. While there are numerous elaborated theoretical conceptualizations of welfare state change, as well as an abundance of

70


sophisticated analytical techniques, it is the missing link between theory and advanced techniques of data analysis which causes concern. If comparative welfare state research fails to bridge this gap, artless sophistication may become a chronic feature. The worst case would be a situation where ever more elaborated theoretical concepts are developed which are by and large resistant to a meaningful translation into measurable empirical concepts and where ever more sophisticated techniques of data analysis sacrifice the real-world complexity on the altar of simplistic model assumptions. An alternative route for comparative welfare state research would be to recouple theory and method. This path would certainly have to move well beyond a purely expenditure based approach to analyse welfare state change.

NOTES 1. 2. 3.

4. 5. 6.

7.

8.

9.

The author is indebted to Jochen Clasen, Olli Kangas and Bernhard Kittel for helpful comments on earlier versions of this chapter. The two main variables determining increases in social expenditure as a percentage of GDP for the 1990–93 period were changes in real GDP and the change in unemployment rates, both, of course, highly correlated. The situation is different for welfare state research that is primarily utilizing micro data, e.g. household panel data. Compared to comparative macroscopic welfare state research, microanalyses often put more emphasis on measurement and conceptualization, one plausible reason being that data analysers are working more closely together with data producers in this sort of social policy analysis. The author pleads guilty to having arrived at this rather cautious and risk averse conclusion (Siegel, 2002: Chapter 12). I owe the phrase ‘assuming away’ to Peter Hall’s critical assessment of the relationship between ontology and methodology in contemporary comparative politics (Hall, 2003: 386). One of the most impressive examples is Castles (2004) as it shows that the author is not exclusively interested in arriving at some parsimonious statistical models but actually provides an in-depth exploration of available expenditure data to investigate a broad range of salient issues related to both descriptive and analytical questions of welfare state change. If it were plausible to assume that time lags between political decisions and spending outcomes were constant across countries and different welfare state programmes, constant time lag terms on the right hand side of regression equations could at least mitigate this problem. However, it is plausible to assume that lag structures do significantly vary in different contexts. See also Plümper et al. (2005: 343–5, 349). Again, time lag effects can be specified in pooled models, but the problem is that large lags between political decisions and expenditure outcomes cannot be dealt with in a reasonable way. For example, the full fiscal long-term effect of major pension reforms introduced in the 1950s started to unfold only 30–40 years later and in interaction with demographic ageing. Hence, the long-term effects of the pension reforms of the 1980s and 1990s can only be estimated and will manifest themselves with a time lag that is difficult to specify and varies across different old-age income systems (Hinrichs, 2001). Significant level effects may indicate that countries with lower spending levels tend to witness higher growth dynamics (usually referred to as catch-up), or that in nations with


10.

71

already above average spending levels social expenditure tends to grow comparatively weaker or decreases. Both indicate what is often labelled -convergence. Beta convergence can reflect ‘catch-up’ (like in Greece, Japan, Portugal, and particularly remarkably, in Switzerland after 1980); but it can also reflect downward adjustments (catch-down) in countries which cluster into the category of high-spending nations: the Netherlands and Sweden for example (and, until German unification, the Federal Republic between 1982 and 1989). The convergent trend in social expenditure is also reflected in a remarkable drop of the coefficient of variation (indicating -convergence) from 0.31 to 0.21 between 1980 and 2001, i.e. a reduction by almost exactly one third. This idea is based on interviews with Paul Conway and Willem Adema, OECD, March 2005.

5. Social expenditure under scrutiny: the problems of using aggregate spending data for assessing welfare state dynamics Johan De Deken and Bernhard Kittel INTRODUCTION Since the late 1980s the governments of most OECD countries have been concerned with reorganizing their welfare states in general, and their retirement systems in particular. Through such reorganization they seek to prepare for – or hope to ward off – the financial crisis of welfare arrangements that has been predicted for the coming decennia. Scholarly contributions have focused on the directions reforms take and the conditions under which they take place (Bonoli et al., 2000; Pierson, 2001). More specifically, cross-national research has put much effort into exploring the extent to which the ideological position of governments influences the size and direction of welfare state reform (e.g. Pampel and Williamson, 1985; Huber and Stephens, 2001; Kittel and Obinger, 2003; Castles, 2004; Galasso and Profeta, 2004). In order to measure reform and change, these scholars have largely relied on the Social Expenditure Database (SOCX) of the OECD, which is considered to be the most reliable source for comparable data on social expenditure. While it is granted that, as has been repeatedly warned, expenditure data contain little information about the substantive content of welfare efforts, they are generally regarded as a valid indicator of overall welfare effort. The results of these research efforts into the political determinants of social expenditure cannot have been more ambiguous. Kittel and Obinger have summarized a variety of studies revealing contradictory coefficient estimates, depending on the period and countries included, the variables included, and the model specification (Kittel and Obinger, 2003: 25–8). In their own analysis of the conditional impact of parties and political institutions, they find that the largest impact results from socioeconomic factors (economic growth, share of elderly people, unemployment rates), and that 72

The problems of using aggregate spending data

73

the contribution of political-institutional factors is marginal in substantive terms, although the coefficients largely exhibit the expected direction and a plausible size. By excavating the detailed series of the OECD Social Expenditure Database (an effort which is unusual for cross-national comparative work), Castles has found that the results also critically depend on the exact composition of the social expenditure indicators used (Castles, 2004). In this chapter, we would like to add one more note of caution. Although the OECD has put an impressive amount of effort into collecting comparable data, the statistical department itself admits that some aspects are simply too complex to be represented in a consistent accounting edifice. It may well be true that the Social Expenditure Database is the best we can hope to obtain. But are the best data available good enough for drawing conclusions about the direction and size of welfare state change, or – more modestly – about the direction and size of social expenditure in general and old-age pensions in particular? After scrutinizing the Social Expenditure Database and comparing it to the European System of Integrated Social Protection Statistics (ESSPROS), a database published by the Statistical Office of the European Union (Eurostat), we tend toward the conclusion that this question has to be answered to the negative. In the following, we first explore some dimensions of the measurement problem. Subsequently we compare models of social expenditure in which we vary the definition of the dependent variable while keeping constant the substantive model specification, i.e. using more or less the same set of independent variables. In the first part, we will focus on the problem of measuring welfare effort consistently. In order to exemplify pitfalls, we focus on a comparison between the Netherlands and Germany, where necessary referring also to some other countries for clarifying our argument and to show that the Dutch–German comparison does not represent a unique but rather a common problem of social expenditure definitions. Nevertheless, the comparison is particularly revealing, because the Netherlands, according to SOCX data, is the country with the most pronounced cutbacks in public social expenditure during the 1990s, while the OECD figures suggest roughly stable levels for Germany. We compare two databases (OECD and EU) and discuss the differences between public, mandatory private, and voluntary private expenditure. In a subsequent step our chapter will focus on problems of measuring pension expenditure. In the second part, we will use the various definitions of social expenditure in general, and pension expenditure in particular, as dependent variables in a regression model in order to explore the impact of conceptual differences on the estimated parameters. The final part concludes.

74


THE DEPENDENT VARIABLE PROBLEM: HOW TO MEASURE WELFARE EFFORT The OECD’s Social Expenditure Database has been heralded as a most welcome tool for comparative welfare state research because it allows disaggregating total social security expenditure and thus offers the possibility of taking account of the structure of social provisions (Castles, 2004). In this section, we explore a variety of measurement concepts and actual scores for social expenditure in more detail for a few countries in order to assess the extent to which different sources of data converge on a particular statement. SOCX versus ESSPROS At the time of writing, the most recent SOCX database includes a historical series of expenditure data on 13 social policy areas for the 1980–2001 period.1 Social expenditure is defined by the OECD as ‘the provision by public (and private) institutions of benefits to households and individuals in order to provide support during circumstances which adversely affect their welfare’. It includes cash transfers as well as direct provision of goods and services ‘provided that the provision of benefits constitutes neither a direct payment for a particular good or service nor an individual contract or transfer’ (OECD, 2001: 9). A distinction is made between three broad categories of social spending: (1) public, (2) mandated private and (3) voluntary private. Public expenditure covers all levels of government (central, regional, local). However, because of the variable quality of the data, the reported government spending may be underestimated for countries where the quality of data at the lower tier of government is inadequate.2 In its definition of the public nature of a programme, the OECD has taken a different approach than Eurostat in its ESSPROS database. ESSPROS defines schemes as public if the decision making power lies with the government in its governing capacity (and not just in its capacity as employer of state employees). The OECD, on the other hand, considers social expenditure as public only if it is made by agents of the government sector. Thus, for the OECD, expenditure is only considered public if it occurs in the context of a statutory scheme. Eurostat on the other hand also considers mandated schemes to be part of public expenditure. The two methods of calculating seem to result in quite different assessments of cross-national social expenditure efforts. The cross-national differences reported by ESPROSS are far more limited than the ones to be found in SOCX. Thus, for example, for the year 2001, Eurostat estimates


75

total social expenditure for the Netherlands at 27.5 per cent of GDP, for Germany at 29.8 per cent and for Denmark at 29.5 per cent (all relatively close to the EU15 average of 27.6). The OECD on the other hand reports only 21.8 per cent for the Netherlands, 27.4 per cent for Germany and 29.2 per cent for Denmark (far more spread out around the EU-15 mean of 24.0 per cent of GDP). What accounts for these differences between the two databases? Public versus Mandated Private and Voluntary Private Social Expenditure One of the reasons for these differences seems to lie in the fact that, in contrast to SOCX’s ‘public social expenditure’, Eurostat’s ‘total social protection’ figures include the costs of administrating benefit systems,3 and a number of expenditure costs that the OECD labels as ‘mandatory private’ and ‘voluntary private’. The ESSPROS category ‘total social benefits’ excludes administrative costs, but still seems to include most of what the OECD categorizes as ‘private social benefits’. These benefits form a grey area between individual non-social private benefits and public social benefits. In an attempt to reduce the direct responsibility of government for social expenditure, during the past decades this grey area has experienced a substantial growth at the expense of programmes directly controlled by the state. In general, private social benefits refer to social expenditures that are not directly controlled by the central state, local governments or publicly regulated social security funds, but that are also not ‘established at market prices given the individual’s risk profile’ (Adema, 2001: 10). Typical examples would include occupational pension plans based on a defined benefit basis, or health care plans financed by employers. What is Private: a Dutch Social Expenditure Miracle? The problem though is that certain characteristics of a programme may be private, whereas other characteristics could point to their classification as public. In this context the statisticians of the OECD give the example of Dutch pension funds. Initially the stipulations of Dutch occupational pensions were voluntary collective agreements. But those arrangements are very often enforced within an entire industry because of the widespread practice of administrative extension.4 For the OECD this implies that those arrangements could be classified as mandatory, i.e. that participation in what formally is a private arrangement is required by (quasi) law. But a closer scrutiny of the SOCX database reveals that in fact the OECD classifies most of those schemes, in particular the Dutch occupational pensions, as ‘voluntary private social expenditure’. In our opinion this is a misnomer: the

76


Dutch government’s policy of administrative extension of collective agreements has led to an exceptionally high coverage of the working population in terms of occupational pensions.5 More than 90 per cent of the employed population are participating in occupational pension plans.6 The following arguments have been made by the OECD statisticians for justifying the labelling of the Dutch pension plans as voluntary private arrangements: (1) Dutch authorities do not have a formal influence on the terms agreed in the initial collective agreements; (2) the government can only use the tool of administrative extensions on request of the parties concerned (i.e. the trade unions and the employers’ associations); (3) most of the companies and parties were (eventually via their association) party to the voluntary initial agreement (Adema and Einerhand, 1998). Although this is a perfectly defensible position, we think it is a very formalistic and legalistic argument: by the same token one could label any neo-corporatist form of social insurance a private voluntary arrangement. One can thus only wonder why the OECD treats these benefits differently from, for example, mandatory sickness payments in Germany. In the Federal Republic employers are legally required to continue to pay the wages of their employees during the first six weeks of work absence due to sickness (Lohnfortzahlung im Krankheitsfall). The OECD correctly classifies this type of benefit as mandatory private expenditure. It is this obligation that forms the prime explanation for the quite high scores of this country in terms of mandatory private social expenditure. Hence it is surprising that a similar scheme, initiated in the Netherlands in 1996 when the government abolished the sickness benefit insurance, is not included in the category of private mandatory in the Dutch case. Instead, it seems to have been classified as voluntary private social expenditure. The 1996 Dutch reform caused public and mandated private social expenditure, as measured by SOCX, to fall without the actual costs or generosity of sickness support in the Netherlands having changed much. Another puzzle in the OECD statistics in this regard is the size of this expenditure item. Despite the fact that the Dutch scheme mandates employers to continue to pay wages, albeit only at 70 per cent,7 for a much longer period (52 weeks instead of six weeks), the Dutch programme appears to be much more modest in expenditure terms. With an average sickness absence rate three times as high as in Germany,8 mandated expenses relative to GDP, even four years after the new system had been in place, remained well below the figures reported for Germany. Again, as becomes evident from Table 5.1, this expenditure item appears to have been classified as a voluntary arrangement in the Dutch case. But even here sickness benefits appear surprisingly low, considering that the Netherlands has about three times as many beneficiaries of sickness

77


Table 5.1 Expenditure on mandatory private and voluntary private sickness benefits in Germany and the Netherlands as a percentage of GDP Country

Mandatory Total

P*

Mandatory Voluntary

Voluntary S

Total

P

S

Total

P

S

1994 1.40 1996 1.38 1998 1.11

0 0 0

1.40 0.64 1.38 0.66 1.11 0.67

0.64 0.66 0.67

0 0 0

2.04 2.06 2.07

0.64 0.66 0.67

1.40 1.38 1.11

1994 0.72 Netherlands 1996 0 1998 0

0 0 0

0.72 2.98 0 4.34 0 4.48

2.75 3.35 3.35

0.23 3.70 0.99 4.34 1.13 4.48

2.75 3.35 3.35

0.95 0.99 1.43

Germany

Note: * P are results based on pensions and S are results based on sickness. The category ‘pensions’ includes both old-age and survivors’ benefits.

Table 5.2 Expenditure on sickness benefits as a percentage of GDP according to SOCX and ESSPROS Country

SOCX Social expenditure on sickness benefits P*

PMP

PMPVP

ESSPROS Paid sick leave

Germany

1994 1996 1998

0.46 0.46 0.32

1.86 1.84 1.43

1.86 1.84 1.43

1.88 1.81 1.46

Netherlands

1994 1996 1998

1.29 1.05 1.02

2.01 1.05 1.02

2.24 2.04 2.15

2.01 1.86 1.89

Note: * P Public; MP mandatory private; VP voluntary private.

benefits as Germany. Even if one adds up voluntary private, mandatory private and public social expenditures on sickness benefits, the Netherlands only appears to spend 50 per cent more than Germany, with about three times as many beneficiaries (see Table 5.2). ESSPROS, on the other hand, seems to label both mandatory private and voluntary private programmes as ‘contractual private schemes’. It defines them as ‘providing social protection decided via bargaining between the social partners (employers and employees)’ and, what is important for the present argument, does include them in its aggregate social expenditure measure (Eurostat, 1996: 22). For Eurostat, the social nature of all these

78


programmes lies in the fact that they are all established by collective agreements and cannot be changed unilaterally by employers. Thus, in the ESSPROS database the social protection interventions of the following institutions are all included in its aggregate social expenditure measures: the central state, local governments, social security funds, autonomous selfadministered pension funds, insurance companies, mutual benefit societies and even direct obligations by employers (to the extent that they are embedded in a mandate by the government). One would thus expect sickness benefit to correspond to the last column of Table 5.1, but unfortunately this is not the case. As Table 5.2 demonstrates, Germany spends about the same amount according to Eurostat data as according to OECD data, while the Netherlands spends less if private expenditure is excluded, or more if it is included. If for Germany the ESSPROS data seem to correspond more or less with SOCX public mandatory private voluntary private, the figures for the Netherlands find no clear match in any of the SOCX series. For pensions, this ambivalent classification leads to a drastic underestimation of the costs of old-age provisions in the Netherlands, as becomes evident in Table 5.3. If one considers only public and mandated expenditure according to the OECD definition, one would get the impression that the Netherlands is not affected by the problem of ageing: pension expenditure even declined during the period 1994–98. It is these kinds of inconsistencies, we will demonstrate, that make pooled time series analysis on available social expenditure data rather futile. Moreover, most of the aggregated data that the OECD publishes in an easy accessible way are only based on public social expenditure. As a consequence, Table 5.3 Old-age and survivors’ pension benefits as a percentage of GDP according to SOCX and ESSPROS Country

Germany

Netherlands

SOCX Social expenditure on pensions

ESSPROS Pensions

P*

PMP

PMPVP

1994 1996 1998

10.26 10.78 10.95

10.26 10.78 10.95

10.90 11.44 11.62

11.5 12.0 12.0

1994 1996 1998

7.77 7.43 7.04

7.77 7.43 7.04

10.52 10.78 10.39

11.0 10.9 11.3

Notes: * P Public; MP mandatory private; VP voluntary private.


79

the majority of comparative welfare state researchers seem to prefer to base their analysis on these very limited data (see e.g. Clayton and Pontusson, 1998; Huber and Stephens, 2001; Kittel and Obinger, 2003; Castles, 2004). At best they also include what the OECD calls mandatory private programmes. We would like to argue that in order to paint a more comparable picture of different countries’ welfare efforts, it is necessary to include mandated private social expenditure too. The OECD classification in this respect is deceptive as, in the case of the Netherlands for example, the OECD classifies mandated occupational pensions and mandated sickness benefits as ‘voluntary arrangements’ (and thus leaves out all the aggregate measures published by the organization). It takes a persistent researcher to recompile aggregates from the original data sources in order to obtain an indicator that takes into account all the necessary components of social expenditure.9 The distinction between public, mandated private and voluntary private social expenditures seems of secondary importance to most comparative welfare state research. If one adopts a so-called ‘functional definition’ of the public sector (SCP, 2004), it does not matter whether the government is directly or only indirectly responsible for expenditure: all social expenditure, irrespective of whether it is ‘public’, ‘mandated private’ or ‘voluntary private’ (in the OECD terminology) has an effect on the income distribution, as well as on such issues as non-wage labour costs. Provided that it is adequately operationalized and measured, the distinction between ‘public’ and ‘mandated private’ on the one hand, and ‘voluntary private’ on the other is, of course, a potentially quite important one. The distinction can have substantial repercussions for theoretically important dimensions or concepts for characterizing welfare states, such as universality, for example. By definition, voluntary private programmes are not universal, and even if embedded in a mandate by the government, private programmes remain less universal than public programmes, as the practice of occupational pensions in countries such as the Netherlands and Denmark shows. In those countries risk redistribution within the earnings related component of the pension system is limited to specific sectors or specific occupational groups. Moreover, in order to make any sensible analysis based on these kinds of nuances in expenditure accounts, spending data need to be reliable over time and classified in a consistent way across countries. The examples of social spending statistics for the Netherlands and Germany have already pointed to severe problems for two countries in this respect. Voluntary Social Expenditure and the Impact of Mandating Changes in public social security programmes cannot be fully understood without taking into account mandatory and voluntary private provisions.

80


This is particularly relevant if public programmes are replaced by private programmes that are mandatory. Shifts from public to voluntary programmes are less likely to reduce the total level of expenditure if such a shift occurs in the context of a coordinated system of industrial relations, as this may explicitly or de facto involve practices of mandating. Hence more centralized systems of industrial relations are said to lead to higher social expenditure than more decentralized ones (Brandl and Traxler, 2005). Again, the Netherlands is a revealing case, as during the 1990s a principle of communicating vessels seems to have been at work in this country: cuts in public programmes were quasi automatically compensated for by private arrangements, which, in the context of the Dutch system of neo-coporatist governance, were not that private after all. This resulted in only marginal changes in both social expenditure and distributional consequences. For example, cuts in the public pension scheme that amounted to a decline of about 25 per cent of the value of the benefits of the basic pension scheme (Haverland, 2001) were compensated (for 90 per cent of the beneficiaries) by the actual and future benefits guaranteed by occupational pension funds. Thus the erstwhile replacement rate of 70 per cent was maintained, regardless of the declining value of the basic pension.10 Similarly, the socalled privatization of the sickness benefit insurance system in 2006 simply created a mandated scheme similar to the one which exists in Germany for the first weeks of sickness absence. Both these policy measures caused public social expenditure, as measured by SOCX, to drop significantly during the second half of the 1990s, while increases in mandatory and voluntary private expenditure during the same period went up by about the same amount. Thus, between 1994 and 1998 public pension expenditure (as a percentage of GDP) increased by 0.7 percentage points in Germany while it decreased by 0.7 percentage points in the Netherlands. However, the sum of mandatory and voluntary private expenditure in the Netherlands increased by 0.6 percentage points, while it remained unchanged in Germany. This means that the social effort for oldage provision developed in a much more similar way than public expenditure figures for the two countries suggest, both in terms of levels and in terms of changes (see Table 5.3). A similar image has been created as a result of the much heralded reform of the Dutch sickness benefit system. While total expenditure for this function in Germany declined from 1.86 per cent of GDP to 1.42 between 1994 and 1998, in the Netherlands this decline was actually weaker (2.24 per cent of GDP to 2.15). If one focused on public and the OECD’s ‘mandatory’ private expenditure, one would arrive at the opposite conclusion: only a moderate decline in Germany (from 1.86 per cent of GDP to 1.43), and a steep decline in the Netherlands (from 2.01 to 1.02).


81

A Brief Digression on Gross versus Net Social Expenditures Both SOCX and ESSPROS are confined to gross expenditure, i.e. the collected data refer to expenditure before taxes have been levied on benefits. This, of course, has major implications for comparability, since a given amount of a benefit not only has a very different distributional impact depending on whether it is taxable or not, but taking account of tax levies can also dramatically reduce the volume of net expenditure. There are basically three ways in which the tax system can affect social protection (Adema, 2001; Adema and Ladaique, 2005): 1.

2.

3.

direct taxes and/or social security contributions on social protection transfers (in countries like Germany, France and Belgium do not exceed 2 per cent of GDP, whereas in Denmark and the Netherlands such levies exceed 5 per cent of GDP); indirect taxes on social protection transfers (in this respect the United States differs sharply from most European countries, and Scandinavia in particular, where substantial proportions of social transfers are clawed back via indirect taxes); and tax breaks for social purposes (in Belgium and Germany, such tax breaks are quite prominent for families, in the United States this is the case for medical care and for employer contributions for private health plans). Tax breaks for pensions are important in countries with funded pension plans such as the Netherlands and the United Kingdom. These include tax exemptions for contributions to private pensions as well as tax relief for investment income on capitalized pension funds. If the first type of tax effect is still relatively easy to assess, the second and the third are far more difficult to calculate. The main problem with measures of ‘fiscal welfare’ (Titmuss, 1963) in the area of old-age pensions is the fact that tax relief can be granted at various stages (contributions, investment and capital gains and/or benefits).11 Because of this difficulty the OECD does not consider these measures in its assessment of ‘net social expenditure’, but merely list them as a ‘memorandum item’.

Often the main problem with all three types of tax effects is that the relevant information is unobtainable. Attempts by the OECD in this respect resulted in tables with many empty cells and footnotes like ‘relevant estimates cannot be obtained: either because information on gross spending for the relevant item is not available or because the information available on taxes and social security contributions is not detailed enough to present relevant estimates’ (Adema, 2001: 18).

82


Towards a Comparable Data Set on Gross Social Expenditure Using again the example of Germany and the Netherlands we will illustrate how we compiled the aggregate and pension expenditure measures we will be using in the subsequent analysis. We will focus on pensions for the following reasons: (1) in terms of expenditure they represent by far the most important welfare state programme; (2) it is an area that is fraught with the type of measurement problems we have been discussing up to now; and (3) it poses a number of additional problems that are related to the intertemporal nature of these programmes.12 Problems inherent in measuring pension expenditure Let us start by elaborating the third point made above. A number of problems have rarely been addressed in comparative studies of pension expenditure. One difficulty concerns the implications of different designs of pension systems, in particular the extent to which national pension provision relies on funding or Pay-As-You-Go (PAYG). In contrast to popular belief, the reliance on funding does not make much difference for the longterm financial equilibrium of a pension system or the financial burden of the pension system for a nation’s economy (Mackenroth, 1957; Orszag and Stiglitz, 1999; Barr, 2001). Nevertheless, the use of the funded approach can, especially in the short term, severely affect the comparability of the costs of national pension systems. During periods in which the returns on investment are high because of booming financial markets, funded schemes are able to spend considerably more than they need in terms of revenue and than unfunded PAYG systems spend. On the other hand, if funded schemes are designed as defined benefit programmes with a back-service obligation,13 revenue will have to be higher than benefit expenditure after a crisis in the financial markets. To illustrate this point, Table 5.4 contains information on total receipts and total expenditure of Dutch occupational pensions. If this second pillar of the Dutch pensions system had been organized on a pure PAYG basis, revenue and expenditure volumes would be roughly equal for every year. But, as is evident, this is not the case. During the final years of the speculative bubble on the stock market, the funds’ expenditure on benefits was much higher than were receipts. In contrast, a few years after the stock market crash at the beginning of this millennium the situation was reversed. In order to compensate for their massive investment losses and to comply with the regulators’ requirements of a minimum funding position, the funds had to collect much more in contribution revenue than they were spending on pension benefits. Thus, if during the late 1990s expenditure data overestimated the cost of financing pension benefits in the Netherlands, by 2003 the situation had

83


Table 5.4 Receipts and expenditure on benefits by Dutch pension funds in million € (and as a percentage of GDP) Category Gross receipts from contributions Gross expenditure on benefits Ratio expenditure/ receipts

1999

2000

2001

2002

2003

10 146 (2.7%) 11 294 (3.0%)

10 928 (2.7%) 13 203 (3.3%)

12 829 (3.0%) 13 347 (3.1%)

18 499 (4.3%) 14 572 (3.3%)

20 552 (4.5%) 15 611 (3.5%)

1.11

1.21

1.04

0.78

0.75

Source: Own calculations based on data from the Pensioenverzekeringskamer (the Dutch Pensions and Insurance Supervisory Authority).

been dramatically reversed and expenditure data no longer adequately reflected the costs and social effort.14 While up to 2001, the ratio of expenditure over receipts was above 1.0, ever since it has been sinking well below that threshold. A second problem related to the intertemporal nature of pension schemes is the way it complicates the assessment of their contribution to fiscal welfare. Tax breaks on occupational and individual pension plans are difficult to deal with because they are aimed at yielding benefits in the future. As a consequence, taxation occurs, and tax relief is given, at various stages in the savings process. Following Adema it is possible to distinguish three stages/areas where tax treatments need to be considered (Adema, 1997; 1999): (1) when contributions are paid either by employees or by employers on their behalf (this can be out of either taxed or untaxed income); (2) when investment returns are realized by the funds which can be taxed or left untaxed (even the capital itself can be liable to a wealth tax); (3) when benefits are being paid to pensioners either in the form of a lump sum or as an annuity, which can be taxed or left untaxed. For all three areas one can ask the question whether the cash flow consequences should be recorded, or alternatively if some sort of present value (accruals) estimate of the tax treatment should be used (Adema et al., 1996). The problem is that most countries have come to adopt the so-called EET model, in which both contributions and funds are exempt from taxation, whereas benefits are taxed.15 In this case it is impossible to assess the future taxation of benefits (which would correct the apparent generosity of the tax credit granted at two of the three stages mentioned above), since the future tax receipts will, amongst other factors, depend upon the ratio between pension income when retired and income from paid work when contributions are made. As we indicated above, these sorts of problems and the

84


resulting lack of comparable data have led the OECD to present tax relief on pensions merely as a memorandum item. But this cannot be considered to be a solution, since the exclusion of tax relief within an aggregate social expenditure measure leads to a significant underestimation of the cost of social security. A third problem concerns the demarcation of pension expenditure. This is, of course, less of a problem if pension expenditure is merely used as one element of total social expenditure. But it can pose serious problems if the study of welfare states is disaggregated to the level of individual programmes. Both SOCX and ESSPROS list old-age benefits and survivors’ benefits in separate spending categories. The problem here is the degree of variation in the structure of national pension schemes, which makes the kind of distinction the OECD has suggested misleading. In some countries, for example, widows will simply be catered for by a basic pension scheme, whereas in other countries they will obtain a derived pension from the deceased male breadwinner husband. Only the country specific programme differences can explain the fact that according to SOCX data a country such as Belgium spends more than 2.5 per cent of its GDP on survivors’ pensions, whereas in Denmark the corresponding figure is only 0.02 per cent (Table 5.5). In its manual, ESSPROS points to the problem that oldage and survivors’ functions are part of a coherent set of benefits ‘which is sometimes instituted as one system’ and concludes that ‘it is recommended to take into account that a strong interdependence exists between these . . . functions’ (Eurostat, 1996: 60).16 In addition, one might have to include part of the expenditure that social assistance schemes in countries like Germany pay out to pensioners who have no or insufficient pension claims on their own.17 The problem is that the SOCX data do not seem to adequately measure the latter category. Given the structure of the pension system in the four countries included in Table 5.5, one would expect the row ‘Low-income assistance’ to show the highest levels in Germany, where the social assistance scheme operates as a back-up to one of the most actuarially orthodox pension schemes in the world, and the lowest in countries with generous basic pensions (such as Denmark and the Netherlands). However, we find the opposite to be the case. Moreover, for some countries, such as Belgium, expenditure on social assistance is not even listed in the SOCX database. For other countries, the data reported by SOCX and ESSPROS again differ substantially. In particular expenditure on old-age pensions in Denmark differs by almost 30 per cent when the two databases are compared. The differences in the category measuring social assistance (which only partly goes to old-age pensioners or recipients of survivor benefits) might be explained by the fact that ESSPROS also includes benefits in kind such as

85

11.15

8.5 3.0 0.5

11.5

Old-age and survivors’ (12)

ESSPROS: 4 Function old age 5 Function survivors 6 Function social exclusion

Old-age and survivor’s (45)

11.5

8.7 2.8 0.4

11.54

8.74 1.37 2.80 0.32 na

1998

1994

8.21 1.11 2.94 0.35 na

Denmark

Belgium

SOCX: 1 Old-age cash benefits of which non-public 2 Survivors’ benefits of which non-public 3 Low-income assistance

Category

11.8

11.8 0.0 1.4

8.36

8.34 0.78 0.02 0.00 1.14

1994

Germany

11.5

11.5 0.0 1.1

7.61

7.59 0.77 0.02 0.00 0.70

1998

Netherlands

11.5

11.0 0.5 0.6

11.60

11.11 0.66 0.49 0.00 0.45

1994

12.0

11.5 0.5 0.6

10.92

10.41 0.65 0.51 0.00 0.43

1998

11.0

9.4 1.6 1.4

10.49

8.86 2.19 1.63 0.53 0.72

1994

10.9

9.5 1.4 1.5

10.39

8.94 2.73 1.45 0.62 0.52

1998

Table 5.5 Public, mandatory private and voluntary private social expenditure on old-age cash benefits and survivors’ benefits in four countries as a percentage of GDP

86


accommodation, basic services etc. In the analysis which follows we will only add the old-age and survivors’ benefits (which also seems justifiable as the social assistance measure does not seem to perform as we expected). It seems the case that ESSPROS already includes social assistance benefits for the elderly in its old-age function. That could in part explain why expenditure for this function is so low in Germany (at least in ESSPROS) and might in part account for the much higher old-age expenditure reported for Germany.18 So far we have illustrated the inadequacies of and differences between the databases of the OECD and the EU by investigating expenditure items for pensions and sickness benefits. In the subsequent statistical analysis we will first examine total social security expenditure before turning to pension expenditure. This is not only because total social spending is a central variable in most comparative welfare state research, but also as an attempt to bypass the problem of cross-national differences in terms of the functional delimitation of programmes. For example, in countries without a basic pension scheme, i.e. where the statutory scheme is more actuarially orthodox, many pensioners may see their pension benefits supplemented by social assistance. In those countries, an important part of the pension bill thus may end up being footed by the social assistance scheme. By only examining the pensions function one would thus underestimate social expenditure for pensioners in such countries if compared with countries where programmes that are classified under the pension function guarantee a basic income upon retirement. Subsequently, though, we will again focus on pension expenditure, as on the basis of our analysis so far, we expect the issue of public versus mandatory and voluntary private expenditure to play a more salient role. Data Sets for Total Social Expenditure From the SOCX database, we used three series of variables that measure total social expenditure: 1. 2. 3.

SOCX public total social expenditure as a percentage of GDP; SOCX mandatory private social expenditure as a percentage of GDP; and SOCX voluntary private social expenditure as a percentage of GDP.

Most comparative welfare state research is based on the first series, which we will refer to at as PTSESOCX. Some researchers also add to this private mandatory expenditure, which results in what we will label PMTSESOCX (Public and Private Mandatory Total Social Expenditure, i.e. the sum of

87


Table 5.6

Coverage of types of social expenditure by the SOCX database

Country Canada Greece Ireland Luxembourg New Zealand Spain Australia Japan Norway Switzerland United States Austria Belgium Denmark Finland France Iceland Italy Netherlands Portugal Sweden United Kingdom

Public

Mandatory

Voluntary

X X X X X X X X X X X X X X X X X X X X X X

– – – – – – X X X X X X X X X X X X X X X X

– – – – – X – – – – – X X X X X X X X X X X

Source: OECD (2001).

12). However, as we have demonstrated above, using the case of the Netherlands, the OECD’s category of ‘mandatory private social expenditure’ fails to incorporate important expenditure items which we consider to be part of mandatory programmes too, in particular occupational pensions which are part of collective agreements and subject to administrative extension by the government. This is the reason for introducing a third variable, PMVTSESOCX, in which we add all three categories of expenditure listed above. In doing so, we were faced with yet another problem: the series in the SOCX database turned out to be incomplete when it came to private social expenditure. For the 22 countries we include in our analysis, this (lack of) coverage is documented in Table 5.6. It is striking that it is primarily the so-called liberal welfare states which tend to be characterized by low public social spending and for which data on private voluntary spending are missing in the SOCX database. This is a

88


serious problem, as one would expect that in these countries, low public social spending is compensated for by voluntary private social spending. This type of social spending does not seem to be measured at all in the countries which are usually characterized as liberal welfare states. As a consequence, even our system of recoding will underestimate total social spending figures for some countries. We also use three dependent variables that express expenditure on pension benefits: 1. 2. 3.

SOXC public pension benefits expenditure as a percentage of GDP; SOXC mandatory private social expenditure as a percentage of GDP; and SOXC voluntary private social expenditure as a percentage of GDP.

Each of these three is actually the sum of old-age pensions and survivors’ benefits. We decided to add these two programmes because of the problems discussed above, i.e. the categorization of programmes in systems of social security that are organized in different ways.19 Similarly to our different sub-measures for total social spending, we developed three dependent variables: PPENSSOCX takes into account only public pension expenditure. PMPENSSOCX is based on the sum of public pension expenditure and mandatory private expenditure and PMVPENSSOCX is based on the sum of public, private mandatory and voluntary private expenditure. From ESSPROS we directly took two series on total expenditures: 1. 2.

ESSPROS total social protection expenditure as a percentage of GDP; and ESPROSS total social benefits expenditure as a percentage of GDP.

The main difference between these two is that the former (which in our analysis will be referred to as TSPESS) includes not only expenditure on various benefits (for which we will use the label TSBESS), but also the costs of administrating the benefit schemes and an item which Eurostat refers to as ‘other expenditure’. The share of the costs of administering the schemes is on average about 3.4 per cent of total expenditure. The category ‘other expenditure’ is on average less than 0.9 per cent20 and ‘usually refers to actual interest payable by the scheme to banks, and other creditors in respect of loans taken up’ (Eurostat, 1996: 35). In principle we expect that TSBESS should more or less correspond with the variable PMVTSESOCX that we derived from the SOCX database (as the OECD does not take into account administrative costs). The main difference between these two is that the OECD based variable, in contrast to the one based on Eurostat data,


Table 5.7

89

Overview of the various dependent variables

PTSESOCX

SOCX based total public social expenditure as % GDP

PMTSESOCX

SOCX based total public and mandatory private social expenditure as % GDP

PMVTSESOCX

SOCX based total public, mandatory private and voluntary private social expenditure as % GDP

PPENSSOCX

SOCX based public pension expenditure as % GDP

PMPENSSOCX

SOCX based public and mandatory private pension expenditure as % GDP

PMVPENSSOCX

SOCX based public, mandatory private and voluntary private pension expenditure as % GDP

TSPESS

ESSPROS based total social expenditure including expenditure on administration as % GDP

TSBESS

ESSPROS based total social expenditure excluding expenditure on administration as % GDP

TPENSESS

ESSPROS based pension expenditure as % GDP

also comprises expenditure on active labour market policies. On the other hand, if one looks at the health expenditure component, PMVTSESOCX only takes into account public expenditure, while TSBESS seems to be based on a broader definition that also takes into account what the OECD would call mandatory private and voluntary private expenditure. Finally we also developed out of the ESSPROS database a dependent variable measuring pension expenditure: 1.

ESSPROS pension expenditures as a percentage of GDP.

Similarly to our SOCX based variable, this variable, which we will refer to as TPENSESS, includes both old-age and survivors’ benefits. In principle it should more or less correspond to PMVPENSSOCX. Table 5.7 gives an overview of all the dependent variables we will use in our analysis. Table 5.8 presents simple correlations between the dependent variable in its various conceptualizations. The lower-left part is based on the variables defined in levels and reveals basically two things. First, the correlation coefficients within the sets of social expenditure and pension expenditure indicators are well above 0.9. Second, the correlations between the social and pension expenditure indicators range from about 0.5 to 0.7, as is to be

90

0.526 0.539 0.602 0.608

PPENSSOCX PMPENSSOCX PMVPENSSOCX TPENSESS

0.541 0.549 0.612 0.628

0.935 0.938 0.950

0.998

PMTSE

0.425 0.433 0.581 0.592

0.940 0.945

0.548 0.570

PMVTSE

0.531 0.530 0.645 0.695

0.992

0.940 0.941 0.603

TSPESS

0.508 0.505 0.616 0.675

0.934 0.936 0.607 0.993

TSBESS

0.991 0.944 0.914

0.898 0.896 0.587 0.839 0.838

PPEN

0.951 0.895

0.999

0.895 0.893 0.586 0.838 0.838

PMPEN

0.929

0.975 0.976

0.849 0.852 0.715 0.838 0.842

PMVPEN

0.936 0.936 0.959

0.838 0.846 0.727 0.837 0.843

TPEN

Note: Shown are Pearson’s correlation coefficients. Those in the lower left part refer to levels 1994, 1996, and 1998 (NT18354 observations) and those in the upper right part refer to first differences 1994–96 and 1996–98 (NT18236 observations).

0.993 0.928 0.936 0.946

PTSE

Correlations between various social expenditure measures

PTSESOCX PMTSESOCX PMVTSESOCX TSPESS TSBESS

Table 5.8


91

expected given the importance of pension expenditure for total social expenditure. The upper-right part presents the coefficients for the differenced variables, which refer to changes in levels of expenditure. The picture is very similar here, except for the finding that now the correlations between the social and pension expenditure indicators – ranging from 0.8 to 0.9 – are hardly lower than those within each set of indicators. There is, however, one exception, and that is the correlation between the change in total social expenditure, measured as the sum of public, mandatory private, and voluntary private expenditure and the change in other social expenditure indicators, which ranges from 0.55 to 0.72 and hence is considerably lower. Hence, we can draw two conclusions from the analysis of correlations between our different measures of social expenditure. Firstly, the stark conceptual differences between various definitions of both social and pension expenditures appear not to matter in statistical terms, as long as the definition remains within the realm of non-voluntary programmes. This finding may be partly due to the fact that mandatory and voluntary expenditure data are incomplete and hence the variables containing these elements are in practice inconsistent. However, dropping all cases for which private expenditure data are missing leads to a correlation coefficient between PTSESOCX and PMTSESOCX (NT 24) of 0.995 for levels and 0.998 for first differences, and the ones of PMVTSESOCX with PTSESOCX and PMTSESOCX become 0.465 and 0.492 respectively. Hence the extent of congruence is remarkable because it implies that the considerable conceptual differences in measurement which we have highlighted in the first part of this chapter appear to affect the relative positions of the cases on the scales only partly. Thus, variation oriented statistical methods are unlikely to capture the differences between public and mandatory private programmes. Secondly, the congruence of total social expenditure and pension expenditure in the results for differenced variables is indicative for the importance of pension expenditure for total expenditure trends. Hence, we can expect the results to be fairly robust across these variables as well.

MODELLING ACCOUNTS OF SOCIAL EXPENDITURE USING DIFFERENT CONCEPTUALIZATIONS OF THE DEPENDENT VARIABLE In this section, we explore the effect of varying the definition of the dependent variable in expenditure analysis on the substantive conclusions based on regression analysis. We expect that the considerable conceptual diversity

92


of the dependent variable should be reflected in substantive variation of coefficient estimates. Our starting hypothesis is therefore the following: given the same model specification, differences in the definition of the dependent variable should lead to substantively different conclusions. For example, if a country shifts a welfare programme from the public to the private sector, public social expenditure should decline, but the sum of public and private expenditure should remain constant. Such manifestations of policy differences should appear in the results as differences in coefficient estimates if social expenditure data are regressed on political indicators in a quantitative macrocomparative research setting. Unfortunately, there is no agreement at all in the literature with regard to the most appropriate model specification. In the traditional welfare modelling business, there is almost unanimous consensus that it is most appropriate to define the dependent variables in levels because political and institutional effects have long-term, cumulative effects on social expenditure (see e.g. Huber and Stephens, 2001). This emphasis does not lend itself as easily to the analysis of aggregate retrenchment effects however, since the concept of retrenchment refers to a change with regard to some previous value, not to differences between countries. Hence, in line with more recent contributions, we decided to analyse differences first (Kittel and Obinger, 2003; Castles, 2004). If one does not focus on long-term, cumulative effects on the size of welfare effort but on retrenchment dynamics, this is the most adequate specification because the dependent variable is then in fact conceptualized in terms of changes. Given the methodological focus of this chapter, we do not wish to add more complexities to the substantive discussion than necessary and hence merely draw on a long-standing debate in political economy. According to Pierson’s analysis of welfare retrenchment, the two crucial factors are the willingness of a government to embark on expenditure cutbacks and its possibilities to implement its preferences (Pierson, 1994; 1996). Hence we put partisan effects and their dependency on the institutional framework centre stage in our empirical analysis. This idea has been formulated repeatedly in the literature, but according to our knowledge the first attempt to explicitly model this conditionality is Kittel and Obinger (2003). Hence we mainly build on their approach. We use the traditional rough indicators for the ideological position of the government and institutional constraints and follow Kittel and Obinger (2003) in estimating a conditional effect between the two. We control for the usual socioeconomic factors, e.g. the share of the population over 65, unemployment rates, and GDP growth (for a more detailed discussion of these ‘usual suspects’ see Castles, 2004; Siegel Chapter 4 and Kangas and Palme, Chapter 5, this volume). Furthermore, we add the lagged level of the


93

dependent variable as an indicator of the reference point of retrenchment aims. We measure government ideology by the share of left parties in government. This is expressed as the ratio of cabinet members from socialdemocratic, green, and more-to-the left parties to the total size of the cabinet, using the definitions from Schmidt (1996). If there are few institutional constraints on policy implementation, we expect the coefficient to be positive, while it should be close to zero if there are many. Institutional constraints are measured by the indicator for institutional rigidity which Kittel and Obinger (2003) have used and which is a combination of Lijphart’s indicators of bicameralism and federalism (Lijphart, 1999). The share of the population over 65 and the unemployment rate should be positively related to social expenditure while GDP growth should impact negatively on total social expenditure. The lagged level of social expenditure should be negatively correlated to the subsequent change as higher levels indicate a higher pressure to consolidate welfare systems or, put otherwise, because countries may have reached their ‘saturation level’ of social insurance (Flora, 1986a; Alber, 1987). The period analysed and the countries included are mainly determined by data availability and cover 18 countries over varying periods between 1993 and 2001.21 Before discussing the results in more detail, a few notes on crucial but difficult specification decisions are required. The first and most ambiguous decision concerns the time horizon of welfare reform. As Pierson has impressively shown, welfare reforms are not usually effectuated within one year, as the usual approaches to pooled time series cross-section analysis assume by default (Pierson, 1996; 2003). Instead, Pierson argues that the effect tends to spread out over many years, and in pension policy up to about 70 years. While this is certainly true, it also overstates the issue because such a time horizon exceeds the reference frame of both political actors and citizens. In principle, any mid-term period should reveal some, at least initial, effects of welfare legislation if it is meant to be serious. However, variation in effect lags across countries, periods, and programmes and therefore makes any attempt to capture effects by one average coefficient impossible. Here we opt for a pragmatic intermediate solution and analyse two four-year intervals, 1993–97 and 1997–2001, which both cover periods of intense debate over welfare reform in OECD countries. This period is also bracketed by data availability for the dependent variables. Secondly, we have lagged the explanatory variables by one year in order to leave one additional year for effects to take off. Hence the models are estimated on a two-wave panel, defining the dependent variables in changes from 1993 to 1997 and from 1997 to 2001. The lagged levels are taken from 1993 and 1997. Changes in the share of the population over age 65 and

94


changes in unemployment rates refer to the periods 1992 to 1996 and 1996 to 2000. GDP per capita growth is the average growth rate for 1992–96 and 1996–2000. Left government is defined as the average share of government seats of social-democratic, socialist, and green parties 1992–96 and 1996–2000. Institutional rigidity is constant over the whole period of analysis. We used a panel setup with time effects and estimated the models using OLS. Given that due to our operationalization of the time dimension, there are just two waves for the first differences, autocorrelation can hardly be consistently estimated. But since we analyse differences first, we do not expect this to be a serious concern, given previous findings with longer time series (Siegel, 2002; Kittel and Obinger, 2003). Total Social Expenditure: SOCX vs ESSPROS Table 5.9 compares the models estimated for four different definitions of social expenditure, as defined and explained above. As far as the substantive model specification is concerned, three critical notes should be made. First, the share of the population over 65 is negatively, though statistically insignificantly, associated with social expenditure. Sensitivity analysis reveals that this is due to the inclusion of the lagged dependent variable, which contains the level effect of the share of the elderly population. Secondly, changes in unemployment, included in accordance with the expectation that increases in unemployment rates would generate a push effect on social expenditure, do not significantly correlate with social expenditure changes – a finding that is equally puzzling. Furthermore, the interaction analysis reveals that under the condition of high institutional rigidity, left government is even negatively associated with social expenditure. This supports the Nixon-goes-to-China logic based on a party systems explanation as suggested by, inter alia, Kitschelt (2001) and Ross (2000b). According to this proposition leftist governments are in a better position to cut back welfare programmes than parties to their right because they are the natural pro-welfare state party and therefore may credibly claim that, regardless of restrictive saving measures, they still have a vital interest in preserving generous welfare state provision. But findings in favour of this proposition remain fairly ambivalent (Kittel and Obinger, 2003). Apart from these somewhat puzzling details, the models turn out to produce astonishingly robust results. In particular, we find a clear conditional effect of partisanship and institutional rigidity, a clear negative effect of the lagged levels of social expenditure, and a clear negative effect of GDP growth. Jointly, these variables capture about 60 per cent of the variation in the dependent variable, a number which is even more impressive in

95


Table 5.9 Social expenditure dynamics, 1993–2001: dependent variables compared Social expenditure (first differences) SOCX SOCX ESSPROS ESSPROS PTSESOCX PMTSESOCX TSPESS TSBESS Public Total Total Total protection benefits Lagged levels Change of population65 Change of unemployment GDP/capita growth Left government (LG) Institutional rigidity (IR) Interaction LGIR Interaction analysis LG if IR at minimum LG if IR at maximum R2 (adj.) NT

0.36*** (0.06) 0.37 (0.46) 0.20 (0.12) 0.13*** (0.04) 0.28 (0.96) 0.15 (0.97) 7.29*** (2.24)

0.38*** (0.06) 0.58 (0.48) 0.21 (0.13) 0.14*** (0.04) 0.26 (0.98) 0.10 (1.01) 6.90*** (2.30)

0.35*** (0.06) 0.58 (0.52) 0.18 (0.15) 0.15*** (0.04) 0.43 (1.20) 0.17 (1.06) 9.09*** (3.11)

0.35*** (0.05) 0.73 (0.47) 0.20 (0.14) 0.14*** (0.04) 0.55 (1.09) 0.47 (1.04) 9.21*** (2.95)

2.78*** (0.77) 4.25** (2.00)

2.64*** (0.77) 4.01* (2.05)

3.39*** (1.05) 5.37** (2.64)

3.32*** (1.06) 5.57** (2.42)

0.62 36

0.61 36

0.61 36

0.62 36

Notes: All results are from fixed time effects models estimated with OLS, standard errors in parentheses are Huber-White sandwich estimators correcting for heteroskedasticity. NT: 18 OECD countries, two periods, 1993–97 and 1997–2001. Since data for voluntary private social expenditure are, if at all, available only up to 1998, we do not include the analysis of this variable here. See Table 6.11. The coefficient of left government reported refers to the condition that institutional rigidity is set to the mean. The coefficient of institutional rigidity reported refers to the condition that left government is set to 0.5. See the table section on interaction analysis for coefficient estimates of left government at extreme values of institutional rigidity. * p 0.10, ** p 0.05, *** p 0.01

the face of a complete failure of those factors which carry the largest explanatory load in models using annual data (Kittel and Obinger, 2003). More disconcerting in methodological terms, however, is the finding that our results do not depend on the definition of social expenditure. This

96


implies that although we use fairly different definitions, which causes considerable divergence in the measures, this variety does not appear to affect the substantive conclusions we might infer from these models. One implication drawn from this might be to emphasize the robustness of the results regardless of different model specifications. However, in our view this is a completely faulty interpretation because it means that our statistical analysis is simply unable to account for serious differences in measurement: although we measure considerable shifts in the composition of social expenditure which should be attributable to government activity, there is no indication in our findings that these shifts matter, although our study revealed clearer and more robust indications than previous studies that government policy mattered, even during the 1990s (cf. Huber and Stephens, 2001; Kittel and Obinger, 2003). Since we were unable to obtain data for voluntary social expenditure, or for pension data beyond 1998,22 we have to shift the period analysed somewhat in order to assess these variables. Therefore, in Table 5.10 we present a variant of the analysis performed in Table 5.9, in which the first difference period has been shifted to 1994–96 and 1996–98. Compared to Table 5.9, the story told by this table does not seriously diverge from the longer time span, except for a clearly worse overall fit, and generally smaller and less statistically significant coefficient estimates.23 However, the results for the new variable included, PMVTSESOCX, are striking. While the fit of the socioeconomic variables appears to be much stronger, the politicalinstitutional interaction model, which contributes most to the fit of the other models, breaks down for this variable. This is most relevant, because it hints at the potential need to nuance the conclusions about the impact of these factors. Apparently, the political factors we use do differentiate between different levels of public effort in social protection but are not related to the voluntary part. At the same time, to the extent that the public effort seems to underprovide for social protection, voluntary programmes are used to compensate for the missing part, thereby jointly matching societal demand, which is captured by the included socioeconomic variables. Pension Expenditure: SOCX vs ESSPROS In order to facilitate comparisons, we use the same model specification for the analysis of pension data (Table 5.11). Comparing the results we find that the model overall performs considerably worse. The overall fit drops substantially, hardly any coefficient estimates remain statistically significant, and the political-institutional interaction effect gets lost in the noise, although its basic structure remains intact. Note that the effect of change in the elderly population now has the expected direction – even though the

97


Table 5.10 Social expenditure dynamics, 1994–98: dependent variables compared Social expenditure (first differences) SOCX Public

0.18*** (0.04) Change of population 0.21 > 65 (0.77) Change of 0.29 unemployment (0.20) GDP/capita growth 0.11** (0.05) Left government (LG) 0.31 (0.75) Institutional rigidity 0.43 (IR) (0.68) Interaction LGIR 3.92* (2.01) Interaction analysis LG if IR at minimum 1.37* (0.73) LG if IR at maximum 2.41 (1.64) 0.44 R2 (adj.) NT 36

Lagged levels

SOCX SOCX ESSPROS Public Public Total Mandatory Mandatory protection Voluntary

ESSPROS Total benefits

0.18*** (0.04) 0.38 (0.79) 0.31 (0.20) 0.12** (0.06) 0.28 (0.77) 0.45 (0.65) 3.97* (2.11)

0.14*** (0.04) 1.14 (1.04) 0.52** (0.25) 0.16** (0.06) 0.23 (1.09) 0.06 (1.02) 0.90 (3.21)

0.16*** (0.04) 0.31 (0.86) 0.26 (0.23) 0.12* (0.06) 0.37 (0.87) 0.33 (0.73) 5.16** (2.37)

0.16*** (0.04) 0.42 (0.82) 0.26 (0.21) 0.11 (0.06) 0.31 (0.83) 0.30 (0.64) 5.28** (2.23)

1.42** (0.72) 2.41 (1.73) 0.44 36

0.62 (0.91) 0.25 (2.64) 0.26 36

1.84** (0.82) 3.14 (1.94) 0.38 36

1.95** (0.86) 3.14* (1.79) 0.43 36

Notes: All results are from fixed time effects models estimated with OLS, standard errors in parentheses are Huber-White sandwich estimators correcting for heteroskedasticity. NT: 18 OECD countries, two periods, 1994–96 and 1996–98. The coefficient of left government reported refers to the condition that institutional rigidity is set to the mean. The coefficient of institutional rigidity reported refers to the condition that left government is set to 0.5. See the table section on interaction analysis for coefficient estimates of left government at extreme values of Institutional rigidity. * p0.10, ** p0.05, *** p0.01.

effect only seems to be relevant for the SOCX public expenditure data. At the same time, the different pension indicators do not entail different results, apart from minor and irrelevant variation in coefficient estimates. This finding is surprising. Given the high correlations between social and pension expenditure data, the degree of divergence in the results is astonishing, but also in line with earlier surprises arising from our analysis.

98


Table 5.11 Pension expenditure dynamics, 1994–98: dependent variables compared Pension expenditures (first differences) SOCX SOCX SOCX ESSPROS PPENSOCX PMPENSOCX PMVPENSOCX TPENESS Public Public Total Total Mandatory private Lagged levels Change of population 65 Change of unemployment GDP/capita growth Left government (LG) Institutional rigidity (IR) Interaction LGIR Interaction analysis LG if IR at minimum LG if IR at maximum R2 (adj.) NT

–0.06 (0.04) 0.51* (0.30) 0.14 (0.10) –0.04 (0.03) –0.16 (0.30) –0.19 (0.43) –1.42* (0.83)

–0.06 (0.04) 0.54* (0.30) 0.14 (0.10) –0.04 (0.03) –0.09 (0.31) –0.11 (0.44) –1.19 (0.85)

–0.07 (0.04) 0.45 (0.30) 0.15 (0.09) –0.05 (0.03) –0.004 (0.33) –0.04 (0.40) –1.09 (0.85)

–0.05 (0.04) 0.17 (0.34) 0.13 (0.10) –0.04 (0.03) –0.18 (0.32) –0.28 (0.42) –1.22 (0.94)

0.44 (0.42) –0.92 (0.58) 0.23 36

0.42 (0.43) –0.73 (0.60) 0.23 36

0.46 (0.63) –0.59 (0.63) 0.29 36

0.35 (0.48) –0.83 (0.64) 0.35 36

Notes: All results are from fixed time effects models estimated with OLS, standard errors in parentheses are Huber-White sandwich estimators correcting for heteroskedasticity. NT: 18 OECD countries, two periods, 1994–96 and 1996–98. The coefficient of left government reported refers to the condition that institutional rigidity is set to the mean. The coefficient of institutional rigidity reported refers to the condition that left government is set to 0.5. See the table section on interaction analysis for coefficient estimates of left government at extreme values of Institutional rigidity. * p0.10, ** p0.05, *** p0.01.

Comparing Tables 5.8, 5.9, 5.10 and 5.11, there are three major findings. Firstly, differences in conceptualization appear to be invisible in the actual research setting. Neither does it matter which elements are included in social or pension expenditure, nor whether the variable is consistently measured or not. The correlations within the sets of social expenditure and pension expenditure variables are extremely high and the substantive interpretation


99

of the model remains unaffected. Secondly, the four-year dynamics in total social expenditure are fairly well captured by the model (Table 5.9). In particular, we find the expected conditionality of the partisan effect on the institutional setup of the polity. Shortening the sub-periods, however, leads to a reduction of the model’s explanatory capacity, albeit without curtailing the conditional effect (Table 5.10). Thirdly, despite the high correlations between the indicators of social and pension expenditure (well above 0.8), a comparison between tables 5.10 and 5.11 reveals that the same model employed for the two sets of indicators yields considerable differences in model fit, although the general picture remains unchanged.

CONCLUSIONS By discussing measurement and modelling issues in the quantitative macrocomparative analysis of social expenditure data, we have highlighted a variety of serious methodological problems, which are rarely considered in customary research in this tradition. We should note, moreover, that these problems are certainly not unknown or even new. Their appreciation needs careful scrutiny of the lengthy, complicated, and not easily available technical notes to the datasets. We summarize our findings and conclusions in a tabular form. The first problem is that even our best sources for internationally comparable data (the OECD social expenditure database and the European Union’s ESSPROS database) are conceptually ambivalent and inconsistent to an extent which seriously undermines the conceptual validity of the indicators used. We have discussed the Dutch case, where the aggregate data does not cover mandatory private pension programmes. If such inconsistencies occur in one country for which we happen to have information, how can we rely on measures for other countries? A second problem is the fact that, despite the impressive amount of work done by the data collecting organizations (OECD and EU), many indicators have missing values for many observations, in particular at the more disaggregate level. This raises the question of the trustworthiness of the aggregate measures, for which more complete time series for longer periods and more countries are available (see also Siegel’s discussion of aggregation problems in Chapter 4, this volume). This problem is often due to the lack of data. For example, there are hardly any data on voluntary pension plans which exceed legal requirements in Belgium. A third and related concern involves measurement differences between data sets which are often considerably larger than differences over time. Hence the uncertainty about the exact score of a country at a certain point

100


in time is larger than the measured changes over time. Since the measurement error may be larger than the measured variation, our ability to interpret the results of time series analyses is seriously undermined (again, see also Siegel’s contribution to this volume). Fourthly, even within data sets, the comparability of observations between countries and over time is compromised by changes in concepts and measurement errors or the inconsistent application of measurement concepts. In addition, different methods of measurement make specific types of expenditure appear higher or lower than they are in reality. As a fifth concern we have noted in passing that fiscal policy constitutes another major stumbling block for assessing the actual magnitude of public welfare efforts. This relates to the whole area of tax regulation for pension programmes, family policy, education, but also applies to a variety of welfare- and income-relevant issues such as social housing policy and mortgage deductibility regulations. Since such measures form a heavy burden on the public purse and have considerable redistributive effects (though often to the benefit of the better-off), they should in fact be included in a functional definition of welfare. While data on these issues are practically unavailable on a cross-nationally comparable basis, it can safely be assumed that their size and hence their distorting effect for comparative measurement is more substantial than all the observed differences between indicators including public, mandatory, and voluntary welfare programmes (see Obinger and Castles, 2006). The final concern in this certainly not all-encompassing list is that in some, albeit important, respects, the above problems appear to matter only marginally in quantitative macrocomparative research. There is too little and too much variation in outcomes at the same time. We find that using conceptually different indicators of either social expenditure or pension expenditure – public versus public plus private mandatory – as a dependent variable leads to practically identical parameter estimates. While one might applaud this result as proving the robustness of the results produced by different model specifications, in our view it is more indicative of the problem that the method employed is insensitive to important conceptual differences in the measurement of indicators. Even if we accepted this as a sign of robustness, we would be faced with another problem: pension data are a major element of social expenditure and hence should be closely related to the latter. They do indeed correlate strongly with social expenditure, but substituting the former for the latter leads to a substantial decline in model fit. Hence, conceptually divergent indicators converge in outcomes, while conceptually related indicators diverge in outcomes. All of the above concerns may be regarded as a debate about half-full and half-empty glasses. However, in our view the problems jointly highlight a


101

serious methodological fallacy which is inherent in the usage of social expenditure data as indicators of welfare efforts and which contributes to the dependent variable problem. As we have shown, a close and detailed examination of the data allows the extraction of information about ongoing shifts in the composition of social expenditure, which may plausibly be attributed to governmental retrenchment policies. Such a careful tracing of the implications of policy decisions for expenditure data, however, is quite the contrary to the large strokes of quantitative macrocomparative analysis ‘rigorously testing’ broad hypotheses about partisanship, institutions, and socioeconomic determinants. They simply miss these subtle shifts due to the enormous amount of noise in the data. Nevertheless, and despite all these concerns, there are some indications which may support a more positive conclusion. Firstly, our selection of examples may overemphasize the problems. The selection of countries for our analysis is due to our prior knowledge of particular cases for which we are able to compare the data contained in the data set to other sources of information. Hence we cannot claim to have presented a representative sample.24 Secondly, it is not inconceivable that many measurement problems may be solved in future versions of the data sets. Conceptual ambiguity, missing data, and erroneous classifications are not irresolvable problems, although their solution may take considerable further efforts. Thirdly, and perhaps most intriguingly, the little information we have on the differences in findings for public and total social expenditure tells an interesting story: politics seems to be able to account for variation in public spending, but not for variation in total spending, including voluntary expenditure. At the same time, socioeconomic factors tend to be more closely related to total spending than to public spending. This suggests the existence of communicating vessels: private provision substituting for public provision in case the latter is more restrictive. Hence the measures for total expenditure seem to follow societal needs, as measured by the socioeconomic indicators in the model, more closely than public programmes alone. But this interpretation is based on a very weak foundation: the measures which we used for total social expenditure are seriously distorted by conceptual ambiguity and missing information, particularly for those countries in which voluntary programmes cover relatively larger parts of the total welfare effort, and by the weakness of the findings, which are to a considerable degree based on the interpretation of statistically insignificant results from a fairly simple model covering little variation in a small data set. Finally, our explanatory model is far from complete. The effects which we identified may thus be distorted due to missing variables. Most importantly, future research should elaborate the empirical analysis of two

102


potentially important factors which we have noted in passing in the course of our argument, but did not include in the model specification. The first is the extent to which collective agreements affect the dynamics of welfare arrangements such as collective pension plans or sickness funds. There are basically two mechanisms which lead collective agreements to affect the size of such programmes. Firstly there is the share of trade union members in employment (union density) and the share of employees covered by legal extension clauses. We have not dealt with these issues here. The second factor is related to the developments in the capital markets: if the cycle of the capital markets differs from the cyclical development of social development (which can be induced by, for example, demographic and by labour market developments), funded schemes can in the short run either perform better (when the returns on capital markets grow faster then social expenditure) or worse (when returns on the capital markets grow slower than social expenditure). In case of the first type of development, funded pension schemes may appear momentarily less costly than PAYG arrangements. It is thus particularly relevant to control these factors in time series analysis where the consideration of the timing of effects becomes relevant. Hence, our overall conclusion is that while political scrunity is reshaping the welfare state, this process goes largely unnoticed by analysts who attempt to draw big conclusions from all-encompassing models without scrutinizing the technical annexes of the data sets which they use. To simply turn back to business as usual and to continue estimating one regression equation after the other on the same ambiguous aggregate data seems, in our view, fairly careless. We have a serious and – at least currently – insurmountable data problem, which curtails our ability to discriminate between contending hypotheses. Thus, despite all hopes and many endeavours to the contrary, we have not yet completed the phase of careful data collection, which, to a considerable extent, will have to be based on systematic, in-depth, comparative case studies, exposing us to the plight and dust of archives. The time of elegant models is yet to come in comparative welfare state research.

NOTES 1. 2. 3. 4.

However, due to missing data, in practice one must settle for considerably shorter periods. For example, expenditure statistics at the municipal level in Belgium are much less detailed than at the federal or regional level. According to Eurostat these costs amount on average to about 3.5 per cent of expenditure in the European Union. Austria scores the lowest with 1.5 per cent while the Netherlands scores the highest with 4.7 per cent. The Minster of Labour declares a collective agreement generally binding for all employees working in an entire sector of the economy.

The problems of using aggregate spending data 5. 6. 7. 8.

9. 10.

11.

12. 13.

14.

15. 16. 17.

18.

19.

103

Interestingly enough, in its discussion of replacement rates in its 2005 Pensions at a Glance (OECD, 2005c), the OECD considers Dutch occupational pensions to be private mandatory schemes. In the US, by contrast, this applies to only about half of the working population, although this country is usually considered to be the reference case for widespread occupational pensions. In some collective wage agreements this is increased to 100 per cent. Employers can privately insure this risk of having to continue to pay wages for this, by international standards, very long period. During the period 1983–2001 this was on average 4.1 per cent in the Netherlands, and only 1.4 per cent in Germany, which in both countries was also the figure for the year 2000 (see Figure 1 and Table 1 in Sisko Bergendorff Sickness Absence in Europe – A Comparative Study, 4th International Research Conference on Social Security, ISSA, Antwerp 5-7 May 2003, available at www.issa.int/engl/publ/2contanvers.htm). Even then one is faced with the problem that SOCX is sometimes ambiguous as to whether data are just missing or the value is actually zero. The shift from the public to the mandatory private scheme, though, may lead to inequalities in the future, as the 70 per cent replacement rate that the occupational schemes seek to achieve is only a convention, i.e. it is not backed up by law. Moreover, the method of financing these benefits can differ substantially between the different industry-wide and enterprise based pension funds. In some cases the employers bear most of the costs; in other cases it is the employees. The funds differ also substantially in terms of their policies towards survivor pensions etc. In other words, the shift from public to mandatory private schemes does seem to imply important consequences in terms of the pooling of risks. For an overview of these arrangements in the European Union, see Fédération des Experts Comptables Européens (2001), Aspects of Taxation of Occupational Pensions in the EU, Brussels (available at www.fee.be/publications/main.htm). See also Chinu Patel (ed.) (2004), Taxation of Occupational Pensions in EU Countries, Oxford: European Actuarial Consultative Group. These would also form an increasing problem for other programmes, such as disability schemes if they were run on a funded basis. Like in a pay-as-you-go scheme, a funded defined benefit scheme guarantees its pensioners predefined benefits, irrespective of the investment performance of the pension fund. In case of disappointing investment returns or of investment losses, these deficits have to be compensated for (‘back-serviced’) by increased contributions to be paid by the current and future sponsors of the plan, without resulting in corresponding higher benefits for future retirees. This is also illustrated by the steep rises of contribution rates that the world’s second largest pension fund, the Dutch ABP, implemented during this period: from 13.2 per cent in 2002 over 15.2 per cent in 2003 to 18.6 per cent in 2005 (an increase unheard of in any PAYG system) The only exceptions to the EET rule seem to be some occupational pension arrangements in Germany and Luxembourg. This interdependence also refers to a third function, that of disability. Similar problems exist regarding early retirement pensions, which in some countries are primarily financed via the unemployment insurance system (e.g. Belgium), while in other countries they are primarily classified as a cost for the pension system (e.g. in Germany). In the absence of a statutory early retirement arrangement, incapacity benefits can even act as a functional equivalent (e.g. the UK). The ESSPROS manual defines ‘old-age pension’ as ‘periodic payments intended to i) maintain income of the beneficiary after retirement from gainful employment at the standard age or ii) support the income of old persons. This could also include social assistance benefits for the elderly’ (Eurostat, 1996: 58). Some early retirement benefits are not included, in particular ‘early retirement for labour market reasons’ which in the SOCX database is classified in the category ‘unemployment’.

104

20. 21.

22. 23. 24.

Measuring and analysing ‘welfare efforts’ A similar coding problem applies to disability pensions, which in most countries form a separate category for those whose working capacity is reduced before reaching the statutory retirement age, but in Denmark largely fall under the category ‘old-age cash benefits’ (SOCX category 1.5.1, which in the Danish case includes expenditure on pensions to those who become incapacitated before reaching the statutory retirement age). It ranges from about 0 per cent in Scandinavian countries over about 0.2 per cent in countries such as Germany to 1.7 per cent for Belgium and the Netherlands. The countries which are available in both data sets and which hence are included are the European countries Austria, Belgium, Germany, Denmark, Finland, France, Greece, Iceland, Ireland, Italy, Luxemburg, The Netherlands, Norway, Portugal, Spain, Switzerland, Sweden, and the United Kingdom. Although the database contains data up to 2001, this is not the case for these series. In passing we note as a potential implication of the finding that the longer periodization leads to a better fit that a sensible conceptualization of time horizons of effects may indeed matter more than is usually acknowledged in panel studies using annual data. Still, how many cases are allowed to contain serious measurement errors before we stop being able to use the data without distorting results?


105

APPENDIX Table 5A.1

A comparison of SOCX and ESSPROS categories

SOCX categories

ESSPROS functions

Issues

Differentiated for • public • mandatory private • voluntary private

No differentiation

Differentiation of SOCX seems arbitrary and not consistent

Functional breakdown and systematic breakdown into • non-means tested cash • lump sum • benefits in kind • means tested

Only functional breakdown in SOCX

1. Old-age cash benefits

Old age

2. Disability

Disability

3. Occupational injury and disease

Sickness/heath care and disability

Not separately defined in ESSPROS

4. Sickness benefits

Sickness/health care

Inconsistencies between SOCX published tables and data retrieved from CD-ROM

5. Services for the elderly

Old age and disability

No distinction between old age and disability in SOCX

6. Survivors

Survivors

7. Family cash benefits

Family

8. Family services

Family

9. Active labour market policies

–-

Not included in ESSPROS

10. Unemployment

Unemployment

Not the same coverage and definitions

11. Health

Sickness/health

Only public expenditures in SOCX

12. Housing benefits

Housing

13. Other contingencies

Social exclusion

7 & 8 in one category in published SOCX tables

ESSPROS does not include programmes like those devoted to immigrants and refugees

Source: In part based on OECD (2001) with additions by the authors.

6. Social rights, structural needs and social expenditure: a comparative study of 18 OECD countries 1960–2000 Olli Kangas and Joakim Palme INTRODUCTION What has driven the expansion of welfare state expenditures? Why do some countries spend more on social security than others? These basic questions have occupied social scientists’ thoughts for decades. Over time researchers have come to shift the focus both geographically and in terms of level of aggregation. Cutright (1965) and Wilensky (1975) started out by analysing cross-sectional variation in social expenditure among industrial as well as developing countries. The focus was then shifted to the variation in expenditure among the most advanced industrial nations, and the variation over time within nations was also empirically analysed (e.g. O’Connor and Brym, 1988; Pampel and Williamson, 1989; Hicks and Swank, 1992; Huber and Stephens et al., 1993; Hicks and Misra, 1993; Huber and Stephens, 2001; Castles, 2004). Another shift occured when researchers started to examine the development in specific programmes such as pensions (Pampel and Williamson, 1985; Palme, 1990; Huber and Stephens, 1993), sickness benefits (Kangas, 1991), family benefits (Wennemo, 1994; Ferrarini, 2006), unemployment insurance (Carroll, 1999) and social assistance (Nelson, 2003; Kuivalainen, 2004). Yet comparative research in this area has still remained surprisingly inconclusive. The importance of economic development and party politics has, for example, been given very different weights as an explanatory factor. The same applies to the more current ‘welfare state retrenchment’ or the ‘new politics’ of the welfare state literature, where results are even contradictory (see Pierson, 1994; Green-Pedersen, 2000; Korpi and Palme, 2003; Castles, 2004; Siegel, 2005). There are at least three, as we see it, fundamental problems that have contributed to this state of affairs, where theoretical controversies persist and 106

A comparative study of 18 OECD countries 1960–2000

107

empirical research shows only a few signs of accumulation. The first problem is to do with the fact that researchers who have analysed the causal factors behind welfare state growth usually have not distinguished between the two different types of factors that, both over time and across nations, produce variations in expenditure: the extension of social rights, on the one hand, and the growth of needy populations, on the other hand (for exceptions, see Saunders and Klau, 1985; Alestalo and Uusitalo, 1992; Castles, 2004). This has led to what we see as the second problem: a misconception of how politics is likely to influence welfare state variation. While it is difficult to deny that social expenditure is in many ways an important indicator of welfare state development, it is more difficult to recognize why politicians would be inclined to increase expenditure as such; a more likely motive would be to improve social rights and the well being of citizens. As a response to this, researchers have attempted to compile data sets on legislated social rights which can be used as alternatives to expenditure data. The results from analysis of these data suggest that variation in social rights cannot be explained in the same way as social expenditure, and, in particular, that the relative importance of political factors is sensitive to the operationalization of the dependent variable. When analysing public social expenditure, the impact of political power constellations is harder to verify (e.g. Castles, 1982a; Castles, 2004), whereas studies based on social rights as the dependent variable indicate that politics does matter (DeViney, 1983; Myles, 1984; Pampel and Williamson, 1985; Korpi, 1989; Esping-Andersen, 1990; Palme, 1990; Kangas, 1991; Wennemo, 1994; Carroll, 1999; Montanari, 2001; Huber and Stephens, 2001). Our hypothesis is that these divergent results can be explained by the combined effects of structural ‘need’ variables, such as the age structure of the population and the unemployment rate, and factors related to social rights, such as the extension of the coverage of social benefit schemes, earnings replacement levels, and qualifying conditions regulating access to benefits. Since most ‘need’ factors are hard to alter through political decision making, the volume of social spending is not directly linked to political variables, yet is greatly affected by other phenomena. For example, the relative size of the aged population has substantial ramifications for spending on pensions, as shown by Pampel and Williamson (1985; 1989). In the case of pension rights the impacts of demography are more negligible, while the impact of political factors is more pronounced (Myles, 1984; DeViney, 1984; Palme, 1990). Another good example is the proportion of children to the total population versus the level of child allowance. The share of children is harder to change by means of political decisions – or rather, the impact takes a couple of decades to become visible – than the amount given in child benefits, which can be adjusted overnight.

108


This suggests that social rights and expenditures can be explained somewhat differently. It also indicates that neither an approach stressing structural variables nor a research strategy based on the concept of social rights alone will appropriately capture the determinants of cross-country differences in the level of social security spending. This is important when we try to understand the expansion as well as the retrenchment of the welfare state. The third fundamental problem is one of measurement. Due to the easy availability of published data, the proportion of gross domestic product (GDP) spent on social security has generally been used as the indicator of welfare state development. As discussed by De Deken and Kittel (Chapter 5), as well as Siegel (Chapter 4) in this volume, there are considerable inconsistencies within and between available statistical data sets.1 The purpose of this chapter is to contribute to the discussion sketched above by analysing the relative importance of factors pertaining to structural need and social rights respectively, in explaining the variation in social expenditures in 18 OECD countries between 1960 and 2000. In order to do this effectively, it has also been necessary to address the measurement problem. Here, it is important to consider the obvious differences in available data sets, as well as how and to what degree the achieved results are contaminated by the peculiarities of the data used. An additional aim of the analysis is thus to highlight some of the methodological problems with comparative welfare state research. The research design is displayed in Figure 6.1. In this chapter we are interested in the relationships indicated by solid arrows. However, at the end of this chapter, we cursorily touch upon the issues between party politics, overall levels of social rights and social spending. The structure of the chapter is as follows. The first section describes and discusses the data. The second section presents the methodological considerations behind the analysis. The results are presented in the third section. In the concluding section, the findings are discussed.

DATA Social Rights Data on social rights have been compiled within the Social Citizenship Indicator Program (SCIP) conducted at the Swedish Institute for Social Research, Stockholm University (for a more detailed description, see Korpi, 1989; Esping-Andersen, 1990; Palme, 1990; Kangas, 1991; Wennemo, 1994; Carroll, 1999). The concept of social rights, based on Marshall’s (1950)


109

Party politics Rights

Economic structure

Public social expenditures Needs

Institutions Background variables

Intervening variables

potential effects cursorily studied effects

Figure 6.1

studied effects

The determinants of public social expenditure

notion about the sequential enlargement of citizenship rights, refers to legislated social provisions aimed at guaranteeing citizens’ economic welfare and security (Marshall, 1950: 11; Korpi, 1989). Since the growth of social rights has come about as a result of modern social policy, and eligibility for benefits is established in legislation, the data collection is limited to statutory or ‘public’ welfare programmes.2 Hence, when calculating the quality of social rights, we concentrate upon schemes characterized by being created via national legislation or involving direct public participation in financing – various private alternatives or functional equivalences have been neglected. The time span for which data has been collected stretches from 1930 to, for the purpose of the present chapter, 2000 – as a rule, with observation points at five-year intervals. The countries included in the data set are Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Ireland, Italy, Japan, the Netherlands, New Zealand, Norway, Sweden, Switzerland, the United Kingdom, and the United States. The data set consists of information on five main social security programmes: old-age pensions, sickness insurance, unemployment insurance, work accident insurance, and family allowances.

110


Since work injury insurance in many countries is coordinated with sickness insurance, we did not analyse it separately here. For each scheme, we have information on the degree of its coverage (the proportion of those who are in principle entitled to benefits); the qualifying conditions regulating access to benefits (minimum membership of a scheme, length of contribution period, residency criteria, income/means testing etc.); the number of unpaid waiting days (for sickness, unemployment and work injury benefits); the length of the benefit period (for sickness, unemployment and work injury benefits); and the amount of benefit paid. Benefits are calculated for a number of ‘typical’ cases with regard to earnings of an ‘average production worker’s wage’ in manufacturing; household composition (single person and four-person family with one wage earner/couple in the case of pensions); and the length of work incapacity (relevant for sickness, work injury, and unemployment. In these cases benefits are separately calculated for one week and 26 week absences). Child allowances are only calculated for a family with two children (aged two and seven) and with one full-time working parent earning an average industrial worker’s wage. Benefit levels for each year in each country have been related to this wage level in order to make the provisions more comparable over time and across countries, i.e. benefits are proportional and the yardstick is the average wage level in industry. Since benefits were previously non-taxable income, but are taxed today with few exceptions, it is necessary to take the effects of taxation into account. Therefore, net benefits – that is, benefits after taxes and social security contributions – are related to the net wage (for details of taxation, see Palme, 1990: 26–36). Structural Needs In order to systematically examine the ways in which different aspects of social rights, certain structural ‘needs’, and social expenditures are associated, the following sections offer separate analyses for old-age pensions, sickness insurance, unemployment insurance, and child allowances. Needless to say, structural factors affecting social spending differ across programmes (see for example Saunders and Klau, 1985: 101–11). For oldage pensions the most important structural variable is naturally the relative size of the elderly population. This demographic pressure is measured as the proportion in the total population that is 65 years of age and over. In a similar way, a natural proxy for the ‘need’ for child allowances is the proportional size of the younger population which is measured as the proportion of persons below 16 years of age in the total population.3 In a parallel fashion, the unemployment rate can be assumed to reflect the ‘need’ for unemployment benefits. Unfortunately, it is more difficult to find structural


111

variables with a direct impact on the ‘need’ for sickness cash benefits. There are some sporadic data mainly based on various surveys for some countries and for some points in time, however no comparable data are available for the whole time period and for all the countries we are interested in. In the absence of a direct and more reflective indicator, we have used life expectancy as a proxy for the overall health status of the population. Data on structural factors employed in the subsequent analyses are mainly derived from OECD publications.4 Although our data on social rights date back to the 1930s, the present inquiry is restricted to the period 1960–2000, as reliable data on disaggregated social expenditure are only available from 1960 onwards. The points in time for which we have data are 1960, 1965, 1970, 1975, 1980, 1985, 1990, 1995 and 2000. Social Expenditure Data on social expenditure are available from various sources. The three main historical databases used in previous studies have been compiled by ILO, OECD, and parts of Flora’s extensive project (e.g., 1986a, 1986b, 1987a). Although the last data set is the most detailed and one appears to be the most reliable and consistent of the three, the problem is that it only covers 12 Western European nations, and figures are only available up to 1980.5 In order to include all of our 18 advanced OECD countries and the years for which we have social rights data, only the ILO and OECD data sets are applicable.6 Only ILO data are disaggregated by expenditure categories that best correspond to our measures of social rights by programme. The problem with the ILO data is that there may be huge discrepancies between consecutive years in the spending levels within a single spending category. OECD also provides data on social transfers where transfers are broken into separate spending categories. Crosschecking against Flora and various national sources suggested that OECD data are more consistent with the other databases. A closer examination of ILO statistics also revealed that occasionally very large changes compared to previous years appear without clear explanations to clarify why. Unfortunately, use of OECD data also leads to problems. The OECD has compiled various social spending data sets. For example, the first data set covers the period 1960 to 1980 and the most recent one 1980 to 2001/02. The problem is that the overlap between these two time series is not perfect. The correlation between the overall spending levels is 0.71 for 1980 data. In order to improve the comparability of data, in some cases we have been obliged to make our own adjustments to OECD and ILO figures and in some cases complemented them with national data (see below). These adjustments were also motivated on a conceptual basis. In some cases the

112


SCIP, for example, classifies sickness wage continuations schemes as ‘public’ (legislated and obligatory), whereas the OECD has excluded them. The consistency between individual expenditure categories is sometimes, but not always, better. Both the ILO’s and OECD’s disaggregated pension expenditure are rather divergent (r0.64 in 1980). For other disaggregated expenditure categories, the available statistics are more consistent. The overlap of OECD sickness data with ILO data is surprisingly good (0.90 in 1990). The greatest discrepancies appear in the cases of Austria, Germany and to some extent Finland and Norway. Both of these statistics show extremely high values for Sweden, with the spending rate at over 4 per cent of GDP compared to 2 per cent or below for the other countries. This calls for some clarification. The discrepancies between the databases are due to the different treatment of legislated wage continuation programmes and the inclusion of varying types of cash provisions. In the Austrian, German and Norwegian cases, the ILO statistics exclude wage continuation programmes in which the employers bear the burden for all payments. In the Swedish case, both data sets include cash provisions for purposes other than income maintenance (e.g. compensation for health care, travel expenses, medicine etc.) which makes Sweden a clear outlier compared to the rest of the OECD countries. In the following analyses, expenditure on the legislated wage continuation practices in Austria and Germany have been added (as in Flora’s data). The Norwegian, Finnish and Swedish figures have also been recalculated. Our data for the two former countries, obtained from statistical reports for the Nordic countries (various years), also include the two-week wage continuation schemes which are obligatory for employers. Derived from national statistical abstracts (various years), Swedish data refer solely to sickness cash benefits. In the case of child allowances, spending figures are available from the ILO database for 1960–98 and from the OECD for 1980–2002. In 1995 the correlation coefficient between the databases was 0.79. The discrepancy is due to the fact that OECD data also include some other transfer categories. Since the ILO figures are conceptually closer to our measurement of the generosity of child allowances, we rely here on ILO data. However, since the ILO data have some breaks in the statistics for Switzerland, the Swiss spending data are taken from the OECD.

METHOD7 To examine whether the effects of the explanatory on the dependent variables have changed from one time point to another, we first examined annual cross-sectional data and performed separate regression analyses for each of the four social insurance programmes for each of the six cross-sections.


113

Here, we used ordinary OLS regressions. To make more efficient use of our data, we also pooled the nine cross-sections. In other words, the data were merged and these pooled data were analysed as a panel which provides multiple observations for each country in the sample. By pooling data we receive additional information on the variation between countries as well as over time (see e.g. Hsiao, 1990; Hicks, 1994; Micklewright, 1994). There are a number of regression techniques available to deal with the particular problems of analysing pooled data – and not so surprisingly, each of them has its weaknesses. The results also seem to be sensitive to the specific method applied (see e.g. Hsiao, 1990; Beck and Katz, 1995; Kittel, 1999; Huber and Stephens, 2000). In this study, it is neither necessary nor possible to go deeper into these methodological problems. By using cross-sectional analyses (and simple visual scatter plots, which are not presented here due to space limitations), and combining them with pooled regression data, we aim to explore the relationships between social rights and social spending. However, in order to test the robustness of our findings, we conducted a number of alternative analyses. We began with a model assuming, seemingly unrealistically, that there is no autocorrelation; thereafter, we tested a model with panel specific autocorrelation and combined both of these runs with and without panel level heteroskedastic error models. By and large, the results were the same, however the standard errors and levels of significance varied, which had some implications for our political variables. Models with no autocorrelation produced highly significant results for the political variables, whereas models with autocorrellation tended to produce larger standard errors and, in some cases, led to statistically non-significant effects for the same variables. Pooled regressions were run by the Stata 9 (Stata, 2005: 226–35) cross-sectional time series package using Prais-Winsten regressions on correlated panels and corrected standard errors (PSCE). To deal with autocorrelation, we could have used lagged variables, in which case we would have lost the effect of the level variables and our results would have been more dependent on the short-term changes (Huber and Stephens, 2000). In order to avoid this, we chose the approach outlined above. The research strategy applied in this chapter represents an analysis in successive steps. First, simple bivariate plots for each cross-section were made, succeeded by OLS models for each year. In the third step, various pooled cross-sectional regression models were estimated and, finally, the results were compared with each other. Due to space limitations, only PSCE estimates (panel specific autocorrelations and panel level heteroskedastic errors) are displayed in the tables that follow. The results from various control runs which produced significantly different results are discussed briefly. Again due to space limitations the cross-sectional results are not shown for each year but rather for each decade of the observation period. As our cases do not

114


form a sample of any defined universe, statistical significance tests should be treated only as a heuristic device for evaluating the results obtained.

RESULTS Total Expenditure In Figure 6.2, an aggregate index of social rights for 18 advanced OECD countries in 1960, 1975, 1990 and 2000 is plotted against total social 1960 (r = .69**)

1975 (r = .77**)

35

30

Social spending, % of GDP


35

25 20 GER 15 AUT FRA BEL NZL UK SWE DEN 10 AUS FIN NOR NL IRE CAN ITA SUI 5 USA JAP 0 –8 –6 –4 –2 0 2 4

30 NL SWE GER BEL FRA DEN IRE AUT NOR NZL SUI UK FIN CAN USA AUS JAP

25 20 15 10 5 0

6

–8

8

–6

Index of social rights

–4

SWE DEN NL FRA FIN BEL ITA NOR AUT GER UK IRE SUI

NZL

20 15

USA

CAN

AUS 10

2

4

6

8

35 Social spending, % of GDP


25

0

2000 (r = .62*)

1990 (r = .77**) 35 30

–2


JAP

5

FRA DEN SWE GER BEL IRE SUI AUT FIN ITA NOR NZL NL UK CAN AUS JAP USA

30 25 20 15 10 5 0

0 –8

–6

–4

–2

0

2


4

6

8

–8

–6

–4

–2

0

2

4

6

8


Note: For all unstandardized partial regression coefficients shown in Tables 6.1–6.6 and for the coefficients in Figures 6.2 and 6.3: * p0.1; ** p0.05, *** p0.01.

Figure 6.2 Social rights and social spending in 18 OECD countries, 1960–2000


115

spending rates according to the OECD data. The index of social rights was constructed as follows: quality measures8 were calculated for each social security programme, separately transformed into standardized z-scores,9 and added together. There are different stories to be read from the figure. If we first look at the social rights index, we can see a remarkable growth dynamic from the early 1960s to the mid 1970s. During that period the so-called Scandinavian welfare model – high social rights attached to high social spending – emerged. Secondly, there was an increase both in social rights and spending levels which, since the 1990s, has become steady in many countries. Taken together, there seems to be a ‘growth to limits’ (Flora, 1986a, b), and a transition from an expansionist welfare state to a recalibration of social policy. Thirdly, there is a clearly positive and significant relationship between the level of social provisions and spending, but the relationship is far from perfect. On the one hand, high spenders do not necessarily guarantee the highest levels of social rights. On the other, the low spenders rarely provide higher levels of social rights. We have countries – such as Sweden – that are on top when it comes to social rights and spending levels, but there are also countries such as Denmark, Germany, and France that display low social rights relative to their levels of spending (since the 1990s). Some other countries score relatively high on the social rights index, and are nevertheless among the medium or low spenders, as exemplified in 2000 by Norway, Finland and the Netherlands. Therefore, the overall association between the index of social rights and social spending is only moderate (r0.63 in 2000). The picture essentially does not change if we use OECD or ILO data solely on social insurance expenditures (r0.33 and r0.55, respectively for all 18 countries in 1990). This indicates that social rights alone cannot properly explain the cross-country differences in social expenditure. We must also consider structural needs as a cause for the variation in this expenditure. Analyses where social expenditure rates have been regressed on the total index of social rights (as described above), and on the structural needs measured as standardized unemployment rates, the size of the elderly population, the percentage of those 16 years of age or below and life expectancy, produced statistically significant positive coefficients for both levels of social rights and the size of the elderly population. The role of the share of children is negligible. As indicated by the variance explained, the model performs less well in the most recent observations, which suggests more ‘noise’ affecting social spending levels. The most important factor in all of the cross-sections seems to be the level of legislated social rights. In the pooled analysis (where various control runs produced more or less congruent results) the growth of the elderly population, increased longevity,

116


levels of unemployment and social rights are important factors. Longevity has no impact in cross-sectional inspections, whereas it becomes significant in the pooled analysis. In a way, this is a logical result if we keep in mind that cross-sectional differences between countries are not that big, but that in all countries life expectancy has increased and consequently the average number of years on pension also increased, as most elderly people take home a public pension. As we have been able to improve the comparability of the data for some of the individual social security programmes, it can be assumed that the disaggregated expenditure data reflect differences in both structural needs and the generosity of social rights among countries more accurately than aggregate data. In the following subsections, we have therefore broken down data for both expenditure and social rights in order to examine the relative importance of social rights and structural factors and, in addition, the relative importance of different aspects of social rights. Unemployment Insurance When it comes to determinants of spending on unemployment protection, the unemployment rate is the most obvious structural need variable. For social rights indicators, we have used the compensation level/replacement rate (proportion of previous wage), duration of benefits period (in weeks), coverage (insured/labour force) and the number of uncompensated waiting days in unemployment insurance. Table 6.2 displays results from regression analyses of unemployment insurance expenditures. When explaining the growth of income maintenance expenditures on unemployment compensation, Saunders and Klau (1985: 105–09) pointed to two important determinants: the level of unemployment and the generosity of the insurance scheme. Both factors also turn out to be the most crucial ones in our regression models. The impact of the level of unemployment is statistically significant in six of the nine cross-sections, while the effects of the replacement level are not that strong in the cross-sectional analyses (significant in four of the years). At several time cross-sections, the insurance coverage rate also has a significant impact, but its relevance is somewhat weaker than that of the replacement level and the unemployment rate. In addition, the influence of duration is, in some cross-sections, positive (in 1990 and also in 1985 and 1995, which are not displayed in the table) and seems to grow in importance towards the end of the observation period. The increasing relevance of duration mirrors the fact that the increase in unemployment during the late 1970s and early 1980s resulted in increases in longterm unemployment spells relative to total unemployment. In 1985 for

117

0.547 ns ns 0.954** ns ns

16 0.366

Constant Unemployment rate 0–16/total population 65/total population Life expectancy Social rights

df R2 adjusted

15 0.675

5.808* ns ns 0.686* ns 0.767**

1970

16 0.525

17.667* ns ns ns ns 1.503***

1980

16 0.565

20.613** ns ns ns ns 1.430***

1990

16 0.350

22.381*** ns ns ns ns 0.933**

2000

157 0.858

36.647** 0.467*** ns 0.875*** 0.560** 0.665***

1960–2000

15 0.564

15 0.538

1.652* 0.235** 0.023** ns ns ns

1980

14 0.719

2.423** 0.243*** 0.032** 0.005* ns ns

1990

16 0.323

0.554 ns ns 0.005** ns ns

2000

157 0.706

0.871 0.179*** 0.007** 0.002** 0.004* ns

1960–2000

Note: For all unstandardized partial regression coefficients shown in Tables 6.1–6.6 and for the coefficients in Figures 6.2 and 6.3: * p0.1; ** p0.05, *** p0.01.

15 0.596

0.357 0.094** 0.006** ns 0.055* ns

0.164 0.047* ns ns 0.004* ns

Constant Unemployment rate Replacement rate Duration Coverage Waiting days

df R2 adjusted

1970

1960

Independent variables

Table 6.2 Regressions of spending on unemployment on aspects of social rights and structural factors, 1960–2000, unstandardized coefficients (ns not significant)

1960


Table 6.1 Regressions of social spending (OECD data) on aspects of social rights and structural factors, 1960–2000, unstandardized coefficients (ns not significant)

118


example, Belgium, Ireland, Italy, and the Netherlands reached the point where over 50 per cent of their unemployed had been out of work for 12 months or more (see e.g. OECD, 1990: 13–15, 203). Previously labelled as having low unemployment, in the 1990s the Nordic countries, and Finland and Sweden in particular, presided over steeply rising unemployment rates. By 2000 overall unemployment levels had decreased, although long-term unemployment remained high. There have been calls for ‘activating’ the unemployed, including a tightening of qualifying conditions and increasing the number of days in waiting. However, as indicated by all equations in Table 6.2, the length of the waiting period is not that decisive (although it has a negative sign) when it comes to spending levels.10 The same goes for the other system characteristics. Only in the pooled data set are coeffecients for the quality of benefits significant. The most important explanatory factor seems to be the level of unemployment, suggesting that changes in unemployment have put the strongest upward pressure on expenditures. The overall performance of the model improved from the early 1960s up to the late 1980s, whereafter it decreased. Sickness Insurance Unlike with unemployment insurance, old-age pensions and child allowances, it is almost impossible to point to a single structural factor which has an impact on sickness insurance expenditure. Firstly, the potential target group (that is, the sick) is impossible to define as simply as the retired, children, or the unemployed who are potentially entitled to old-age pensions, child allowances and unemployment benefit respectively. Secondly, comparative analysis has shown that changes in the age composition of the population tend to lead to an increase in sickness absenteeism, particularly costly long-term absences from paid work (Kangas, 1991: 124–7). Thirdly, the effects of structural ‘need’ factors are not directly transformed into compensated absenteeism, but are filtered by national sickness insurance and other social insurance programmes. Programme characteristics determine whether these effects are strengthened or weakened. The relative impact of the structural needs and the characteristics of the sickness insurance schemes is evaluated in Table 6.3. Life expectancy, pertaining to the health status of the population, yields insignificant coefficients with varying signs in cross-sectional analyses. Thus, it may be that there is an underlying trend towards higher rates of sickness absence despite the improved health status of the population (Bäckman, 1992). As expected, the replacement rate has a positive effect, increasing the impact on expenditures; the impact is also significant in all of the crosssections. The result is fortified in the pooled analyses. Thus, the quality of

119

0.037 ns 0.014*** ns ns ns

16 0.508

df R2 adjusted

1960

Constant Life expectancy Replacement rate Duration Coverage Waiting days


16 0.356

0.413 ns 0.003** ns ns ns

0.073 ns 0.015*** ns ns ns 16 0.354

1980

1970

16 0.327

0.066 ns 0.017** ns ns ns

1990

16 0.387

0.021 ns 0.015** ns ns ns

2000

159 0.415

0.060 ns 0.011*** 0.002*** ns ns

1960–2000

Table 6.3 Regressions of spending on sickness benefits on aspects of social rights and structural factors, 1960–2000, unstandardized coefficients (ns = not significant)

120


the benefit is the most crucial explanatory factor for the variation in spending on sickness allowances. As in the case of unemployment insurance, the length of the benefit period (duration) also has some importance, although the coefficients are not significant in cross-sections but they are significant in pooled data. Contrary to expectation, the number of waiting days has only a minor and statistically non-significant negative impact on the spending rate in cross-sections. The results reflect that the heavy costs of sickness, which are caused by long-term illnesses, are not especially sensitive to incentive structures created by uncompensated waiting days. The results presented in Table 6.3 are based on the adjusted ILO spending data as described above. If we perform identical analyses for unadjusted ILO and OECD figures as done for the year 1990, the results are not strikingly different. In the uncorrected ILO and OECD data the effects of duration are stronger. This is mainly due to the extreme Swedish figures: an unlimited benefit period coupled with an extremely high level of spending. Pensions As most studies (e.g. Wilensky, 1975; 2002) indicate, in the case of pensions the characteristics of the population may be the most important structural factor. Table 6.4 shows the results from analyses where pension expenditures are regressed on the proportion of elderly in the population (persons 65 years and over/total population), life expectancy and a number of pension rights variables. The coverage of the old-age pension scheme denotes the sum of entitlement and the recipiency rates; the former refers to the proportion of insured persons in the total population between the ages of 15 and 64, while the latter pertains to the percentage of those above normal pension-age who actually receive old-age benefits (Palme, 1990: 32). Basic security11 refers to the level of minimum security guaranteed to the elderly, while income security12 measures the adequacy of earnings related pensions. In cross-sectional analyses, the size of the aged population has a significant effect on the spending volume as a rule, and the estimates from the pooled analyses are strongly significant (regardless of the model specification), while life excpectancy is significant only in the pooled time series specification. The two variables pertaining to replacement rates are only occasionally of importance in the cross-sectional regressions, whereas only income security is significant in the pooled analyses. Various control runs yield rather similar results. If we use ILO spending data, the importance of income security is accentuated, and it is clearly more important than the level of basic security. In sum, the robust message coming out of these unrobust results is that the quality of the pension has an important effect on the spending levels; however, in comparison to the demographic

121

df R2 adjusted

Constant 65/total population Life expectancy Basic pensions Income-related pensions Coverage


16 0.625

0.678 ns ns ns 4.436*** ns

1960

16 0.504

1.484 0.589*** ns ns ns ns

2.946 0.479** ns ns 2.551** ns 15 0.658

1980

1970

16 0.471

2.841 0.722*** ns ns ns ns

1990

16 0.667

7.930 1.050*** ns ns ns ns

2000

Table 6.4 Regressions of spending on pensions on aspects of social rights and structural factors, 1960–2000, unstandardized coefficients (ns not significant)

158 0.705

10.820*** 0.438*** 0.136* ns 1.288** ns

1960–2000

122


factors, the qualitative indicators are slightly more marginal. Contrary to other variables, coverage also remains insignificant in the pooled analyses, indicating that cross-country differences in spending levels cannot be properly explained by differences in that dimension. This may be seen as an effect of the convergence in coverage of social insurance programmes (Montanari et al., 2006). All in all, the results in this section suggest that both the size of the elderly population and the adequacy of pension rights have a substantial impact on the extent of cross-national variation in public pension expenditures. Moreover, the results suggest that needs are somewhat more important than rights when considering the size of the elderly population. The results from the pension analyses again reflect the problems involved in using different existing statistical data sets. They indicate that when discussing and comparing results from different studies, we must always carefully consider the contamination caused by measurement differences and errors in the data sets. The same label on an expenditure category given in the statistical source does not guarantee that the concepts are phenomenologically equivalent, as De Deken and Kittel show in Chapter 5 of this volume. Child Allowances In the case of child allowances, the number of waiting days or the duration of the benefit period is not relevant, as is the case for unemployment or sickness insurance. Therefore, when analysing the determinants of public outlays on families, the only variable that reflects the extent of legislated social provisions is the replacement level of family related allowances (allowances/net average industrial wage). The structural need variable pertains to the proportion of children below 16 years of age in the population. Table 6.5 shows that the benefit generosity is clearly the most decisive explanatory facor in the cross-sections and the variance explained is very high at the beginning of the period, but then it begins to decrease towards the year 2000. Interestingly and somewhat surprisingly, the need variable has statistically significant impacts on expenditures only in a very few crosssections. The quality of provisions paid seems to be the most important explanatory factor for cross-country differences over time and in explaining cross-sectional differences.

DISCUSSION There has been a substantial variation in the level of social expenditure among the 18 most advanced OECD countries over time. The results

123

df R2 adjusted

Constant 0–16/total population Quality of allowances


16 0.742

0.196 ns 8.542***

1960

15 0.796

1.142 0.504* 10.625***

1970

16 0.485

0.241 ns 9.348**

1980

16 0.466

0.379 ns 8.540***

1990

16 0.323

0.421 ns 9.005**

2000

160 0.642

0.245 ns 9.039***

1960–2000

Table 6.5 Regressions of spending on child allowances on aspects of social rights and structural factors, 1960–2000, unstandardized coefficients (ns not significant)

124


presented in this study suggest that rights and needs both make a contribution in explaining the variation in overall social expenditure levels. The results from the analysis of individual insurance programmes show that the relative importance of structural needs and social rights related variables differs. In the area of unemployment benefits, the unemployment rate has clearly become the most important factor more recently. Hence, levels of unemployment benefit expenditures have become more needs driven than in the past. Also, regarding old-age pensions, the age structure of the population is an important determinant of cross-country differences in expenditures, while the structural variables seem to play a less important role in explaining differences in expenditures on sickness insurance – this partly reflects the fact that it is hard to find precise health indicators which could be used as needs proxies for this benefit type. The relative importance of different dimensions of social rights also varies among programmes, and for different points in time. As a rule, the replacement level is significantly and positively associated with spending levels, particularly in family benefits and pensions (in pooled time series analyses). The duration of the benefit period plays a marked role for sickness expenditures and, since the mid 1970s, also for unemployment provisions, reflecting the increase in long-term unemployment. With the possible exception of unemployment, the importance of coverage is marginal for spending levels. The length of the waiting period tends to dampen expenditures on sickness insurance, but the impact is weaker. Our review of the existing statistical data sets indicated that in some cases, there are remarkable inconsistencies in data that greatly contaminate the results achieved. Hence, the choice of a database partially dictates the results. This suggests that, especially if we are operating with disaggregated data, we must consider extremely carefully whether our observations on the dependent variable are comparable over time and across countries. Although various structural influences can be seen to be beyond political control, this is not necessarily always the case: family allowances may affect the birth rate; improving living standards for the elderly by providing public pensions may increase life expectancy and hence the size of the elderly population; and the unemployment rate may be affected by national political strategies, to give just a few examples. The case of unemployment insurance, in particular, illuminates the complex interplay between political decision making, structural needs, and social spending. A number of studies on the political determinants of unemployment rates have shown that rates are lower in countries with left-wing dominance (Korpi, 1991), whereas some later developments in the unemployment rates (particularly the increases in unemployment in Sweden and Finland) may challenge this view. Our study shows that the most important single

125


Social rights .606***

Cumulative share of confessional parties in the cabinet 1960–2000

.528** .854*** .545**

Social spending .974*** Cumulative share of leftwing parties in the cabinet 1960–2000

Life expectancy

.338**

.522*** The proportion of 65+ .486*** Unemployment rate

Figure 6.3 Politics, social rights and social spending in 18 OECD countries, 1960–2000 factor affecting spending on unemployment provisions is precisely the unemployment rate. Therefore, if the links between partisan politics and unemployment are not clear-cut, it is not all that surprising that analyses of the relative significance of political mobilization for expenditure levels tend to produce inconclusive results. This suggests that while structural factors in previous periods of welfare state development might have been largely exogenous, there is evidence to suggest that they, in some areas, have become at least partly endogenous. This is illustrated in Figure 6.3 which shows the connections between politics, social rights, structural needs and spending on social security. In principle there are a number of ways to oprerationalize political variables. Firstly, we could use cabinet seats occupied by each political block (in our cases left-wing parties, denominational parties and secular non-socialist parties); for example, consider the seats for the preceding five-year period of each year

126


of observation, or secondly we can construct cumulative scores to catch the long-term incumbencies of political power relations (Huber and Stephens, 2000: 329).13 The five-year averages would capture the immediate (or short-term) effects of policy making, whereas the cumulative scores capture the impact of longer political hegemony. As most social policy schemes have their own inbuilt inertia we decided to follow Huber and Stephens (2000) and concentrate on the long-term effects and use the cumulative index of political power. In a pooled regression model, where social expenditure is regressed on all the dependent variables and where political variables pertain to the accumulated power, there seems to be some impact from confessional parties on spending, while left-wing dominance has a stronger tendency to increase spending. Both left-wing power (highly significant) and the relative strength of confessional parties (significant) are positively linked to the generosity of social rights. The demographic factors are not related to politics, and their impact on spending levels, when the relative roles of the other factors are controlled for, is highly significant. Thus, all in all, in the ‘end product’ – social security spending rate as a percentage of GDP – the impact of these different factors is blurred in such a way that the relative importance of politics is more difficult to establish than that of social rights, which are the objects of political decision making. It is the task of future studies to analyse in more detail how political and structural approaches can unravel the interplay between social rights and various factors of need. However, this study indicates that previous contradictory results regarding the relative importance of policy related factors for levels of social spending can partly be explained by the use of poor data and the lack of attention to the independent effects of structural ‘need’ variables. Our discussion also highlights the fact that the use of different statistical sources can lead to different results and conclusions. To further complicate the picture, Richard Titmuss’ (1974) classical division between social, occupational and fiscal welfare is worth keeping in mind. Social welfare pertains to legislated social policy. Most official statistics, such as published by the ILO or OECD, primarily contain data on this aspect of social policy. In addition, our definition of social rights is based on legislated ‘social’ welfare. Although this aspect is the most important part of income maintenance, it is not the only one and in some cases the concentration on legal rights to public provision gives a distorted picture of welfare state realities. In many countries with flat-rate or low-level earnings related benefits, occupational schemes have mushroomed to complement the gaps in compensating income loss. Danish and Dutch pensions are good examples in this respect (see also Jochem, Chapter 12 in this


127

volume). In both countries the statutory pensions are not tied to the claimant’s previous income. Therefore, a huge occupational and fiscal/ individual pension sector has evolved. The same pattern is visible in sickness benefits where employers guarantee sick-pay provisions on various scales. There have been a number of attempts to measure the total level of social spending (e.g. Adema, 1999). These measurements accentuate specific points that have important ramifications for exercises like ours. Firstly, differences in total public and mandatory private spending are much smaller than differences in public spending on welfare. Secondly, it makes a significant difference whether we use gross spending figures – as we did here – or net figures. Usually big spenders, such as Denmark and Sweden, have big taxes; as a result, there is a claw-back effect whereby the gross spending of 30 per cent of GDP or so is reduced to 20–25 per cent, which is not drastically more than what small spenders de facto spend on their total social budgets. Changes in the taxation of social benefits may have important consequences for our understanding of the developmental patterns in social rights. For example, the generosity indicators of public pensions showed a downward trend from the 1980s. This trend has been taken as an indication of the demise of pension rights. It is important to separate declining replacement rates caused by real or nominal cuts in benefits as such, from those that are due to changes in taxation. In some cases absolute improvements in pension benefits do not result in increased replacement levels but rather the opposite. It is also the case that a number of countries, among them Sweden, recorded increased replacement rates in the early 1980s as a result of poor real-wage developments, which, with the then existing benefit formulas, meant that earnings in earlier years boosted replacement rates calculated in relation to the final wage. And when real wages later started to grow, this was ceteris paribus translated into declining replacement rates. Thus, if we look at the totality of social spending or welfare efforts, the big-spending ‘social-democratic’ welfare states are perhaps not as superior at providing social security as they are usually believed to be, and the lowspending ‘liberal’ countries are perhaps not ‘as bad’ as usually argued. Consequently, the social-democratic welfare states are not as expensive, nor the liberal welfare states as cheap, as often claimed. At the end of the day, the advanced rich countries seem to use roughly the same amount of their GDP on the three forms of welfare. In spite of similar spending levels, distributional consequences are greatly divergent (see e.g. Kangas and Palme, 2000). Therefore, the measurement and analysis of different aspects of social rights – preferably supplemented by analyses on occupational welfare – offer very valuable tools for understanding the qualitative differences in public and private provisions of social welfare.

128


NOTES 1.

2.

3.

4. 5.

6.

7.

Statistical concepts are not comparable among countries, as data, to a great extent, follow country specific institutional arrangements. Nor are data reliable over time within single countries; in some cases methods of measurement undergo substantial changes from one observation point to the next. Despite the evident awareness among comparativists of the deficiencies in available statistics provided by international organizations such as the ILO and the OECD, surprisingly little as been done to reduce the measurement error. An additional problem is the distinction between gross and net spending (Adema, 1999); for example, the Nordic countries may be big gross spenders but, since they also tax social transfers, their net social spending is not always higher than it is in other European countries. The SCIP database thus excludes all labour market based collective and individual insurance programmes. Hence, in social protection areas where collective schemes are common – as with collective pensions and sickness benefits in various forms of wage continuation – our measures do not give a full picture of social protection. Therefore, the level of ‘real’ social protection for countries that rely on ‘social policy by other means’ (Castles, 1989) in their social security production is underestimated. This especially applies to the Antipodean countries and also to many other countries. For example, legislated pension rights and sickness benefits are much more generous in Sweden than in Denmark. However, if we take into consideration that most Danish employees are covered by lavish occupational pensions schemes and sick pay (100 per cent of income), the differences between these Scandinavian countries will disappear, or Denmark may display higher overall social security levels than Sweden. Thus, the picture we are able to draw here is fragmentary and by no means complete. Currently, we neither have sufficient data on private individual and occupational social benefits nor on the spending on these items. However, the statutory pension age varies across nations, and due to flexible retirement schemes the average retirement age has been falling in most of the OECD countries (OECD, 1988: 77–82). In many countries early retirement takes up a considerable part of total pension spending. The age limits also vary in family allowance programmes (see e.g. US Department of Health and Human Services, 2005). Therefore, the measures used in this study must be regarded as more or less accurate proxies of the demographic pressures affecting spending rates. Historical Statistics (various years), Labour Force Statistics (various years), and for life expectancy: Financing and Delivering Health Care (OECD, 1987) and www.mortality. org. In addition, decomposing expenditures into separate variables of different insurance programmes on the basis of the Flora data is not always straightforward. The reason is that the compilation of data has followed country specific institutional variation rather than ‘a variable approach’ where the national figures have been standardized to make them comparable across nations. Figures provided in the Flora database were used to cross-check the reliability of the other statistical sources in the early 1980s. In the 1980s the correlation coefficients for total social spending were .77 for Flora-ILO; .82 for Flora-OECD, and .89 for ILOOECD. In the case of pensions, correlations were as follows: Flora-ILO: .52; FloraOECD: .86; ILO-OECD: .64. The corresponding figures for unemployment were .92; .92 and .93; for sickness .76, .82 and .71; and for child benefits .93, .95 and .73. In previous analyses (e.g. Saunders and Klau, 1985; Flora, 1986a; 1986b; 1987a; Alestalo and Uusitalo, 1992), determinants of growth in individual programmes have been broken down into three broad categories: (1) the demographic ratio, which refers to changes in the proportional size of the population that is in principle entitled to benefits; (2) the coverage ratio, which denotes to what extent the potential beneficiaries are actually in receipt of allowances; and (3) the transfer ratio, which pertains to the increases in benefit levels. We have data on (1) and (3) but not (2), which made it inappropriate to


8.

9. 10.

11.

12.

13.

129

apply their approach. Moreover, we have an underlying purpose for using the regression technique for testing other kinds of hypotheses in another context. The quality of unemployment insurance and sickness insurance refers to the sum of the replacement rates plus coverage. The quality of child allowances refers to the replacement level, while the quality of old-age pensions consists of coverage, basic security, and income security (cf. Palme, 1990). The compositions of the indices are defined in more detail in the main text. Z-scores rather than raw data were applied in order to avoid problems caused by differences in scales used to measure various aspects of social rights. The variable was significant (sig..004) only in pooled regessions without autocorrelation assumptions. Coefficients for all the other variables were almost the same, whereas there were some fluctuations in significance. However, the interpretation is precisely the same. Basic security refers to the pensions that are guaranteed for elderly people without work records or with only a minimim contribution record. In a way it is the level of the minimum safety net guaranteed in a country for a pensioner without other means (Palme, 1990: 34). Income security is the sum of three components: worker pension, full pension and maximum pension. The quality of the worker pension has been calculated assuming that the claimant has worked 35 years on an average industrial worker’s wage. The full pension level is calculated in the same way as the worker pension, only now the claimant is assumed to have worked the maximum number of contribution years required in the qualifying conditions. The maximum benefit level has been calculated assuming that the pension recipient has worked for the maximum contribution period, and has an income at the maximum level taken into account for benefit purposes (though no more than 1.25 times the average industrial worker’s wage in cases where no ceiling exists – Finland and Italy in a few of the years). Period averages for cabinet participation were calculated as follows: when we explain for example social rights for 1960, our political variables pertain to average cabinet shares of left-wing, denominational, secular, non-socialist and green parties in government during the period 1956–60; respectively, politics during 1961–65 is used in the regressions for the levels in 1965. The cumulative scores for cabinet participation are the sums of cabinet shares held by various political blocks since 1946. The score varies from 0 to 11. For example, ‘0’ for left parties indicates that the party block has never had a position in the government, whereas ‘11’ indicates that all of the cabinet chairs in all of the years were occupied by ministers belonging to left-wing parties. In practice the left-wing score varies from 0 in the US to the highest value of 8.81 in Sweden. The interpretation of index values for other political blocks is the same.

PART III

Beyond spending: welfare state generosity, social rights and obligations

7. Welfare state generosity across space and time Lyle Scruggs INTRODUCTION Comparative analyses of welfare state reforms have relied overwhelmingly on public spending data as the indicator of welfare state commitment and change. However, scholars have long emphasized the problems with spending outputs and also stressed the importance of programmatic elements of welfare state policies. One particular focus has been on national commitment to social citizenship rights. This general line of research has offered few alternative measures that are compatible with comparative analysis across many countries and long periods of time. Those that do exist, such as the Social Citizenship Indicators project at the Swedish Institute of Social Research, are not generally available to scholars (see Kangas and Palme, Chapter 6, this volume). This has led to two reinforcing cleavages in welfare state research: largen comparisons of many countries testing general theories but relying on spending data as proxies of welfare state generosity (or effort), and smalln comparisons with sui generis data sets that propose and demonstrate, but seldom really test, hypotheses. Both sides of the cleavage are subject to specific scientific limitations and advantages. For its part, the ‘large-n camp’ focuses on honing a well developed (if not uniform) set of empirical data to test analytical models and theories, but may be consistently misspecifying what the welfare state is. The more qualitative camp, by contrast, can often claim a more nuanced set of concepts, but lacks quantifiable leverage for scientific verification, or, in the case of single country studies, any direct comparisons. Qualitative researchers are probably right about aspects of conceptual validity, but the quantitative approach is probably right about the requirements for theory validation. By both being half right, both approaches cannot help but be half wrong. Moreover, neglected in both approaches is the construction of valid empirical measures of concepts that are to be tested. Qualitative research generally fails to get beyond defining the concept and applying it to a case 133

134

State generosity, social rights and obligations

or two, thus failing to direct research to comparable, systematic data collection. Quantitative research often fails to get beyond the technical aspects of the empirical analysis, thus failing to ask whether the most used, or available, data sets are worth analysing in the first place (see also Chapter 2 by Green-Pedersen in this volume). Collecting comparable data on non-spending aspects on welfare state programmes is an important step in bridging this gap. Such data can add fundamentally to the evaluation of changes in welfare state policy. One of the main goals of this chapter is to introduce and discuss one data set, the Comparative Welfare Entitlements Dataset (CWED), which details changes in the major components of key social insurance programmes in 18 countries from the 1970s to 2002. The other major goal of the chapter is to suggest ways in which data collection efforts on ‘qualitative’ dimensions of the welfare state should be augmented in order to promote the development of more general theories about welfare state development and change. The rest of the chapter is divided into five sections. The next section addresses why cross-national and historical data sets of social entitlements are necessary in order to think theoretically about welfare state change. This section also critiques the widespread use of comparative spending data. The third section presents the CWED data on trends in social insurance programmes. The fourth section develops a benefit generosity index, based on a refinement of Esping-Andersen’s popular decommodification index, and using the data presented in the third section. The evidence in the third and fourth suggests that welfare states became more generous during the 1970s and 1980s, but experienced some considerable retrenchment in the 1980s and beyond. That retrenchment has been greatest in the more generous countries. The fifth section acknowledges that the CWED data set is incomplete, discusses possibilities for expanding it and, reinforces the central theme of this volume: the importance of reliable and systematic data in developing and evaluating explanations of welfare state development and change.

CROSS-SECTIONAL, TIME SERIES QUANTITATIVE MEASURES OF BENEFIT GENEROSITY AND WELFARE REFORM Time and Space in Theories of Reform Any comparative empirical analysis of reform requires detailed data in both space and time.1 The importance of a spatial dimension in comparative

Welfare state generosity across space and time

135

public policy research almost goes without saying: comparative analysis requires multiple objects of comparison. Nevertheless, many ‘comparative’ studies focus one country, or even one programme in one country. The value of this latter type of approach, when nested among other case studies that isolate similar programmes in other countries, and investigate common causal processes, is beyond dispute. Most of the time, however, case studies (or other small-n comparisons) are undertaken with only loose attention to comparability with other studies. So case studies often just accumulate. Attempts to compare findings from such studies exacerbate a classical data problem in social science. Tests of social theories are usually based on data collected for purposes other than testing social theories; theory evaluation with ‘data’ comprised of a hodgepodge of accumulated case studies further compounds the problem.2 More qualitatively oriented researchers do have a point when suggesting that their work can avoid a fundamental social science data problem: it can derive and collect the appropriate data to test a theory. Because analytical comparisons (particularly quantitative comparisons) must assume that their data constitute the correct measurements of the underlying phenomena they are evaluating, ‘qualitative research’ is essential to the comparative evaluation of theories. Anyone with doubts about how critical correct measurement of concepts is for statistical inference should review contemporary texts in econometrics (e.g. Gujarati, 2003). So there are really two problems in comparison. One is using invalid conceptual measures, the second is ensuring that you have consistent measurement across units. Tackling these two problems is important in developing a ‘large-n’ data set to study any question. In many comparisons, this probably does require a set of heavily structured ‘case studies’, demanding detailed knowledge about specific features of individual units. But this has to be undertaken systematically in order to maximize comparability across the cases. Such work is surprisingly rare in social policy research. Even for topics like comparing national incomes or price levels – big topics with big budgets – there are surprisingly few. In social policy (and many other questions in comparative politics) there are precious few data sets that really attempt to do this in order to answer research questions. The importance of time when comparing welfare state reforms is also obvious, yet absent from much social policy work. Identifying a reform requires at least two time points. Korpi’s (1989) discussion of social rights and welfare development, for example, is quite explicit in its call for considering welfare state developments in time. And, as historically oriented researchers often point out, theories of macrosocial phenomena imply a lot of historical inertia with change in the presence of much noise. But precisely because there is so much noise in historical processes, social scientists

136


must look at many similar histories in order to be able to verify empirical regularities. The data demanded to evaluate historical path dependence expand exponentially if we want to have much confidence in any claim (de Marchi, 2006: 49–53). History makes information across both time and space (lots of it) vital. Comparative Social Spending The preceding discussion is perhaps too abstract. Knowing that information across time and space is essential to understanding reform tells us nothing about the content. There are, after all, a few widely used crosssectional time series indicators of welfare states available, and they are widely used in the literature. Why are they inadequate? Why is a measure of welfare programme generosity or social rights necessary? The most widely used series of data on welfare reform are probably national time series of social spending. Probably the most popular measures are general government spending, transfer spending and consumption spending (Garrett and Mitchell, 2001). Such measures of spending enjoy a long history in the comparative analysis of social policy and political economy (Wilensky and Lebeaux, 1958; Wilensky, 1975; Cameron, 1978; Korpi, 1983; Garrett, 1998; Swank, 2002b). More recently, the OECD’s social expenditure database represents a comprehensive effort to make comparable specific categories of social spending (Castles, 2002; Lindert, 2004; cf. De Deken and Kittel, Chapter 5, this volume).3 While it is true that reliance on this spending data is prominent in statistical analyses, many qualitative accounts also utilize a great deal of spending based quantitative information. Undoubtedly, researchers rely on spending measures because they are available, (relatively) comparable, and vary across countries and time.4 Yet researchers widely acknowledge that spending data have significant drawbacks, in particular in evaluating welfare retrenchment (Gilbert and Moon, 1988; Esping-Andersen, 1990; Castles and Mitchell, 1993; Clayton and Pontusson, 1998; Goodin et al., 1999). As a guide to understanding the impact of the welfare state on individual life-chances, for example, total spending reveals little about who benefits. As Esping-Andersen (1990: 21) famously remarked, ‘it is difficult to imagine that anyone struggled for spending per se.’ Several other shortcomings of spending data are also worth considering. Welfare Dependency Structure Spending measures do not typically account for the size of the dependent population (but see the contribution of Kangas and Palme, Chapter 6 of


137

this volume). This is a problem in cross-national comparisons as well as in historical accounts of welfare state commitments. To take the example cited by Esping-Andersen (1990: 20), expenditures on unemployment benefits in the United Kingdom grew sharply in the 1980s under Thatcher, while replacement rates were sharply cut. Rising expenditure was a product of unemployment rising faster than spending cuts. The old-age ratio is perhaps even more relevant in terms of overall social spending. Every OECD country, except Ireland, experienced considerable growth in the over 65 to working-age ratio in recent years. Faster growth in the age ratio implies higher aggregate spending growth and lower aggregate GDP growth, ceteris paribus. Indeed, those studies that have adjusted spending for some indicator of the dependent population are less sanguine about the future generosity of the welfare state; though they often remain sanguine about its overall size (Clayton and Pontusson, 1998, Siegel, 2002; Castles, 2004; Lindert, 2004). Different Rates and Sources of Economic Growth and Tax Structures Another problem with most spending ratio data is that differences in economic growth rates lead spending ratios (spending/GDP) to diverge from real spending itself. Public spending ratios thus underestimate or overestimate inflation adjusted welfare expansion. This effect may be biased towards lowering spending ratios in less developed countries (Abramowitz, 1984). Differences in the tax treatment of transfers (either due to special credits and exemptions, or simply to different tax rates) can distort the degree to which spending ratios translate into different real levels of state commitment to welfare. The role of taxation as an avenue for social transfers has been given more attention in recent years, but while the tax system is increasingly being used as a transfer mechanism – the United States’ Earned Income Tax Credit and the United Kingdom’s Working Families Tax Credit being notable examples – it can also be used to claw back social spending (Howard, 1997). Adema and Ladaique (2005) discuss extensively how gross expenditure can be offset considerably by the tax treatment of transfers (e.g. making benefits taxable), or by increasing consumption taxes. Ireland and Sweden Compared Based on various spending ratios, Ireland has experienced much greater welfare retrenchment after 1980 than has Sweden. The total spending/ social expenditure ratio between 1980 and 2002 in Ireland fell from 47/17

138


per cent of GDP to 29/14 per cent. In Sweden the total spending ratio fell from 58 to 53 per cent of GDP, and the figure for social expenditure was stable at around 29 per cent of GDP. Important changes in demographics, economic performance, and tax structure suggest just how distorted spending data may be as an indicator of generosity. First, in Ireland the dependency ratio dropped dramatically, while it increased dramatically in Sweden. The share of the Irish population over 65 was unchanged over the period 1980–2002 (at 11 per cent), while it grew slightly in Sweden (from 16 to 17 per cent). The share of under 15s in the population, moreover, declined from around 30 per cent to around 20 per cent in Ireland, and was largely unchanged in Sweden (around 19 per cent). Meanwhile, the employment rates of the two countries converged dramatically. In Ireland, non-employment as a share of the 15-64 population decreased from around 43 per cent to 34 per cent between the early 1980s and 2002, while in Sweden it increased from 21 per cent to 27 per cent. Thus, just adding these percentages together (admittedly a very rough gauge), Ireland went from experiencing an 84 per cent (!) non-employment rate in 1980 to 65 per cent non-employed in 2002, while in Sweden this proportion rose from 56 per cent to 63 per cent. So, for a constant share of GDP per non-employed person (again a crude indicator), we would expect the Irish spending ratio to fall by about one-third, while we would expect the Swedish ratio to rise by around 12 per cent. What we observe is a 38 per cent drop in total spending in Ireland (18 per cent social) and an 8 per cent drop in total (nil social) spending in Sweden. Second, the difference in economic performance between the two countries during this period is also very large. The Irish economy grew at a real rate of 5 per cent per year in this period, while the Swedish economy grew at around 2 per cent per year. This partly follows from the demographic and employment trends cited above – a rising share of engaged labour should increase the economic growth rate – but growth has its own implications for what one might call fiscal capabilities. Compared to 1980, real public spending in Ireland was about 180 per cent higher in 2002, while it was just 140 per cent higher over the same period in Sweden. Hence, real spending actually grew twice as fast in Ireland, but since the economy grew about 2.5 times faster, the spending ratio declined. Finally, the evolution of the tax burden on welfare benefits differs somewhat across the two regimes. Effective income tax rates and VAT rates are lower in Ireland, and those rates have declined since 1980 on lower earners (which is assumed to include benefit recipients). In Sweden, income tax rates on earners have also declined, though the overall system is somewhat less progressive than in the early 1980s. While VAT was reduced from 25 per cent to 21 per cent in Ireland, it has increased slightly (from 23.5 per cent


139

to 25 per cent) in Sweden. Reduced or zero VAT rates apply to many more necessities (e.g. food, social housing, utilities, public transit) in Ireland than Sweden (European Commission, 2005b). Adema and Ladaique’s (2005) analysis seems to suggest that net spending in Sweden is considerably lower than gross spending due to tax clawbacks, while there are fewer clawbacks in Ireland. The point of this example is not to suggest that Ireland has or will become a more generous welfare state than Sweden. It is worth noting that the gap between gross and net public transfer spending in the two countries has closed, particularly in Ireland, which may suggest some reduced generosity. These facts do suggest, however, that very large biases may be hidden in an indicator that is so often implicitly accepted as synonymous with the generosity of the welfare state. From Welfare Spending to Welfare Rights An alternative way to evaluate welfare commitments is an entitlement, or social rights, approach. This is essentially the approach advocated by Korpi (1989), promoted in Esping-Andersen’s Three Worlds of Welfare Capitalism, and which has been widely embraced in comparative social policy (Hicks, 1999; Kitschelt, 2001; Castles, 2002; Green-Pedersen and Haverland, 2002; Armingeon and Griger, 2006; Kenworthy, 2007). The approach is critical of spending as too unidimensional. However, it is still compatible with systematic data collection on more qualitative dimensions of welfare policies. The essence of the rights or entitlements approach is its consideration of the generosity and universalism of social insurance. One can think of social insurance as a political commitment which alters the balance of power between segments of the population or economic classes. Such commitments have behavioural effects that extend beyond government outlays at any point in time. This is an important reason for presuming that spending captures only a small part of how welfare policy structures political outcomes. Welfare state generosity could thus be considered more important than spending when trying to understand the political economy of the welfare state. Viewing social insurance as a commitment is also useful for understanding more traditional questions in the sociology and economics of labour markets. Generosity of welfare commitments may affect things like quit rates, unemployment duration, labour market matching, and other micro-level phenomena that the level of spending will not. There are various sources of data which systematically compare welfare state commitments at some level. However, all contain some important shortcomings compared with the approach in CWED. First, the OECD has

140


collected data on unemployment benefit programmes, starting with its Jobs Study in the mid 1990s (OECD, 1994). One of its data sets contains estimates of unemployment insurance replacement rates for a typical worker under a variety of scenarios. These data have been used in a number of large-n empirical studies of welfare retrenchment (e.g. Huber and Stephens, 2001). A major shortcoming of this data set is that the replacement rates ignore income and social security taxes, which considerably distorts comparative generosity. This data set also does not take into account important features of benefits, like population coverage or the qualifying conditions. More recently, the OECD has compiled net-of-tax replacement rate data, but these are available only for the mid to late 1990s onwards (OECD, 2005b). These data are also limited to unemployment insurance only, and thus provide a limited, albeit economically important, perspective on welfare insurance programmes.5 Other sources of replacement rate data are scattered in terms of countries and coverage. Hansen’s Elements of Social Security (2002) includes other programmes (e.g. work injury) and income profiles, but only covers a few countries and only since the early to mid 1990s. The Nordic Social Statistical Committee publishes Social Protection in the Nordic Countries but this only covers relatively recent years, and only for the Nordic countries. A comparative study of old-age pension replacement rates was provided for the late 1980s by the European Commission (1993). Bambra (2005) also produces an index of ‘decommodification’, circa 1998, that incorporates recent data on replacement rates and qualifying conditions plus health data. Finally, SOFI’s Social Citizenship Indicators Project covers the same programmes as CWED, plus work injury programmes. It is more historical, going back to the 1930s, but contains data only for every fifth year. Most importantly, however the data are not available to the scholarly community, despite the fact that they have been in existence for more than two decades. The Comparative Welfare Entitlements Dataset, in contrast to these, contains annual information on replacement rates for unemployment, sickness and public pensions, as well as other eligibility criteria and the size of the insured population. (Plans for the future are to extend the data to cover child care and maternity benefits.) It currently covers 18 countries: Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Ireland, Italy, Japan, the Netherlands, New Zealand, Norway, Sweden, Switzerland, the United Kingdom, and the United States, from around 1970 to 2002. Additional information for most countries is available for the 1960s, except for net replacement rates. The data are collected annually, although some extrapolations from bi-annual data have been made where annual data are not present.6 Table 7.1 provides a more specific breakdown of the particular characteristics in the data set that are used in the rest of the chapter.

141

After tax replacement rate at retirement for single with no work history (or income) After tax replacement rate at retirement for couple with no work history (or income) After tax replacement rate for a single with a full work history (max 45 years) at APW wage

Single replacement rate Family replacement rate Qualifying period Waiting days Duration of benefit Coverage ratio

Minimum replacement rate (single)

Minimum replacement rate (couple)

Standard replacement rate (single)

Sickness benefit

Retirement pension

See definitions under unemployment insurance

After-tax benefit for a family of four (one APW earner, non-working spouse and two children) divided by after tax wage of employed APW Weeks of insurance/employment required to qualify for benefit Number of days before benefits start Weeks of benefits payable for fully insured (single) 40 year old Percentage of the labour force covered by unemployment insurance

Family replacement rate

Qualifying period Waiting days Duration of benefit Coverage ratio

After tax benefit for single, fully insured 40-year old earning average production worker (APW) wage divided by after tax wage of employed APW

Single replacement rate

Unemployment

Definition

Programme characteristic

Dimensions of the decommodification/generosity index

Core programme

Table 7.1

142

Definition After tax replacement rate for a couple with one full work history earner and spouse without work history Years of insurance needed to qualify for single standard pension (defined above) Employee/Employer Employee ratio of payroll taxes (at time pension is claimed) Portion of population above retirement age receiving pension

Programme characteristic

Standard replacement rate (couple)

Qualifying period

Contribution ratio

Take-up rate

(continued )

Core programme

Table 7.1


143

The indicators included in the data set are combined to form indices following an aggregation procedure discussed later in the chapter. Conceptually, the measure is very similar to Esping-Andersen’s decommodification index, but also contains some important modifications to the scoring which are detailed below.7

MEASURING WELFARE STATE GENEROSITY This section outlines the basic approach to measuring generosity in CWED by discussing specific elements outlined in Table 7.1. Programme replacement rates One of the most popular institutional indicators of the generosity of programme benefits is the income replacement rate, that is, the portion of income replaced by a social welfare programme (Esping-Andersen, 1990; cf. Whiteford, 1995; Korpi and Palme, 2003; Allan and Scruggs, 2004; OECD, 2004). Replacement rates provide some measure of the level of well being that is compensated by income transfer programmes. CWED defines replacement rates as (Cash BenefitsIncome Taxes)out of work/(WagesIncome Taxes)in work where income taxes include net social charges (compulsory contributions to social insurance programmes less cash transfers), and APW refers to the ‘average production worker’.8 Since most working people in OECD countries live in families, CWED provides benefit replacement rates for a single worker and for a family, the latter defined as a household with a dependent spouse, two children and a head of household drawing the specified benefit. Benefits for families include child benefits, including means tested benefits. Unemployment replacement rates Table 7.2 shows the development of replacement rates for unemployment benefits. The family types are (a) a fully insured single APW wage and (b) a fully insured head of household as described above. The worker receiving the benefit is assumed to be 40 years old, with 20 years of social insurance contributions. The table organizes countries based on (more or less) their traditional classification into Esping-Andersen’s Three Worlds typology (cf. Scruggs and Allan, 2006a). In the 1970s and into the 1980s, there is a great deal of diversity across countries, not only for the level of the replacement rates, but also with respect to long-term trends. After the 1980s, replacement rates tend to be stable or decline as a share of APW wages. It is notable that the traditional Three Worlds perspective is not a great predictor of replacement rates (see Hu et al., 2006). There is considerable variation within regime types (calling into question the idea of regimes).

144


Table 7.2

Unemployment benefit replacement rates Single replacement rates

Family replacement rates

1971

1980

1990

2002

1971

1980

1990

2002

21 32 24 26 55 66 69 25

26 60 60 31 46 69 69 69

30 66 35 34 20 58 57 72

26 60 29 26 18 58 64 72

40 44 47 53 72 59 68 38

52 64 85 62 63 61 65 83

66 70 64 72 36 60 53 82

65 72 58 57 55 55 61 82

40

54

47

44

53

67

63

63

52 53 43 63 10

58 67 68 68 4

58 64 70 63 20

55 66 70 60 45

61 63 35 78 20

74 67 61 70 13

72 60 63 70 29

67 61 73 72 61

Mean Social democratic Denmark Finland Netherlands Norway Sweden

44

53

55

59

51

57

59

67

87 40 87 52 75

78 34 86 70 82

68 63 74 68 85

59 57 78 65 75

90 48 94 68 82

81 48 89 75 85

73 73 78 73 81

64 67 77 72 78

Mean

68

70

72

67

77

75

75

72

Liberal Australia Canada Ireland New Zealand United Kingdom United States Japan Switzerland Mean Conservative Austria Belgium France Germany Italy

For the single-worker replacement rate, trends for the conservative and social-democratic welfare states converge, while both diverge from the liberal regime average.9 For family replacement rates, there is convergence among all three regimes, with considerable declines in the social-democratic countries in recent years, small declines in the liberal countries, and an increase in income replacement in the conservative countries (mostly due to Italy). Sick pay replacement rates Table 7.3 shows trends in sick pay replacement rates for singles and families which correspond to those provided in the previous paragraph.10 A number of countries, especially among the socialdemocratic and liberal welfare regimes, have sickness replacement rates similar to those for the unemployed. However, it is generally true that the liberal welfare states tend to have lower sickness replacement rates than the

145


Table 7.3

Sickness benefit replacement rates Single replacement rates


Family replacement rates

1971

1980

1990

2002

1971

1980

1990

2002

21 0 24 26 55 0 50 80

28 60 60 36 46 0 52 83

30 66 35 39 28 0 57 81

26 60 29 26 22 0 63 79

40 0 47 53 72 0 47 82

56 64 85 62 63 0 48 83

66 70 64 72 32 0 53 81

65 72 58 57 26 0 60 79

32

46

42

38

42

58

55

52

71 66 58 100 74

75 88 59 100 68

78 89 63 100 74

79 85 62 92 77

90 66 61 100 80

88 84 61 100 72

89 90 66 100 80

86 88 63 93 87


74

78

81

79

79

81

85

84

65 68 87 52 84

78 41 86 100 97

68 88 74 100 84

59 72 78 100 82

70 83 94 68 84

81 54 89 100 97

73 88 78 100 86

64 75 77 100 84

Mean

71

80

83

78

80

84

85

80

other countries, particularly if we count Japan, as Scruggs and Allan (2006a) and Bambra (2005) suggest, as a liberal country. While the majority of other countries have higher replacement rates for sick pay (e.g. Austria, Belgium, Finland, Germany, Italy, Norway, Sweden and Switzerland), a few have higher unemployment replacement rates. Over time, trends are also similar to trends for unemployment replacement rates. There are a few cases where benefits increased in the 1970s (Canada, Norway, Ireland, Finland), but since the mid to late 1980s, declining replacement rates dominate. Not surprisingly, patterns for family replacement rates also resemble what we saw for unemployment. Preferential benefits for male breadwinner households are no more predominant in conservative welfare states than they are in the countries in the other two regime types.

146


Table 7.4

Evolution of social pension replacement rates Single replacement rates


Couple replacement rates

1971

1980

1990

2002

1971

1980

1990

2002

24 28 23 6 26 37 31 0

28 30 31 19 40 39 37 40

32 42 36 17 41 40 31 42

30 40 35 23 39 38 37 39

39 49 20 11 43 43 35 0

42 49 36 36 57 44 43 47

49 59 43 32 65 45 36 49

45 57 45 45 59 43 42 45

22

33

35

35

30

44

47

47

47 15 23 18 12

42 27 35 17 19

41 37 45 18 27

50 39 42 18 30

51 19 37 29 24

49 30 60 27 34

47 36 68 26 51

56 39 65 23 53


23

28

33

36

32

40

46

47

47 33 46 39 35

42 42 55 42 43

51 38 48 43 43

44 31 51 44 35

53 48 55 52 45

57 61 67 56 53

61 53 58 56 53

57 48 60 66 43

Mean

40

45

45

41

51

59

56

55

Social pension replacement rates Table 7.4 shows trends in the replacement rates for social pensions. Social pensions here are defined as the pension benefits for someone without a work history. Where there is not an explicit minimum-age pension, we substitute social assistance. We assume that a ‘family’ consists of two people of retirement age with no children at home. The results indicate that social-democratic regimes have generally provided the most generous social pensions. However, the generosity of these benefits has fallen considerably since the 1980s in Finland and Sweden, and, to a lesser extent, in Denmark and the Netherlands. (Norway is the exception: benefits went up in the late 1990s.) In the early 2000s, Canadian and New Zealand’s social pensions replace as large a percentage of work income as do the social pensions in four of five social-democratic countries.


147

Among the conservative regimes, replacement rates do not uniformly support the idea of strictly income linked benefits. While social pensions are low in some countries (Germany, Italy and Belgium), they are among the highest in others (Austria and France, particularly in recent years). Except for Germany, replacement rates have tended to either remain stable or increase. Replacement rates in countries which belong to the liberal regime are also not uniformly low. As mentioned, Canada and New Zealand have comparatively high replacement rates for those relying solely on public pensions. Differences between single’s and couple’s rates are large in most social democratic countries, as well as in Canada, France and the Antipodes. Across regime types, there is convergence in replacement rates. Benefits in conservative and liberal countries have tended to rise toward the socialdemocratic countries. Replacement rates in social-democratic countries have declined since the 1980s. Standard pension replacement rates The last set of replacement rates presented is for standard public pensions (Table 7.5). These benefits are typically based on a long history of earnings, and are computed using rules governing the treatment of past wages, accrual rates, and calculation formulas. We make a common set of assumptions about career earnings. The basic assumptions are that the notional worker retired on 1 April of the year in question, worked a full career (or from age 21 to retirement age), and earned the APW in each year.11 Family benefits refer to a retired couple with one lifetime APW wage beneficiary and a spouse with no work history. Replacement rates and conditions refer only to public pensions, which excludes some pension benefits that are based on sectoral or industrial relations agreements. (For example, supplementary pension regimes in France are excluded.) With a few exceptions, such as the United States, Italy, and Norway, public pension systems fully matured in the 1980s. Since then, replacement rates have either declined or levelled off. Conservative countries are typically the most generous. Liberal countries have converged only slightly up toward social-democratic countries. This is due to Japan and Switzerland, which both start the period with very small, public pensions. In the social-democratic regimes, there was general replacement rate growth in Norway and Finland, and some expansion of couple’s pensions, specifically in the Netherlands. In these countries we can observe that first retirees who have paid the expected number of years into second pension pillars have started to be rewarded by higher replacement ratios in the more recent past. Across the three regimes, standard pension replacement rates for singles have not converged much over time. The picture with respect to couples,

148

Table 7.5


Evolution of standard pension replacement rates Single replacement rates

Liberal Australia Canada Ireland New Zealand United Kingdom United States Japan Switzerland Mean Conservative Austria Belgium France Germany Italy Mean Social democratic Denmark Finland Netherlands Norway Sweden Mean

Couple replacement rates

1971

1980

1990

2002

1971

1980

1990

2002

24 33 27 16 26 22 27 44 27

28 40 36 51 40 41 37 56 41

32 54 42 51 41 39 47 57 45

30 52 38 50 39 42 55 60 46

39 54 40 21 43 33 35 59 40

42 58 49 72 57 55 48 74 57

49 69 58 73 65 51 56 78 62

45 67 53 77 59 55 57 77 61

85 76 46 75 69 70

80 81 59 77 61 72

82 80 60 75 72 74

85 74 52 72 93 75

75 79 37 68 64 65

68 82 68 65 55 68

70 77 69 62 69 69

67 72 58 52 81 66

51 40 42 42 50 45

46 57 55 55 64 55

54 67 48 63 65 59

52 63 51 61 59 57

58 42 51 51 61 53

59 57 67 56 68 61

64 64 58 61 68 63

63 64 60 65 61 62

however, is different. There has been considerable convergence. This can be explained by the absence of dependant’s supplements in conservative countries, and an increase in such supplements in liberal and social-democratic countries. Despite its liberal welfare state label, the US Social Security programme is a curious blend of non-liberal attributes. APW replacement rates compare favourably with other countries, the benefit formula is very progressive (those with lower lifetime earnings receive larger replacement rates), and the supplement for a non-insured spouse is 50 per cent of the breadwinner’s pension. (This supplement is much larger than the supplement in any other country.) The main shortcoming of Social Security is that pensions in force have always been indexed to prices rather than wages, something that many pension reforms are moving increasingly towards. The progressive


149

replacement rate formula and earnings ceiling combine to provide a relatively low maximum (compared with many continental systems), which is consistent with company schemes for middle and upper-income households. Across these four programmes and two household types, there are some obvious patterns. First, there is considerable convergence in replacement rates for single-earner families. In general, benefits have converged upwards over the entire period since the 1970s, but with clear evidence of cuts after the mid to late 1980s, particularly in previously high-benefit systems. Whether programmatic cuts constitute true retrenchment is, of course, partly dependent on the definition of retrenchment and the historical time scale used. For example, virtually all programme replacement rates were more generous in 2002 than they were in the early 1970s in terms of expected worker pay. In terms of real purchasing power (adjusted for inflation), they are even more generous. Finally, the results consistently suggest little evidence for the conventional wisdom that social insurance is much more ‘family friendly’ in conservative, continental European countries. Social Insurance Coverage Coverage is considered an important indication of the universalization of benefit eligibility, something that Korpi (1989) argues characterizes mature welfare states. In their seminal article on the development of welfare states, Flora and Alber (1982) included the portion of workers covered by social insurance (see also Flora, 1987b). This subsection discusses coverage trends for unemployment and sickness schemes, and the portion of the population above retirement age in receipt of a public pension. The available data for many countries go back to 1960, so we can assess longer trends than we could when discussing replacement rates. Generally speaking, the results suggest that coverage increased substantially in the 1960s and 1970s, and has not declined since then (with the exception of two aberrant cases). Unemployment and sickness insurance coverage Table 7.6 provides trends in the coverage of unemployment and sickness insurance as a portion of the labour force in 1960 and 2000, as well as changes by decade. A few important notes about these data are worth mentioning. First, coverage for the Antipodean countries is listed as 100 per cent here even though these benefits are means tested and subject to a residency requirement.12 Second, coverage in some countries with Ghent unemployment insurance systems (Denmark, Finland, and Sweden) is based on enrolment in union unemployment funds, even though these are not technically

150

65 29 0.44 55 75 40 59 37

53 15 0.29 35 13 82

Mean sd COV Austria Belgium France Germany Italy

Mean sd COV Denmark Finland Netherlands

100 63 58 31 100 23 81 61

1960

67 11 0.16 83 74 89

86 17 0.20 67 84 58 68 59

100 79 100 50 100 84 86 88

2002

49 61 7

12 10 18 9 22

0 15 42 19 0 61 5 27

1960– 2002

0 23 5

7.0 7 17 12 8

0 1 9 11 0 6 5 4

1960– 1970

27 26 1

7 3 0 4 1

0 16 2 3 0 75 15 17

1970– 1980

10 5 1

1 1 1 2 6

0 3 4 5 0 4 10 3

1980– 1990

Unemployment Change:

12 7 1

1.6 0 2 2 6

0 2 29 0 0 4 5 5

1990– 2002

56 38 0.69 84 83 95 88 66

100 79 100 58 100 18 86 16

2002

78 83 15 11 0.19 0.13 77 99 0 100 82 89

66 36 0.55 75 85 95 82 56

100 0 58 41 100 93 81 16

1960

Evolution of unemployment and sickness insurance coverage

Australia Canada Ireland Japan New Zealand Switzerland United Kingdom United States

Table 7.6

22 100 7

9 –2 0 6 10

0 79 42 17 0 –75 5 1

1960– 2002

8 100 5

4 5 0 3 8

0 0 9 13 0 16 –5 1

1960– 1970

12 0 1

10 5 –2 5 3

0 80 2 2 0 10 15 0

1970– 1980

1 0 1

–3 10 0 –2 –3

0 –3 4 3 0 –25 –10 0

1980– 1990

Sickness Change:

1 0 0

–1 –23 0 1 2

0 2 27 –1 0 –76 6 –1

1990– 2002

151

85 7 0.09

68 31 0.46

45 26 0.59

49 28 0.58

Mean sd COV All countries Mean sd COV

93

Sweden

85

59

36

Norway

34

49

3 19

24 17

3 6

4 8

74

59 36 0.61

66 37 0.57

94

68 35 0.51

96 5 0.05

93

100

26 –1

6 8

20 –1

0 2

0 –10

152


compulsory insurance systems. Third, the measure of ‘coverage’ in the population used here is often higher than the portion of the unemployed receiving benefits. While it does reflect the degree of universalism, it can be criticized for being insensitive to the fact that being in unemployment is not a random draw from the covered population, and those in unemployment may have exhausted insurance benefits or failed to qualify due to incomplete contribution records. Unemployment insurance coverage increased considerably during the 1960s and early 1970s in most countries, and has gradually converged towards full coverage over time. This can be seen in the rising average coverage rates and falling coefficients of variation (standard deviation/mean) in the table. The extension of coverage was typically the result of major legislative changes. Coverage in a number of conservative regimes tended to stagnate after the 1970s. Coverage for sickness cash benefits resembles unemployment insurance coverage in the liberal countries (recall that the two systems have the same or similarly structured replacement rates), except for the United States.13 In the non-liberal countries, coverage is generally higher, sometimes much higher, than unemployment insurance coverage. Finland instituted sickness insurance only in the 1960s, and Canada did so only in 1971. Switzerland and Belgium both saw considerable declines in coverage in the 1990s. The reason for the decline in coverage in Belgium is not clear. In Switzerland, excluding sickness coverage in health insurance policies became easier, and coverage rates (official ones at least) declined considerably. Pension take-up rates Unlike coverage of sickness or unemployment insurance, the indicator of the universalism of the pension system is the portion of people above retirement age drawing a public pension.14 Surprisingly, estimating the portion of those over retirement age who are in receipt of a pension presents considerable challenges in many countries. Most countries report the number of pensions paid in their statistical yearbooks. However, many countries have more than one large public pension scheme, and statistics are often compiled for the number of pensions being paid, not the number of people getting pensions. This problem is considerable in conservative regimes such as Italy and France (and even Germany) which have multiple sectoral pension funds. A second complication in determining pension take-up rates is that some countries have administratively separate pension systems for civil servants. The number of civil servant pensions is published for most of these countries, but civil servants are often allowed to retire early and may be entitled to other public pensions (through different jobs), making it difficult to consistently avoid under- or over-counts. For example, in the United States, federal employees hired since 1984 have been insured in

153


Table 7.7

Pension take-up rates

Australia Austria Belgium Canada Denmark Finland France Germany Ireland Italy Japan Netherlands Norway New Zealand Sweden Switzerland United Kingdom United States

1960

2000

52 59

16 28 n.a. 2 19 3 n.a. n.a. 9

11 19 n.a. 7 2 0 n.a. n.a. 1

21 8 10 3 14 4 n.a. 4 2

32 80 88 71

68 87 101 97 101 100 100 100 96 100 127 107 102 95

95 27 14 24

23 0 15 2

100 100 79

100 100 104

5 0 25

72

93

21

99 82 97

87

Change Change Change Change Change 1960–2000 1960–1970 1970–1980 1980–1990 1990–2000

39 2 4 22

15 0 1 0 2 0 0 5 8 0 3 24 1 2

1 1 11 1 1 1 0 10 2 0 35 1 2 3

3 0 10

2 0 5

1 0 6

3 0 5

18

4

0

1

the Social Security system, and many (but still not all) state and local government pension systems have also been integrated into the national Social Security system. The final complication involves the treatment of spouses. Where second earners have had sufficient labour force participation, they will receive a public pension like any other worker. Where spouses have not worked, they would presumably be entitled to a social pension save for the fact that their household pension income (i.e. from their spouse) is higher than the income test. Alternatively, they may be (indirect) beneficiaries of a pension supplement in the name of their spouse. These complications should generate some caution at the margins when interpreting the results in Table 7.7. We attempted to compile total pensioner information that includes as much information as possible and avoids double counting. In some cases, we rely on information from public surveys that ask about sources of income. Our results suggest that most of the countries considered here attained something approaching universal take-up, once social pensions, public pensions and civil service pensions are all considered. With a couple of exceptions – e.g. Australia moved towards and then away from making its state pension universal – pensions

154


are more universal today than they were 30 or 40 years ago. Most of the coverage expansion occurred prior to 1980. Other Eligibility Conditions Here, briefly, I present and introduce the definitions of the other main conditions of eligibility for social insurance included in CWED. Duration of benefits for sickness and unemployment refers to the number of weeks for which a qualified worker is allowed to collect regular benefits. In some countries, the duration of benefit is a function of the length of insurance coverage and age. The period given in the data refers to benefits available to the ‘notional worker’: age 40, and 20 years of insurance. There are a few cases where the unemployed are entitled to smaller benefits after the full benefit period (e.g. France). Coding refers to the full duration, including the ‘reduced benefit’ period. Qualifying conditions for unemployment and sickness benefits refer to the number of weeks one must be insured to qualify for cash benefits of the specified duration. In some cases, this qualifying period is longer than the minimum insurance period required to receive some benefit.15 Waiting days refers to the number of weekdays after becoming unemployed/sick that one must wait for specified benefits. Thus, for example, it counts the waiting period for earnings related benefits in the UK in the 1970s, even though the (much smaller) flat rate portion of the benefit was payable earlier. The qualifying conditions for pension benefits are based on a slightly different set of criteria than those for unemployment and sickness. First, the qualifying period refers to the number of years of insurance necessary to qualify for the standard pension. This is coded to be consistent with the maximum number of years of actual contributions (assuming a full working life and the age of the public pension scheme), since pension replacement rates are based on what one actually receives based on the programme rules. An example may be helpful. A pension system pays 2 per cent of final salary for each year of coverage up to 40 years of coverage. The pension law is passed in 1960, going into effect in 1961. Someone retiring in 1970 would retire with 18 per cent of final salary (nine years at 2 per cent per year), and has a qualifying period of nine years. Someone retiring in 1971 has a 20 per cent replacement rate and a qualifying period of ten years. Someone retiring in 2002 would have an 80 per cent replacement rate and a maximum (40-year) qualifying period. If the pension system grandfathered in workers (a common occurrence), it might, for example, guarantee 20 per cent of final salary for anyone paying in all years that the


155

programme existed. Applying that rule to the above example, someone retiring in 1970 would have a 20 per cent replacement rate, and a nine-year qualifying period, while someone retiring in 1971 would have 20 per cent replacement rate and a ten-year qualifying period. The final eligibility condition is the funding ratio: employee social contribution to the combined employee/employer contribution. If the contribution is borne solely by employees, the funding ratio is 1, if it is borne equally by employees and employers it is 0.5.16

COMPARATIVE INDICES OF WELFARE GENEROSITY A widely used empirical indicator of the generosity of welfare benefits is Esping-Andersen’s (1990) decommodification index. It is based on several important characteristics of unemployment, sickness, and pension programmes, including the replacement rate for singles and coverage/take-up ratios discussed in the last two sections of this chapter. The decommodification index has been used as an indicator of benefit generosity in numerous empirical studies (Huber et al., 1993; Pampel and Gartner, 1995; Rosenfeld and Birkelund, 1995; Mitchell, 1996; Western, 1996; Messner and Rosenfeld, 1997; Crepaz, 1998; Iversen, 1998; Hicks, 1999; Kenworthy, 1999; Huber and Stephens, 2001; Pampel and Williamson, 2001; Radcliff, 2001; Swank, 2002b). The original index was created by assigning each country a score for each characteristic in Table 7.1 based on the cross-country mean and standard deviation for that characteristic, except for coverage and take-up: ‘1’ if less than a standard deviation from the mean, ‘2’ if within one standard deviation, and ‘3’ if greater than one standard deviation. (The points are reversed for characteristics where a higher score implies a more stringent condition.) For each of the three main social insurance programmes, replacement rate scores are doubled, and the other characteristic scores summed. That total is multiplied by the coverage (take-up) rate. CWED data allow us to compute a decommodification score for each year: 1971–2002. For each characteristic, the 18-country mean and standard deviation in 1980 is used as the benchmark against which values for all other country-years are scored. The choice of 1980 as a benchmark is ultimately arbitrary, but since this is the year used in the original index, it is convenient for comparative purposes. The overall index, and the index for each of the three component programmes, are presented in Figure 7.1 (the overall score uses the right-side scale). The figure suggests remarkable stability over the period with little evidence of retrenchment anywhere except in Switzerland,

156

swi

jpn

fin

aus

uk

nld

fra

aut

us

nor

ger

bel

50 40 30 20 10

70 19

swe

ita

den

80 90 000 19 19 2

decom

sickscorecov

penscorecov

uescorecov

80 90 000 970 2 19 19 1

nzl

ire

can

Figure 7.1 Comparative welfare state entitlements decommodification indices (Esping-Andersen scoring method)

70 980 990 000 970 980 990 000 970 980 990 000 19 1 1 2 1 1 1 2 1 1 1 2

15 10 5 0

15 10 5 0

15 10 5 0

15 10 5 0

50 40 30 20 10

50 40 30 20 10

50 40 30 20 10


157

Sweden and New Zealand. This stability would seem to back up claims that welfare states have been very resilient to pressures for retrenchment. How do we explain this apparent stability? One explanation is simply that welfare state programmes have, in fact, not changed much. Another is that the system of scoring welfare state characteristics is flawed. The scoring is, in fact, insensitive even to very large changes in the values for underlying characteristics. For example, Belgium’s unemployment replacement rate moved up from 55 per cent to 70 per cent (0.8 standard deviations (sd) spanning the mean) between the early 1970s and mid 1980s. It then falls to around 62 per cent by 2001 (a decline of about 0.4 sd). This would seem like a clear case of expansion and retrenchment over time. However, because the replacement rate never exceeds one standard deviation above or below the cross-country mean in 1980, the scoring procedure from Three Worlds scores it a ‘2’ for all years. A related problem with the scoring procedure is that, in a given year, a country with consistent, but moderately generous, programme values is scored the same as a country with consistently low values. For example, assuming full insurance coverage for the working population, a country that is 0.9 sd for all of four characteristics of sickness insurance would receive a sickness decommodification score of 10; this is the same score that a country scoring 0.9 sd from the mean on all characteristics would receive. Meanwhile, two countries scoring only marginally higher/lower – e.g. 1.05 sd from the mean – for all characteristics would receive the maximum/minimum score (15 and 5 respectively). A more appropriate, and perhaps more intuitive, scoring method is to use the standardized values for each characteristic (except for coverage). Comparing countries on a continuous scale eliminates most of the problems referred to above. First, it avoids arbitrary assignment of cut-off points. While one might argue that this artificially creates a continuum, this is superior to arbitrary assignment of cut-off points at / one standard deviation, and prevents any accidental bias in the creation of cutpoints (see Scruggs and Allan, 2006a, b). The revision scores a country that is ‘moderately high’ on many characteristics higher than one that is moderately low on all, while countries that are ‘a bit high’ on some and ‘a bit low’ on others score like those with a programme that is in the middle on all attributes. The upper and lower bounds of the scores are also truncated to be / 2 in order to prevent an outlying score from distorting the overall distribution of results. Furthermore, in order to make all scores positive, we add 2 to each characteristic score so it varies between 0 and 4 rather than 2 and 2. The final major change to the decommodification scoring is the use of replacement rates for the family household type. This is justified largely on the basis that the vast majority of workers are members of families, not

158

Table 7.8


Decommodification and generosity scores, 1980

Decommodification score from Three Worlds* Sweden Norway Denmark Belgium Netherlands Austria Switzerland Finland Germany France Japan Italy United Kingdom Ireland Canada New Zealand United States Australia *

39.1 38.3 38.1 32.4 32.4 31.1 29.8 29.2 27.7 27.5 27.3 24.1 23.4 23.3 22.0 17.1 13.8 13.0

Benefit generosity and difference in ranks Sweden Norway Denmark Netherlands Belgium Switzerland France Germany Austria Finland New Zealand Canada Ireland Australia United States United Kingdom Italy Japan

42.3 (0) 38.4 (0) 37.2 (0) 35.9 (1) 31.3 (1) 31.2 (1) 30.3 (3) 29.1 (1) 27.8 (3) 27.4 (2) 26.2 (5) 21.2 (3) 21.2 (1) 19.3 (4) 19.3 (2) 18.7 (3) 17.8 (5) 17.4 (7)

Scores in our replication of the decommodification index differ from these.

single workers. To make the relative weights of replacement rate scores similar to those in the decommodification index, the replacement rate scores for singles and couples are no longer weighted twice. It is not clear whether Three Worlds relied only on single-worker replacement rates because the family rates were not available in the SCIP data at the time, or due to a fear that generous spousal supplements in ‘conservative’ welfare systems in continental Europe would conflate state support for traditional family structures with decommodification. In the latter case, there are two problems. First, it is not clear that support for traditional families does not in fact constitute decommodification, albeit not the postfeminist social-democratic one. Second, based on the replacement rate data in the previous Tables 7.2–7.5, there is really no evidence that conservative regimes have higher family replacement rates due to family supplements. The revised decommodification scores, which we label the benefit ‘generosity index’, are plotted with scores computed using the original methodology (right-axis scale), and as presented in Figure 7.2. Table 7.8 shows the generosity index scores in 1980 alongside Esping-Andersen’s decommodification data in order to provide a direct comparison between

159

70

swi

19

90 20

00 19

70 19

80

90 19

us

nor

ger

bel

20

00

70

19

80 19

90 19

nzl

ire

can

Revised (and original) decommodification scores for 18 OECD countries

19

80

uk

nld

fra

fin

jpn

aut

aus

80 990 000 970 1 19 1 2

Figure 7.2

19

50 40 30 20 10

50 40 30 20 10

50 40 30 20 10

50 40 30 20 10

70 19

80 19

19

decom

decomsdbasisf

00 20

90

swe

ita

den

20

00

160


the CWED results and the original decommodification index from the Three Worlds of Welfare Capitalism. Table 7.8 suggests, unsurprisingly, some basic correspondence with the original decommodification index. First, three paradigmatic cases – Germany, Sweden, and the United States – all fit squarely with the Three Worlds ranking. A number of other cases also fit: Australia, Austria, France, and the other Scandinavian countries. Several anomalies in the original text – Switzerland and Finland – also appear anomalous in our ranking. On the other hand, Japan and Italy emerge in our rankings as extremely liberal countries, with low decommodification scores, while New Zealand appears much more decommodifying than suggested by the decommodification index. The Netherlands’ score is also much closer to those in the Scandinavian countries in the generosity index than it is to scores among the conservative regimes like those of France and Germany. Finally, the overall variation among countries, particularly the distance between the best and worst, is much less in the generosity index than the decommodification index. Such an exaggerated variance between liberal and social-democratic countries suggests that existing empirical relationships employing decommodification as an explanatory variable, or as a variable thought to determine decommodification, are overstated. Are these differences very significant substantively? Consider that the original decommodification index suggested anomalous results for two of the 18 cases – Switzerland and Finland. Our results suggest anomalous results for at least four additional countries – Japan, Italy, the Netherlands and New Zealand. Since our approach replicates the same underlying data that were used in Three Worlds, the results are not a case of employing different indicators. Presumably, our results should be almost identical. This implies a much more tenuous fit between theory and reality than comparative social policy scholars have been led to believe. Compared with the original scoring method, our revised measures show a much greater degree of variation within many countries over time. Figure 7.2 suggests that the 1970s were an important period of increased benefit generosity in many countries, that the 1980s saw somewhat more expansion in generosity, and that the 1990s were a period of reduced generosity. All of the social-democratic regimes, except for oil-fed Norway, were considerably less generous in 2002 than they were in the mid 1980s. Germany, France, and, to a lesser extent, Belgium, also appeared to be substantially less generous in 2002 than they were in the mid 1980s. This contrasts with spending figures; which were higher in 2002 in all three countries. The trends also indicate that a number of countries were less generous in 2002 than at any other time in the last generation. Indeed, Germany, despite its reputation as a welfare reform laggard, is probably the only country that has been consistently less generous

161


40 35 30 25 20 15 1970

Alld Conservative 1980

1990

Liberal Social dem 2000

Year Figure 7.3

Revised decommodification scores: mean values, 1970–2001

over the period covered (see Siegel, 2004 for corroborating evidence). Explanations of this greater within-country variation should be of intrinsic interest to researchers. One immediate implication for the comparative social policy field is that one should be careful in assuming that cross-sectional variations at a single point in time are valid much earlier or later in time. Unfortunately, a great deal of work since Three Worlds has tended to do that. Figure 7.3 provides the average annual scores for all countries and annual averages in each welfare regime (which we have defined here according to the conventional classification, not the one implied by our results in Table 7.8). It mirrors patterns described in earlier sections of the chapter. By the early 21st century, generosity is higher than it was in the 1970s. Programme generosity tends to increase until the mid to late 1980s, after which there are some clear signs of retrenchment. This is especially true in the more generous social democracies. Finally, after diverging throughout the 1980s, scores seem to converge through 2002.

CONCLUSION: WHY AND HOW TO IMPROVE THE CONCEPTUALIZATION AND MEASUREMENT OF WELFARE STATE PERFORMANCE The CWED project data set is far from a complete or ideal set of data on welfare programme commitments. What the project has attempted, and

162


what should be repeated in all future empirical research on comparative welfare states, is to expand the scope of comparison by trying to measure things in a systematic way across as many countries and years as possible. Here we make several specific suggestions that can further improve the depth, quality, and comparability of social rights and welfare state institutions across countries and time. First, social entitlements need to consider two-earner households and single-parent households. Two-earner households have become the norm in most countries, and single-parent households are a particularly vulnerable group in all countries. We have a very patchy view of how the tax and benefit systems have interacted to affect such groups over time. In addressing this issue, there are several complications that have to be confronted. The experience with the CWED suggests that computing replacement rates and benefit conditions for these alternative households will be a considerable task, but that it can be accomplished more easily if it builds on the current project. Second, the distributional effects of welfare programmes across income and social strata are not well examined. The generosity measure is computed for a single point in the wage distribution (the average production worker). However, two policies affecting a typical worker in the same way may affect lower (or higher) earners differently.17 Are cuts affecting some income groups and not others? Are those affected actually groups that should be targeted for cuts? Third, there is still a paucity of systematic comparative data on alternative types of social rights. Indeed, virtually none of the literature considers the adequacy of social assistance, arguably the ultimate social floor of the welfare state. Other substantive improvement in data might stress the role of non-cash benefits and services. Still other improvements would take up programmes that provide a more balanced gender orientation to welfare state programmes, such as assistance programmes targeted at single parents. While most of the above criticisms are not new, work on the next step – specifying new conceptions and associated observables – tends to lag. There is, for example, little work comparing social assistance programmes in the context of social rights and income replacement (Gough et al., 1997 is an exception, but even it primarily looks at spending levels). ‘Universal services’ are often casually asserted as affecting outcomes, but a systematic comparative evaluation of the distributive implications of such services is needed to include them in the analysis alongside cash benefits. Finally, and apropos of the three previous suggestions, we simply need more data across space and time using the main concepts that we already focus on in the literature. Too often, interesting contemporaneous indicators


163

are developed to investigate (or define) some ‘new or ‘nationally unique’ social problems. Appellations such as ‘new’ and ‘unique’ imply comparison to other things, yet those comparisons tend to stop at the preliminary stages, just when the necessary work to be done is identified. The importance of establishing appropriate measures of generosity is vital. Social scientists tend to focus overwhelmingly on creating ‘new theory’ (or modifying old theories). Developing tests of theories is typically an afterthought, or worse, ‘test development’ becomes an exercise in a search for measures to show that this is consistent with a theory. Why it is so would fill its own volume (or, on second thoughts, may be all too obvious) but it is scientifically pathological. If concepts in a theoretical model are poorly operationalized and measured, then we are not testing the theoretical model. One might argue that social science contains very abstract, hard to operationalize concepts and theories, so asking for precise correspondence (compared with, say, physics) is unreasonable. This has it completely backwards. Errors probably proliferate exponentially: sloppily operationalizing sloppily theorized relationships does not make a ‘good test’. The time spent conceptualizing and operationalizing should vary inversely with theoretical specificity (allowing of course for some minimal standard of both). Vague theories demand more attention to concept validity and better measurement, they do not justify worse concept validity and measurement. Mis-measurement produces well known problems in statistical analysis, including bias and inconsistency of parameter estimates (Gujarati, 2003: 526–7). Since many claims about causal relationships are based on inferences drawn from statistical analysis, measurement problems may fundamentally undermine existing ‘facts’. For example, a great deal of recent empirical research relies on spending data to infer that the relationship between political partisanship and welfare state generosity has largely disappeared in the ‘era of permanent austerity’. If, however, as many (perhaps most) social policy scholars suggest, spending data have always been poor indicators of welfare generosity – or if previous non-spending indicators of generosity (e.g. the decommodification index) are mis-measured – then empirical results are ‘biased’ against finding any effects. Indeed, results with non-spending programme data as discussed by Allan and Scruggs (2004) or Korpi and Palme (2003) suggest that partisanship continues to play a big role in explaining variations in welfare state generosity in the era of austerity.18 The emperor has no new clothes. Mis-measurement presents even more serious problems when it occurs on the right-hand side of a regression equation. There it produces not just inefficient but also biased estimates of the effects of generosity as a causal variable. This suggests that the estimated effects of social spending in numerous statistical models throughout comparative political economy

164


more broadly may provide biased and inconsistent impressions of the actual effects of welfare generosity on a wide variety of social processes. The size and direction of these biases is unknown until those results are corrected with a more accurate operationalization of generosity.

NOTES 1.

2. 3.

4. 5. 6.

7. 8.

9.

10.

The most obvious shortcoming of current data sets and theory evaluation (including the current one) is that the ‘sample’ remains so selective: less than 20 of the same Western, industrial countries. This tends to make the idea that any new explanation is really ‘tested’ (rather than simply derived from them given how much we already know) quite problematic. There is, furthermore, almost no effort to use results to make any predictions which are evaluated against evidence. If the explanations were very good, for example, then they should help us (a) predict results in other countries, (b) predict future trends in the current countries under study or even (c) ‘postdict’ the past. The problem is more than simply that the social science data are observational; the available data are generally collected for some alternative purpose than testing the theory that the social researcher wants to test. Pierson (1994), to cite one of the most influential examples, exemplifies the practices alluded to here. First, his argument often relies considerably on spending trends to bolster his central argument that there has not really been welfare state retrenchment. Second, the book provides ‘new’ political explanations with a selective consideration of ‘old’ explanations which might predict results with the same success (Scarborough, 2000). Third, and perhaps most problematic, this line of explanation bears no resemblance to a progressive research programme. The ‘new politics’ explanation makes no real effort to explain both past and present under the same model, it asserts that the present requires a different one. Kittel and Winner (2005) suggest that this variation is limited, making pooled statistical analysis of these data potentially problematic. The need for cross-sectional, historical data sets exists independently of questions about how to analyse them. One advantage of the OECD indices is that, in recent years, they incorporate housing benefits and social welfare (the latter available after insurance benefits are exhausted) replacement rates. Detailed sources and procedures are available in the data set codebook. Some uses of the SCIP data (e.g. Sjöberg, 2000b) extrapolate between the five-year interval data points to get ‘annual’ data, or else (e.g. Korpi and Palme, 2003) rely on some other extrapolation procedure to get annualized data. A comparison of our and Esping-Andersen’s results is provided in Scruggs and Allan (2006a). Analyses using the index are Armingeon and Griger (2006), Pontusson (2006), Hu et al. (2006), Scruggs and Allan (2006b), Scruggs (2005). APW wages and taxes used here are provided, with modifications, by OECD publications like Taxing Wages. We use net benefits payable in the first six months of receipt. Further details of the calculation of replacement rates computed in each country and year are provided online in the data set codebook and replacement rate spreadsheets at the CWED website. The current placement of some countries – especially Ireland, Japan, the Netherlands and Japan – is very ambiguous (compare, for example, Esping-Andersen, 1990; Castles and Mitchell, 1993; Esping-Andersen, 1999; Scruggs and Allan, 2006a; Shalev, forthcoming) but not substantively important here. It is important to note that we include periods in which employers are legally obligated to pay wages to absent employees as insurance payments in computing replacement rates. We do so based on the assumption that such payments constitute legal rights.

Welfare state generosity across space and time 11.

12. 13.

14.

15.

16.

17. 18.

165

Using the wage in the year of retirement as the ‘historic insured wage’ leads to a considerable overstatement of actual benefits. Many pension systems were designed to be based on a long earnings history, a fact obscured by comparatively long transitional periods which were especially favourable to transitional workers. This describes inter alia the (pre-reform) Swedish and American systems. Further details of specific cases are available in the CWED data (also see Scruggs and Allan, 2006a). When we discuss the generosity index, we will assume coverage is 50 per cent to account for the fact that the benefits are purely means tested. The non-zero values for the United States in the table reflect the fact that five states, including California and New York, have sick pay insurance modelled on their unemployment insurance systems. Another important factor to consider in interpreting all of the sickness benefit data is that most (though certainly not all) Americans receive some paid sick leave each year from their employer. Coverage rates for public pensions (the proportion of people in the work force who are insured) are hard to evaluate, because many people not in the labour market at a given time can be entitled to a pension based on their previous work history. Thus, the number of people paying social contributions for pensions in a given year (or counted as entitled to a pension) may be a considerable undercount. Empirically, our coding choice tends to give ‘length of service’ rules comparatively long contribution periods and benefit durations. This seems to be consistent with how such systems are comparatively characterized in the literature. However, this obviously fails to fully capture some differences in the way the ‘length of service’ features of this insurance model function. Choosing differently, e.g. taking the minimum qualifying period and duration would also be imperfect and correspond less closely with distinctive features according to social policy scholars. This procedure was adopted to be consistent with Esping-Andersen’s decommodification index. The measure has two shortcomings. First, it ignores the general tax system’s role in funding pensions. Second, it assumes a certain ‘paycheque illusion’. Whether a social contribution is paid out of each worker’s salary or is based on a percentage of the employer’s total wage bill makes no real difference to take-home pay. Some of this type of work has been done (e.g. Hansen, 2002). However, the countries and years of coverage are limited, making it difficult to discuss very broad trends. Because generosity measures, like spending, trend upward during the 1980s they are theoretically and empirically compatible with a role for partisanship in explaining welfare expansion and retrenchment. Unfortunately, the ‘new politics’ literature fails to explain the rise of the welfare state via the same causal processes invoked to explain retrenchment. If new theories only need to explain new events – and not also ‘cover’ new and old events – there is not much recourse to external verification or theory builiding: one can safely have a different theory of each event. Not an impossible state of affairs, but one which seems to command that we give up the candle of social science.

8. Levels and levers of conditionality: measuring change within welfare states Jochen Clasen and Daniel Clegg INTRODUCTION Bold claims are often made about the current development of welfare states, both by critical theorists of social policies and by the politicians that are reforming them. But characterizing the nature and magnitude of the changes that welfare states have undergone in recent decades seems to pose major problems for empirical – and particularly comparative empirical – analysis. The lively debate concerning the range of factors that may result in (more or less) change in social protection arrangements – including most importantly structural socioeconomic forces, changing power resources, new ideas, party competition, institutions, policy legacies and path dependence (for overviews, see van Kersbergen, 1995; Amenta, 2003) – is complicated by the fact that analysts struggle to agree on what, exactly, is the real character and extent of change to be explained. Controversies and contradictory readings abound in the comparative social policy literature. Have we, as some maintain, witnessed a ‘paradigm shift’ in the techniques and strategies for managing social risks, or merely a series of adjustments at the margins? And if recent reforms are leading to the emergence of a distinctively new ‘type’ or ‘form’ of social policy, is this equally true in all developed welfare states, or only (to date) in some? While in social sciences there is always scope for differing interpretations, convincingly and consistently answering these kinds of questions arguably turns first and foremost on the identification of the most appropriate data for examination. Differently put, the key challenge for assessing the extent of certain hypothesized or postulated changes in welfare state programmes is not so much one of accurate measurement, but rather of developing more adequate conceptualizations and operationalizations of the possibly variable qualities of welfare state programmes. Arguments about new forms of social policy provision or regulation speak to certain welfare state properties in 166

Measuring change within welfare states

167

particular, and not necessarily those for which comparative indicators are most readily available. Only by focusing on the most salient dimensions of social policies, and developing truly relevant comparative benchmarks, will compelling cross-national assessments of arguments about the direction and implications of ongoing welfare reforms be possible. Without pretending to present a ‘best’ or ‘definitive’ solution to the inevitably thorny problem of characterizing dynamic patterns of policy provision and regulation in welfare states, this chapter advances one possible analytical framework which, we argue, can usefully complement indicators more habitually used in cross-national comparisons of welfare state arrangements. This framework is based on the analysis of the degree of conditionality that is written into the design of all social programmes. By paying close attention to changes in time across what we call the levels and levers of conditionality, we argue that it is possible to identify the extent and pace of changes in the management of social risks that are implicit in the reforms to welfare state programmes that have been implemented in some – but not all - countries in recent decades. The argument is organized in three sections. In the first of these, some of the measures and indicators that are most widely used in cross-national welfare state analysis are discussed to highlight their limitations in capturing the more qualitative shifts that are widely argued to be occurring in the contemporary period of welfare state reform and change. As a possible complement, rather than alternative, the second section then presents our framework for conceptualizing and analysing (changing) patterns of conditionality in benefit provisions. This framework distinguishes notably between three discrete levels of conditionality which we call conditions of category, conditions of circumstance and conditions of conduct, each having distinctive levers in legislative provisions for social security. Drawing on data from four European countries, the third section applies the framework to an analysis of change within unemployment protection since the early 1980s, measuring the extent to which each of the different conditionality levers has been loosened or tightened in recent reforms, and examining the patterns across all three levels. We conclude with a brief summary of the merits and limits of the approach.

THE PROBLEM OF CHANGE IN COMPARATIVE WELFARE STATE RESEARCH While much academic debate in recent years has been focused on the extent to which the welfare state has been rolled back – a veritable ‘retrenchment business’, as Hinrichs and Kangas (2003: 574) have called it – other scholars

168


have been exploring more complex shifts in the profile of social provision. Contemporary welfare state change is not (only) about retrenchment, but (also) about the more structural agendas that terms such as ‘modernization’ and ‘recalibration’ seek to capture (cf. Pierson, 2001). Some have argued that the welfare state is being superseded by an ‘enabling state’ (Gilbert, 2002), while others have spoken of the emergence of a ‘social investment state’ (Esping-Andersen et al., 2002) or an ‘active social state’ (cf. Vielle et al., 2005). These leitmotifs of welfare state change suggest not merely a modification in the scope or reach of the welfare state, but instead a structural transformation – or paradigm shift – in ways of managing and regulating social risks. Although they have stimulated a lively normative debate, it has proved more difficult to provide systematic evidence of these leitmotifs influencing actual social policy reform. For if disagreements and disputes over the most appropriate indicators has been a characteristic of the retrenchment business (see Green-Pedersen, 2004 and Chapter 2 in this volume), the ‘dependent variable problem’ is arguably even more acute when it comes to the comparative analysis of more qualitative and structural changes in social policy. Considering the three main types of indicators that are most widely used in comparative research on social policy and welfare states – social expenditure, social policy institutions and ‘social rights’ – helps to highlight the problem. The first of these, social expenditure, is the most traditional measure of comparative social ‘effort’, at either the aggregate (welfare state) level or at the level of particular social risks and individual social programmes. Having been somewhat discredited as an inadequate proxy for the development of social rights in earlier comparative work (cf. infra), the analysis of levels of social expenditure has been revitalized by the retrenchment business, and is today used in increasingly sophisticated statistical analyses of welfare state development (e.g. Kittel and Obinger, 2003). Furthermore, closer attention to the impact and extent of tax breaks, direct and indirect taxation and mandatory private expenditure (e.g. Hacker, 2002; Adema and Ladaique, 2005), on the one hand, and the integration of cross-national variations in risk intensity (e.g. Siegel, 2002 and Chapter 4 of this volume), on the other, has helped to enhance the descriptive accuracy of social expenditure comparisons. It remains the case, however, that even the most sophisticated indicators of overall effort capture only the quantity of investment in collective social protection, are still onedimensional and can say little about the risk management strategies underpinning social interventions on which the leitmotifs of welfare state change often turn. Focusing on the structure rather than the level of social expenditure may offer a potentially more fruitful strategy. As Castles (2002) has pointed out,


169

certain changing priorities of policy makers may well be captured by examining the (appropriately ‘adjusted’ or ‘standardized’) share of expenditures on the different social programmes, covering different social risks, in disaggregated data sets such as the OECD’s social expenditure database (SOCX, OECD, 2005a). For example, an increasing share of spending on rehabilitation programmes or active labour market policies relative to unemployment or disability benefits could be seen as indicators of a shift to an ‘active welfare state’, and a shift from expenditure on pensions to child care or education proxies for the development of a ‘social investment state’. It is not certain, however, that this kind of macro-level, expenditure based approach is really capable of capturing ongoing changes in a timely fashion. There is potentially quite a long causal and temporal chain between the legislative changes that concretely reflect changing public priorities and shifts in the structure of social expenditure (see also Chapter 4 by Siegel, Chapter 5 by De Deken and Kittel, Chapter 6 by Kangas and Palme, this volume). This may explain the fact that although data of this sort suggest that there has been minimal ‘structural change’ in, for example, the German welfare state (Castles, 2002), recent research using more case oriented approaches has arrived at very different conclusions (e.g. Bleses and Seeleib-Kaiser, 2004; Clasen, 2005). The second important family of indicators that are increasingly used in comparative welfare state research are derived from broad institutional characteristics of welfare state programmes. For example, following Ferrera (1996), Palier (2002) has suggested that national profiles of social policy arrangements can be captured according to the dominant values taken on four institutional variables in social protection: the financing system (social contributions, general taxes, co-payments), the management structure (state, social partners, local authorities), the type of benefit (flatrate, earnings related, means tested) and the mode of access to benefits (citizenship based, contribution based or needs tested). Considered at either the whole system or sectoral level, these variables combine to demarcate different ‘ways of doing’ social policy in different national contexts. For Palier, evidence of significant qualitative change in social policy arrangements is accordingly most convincingly highlighted by changes in the values taken – measured according to the appropriate structural indicators – on one or more of these variables. This framework for analysis has proved very useful in capturing some developments in social protection systems, not least because it integrates ‘input’ dimensions of social policies (such as financing or management) that were often neglected in previous comparative research. It establishes very demanding conditions for the identification of change, however. Thus, a change in the type of benefit deployed to manage a given social risk would

170


only be identified on the basis of a shift from, say, a primarily earnings related (social insurance) to a primarily means tested (social assistance) approach. Though of course possible in the longer term, such wholesale ‘regime’ shifts remain rather unlikely in the short term. Furthermore, as Leitner and Lessenich (2003) have argued, the nature and quality of a particular social insurance benefit can vary quite considerably over time according to the setting and resetting of its lower-level institutional parameters. Perhaps more importantly for current purposes, a related risk of the approach is that it seems to reduce possibilities for identifying change to shifts within the standard palette of already existing institutional ways of doing social policy, and may thus fail to capture forms of innovation that introduce entirely novel arrangements. Thus, if Palier’s (2002) suggestion may prove useful in showing that country x is increasingly using a risk management strategy conventionally associated with country y, it would have difficulty showing that either country x or country y is groping towards a quite new model of risk management. Because they cannot easily capture either subtle but potentially significant shifts within relatively stable macroinstitutional parameters or forms of institutional innovation that touch only marginally on the standard (‘golden age’) variables of welfare state structure, the use of only broad institutional indicators of welfare state quality may lead to an exaggeration of the extent of stability in welfare state programmes. The final family of commonly used approaches to comparative social policy mapping and analysis are based on what Siegel (2003) calls ‘social rights’ indicators. Though indicators such as benefit levels and replacement rates also belong in this family, the most sophisticated and probably best known social rights indicator is the measure of ‘decommodification’ first devised by Esping-Andersen (1990; see also Scruggs in Chapter 7 of this volume). Criticizing expenditure indicators for failing to capture the essence of social policies, and the political struggles that surround them, Esping-Andersen combined information from a range of quantifiable micro-institutional parameters of social policies (e.g. replacement rates, contribution conditions, waiting days) to develop a measure of the extent to which welfare states decommodify individuals, that is free them from dependence on the sale of their labour power in the market. He was then able to place countries on decommodification scales, again in respect of both particular sectors of social policy and their entire welfare states. Decommodification specifically, and measures of the generosity of social rights more generally, remain crucial benchmarks of the quality of different social programmes and welfare states. Moreover, their ability to quite directly tap into the question of social citizenship – the role of the state in furnishing individual opportunities and resources – in many respects gives


171

them a better ‘handle’ on the themes at the centre of more normative discussions about the ‘new welfare state’ than expenditure or broader institutional indicators. That said, however, these indicators are arguably also increasingly wanting. Political discourse within social policy has in recent years been progressively shifting away from rights and towards ‘responsibilities’, and an assessment of the changing ‘social citizenship relationship’ requires attention to both. Although measures such as decommodification are flexible enough to incorporate measures of changes in certain responsibilities (or conditions) for benefit receipt, they cannot capture all of these. Most importantly, measures of social rights take the notion of social risks largely as given, while much current political discourse concerns the need to reconceptualize and reorganize the boundaries of risk categories, and to make their collective treatment more individual and less social. Capturing the extent and progress of these themes in actual policies arguably also requires additional indicators (see Scruggs in this volume). To summarize the discussion above, the main kinds of indicators used in comparative welfare state analysis capture different dimensions of welfare state variation, and each is valuable for understanding of particular kinds of social policy performance and development. Each, however, also has more or less severe weaknesses as empirical indicators of change in the structure and form of welfare provision and regulation that dominate much of the more normative or critical discussions of contemporary social policy reform dynamics. Though social rights approaches probably operate at the level most apt to analyse the changes in the social citizenship relationship that this literature prescribes, describes or decries, they lack the conceptual focus necessary to empirically capture salient dimensions of this relationship. It is in this context that the analytical framework that we present below can perhaps enhance our capacity to comparatively assess both the scope and the implications of ongoing transformations in the architecture of social programmes.

A CONDITIONALITY APPROACH This framework or approach is centred on the notion of conditionality. Conditionality has been something of a ‘buzzword’ in recent welfare state discourse, and some scholars have even presented ‘conditional aid’ as representing a new paradigm for social policy (cf. Dufour et al., 2003). Our approach to this issue is somewhat different. We depart from the observation that, firstly – and contrary to the impression that is given by some more recent discussions of conditionality – individual rights to social benefits have always and everywhere been conditional in some ways, and conditionality is

172


as such a cornerstone and basis of risk management in welfare states (cf. also Goodin and Rein, 2001). Secondly, however, there are several dimensions as well as a range of possible ‘levels’ and ‘levers’ of conditionality in social policy, and differing balances between these levels and levers can tell us something potentially important about the pattern of risk management that is institutionalized in and through social policies. Evidence of variation in this balance, therefore, can – thirdly – generate benchmarks for assessing the direction and scope of change in the comparative analysis of welfare reform. It should be emphasized that assessing modifications in the conditionality underpinning welfare provision is not intended to serve as a substitute for other indicators but as a complementary measure, albeit an important one. As will become apparent, the conditionality of a cash benefit, for example, says nothing about its generosity (which might be measured as a replacement rate) or about the institutional structure through which it is delivered. Nevertheless, a focus on the conditionality of welfare state programmes does help to capture changes in the relationship between rights and responsibilities, and thus provides an empirical basis for gauging the reality of ‘transformations’ in social citizenship often heralded in welfare state literature and debates. For this to be possible, it is first necessary for the various ways that conditionality can be adjusted in welfare reforms to be defined and distinguished analytically. Conditions of Category The first, or primary, condition for the receipt of social security is always membership of a defined category of support; being past retirement age for retirement pensions, having some form of disability for incapacity benefit, being unemployed for unemployment benefits and so on (Bolderson and Mabbett, 1996). Even so-called universal benefits do not entirely abolish ‘categorical gateways’ to support (cf. Gal, 1997; Goodin, 2000; Kiddal and Kuhnle, 2005); universal health care is of course only available to the sick, and even a basic income ‘for all’ is confined by citizenship or residency conditions. Though risk or membership categories are often taken as given in measures of social rights, it should be remembered that they are in fact socially constructed and politically managed, and thus subject to potential change, either in a more restrictive or more expansive or encompassing direction. An obvious case is pension age, which has been the subject of pressures for (mainly restrictive) reforms in most countries in recent years, and actual reforms in many. But governments have also changed definitions of risk categories such as unemployment, for example by excluding certain social groups or categories of worker (single parents, older people, students . . .)


173

at certain times and reintegrating them at others. Even where categorical gateways have professional gatekeepers – for example doctors, as in the cases of health care or disability benefit programmes – politicians have often changed directives and guidance they give to these professionals in an attempt to use the lever of category definition to impact on inflows into benefit receipt. Conditions of Circumstance Analytically secondary to conditions of category are conditions of ‘circumstance’ or, in more common social security parlance, eligibility and entitlement criteria. The fact that this type of conditionality has always underpinned welfare state provision is most obvious from the fact that some of their (potentially varying) settings and codifications have traditionally been integrated in the calculation of synthetic indicators of social rights, such as Esping-Andersen’s decommodification index referred to earlier (again, see Scruggs in Chapter 7 of this volume). For social insurance benefits, for example, legislation has long included rules regarding the extent to which and how a claimant’s work history determines access to benefit rights. The pertinent and potentially varying considerations in social security legislation include the number, or value, of contributions and/or full days of labour that are required in a given time period to open access to benefits, as well as the extent to which these contributions must be real (from paid work) or can also be fictitious and credited (i.e. related to non-work activities such as child rearing, home work, education etc.). Accordingly, individual access to a social insurance benefit can be more or less conditional, with the polar extremes of the range sometimes captured in the distinction between ‘Beveridgean’ and ‘Bismarckian’ social insurance, i.e. loosely or tightly conditional on paid employment (cf. Clasen and van Oorschot, 2002). Though it is less often remarked (and far less well integrated in crossnational indicators of social rights), it should be pointed out that means tested (or assistance based) benefits have also always incorporated potentially varying degrees of a structurally similar – though necessarily operationally different – form of conditionality. Though still conditions of circumstance, the circumstances in question are not – as with classic social insurance benefits – work histories, but instead the claimants’ degree of financial need. But the definition of need can vary in both extensity (narrow or wide definition of family obligations to support; many or few types of income and wealth taken into consideration) and in intensity (level of income and wealth disregarded; level of marginal tax applied) (cf. Eardley et al., 1995). Following from this, the risk management in assistance based

174


benefits can serve either to ‘screen in’ only the poorest and most needy population or to ‘screen out’ only the richest and least needy (Castles and Mitchell, 1992; Mitchell et al., 1994). Indeed, it is only in a ‘pure’ case of universal benefits that we find no conditions of individual circumstance whatsoever in the design of social security benefits. In the two other main kinds of social security programme, the various levers of work- or need-related eligibility criteria can be tightened or loosened to make social rights more or less accessible for the individual and thus more or less encompassing for society as whole. For social insurance benefits this can happen independently of changes impacting upon the benefit’s formal generosity (or level of entitlement) as measured through e.g. replacement rates. Conditions of Conduct The third and final level of conditionality is for its part logically subsequent to the others, intervening only after eligibility for benefit has been otherwise established, and having the function of regulating the ongoing benefit receipt. It pertains to what could be called conditions of ‘conduct’, with the policy levers being the tightening or loosening of behavioural requirements and constraints imposed upon different kinds of benefit recipients through legislation or administrative guidance. This is the form of conditionality that has featured most prominently in recent discussions of the ‘new welfare state’ and the changing nature of risk management and the social citizenship relationship. The most obvious case of a growing emphasis on this tertiary level of conditionality in policy reforms are so-called activation policies for the unemployed, under which unemployment benefit and unemployed social assistance recipients are, for example, obliged to provide evidence of job search activities, participate in training programmes or agree to specialized counselling (cf. Lødemel and Trickey, 2001; Dufour et al., 2003; Barbier and Ludwig-Mayerhofer, 2004). But this kind of conditionality has also been sporadically mooted, in a variety of different countries, in relation to policy themes as different as the use of family benefits to promote good parenting, the modification of conditions for access to certain health benefits to produce healthy lifestyle choices, or the conditioning of housing benefits on good neighbourliness and the avoidance of ‘anti-social behaviour’. Our overall specification or conceptualization of the universe of levels and levers of conditionality in social programmes is summarized in Figure 8.1. As the two closed arrows in the figure are intended to suggest, these three types of conditions and conditionality can further, and importantly, be understood as occupying distinctive positions on the spectrum of underlying

175


STATUS Level of conditionality

I. Category

II. Circumstance

III. Conduct

Levers of conditionality

Category definitions

Eligibility and entitlement criteria

Behavioural requirements

BEHAVIOUR Figure 8.1

Levels and levers of conditionality in social programmes

bases for risk management in welfare states. While an emphasis on risk or membership category definitions thus confirms the basis of the risk management in socially defined status, an emphasis on constraints or requirements related to conduct gives more weight to individual behaviour as the basis for access to social resources and goods. The standard eligibility criteria for social benefits occupy a somewhat intermediate and indeterminate place on this spectrum; though eligibility criteria are largely impersonal and their fulfilment gives automatic access to a status, they have often been defined in practice – as the traditional basis of social insurance in past work performance best demonstrates – to simultaneously but indirectly place certain kinds of behaviour (i.e. ‘good work habits’) at the centre of the social citizenship relationship, too. Notwithstanding such ambiguities, we can nonetheless hypothesize that a shift in the nature of the social citizenship relationship should in principle be reflected in a changing balance between category, circumstance and conduct conditions, and thus be identifiable by scrutinizing actual changes in the setting of category definitions, eligibility and entitlement criteria and behavioural constraints in a particular social programme. It is to applying the conditionality framework to look for shifts of this kind in reforms to provision for the unemployed in four European countries that the remainder of this chapter is largely devoted.

APPLYING THE FRAMEWORK: THE CASE OF UNEMPLOYMENT PROVISION The case of unemployment provision is likely to offer a good test of the utility of our framework for a number of reasons. Firstly, unemployment

176


is perhaps the social risk which calls for, and critiques of, a ‘new paradigm’ of social protection have been most insistent in recent years (see also Kvist in Chapter 9 of this volume). It is also an area where there has been – in a context of historically high unemployment and (in part a result of other reforms) rapidly changing labour markets – quite considerable legislative activity, but where the real impact and thrust of changes in different national contexts often remains unclear (e.g. is activation a form of retrenchment?). Finally, unemployment protection is a sector in which legislators could plausibly adjust all three levels and levers of conditionality, and where any cross-national similarities and differences in patterns of readjustment and deployment should thus be particularly evident. This will be illustrated below. To capture such patterns, a relatively long time period is required. Our analysis covers roughly the quarter-century from the beginning of the 1980s to around 2004 or 2005. It includes four European welfare states, Denmark, France, Germany and the UK. Though we acknowledge that certain levers of conditionality can in the sphere of unemployment policy only be effectively turned at the ‘street level’ – one thinks here of the effective implementation of behavioural conditions by employment service officers or unemployment benefit administrators – for feasibility the data used are based only on major legislative changes in unemployment insurance.1 Below we sketch profiles of legislative trajectories oriented to our framework on the basis of these data, which are listed in full in the chapter appendices. Two points should be noted here. First, the following discussion is based on legislative activity that affected one or several aspects of conditionality within unemployment support in four European countries (see Appendices A to D). The appendices therefore are not intended to cover all major types of legislation in the field. Once again emphasizing the complementary nature of the conditionality approach suggested here, any decisions regarding the level of benefit, for example, are excluded because changes therein do not affect the conditionality as defined above. Second, the appendices document legislative changes, and the direction of changes (tightening or relaxing benefit conditionality). They thus document politically binding decisions over time within a particular country, rather than actual degrees of conditionality that could be compared across countries. In other words, legislative activity monitored here allows statements about the trajectory and frequency of changes in, for example, eligibility requirements, but does not provide information on the actual specification of such requirements in the four countries. In principle such information, as well as data relevant for other levels of conditionality, could also be collected, allowing comparative judgements to be made regarding the strictness of national welfare state


177

programmes at a particular point in time, as well as relative to the direction and nature of change over time. While such an endeavour is beyond the remit of this chapter, the following analysis can be regarded as a first step for such a comprehensive cross-national analysis of conditionality. The United Kingdom A conditionality framework sheds light in a particularly interesting manner on the nature of the British trajectory of reform in unemployment protection over the past 25 years. At a time of very high and persistent unemployment during the 1980s and early 1990s, UK policies concentrated mainly on tightening conditionality at the first and second levels (see Appendix A). For example, important category redefinitions reduced the potential pool of people eligible for unemployment support through the introduction of incentives to the over 60s to retire early (1981), the exclusion of almost all those under the age of 18 from eligibility to unemployment support (1988), and the removal of students from the unemployment risk pool a few years later (1990). The second level of conditionality (governing eligibility) was also frequently and consistently tightened between 1980 and 1996 when the Jobseekers Allowance was introduced. Numerous adjustments resulted not only in the abolition of (limited) ‘Bismarckian’ elements of British unemployment insurance (notably earnings related supplements), but also in a substantial erosion of the traditional ‘Beveridgean’ insurance logic. This has been achieved through a series of curtailments and by making the system less accessible for those with less stable employment and contributory records. As a result, means tested benefits have become the clearly dominant form of support available to unemployed people in the UK today. Whereas in the late 1970s more than half of all unemployed received contribution based support, 20 years later the rate had declined to about 16 per cent (Clasen, 2005: 59). Interestingly, since 1997 Labour governments have refrained from further adjustments of the second level of conditionality, i.e. leaving benefit eligibility and entitlement criteria unchanged. However, both the first and third levels of conditionality have moved to centre stage. Regarding the third level (conditions of conduct), Labour continued and reinforced a trend which had already begun under Conservative governments in the late 1980s, namely influencing the behaviour of unemployed benefit claimants by tightening the conditions of active job search, relaxing the definition of appropriate employment and toughening sanctions for non-compliance (Finn, 1998). More ‘positive’ forms of activation were essentially limited to providing increased help with job search, through initiatives such as ‘Restart’ interviews. As well as making compulsion in labour market policy

178


more explicit, Labour’s New Deal initiative since 1997 has built on the Restart concept with respect to intensive counselling and employment guidance for the unemployed; but it has also reintroduced a few investment intensive training and job subsidy programmes. This growing emphasis on activation has, in turn, encouraged further reform initiatives at the first level of conditionality. In a de facto reversal of the tightening of category membership during the 1980s and early 1990s, the Labour government has increasingly widened the remit of New Deal programmes, from the unemployed originally to potentially all workingage beneficiaries of social support (the disabled, lone parents, jobless partners of the registered unemployed) today. At the delivery level, the so-called Jobcentre Plus is now the point of contact not only for the registered unemployed, but for all working-age benefit claimants. We see a loosening of the definition of unemployment, or at least blurring of the boundaries between unemployment and other social risks. This has allowed UK labour market policy to gradually shift from an emphasis on unemployment to an increasing emphasis on ‘worklessness’ (Clasen, 2005). In other words, having settled on a new and seemingly stable level of secondary conditionality, with strict access to insurance and a dominant needs based approach, recent activities have focused on the first and third levels of conditionality. Notably, this means that past activities – the status and origin of claimants (contributions, employment status) – have all but become irrelevant. Instead, the new British conditionality logic is both wider (beyond unemployment) and more focused on work tests and employability criteria than previously. Over the past 25 years, the overall pattern could thus be described as a gradual progression down through the levers of conditionality, from adjustments at the primary level (category definition), accompanied and gradually superseded by successive initiatives tightening conditionality at the secondary level of conditionality, before finally a concentration on tertiary conditions (activation) took over. This activation logic has seen a small move ‘back up’ the levels of conditionality, with a partial reversal of earlier changes to primary-level (category) criteria. Germany Also faced with fast rising levels of unemployment during the early 1980s and even more so the 1990s, the reform profile in Germany differed considerably from that in the UK. As a more Bismarckian social insurance oriented country, German unemployment insurance offers a multitude of levers at the second level of conditionality, i.e. those that govern eligibility and entitlement criteria. In addition, the governance of unemployment support as a whole (until recently a de facto three-tier system) has provided


179

particular incentives for adjusting levers that would result in cost transfers between the federal level, the unemployment insurance fund, other social insurance funds and local authority budgets (Clasen, 2005). The annual budget of the unemployment insurance fund triggered several adjustments at the second level of conditionality during the 1980s and 1990s. Some of them relaxed conditions of benefit receipt, through for example the repeated prolongation of entitlement for older unemployed with longer contribution records between 1985 and 1987 (see Appendix B). By contrast, incremental legislative change tightened eligibility conditions of unemployment protection at the margins of the labour market, e.g. for those with shorter contribution periods such as job starters and those with repeated spells of joblessness (1982, 1994, 1998). Unlike in the UK, the first level of conditionality remained largely off the reform agenda during the 1980s, at least within unemployment insurance and notwithstanding the de facto category redefinition of older benefit claimants who made use of extended unemployment benefits as a form of quasi pre-retirement (Trampusch, 2005a). As for the third level of conditionality, tighter suitability criteria (governing regulations of job offers which can be turned down without risk of benefit sanction) were introduced as early as 1979 (Clasen, 1994: 153), and redefined again in later years (1982, 1998). There was, however, little emphasis on behavioural job search or activation criteria at federal level until the late 1990s, and the activation type policies were introduced with much less vigour than in the UK. While local authorities have had the right to ‘activate’ long-term unemployed persons for some time, and some have made extensive use of this not least on the basis of financial considerations (Buhr, 2003), prior to 2005 claimants of unemployment insurance have not been subjected to a routine form of mandatory activation. The major recent labour market reform in Germany (the so-called ‘Hartz’ reform) has however had a considerable impact on all three levels of conditionality. Apart from other changes (see Kemmerling and Bruttel, 2006) legislation merged unemployment assistance and social assistance into unemployment allowance II (ALG II) for ‘employable persons’, thereby expanding the category of unemployed welfare state clients (firstlevel conditionality). This change resulted in a steep and sudden rise in official unemployment in January 2005 when the legislation came into effect. Recipients of ALG I were affected by a shortening of benefit entitlement while the means test for claimants of unemployment assistance was tightened significantly (second-level conditionality). Moreover, benefits payable under the pre-existing unemployment assistance arrangements were earnings related, whereas the new ALG II is flat rate and no longer related to former earnings. Finally the reform has made steps towards an

180


activation oriented policy regime that regulates behavioural aspects of job seekers more strictly. This applies to ALG II recipients and benefit claimants under the age of 25 in particular, who are supposed to be offered, and to accept, a job or an ‘activation’ programme instead of benefit receipt. At the time of writing it remains unclear whether this regulation is being implemented accordingly. France The recent trajectory of unemployment benefit reform in France in many respects parallels that seen in Germany, not least in the overwhelming concentration of reform initiatives at the second level of benefit conditionality. The contribution financed French unemployment insurance system has, furthermore, in the main coped with a context of consistently high unemployment since the early 1980s by tightening eligibility conditions rather than reducing benefit rates. Important reforms over the period 1982–84 tied benefit eligibility and entitlement more closely to prior contributions, and created a separate tax financed and less generous system (the ‘solidarity system’) alongside unemployment insurance for some of those with inadequate contributory records. Further reforms in 1992 tightened the contributory requirements for unemployment insurance once again, and deepened the dualism of the unemployment protection system. Although some of these cuts have been partially repaired by other reforms, enacted notably in the period of stronger economic growth at the end of the 1990s, the dualism has not been challenged. Indeed, spare resources have often been devoted to improving benefits for certain groups of insured claimants, as was the case for example of a 1997 reform which created a special, more generous, unemployment benefit payable up to the age of 60 for workers with long contribution records. Meanwhile, coverage of unemployment insurance fell by around 10 percentage points in the 1990s. In sum, similarly to Germany, the main story of French unemployment benefit reform in recent decades has been one of using the levers of secondary conditionality to increase the differentiation in rights to unemployment benefits (cf. Daniel, 2000). Explicit reforms to the first level of conditionality have been far more rare, though – as in Germany – the unemployment insurance system was used in the early 1980s to encourage the effective withdrawal of older workers from the labour market (Daguerre and Palier, 2006). These incentives remain in place today, and were in fact somewhat strengthened by a reform in the late 1990s. Another form of category redefinition has been more implicit, with the progressive institutionalization of the Revenu Minimum d’Insertion (RMI), introduced in 1988, as the ‘third tier’ of the


181

French unemployment protection system (Audier et al., 1998). Though the recipients of the RMI may remain registered as unemployed, unemployment is not itself a condition for the receipt of the RMI. Furthermore, this benefit still rests on a mix of rights and responsibilities that are quite distinct from those found in formal unemployment protection. This is notably the case with respect to tertiary forms of benefit conditionality. Although the RMI has always been – at least formally – conditional upon the beneficiary agreeing to individualized reintegration activities, this principle had until very recently gained little ground in formal French unemployment benefit policy. Only in 2000 was a bolder proposal for the systematic conditioning of unemployment insurance benefits on positive job search activities brought forward, with the creation of the so-called PARE (Plan d’Aide au Retour à l’Emploi) (Freyssinet, 2002; Kerschen, 2005). Though there have been difficulties in the implementation of the latter measures, and the corresponding capacity of the unemployment insurance system to oblige participation in training or job search activities remains uncertain as a result, initiatives outlined in the recent ‘law on social cohesion’ and intensified by the new de Villepin government suggest that tertiary conditionality may belatedly be – again, as in Germany – becoming a more important lever in the reform of French unemployment benefits. Denmark As in the UK, a clear Danish reform profile can be detected since the early 1980s, with similarities in both the first and the third level of conditionality. However, unlike in the UK there has been no effective shift in distributive principles as a result of tighter conditionality at the second level. Instead, the social insurance notion within unemployment support has been maintained, albeit in a form that is still distinctively different – and much more ‘Beveridgean’ in character – than the one in Germany or France. Danish unemployment benefit entitlements are thus still among the most generous in Europe, though not for short-term unemployed persons on above average earnings since benefit thresholds and ceilings make transfers resemble a flat-rate system for a large section of unemployed persons. Danish governments tightened the definition of unemployment in the late 1970s and early 1980s, making considerable use of labour market exit strategies in the face of high unemployment. The early retirement programme efterløn was introduced in 1979 with the explicit aim of redrawing the category definition of ‘unemployment’ for workers of age 60 and over (first-level conditionality). However, unlike in the other three countries, thereafter several other levers were adjusted at the first level of conditionality which were aimed at widening or at least consolidating the potential membership

182


in unemployment protection, particularly for long-term unemployed (1985) and younger age groups with the introduction and later extension of the ‘youth benefit’ and ‘youth allowance’ (1990–92), albeit coupled with the right and obligation to accept activation offers. Thus, in common with the UK, Denmark tightened the third level of conditionality as early as the late 1980s and early 1990s, and concentrated increasingly on this strategy from 1993 onwards. Also as in the UK after 1997, the focus on activation was coupled with a reversal of the previous policy direction at the first level of conditionality, making efterløn financially less attractive (1998–2000) or extending the activation principle to social assistance (1997). There has been rather less reform activity at the second level of conditionality in Denmark than in the other three countries. Furthermore, the direction of change has not been unequivocal, as in the UK. Thus, though eligibility conditions for unemployment insurance were tightened in 1985, and a lower-rate benefit was introduced for those who no longer had access to full benefits, this reform was overturned in 1988, when full-rate benefits were reinstated. In the second half of the 1990s, Danish governments made considerable cuts in unemployment benefit rights. The bulk of these, however, have come neither through selective reductions in benefit levels nor in tighter eligibility criteria. The latter became stricter in a reform in 1994 (Kvist, 2002), but remain relatively loose in comparative perspective. Supporting about 85 per cent of all unemployed in 2004 (Danmark Statistik, 2005: Table 160), the Danish unemployment insurance system remains one of the most encompassing programmes in Europe. Nevertheless, secondlevel conditionality was tightened via several changes in the maximum duration of benefit entitlement, which was reduced in successive stages from seven to four years between 1993 and 1998. Danish reform efforts have concentrated not so much on preventing inflows into unemployment benefit but more on speeding up outflows, through considerable reform activity at the third level of conditionality, i.e. the turn towards activation principles in the 1990s. Since 1993 there has been a vast expansion of activation measures, with a particular emphasis on training programmes, which are both a ‘right and a duty’ for the unemployed (e.g. Goul Andersen, 2002). The length of time that it is necessary to be unemployed before being subject to this ‘right and duty’ has been reduced in successive reforms, and since 1998 is one year for the adult unemployed and six months for the young. Summary How can these stylized and necessarily simplified reform trajectories be compared? Based on the discussion above and information provided in

183

Measuring change within welfare states Countries

1980–85

1986–90

1991–95

1996–2000

2000–05

UK

I II III

– – /

– – –

/ – –

+ / –

+ / –

Germany

I II III

/ – –

/ (–; +) /

/ (/; –) /

/ (/; –) –

+ – –

France

I II III

– (–;+) /

– / (–; /)

– – /

– + /

/ (– ;/) –

Denmark

I II III

– – /

(–; +) + –

– – –

+ / –

+ / –

Notes: I, II, III primary, secondary and tertiary levels of conditionality (see text); tighter/more intensive conditionality; looser/ less intensive conditionality; / no significant change, or major changes in different directions cancelling each other out within period (; ) differences between groups of unemployed.

Figure 8.2 Conditionality shifts in unemployment benefit reforms in four countries Appendices A to D, Figure 8.2 provides a rough picture of legislative profiles. It shows that in all four countries reforms can be identified which altered all three different levers of conditionality, i.e. category definitions (first level), eligibility and entitlement criteria (second level) and behavioural conditions (third level) attached to continuous benefit support. A chronologically sequential path is conspicuous in the UK where governments focused on the first and second levels during the 1980s, on the third level in the late 1980s, and completed the second level in 1996. Since then Labour governments have almost exclusively altered the first and third levels in a drive towards what is now called ‘activation’. The Danish reform profile is very similar to the British at the first and third conditionality levels, but diverges at the second level. Unlike in the UK, there has been little tightening of the conditions of circumstance required for access to the unemployment protection system and the traditional notion of unemployment insurance has remained dominant, albeit coupled with a stronger conditionality at the third level. Very much in contrast to both the UK and Denmark, French and German reforms were distinctive in three respects. First, the second level of conditionality (eligibility and entitlement criteria) remained the privileged policy lever in these two countries. Second, within this domain, policies had a ‘differential’ focus in France and Germany, altering access to unemployment support relatively little for some groups (typically workers with long

184


contributory work records) but tightened eligibility conditions for groups which are less firmly entrenched in the labour market. Both Germany and France have thus deepened the dualism of their respective unemployment protection systems in recent decades. By contrast, both Denmark and the UK have maintained or created much more uniform conditions within their respective systems (see Clasen and Clegg, 2006a). Finally, while all four countries made use of labour market exit strategies by (de facto or de jure) changes in the category definition of unemployment support (first-level conditionality), France and Germany maintained such approaches for much longer during the 1990s, and also turned to ‘activation’ strategies (tightening third-level conditionality) much later than Denmark and the UK. This framework for analysing change in the conditionality of unemployment support helps to highlight a certain number of potentially salient cross-national similarities and differences. Firstly, until 1995 there appears to be no particularly strategic reorientation in the management of the risk of unemployment in any of the countries, with the main focus being a general tightening of available conditionality levers – retrenchment – across all cases, albeit with a generally broader range of conditionality levers being used to this effect in Denmark and the UK. Since 1995, however, it is possible to identify a clear change in the way that unemployment is managed in these two countries, with levers of eligibility or entitlement increasingly untouched on the one hand, and with status based risk management conditions becoming less, and behaviour based risk management conditions more, relevant. Over and above the continuing and often highlighted differences between the approaches to unemployment policy in the UK and Denmark (e.g. Torfing, 1999; Barbier, 2004), the framework shows a certain structural affinity in the way that the management of the unemployment risk is being reconceptualized in these two countries. This distinguishes them very clearly from France and Germany, although the recent reforms in Germany – and notably the broadening of the definition of unemployment and concomitant tightening of behavioural requirements – suggests that the Federal Republic too may be increasingly shifting towards a new model of risk management in unemployment, rather than merely adjusting the existing framework.

CONCLUSIONS The aim of this chapter was to try to elaborate a framework through which the extent of certain changes in welfare states, and notably the widely touted notion of an emergent new ‘model’ or ‘paradigm’ within the social citizenship relationship, could be empirically investigated and scrutinized


185

in cross-national analysis. As argued in its first part, these kinds of changes are difficult to capture with the standard expenditure, macro-institutional or ‘social rights’ based dependent variables that are commonly deployed in welfare state research. By identifying lower-level institutional features that are more pertinent to the dynamics expressed by the qualitative leitmotifs of welfare state change, we have shown that it is possible to develop comparative benchmarks against which policy trajectories can be judged, concentrating on particular policy programmes and their inherent dimensions and dynamic balance of conditionality. As its application in the analysis of unemployment benefit reforms in four countries has shown, such benchmarks are capable of systematically identifying patterns of similarity and difference in legislative activity and direction of welfare state change. Viewed in isolation, such a framework of course has a number of limitations. As emphasized earlier, it is mainly a device for comparing the emphasis, direction and structure of national reform trajectories over time. It thus can obviously not help analysts interested in a direct comparison of policy outputs. The conditionality framework could not tell us, for example, that unemployment benefits remain much more generous and encompassing in Denmark than in the UK. Nor could it tell us that levels of investment in active labour market policies for every unemployed person are far lower in the UK than in Denmark, and indeed than in Germany or France. Our motivation here, however, has not been to develop a ‘competitor’ to the more habitual indicators that can better inform us of these dimensions of cross-national variation (in these two examples, social rights indicators and ‘adjusted’ social expenditure indicators, respectively). Rather, it has been to develop a framework that could complement these more standard indicators by capturing other dynamics of change within which other dimensions of variation can be nested. Though, say, the British situation seems as unlike the Danish as it is possible to be on indicators of the generosity of social rights for the unemployed or investment in active labour market policies, it is important for our explanatory perspectives and causal analyses to know – and be able to show – that the UK is in certain structural respects following a much more similar reform trajectory to Denmark than is either France or Germany. The above indicates another advantage of the framework suggested here, namely that of giving rise to new research perspectives or questions. In our example, it would lead us to ask about the causes of cross-national similarities and variation, linking the Danish and British reform profiles in contrast to the French and German. In keeping with Part I of this volume, directed at the ‘dependent variable’ problem in comparative welfare state analysis, this chapter has deliberately not engaged in explanatory analysis (for this, see Clasen and Clegg, 2006b). Instead, it has sought to highlight

186


the substantive advantages of a framework that can be used to tease out and demonstrate the patterns and logics at work in social policy reform that are either missed or not systematically treated within more standard analytical approaches. Though we have focused here on unemployment benefits, our framework could indeed usefully be applied in a similar way to enrich cross-national analysis of a number of other areas of welfare state reform – such as family benefits, disability provision etc. – where similar or equivalent qualitative shifts in risk management, or the social citizenship relationship, seem likely. At a minimum, and more methodologically, this kind of framework can help to steer and discipline qualitative crossnational analysis, limiting the number of potentially relevant observations and organizing findings on a common analytical grid. This in itself would help to promote the generation of more systematic cross-national knowledge on the very many areas of welfare state variation and dynamics of welfare state change that are easily overlooked when it is the available dependent variables that guide the questions we ask in comparative analysis, rather than the other way around.

NOTE 1. The data were collected as part of a UK ESRC commissioned research project (project no: R000223983) that investigated legislative changes in three social security branches (family, pensions and unemployment) in five European countries. National legislative and secondary sources were used to compile the data, with the help of respondents from each of the countries.


187

APPENDIX A Table 8A.1 Major legislative changes in the conditionality of unemployment support, 1980–2005 (United Kingdom) Year

Level* Direction** Measure I

III

1982

II

Abolition of Earnings Related Supplement (ERS) to Unemployment Benefit (UB).

1985

III

Exemption from disqualification provisions in UB for those accepting voluntary redundancy.

1986

III

II

III

Increase in maximum disqualification period in UB from 6 to 13 weeks. Abolition of 1⁄4 and 1⁄2 UB rates for those with incomplete contribution records. Introduction of ‘Restart’ programme.

II III

I

1989

III

Introduction of ‘Actively Seeking Work’ test. After 13 weeks of unemployment, conditions defining suitable work (i.e. limitation of possibility to place limitations on ‘suitable’ or ‘acceptable’ work) abandoned.

1990

I

III

Students no longer eligible for unemployment support. Reductions in Income Support made possible for unemployed claimants failing to attend Restart interviews.

III

II

1981

1988

1992

1996

Higher Supplementary Benefit for unemployed over 60-year-olds who chose to retire early. Registration at Job Centre becomes voluntary for the unemployed.

Tighter contribution requirement for UB. Increase in disqualification period for UB from 13 to 26 weeks. Exclusion of 16- and 17-year-olds from UB and Income Support (IS), except in special circumstances.

Further tightening of disqualification conditions in UB. Reduction of UB for recipients of occupational pensions over 55. Introduction of Jobseekers Allowance (JSA), consisting of ‘contributory JSA’ and ‘incomerelated JSA’.

188


Table 8A.1 (continued ) Year

Level* Direction** Measure II

II

III

III

III

1998

III

Introduction of New Deal programmes for young people under 25 (NDYP) and long-term unemployed (NDLTU) (out of work for 2 years).

1999

III

I

New Deal for Partners (NDP); joint claim required for those with partners claiming JSA for over 6 months (for those without children and under 25).

2001

III

I

I

I

III

2002

2004

Reduction of maximum duration of contributory benefit from 1 year to 6 months. Reduction of contributory benefit rights for unemployed recipients of occupational pensions of all ages. Introduction of requirement to sign a jobseeker’s agreement; and obligatory jobseekers’ directions. Introduction of a ‘permitted period’ of 13 weeks for restriction of job search. Introduction of ‘project work’ pilots for long-term unemployed, (13 weeks compulsory supervised job search followed by 13 weeks work experience).

NDLTU revamped as ND 25 (compulsory after 18 months unemployment within past 21 months). Introduction of NDDP – New Deal for Disabled People (voluntary). NDLP: introduction of compulsory job related interview and interview with personal advisor every 6 months. NDP mandatory for JSA claimants under 45 years of age (if no children). Entry into NDYP and ND25 after 3 months of unemployment (piloted in certain areas).

Notes: * first level: condition of membership; second level: conditions of eligibility and entitlement; third level: conditions of behaviour/conduct; ** ‘–’ represents a tighter/more intensive conditionality: ‘’ represents a more relaxed/less intensive conditionality. Sources: Journal of Social Policy, ‘Social Policy Review’; CPAG, Welfare Rights Bulletin (1980–2003); Clasen (2005).


189

APPENDIX B Table 8A.2 Major legislative changes in the conditionality of unemployment support, 1980–2005 (Germany) Year

Level*

Direction**

Lever

1982

II

II

III

1983

II

Stricter differentiation of duration of entitlement (ALG) in accordance with contribution record (ratio of former to latter changed from 1:2 to 1:3).

1985

II

For employees older than 49: increase in ALG entitlement (max. 18 months; dependent on contribution record).

1986

II

For employees older than 43: increase in ALG entitlement (max. 24 months for those older than 53); dependent on contribution record.

1987

II

II

For employees older than 42: increase in ALG entitlement (max. 32 months for those older than 53); dependent on contribution record. Ratio of entitlement to contribution period reverts back to 1:2.

1994

II

Entitlement to ALH limited to 1 year for those without prior receipt of ALG (indefinite before).

1998

III

Tighter suitability criteria (suitability of job offers defined merely in monetary terms; previous qualification irrelevant; after 6 months: any job deemed suitable with net earnings higher than benefit). Proof of active job search required; stricter benefit sanctions introduced.

Increase of minimum contributory period (Arbeitslosengeld, ALG) from 6 to 12 months. Increase of contribution period (from 70 to 150 days) for means tested unemployment assistance (Arbeitslosenhilfe, ALH). Increase of benefit suspension period (4 to 8 weeks) and tighter suitability conditions.

190


(continued )

Table 8A.2 Year

Level*

Direction**

II

II

III

III

II

2000

II

ALH (for those without prior receipt of ALG) abolished.

2002

III

Job-Aqtiv legislation (more emphasis on activation: job placement, profiling, job search vouchers, temporary work options, job rotation, also training).

2003

III

Tighter suitability criteria for younger unemployed; new job placement and counselling instruments; new options for business start-ups; temporary work placements. In disputed cases, proof of acceptability of job offers transferred from employment office to job seeker.

I

II

I

ALG duration fixed at standard max. 12 months (max. 18 months claimants aged 55 or older). Introduction of ALG II: ALH and social assistance (for employable claimants) to merge into a single scheme

1998

2004/05

Lever Longer ALG duration restricted to over 45-year-olds (previously over 42); max. 32 months only for 57-year-olds (previously 54). Participation in approved training no longer recognized as equivalent to insured employment (i.e. no longer establishes benefit eligibility). Stricter work test imposed on ALG recipients (for ALH claimants since 1996). Employment office can request temporary participation in low-paid seasonal jobs. Introduction of ‘reintegration contract’ stating responsibilities of job seeker and employment office. Improved entitlement for claimants who accept less well paid job (and then become unemployed within 3 years) and for those who lose part-time job.


Table 8A.2 Year

191

(continued )

Level*

Direction**

II

III

III

Lever (wider definition of ‘unemployed’; tighter means test). Tighter job suitability criteria for ALG II recipients (any legal work and wage level suitable even if below collective wage agreement or standard wages paid in locality). Young unemployed (under 25) eligible for ALG II only if they accept offers of training, suitable employment or other job integration measure.

Notes: *: first level: condition of membership; second level: conditions of eligibility and entitlement; third level: conditions of behaviour/conduct; **: ‘–’ represents a tighter/more intensive conditionality: ‘’ represents a more relaxed/less intensive conditionality. Sources:

Information derived from Clasen (2005) and www.bundesregierung.de.

192


APPENDIX C Table 8A.3 Major legislative changes in the conditionality of unemployment support, 1980–2005 (France) Year

Level*

Direction**

1981

II

1982

1984

1988

1991

II

II

II

I

I

III

I

1992 II

II

Measure Creation of an Allocation Exceptionnel de Secours (ASE), for unemployed having exhausted rights to contributory benefits. Based on conditions of age, contribution and income. Introduction of filières d’indemnisation, linking duration of entitlement to contribution conditions (instead of age previously). Benefit duration increased for longer contributors. Benefit durations reduced for some shorter contributors. Reduction of entitlement periods for unemployed with less than 12 months of affiliation to the system, and for those aged 50-55. Introduction of a mechanism allowing individuals over 55 to maintain their unemployment benefit entitlement, without needing to look for work, by withdrawing from the labour market. Introduction of Revenu Minimum d’Insertion, minimum income benefit subsidiary to the others, paid to ‘the excluded’. RMI includes formal obligation to sign an ‘insertion’ contract. Reduction of minimum age to benefit from higher rate of Allocation de Fin de Droits (AFD) (flat-rate minimum benefit in insurance regime), for those deemed effectively out of the labour market: from 55 to 62. Replacement of existing insurance benefits by Allocation Unique Degressive (AUD) Increase in minimum period of contribution from 3 months in last 8 to 4 months in last 8. For each of the different filières, there is an increase in the number of contributions required


193


1997

1998

Level*

Direction**

II

I

II

I

I

2001

Measure for an equivalent maximum period of entitlement. Significant reduction (from 21 months to 7 months) in the duration of benefit entitlement for those with between 6 and 12 months of contributions. AI unemployment assistance benefit no longer available to young job seekers aged 16-25 or to single mothers who have been seeking work for less than 5 years. Degressivity mechanism (whereby benefit levels are reduced periodicially during an unemployment spell) is suppressed for those with very short contribution histories. Introduction of Allocation des Chômeurs Agés (ACA) maintaining full rate AUD until age 60 for those with 40 years of contributions. Introduction of Allocation Equivalent Retraite (AER), a more generous ASS for people with 40 years of pension contributions. Replacement of AUD with new Allocation de Retour à l’Emploi (ARE). Minimum contribution period changed from 4 months in previous 8 to 4 months in previous 18. Introduction of an individual return to work plan (Plan d’Aide au Retour à l’Emploi – PARE), which is compulsory for new claimants of ARE.

II

III

2003

II

Across the board reduction in maximum ARE entitlement durations, varied according to age of claimant.

2003/ 2004

II

Introduction of a maximum duration for previously unlimited Allocation Spécifique de Solidiarité (unemployment assistance, means tested but contributory) benefit, of 3 years (for those currently in receipt) or 2 years (for new claimants). Only applies to those under 55 years of age.

194



2005

Level*

Direction**

Measure

III

New Revenu Minimum d’Activité (RMA), turning RMI into an employment subsidy for those who are in receipt of it for more than 2 years and (likely) for those newly excluded from the ASS.

III

Decree on the control of the unemployed, tightening of sanctions for insufficient job search.

Notes: *: first level: condition of membership; second level: conditions of eligibility and entitlement; third level: conditions of behaviour/conduct; **: ‘–’ represents a tighter/more intensive conditionality: ‘’ represents a more relaxed/less intensive conditionality. Sources: Liaisons Sociales (various years); L’Année Politique, Economique et Sociale (various years); Daniel and Tuchszirer (1999).


195

APPENDIX D Table 8A.4 Major legislative changes in the conditionality of unemployment support 1979–2005 (Denmark) Year

Level*

Direction**

1979

I

Introduction of efterløn; a new benefit giving the unemployed the chance to retire after age 60 and remain on benefit until retirement age, at 67.

1985

I

I

De facto shortening of duration of benefits by ending the possibility for people to requalify for unemployment benefit by participating in public job schemes. Introduction of special extended benefits (lowerrate) for those who have run out of entitlement to normal unemployment benefit due to reform above.

1989

I

Introduction of youth allowance within social assistance for those aged 16 and 17; paid at a lower rate than normal social assistance.

1990

I

III

Introduction of a ‘youth benefit’ in unemployment benefit system for 18- and 19-year-olds. Claimants obliged to accept an offer of work or education from the municipality to receive benefit (50 per cent of unemployment benefit).

I

III

I

I

II

1991

1992

1993/4

Measure

Youth benefit in unemployment benefit system extended to 20-year-olds. ‘Education offers’ (normally available only after a second exhaustion of benefit entitlement, and after completion of a work offer) extended to non-educated unemployed after 12 months (before first work offer). Extension of the ‘youth allowance’ (lower rate) in social assistance to 18-24 year olds. Introduction of transition allowance for the elderly long-term unemployed, giving possibility for effective retirement (on unemployment benefit) at age 55. 1994 Labour Market Reform Duration of unemployment benefit reduced to 7 years (4 years 3 year ‘active period’)

196



Level*

Direction**

Measure (possible extensions: of up to 2 years extension for educational leave , up to one year in the case of parental leave, and 6 months for maternity leave). Limitation of possibilities to requalify for benefits through supported employment: 6 months non-supported ordinary employment now necessary. Introduction of individual action plans for the long-term unemployed and the right and duty to activation after 4 years unemployment. Extension of rights to unemployment benefits for 50- to 59-year-olds.

II

III

I

II

II

III III

III

1996

III

Activation principle extended, for the first time, to people aged 50-59.

1997

III

Act on ‘Active Social Policy’ in social assistance. Compulsory labour market activation or ‘social activation’ for recipients of social assistance. Target group for priority activation shifted from recipients of youth allowance to all unemployed receivers under age 30.

1998/ 99

II

Duration of unemployment benefit receipt limited to a maximum of 4 years (1 year 3 year ‘active period’).

1994

Unemployment benefit receipt limited to a maximum of 5 years (2 years 3 year ‘active period’). Eligibility for unemployment benefit requires 52 weeks ordinary employment in 3 years (instead of 26 weeks in 3 years previously). Right and duty to activation after 2 years. Youth benefit extended: unemployed people up to age 25 must participate in education after 26 weeks. More restrictive definition of acceptable work introduced for those unemployed for more than 6 months. Strengthening of sanctions for those refusing activation.


197


Level*

Direction**

Measure

II

III

III

2000

III

Activation principle extended to 60-year-olds.

2002

I

Reforms to social assistance and ‘introduction allowance’ (for non-EU nationals). Availability for work requirements strengthened.

2003

III

Tighter suitability for work criteria (travel to work) introduced.

Abolition of longer rights to unemployment benefits for 50- to 54-year-olds. Right and duty to activation after one year. Right and duty to activation for all young unemployed after 6 months. More restrictive definition of acceptable work for those unemployed for more than 3 months. Registration at unemployment office required from first day of unemployment.

Notes: *: first level: condition of membership; second level: conditions of eligibility and entitlement; third level: conditions of behaviour/conduct; **: ‘–’ represents a tighter/more intensive conditionality: ‘’ represents a more relaxed/less intensive conditionality. Sources:

Goul Andersen (2002); NOSOSCO (various years).

9. Exploring diversity: measuring welfare state change with fuzzy-set methodology Jon Kvist1 INTRODUCTION Is the glass half-empty? Is it more empty than full? Such questions are often linked to judgements which concern qualitative states and changes in degree and kind. Abounding in comparative studies, such judgements bring forward issues of how best to conceptualize and measure. In comparative studies of the welfare state they prompt reflections on what constitutes the welfare state (see Bonoli, Chapter 3 in this volume), how to operationalize it and how to measure change over time and space. Comparative welfare state research has made significant progress in the theoretical understanding of the welfare state itself, not least due to a dialogue between qualitatively and quantitatively oriented studies (Amenta, 2003). Since 1990, when Gøsta Esping-Andersen published Three Worlds of Welfare Capitalism, a common starting point has been the distinction between different types of welfare state regime: identifying a liberal, conservative and a social-democratic welfare state regime. In short, diversity – the co-existence of similarities and differences – characterizes different welfare states. Comparative research however has made much less progress in the measurement of welfare state and welfare state change (see Clasen and Siegel, Chapter 1 of this volume). A lack of consensus about how to measure either is the main reason why scholars disagree on the direction and magnitude of recent change in social policy, i.e. whether reforms amount to fundamental or marginal change (Clayton and Pontusson, 1998 with Pierson, 1996; or Gilbert, 2002 with Kvist, 1999). Of course, neglecting issues of measurement is not unique to comparative welfare state research. In a review of macro-level comparative studies, Bollen et al. (1993) found that although researchers acknowledge problems of measurement, they largely ignore their consequences. And yet, because 198

Measuring welfare state change with fuzzy-set methodology

199

researchers are often unable to apply statistical tests of data validity and reliability because of the small number of countries (the small-n problem), they have a particular need to reflect on and tackle measurement problems in alternative ways. Otherwise, they run the risk of making (false) heroic conclusions on small n (Lieberson, 1991). Concentrating on the connection between theory and data, a relationship also known as measurement validity (Adcock and Collier, 2001), the subsequent discussion could apply to a large number of substantive areas. However I concentrate on problems of measurement in comparative welfare state studies not only for illustrative purposes but also because this is a large area of research which has paid relatively little attention to methodological aspects of this kind, with some notable exceptions, such as Castles (2002), who defends the use of social expenditure as an indicator for measuring welfare state change. The key question is whether measurement meaningfully captures the ideas contained in concepts and ideal types. Although social expenditure is ‘widely seen as providing misleading indicators of the nature and extent of welfare state activity’, Castles (2002: 618) argues that we should refine the approach as a ‘second best solution’. This chapter offers an alternative approach to measurement and a very different strategy, that of formulating a new way of going about measurement by using fuzzy sets and axioms in fuzzy-set theory. The aim is to advance the application of fuzzy-set theory as a new method for conceptualization and measurement (see Ragin, 2000 for a broad introduction to fuzzy-set social science). I argue that the fuzzy-set approach is particularly useful for assessing diversity and change across a limited set of cases, and that it can overcome some of the problems typically related to measurement validity and precision. In other words, using fuzzy sets help to assess whether the glass is half-full or half-empty, or how, if at all, the welfare state is retrenched or restructured. Introducing the issue and focusing on key theoretical concepts, the subsequent section concentrates on welfare state diversity (1). The following sections argue that cases and ideal types can be viewed as configurations of concepts (2) and discuss how concepts can be conceived and operationalized as fuzzy sets (3). Finally the chapter demonstrates how to formally examine concepts and ideal types with fuzzy-set theory.

WELFARE STATE DIVERSITY AND SOCIAL CITIZENSHIP One set of burning questions in comparative welfare state research concerns whether the welfare state is undergoing retrenchment or restructuring or

200

Residual Figure 9.1


Institutional The dichotomy of residual and institutional welfare states

whether it is resilient to change (Mishra, 1990; Kvist, 1997; Taylor-Gooby, 2002). Before 1990 it was common to distinguish between residual and institutional welfare states (Titmuss, 1958; Wilensky and Lebeaux, 1958; Alber, 1988), locating real welfare states on a continuum stretching from a residual welfare state at one end of the spectrum to an institutional welfare state at the other (Figure 9.1). According to the dominant thinking for the last 30 years, cash transfers in the institutional welfare state were typically universal and generous, whereas benefits in the residual welfare state guaranteed a minimum but were targeted and reserved for the deserving poor. Welfare states, therefore, were perceived as moving only in two directions: either expanding to become more institutional, or retrenching and becoming more residual. Such views regarded historical trajectories of welfare states as initially going through a phase of expansion, eventually reaching a point of maturity (Flora, 1987b) or turning point leading to retrenchment (Mishra, 1990) from the mid 1980s onwards. Advanced by Esping-Andersen (1990), the idea that welfare states come in three types, and not two, did not alter the notion of welfare states moving either in an institutional or a residual direction. Indeed, Paul Pierson’s (1994) influential text on change and politics of the welfare state mainly added a point of no return, i.e. welfare states’ resilience to change. However, since the mid 1990s a growing number of scholars have argued that we are witnessing changes which can not be captured by unilinear, onedimensional conceptions of more, the same or less welfare state (e.g. Kvist, 1997). Revisiting Esping-Andersen’s (1990) Three Worlds of Welfare Capitalism, researchers pointed out that welfare state ideal types reflect different political ideological notions of social citizenship constituted by social rights and obligations (Marshall, 1950) which, in turn, are manifested in specific configurations of benefit characteristics, such as generosity and eligibility. To illustrate, the liberal welfare state model depicts an ideal type where benefits are meagre and targeted at the needy, positioned in the lower left hand quadrant of Figure 9.2, while generous benefits in the conservative model are selective, favouring labour market insiders, thus placing its ideal typical position in the upper left hand quadrant. The ideal typical socialdemocratic welfare state model can be found in the upper right hand quadrant, granting both universal and generous benefits. Finally, the combination of easily accessible but not generous benefits (lower right hand quadrant – and not caught by Esping-Andersen’s trilogy) – is perhaps best


201

described as the Beveridgean, the lib-lab model (Room, 1979), or, simply, the labour model. Methodologically, the above describes welfare state diversity on a lower level of abstraction, i.e. that of social rights. With reference to Robert Adcock and David Collier (2001), the welfare state is regarded as a ‘background concept’ with broad constellations of meanings and understandings whereas social citizenship is a ‘systematized concept’ that entails a specific formulation. Depending on specific research interests, systematized concepts other than ‘social citizenship’ might have been employed to inform the study of the welfare state. Whereas ‘social citizenship’ (or ‘social rights’) are theoretical concepts relating to the output, or policy, side of the welfare state, more outcome oriented research might opt for concepts such as ‘autonomy’ or ‘equality’, whilst more input oriented studies might make use of concepts such as ‘welfare effort’ or ‘popular support’. The use of social rights as a systematized concept allows for the identification of different combinations of ‘less’, ‘the same’ or ‘more’ of the two constitutive dimensions of social rights, accessibility and generosity. In turn, this facilitates the investigation of multidimensional change, or a process described as ‘restructuring’. For example, if the generosity of a particular benefit (I) improves between one point in time (t1) and another (t2) while it simultaneously becomes more difficult to access (see Figure 9.2), it could be argued that the direction of change is towards a conservative welfare state model. The same may also be true in instances where one dimension remains stable and the other dimension enhances a trait which is characteristic of the conservative welfare state model (see II in Figure 9.2). Whether the observed alteration amounts to a qualitative change depends on the start and end points of the benefit trajectory. For example, in Figure 9.1, benefit I is subject to a change in degree, not in kind or type. Put differently, change here means that I has moved closer to the ideal type of the conservative welfare state model. The closer to the corner, the more the benefit reflects the ideal type. The change within benefit I illustrates a situation where it belongs more strongly to the ideal type of the conservative welfare state regime at t2 than at t1. The change, therefore, is quantitative. Benefit II is subject to both quantitative and qualitative change. The quantitative change implies that benefit II becomes much less accessible. Moreover, as Figure 9.2 shows, benefit II is in the upper right hand quadrant at t1 and in the upper left hand quadrant at t2. This shifting of corners reflects a qualitative change. The example demonstrates how benefit II moves from belonging to a social-democratic welfare state model to belonging to that of a conservative welfare state model. Starting points also matter for the assessment of change. Although the change in benefit II may be argued to be stronger and of a more qualitative

202


Generosity Fully generous t2

t2

(II)

t1

(I) t1 Accessibility Not easy

Easy

Fully notgenerous Figure 9.2

Accessibility and generosity of social rights

nature than the change in benefit I, Figure 9.2 still shows that benefit I, rather than benefit II, is closer to the ideal typical corner that symbolizes the conservative welfare state model.

CONFIGURATIONS OF CONCEPTS Social citizenship is constituted of both rights and obligations, and so far we have only set out two dimensions of rights. Looking only at rights and neglecting obligations is in line with most conventional analyses of social citizenship as exemplified by the Social Citizenship Indicator Project (SCIP) at SOFI, Stockholm University, a database which has information on coverage and generosity, but none on obligations. Twenty years ago, researchers and politicians might have been able to justify this neglect in theory and political practice. This is no longer the case. With the general shift in welfare policies towards more active, employment centred objectives, there has been an increasing emphasis on individual obligations, both prior to and during the receipt of benefits. In Figure 9.3, a third dimension, obligations, has thus been added to the core analytical concept. The further a benefit is situated towards the back of this cube, the stronger the attached obligations and vice versa. As


203

Obligations Accessibility

Generosity

Figure 9.3

Analytical property space for social citizenship

a consequence of adding obligations as a third dimension, four new ideal types emerge which may be labelled ‘new’ in contrast to the original four ‘old’ ideal types. For example, the ‘old social-democratic’ welfare state regime is located in the upper front right hand quadrant, while the ‘new social-democratic’ welfare state regime, with a stronger emphasis on individual obligations, is closer to the upper back right hand corner. In a Weberian sense the corners of the cube constitute eight ideal types and can thus be regarded as yardsticks, measuring how close or distant given empirical phenomena are to or from these ideal types and to or from each other (Weber, 1949; Kvist, 1999). In the analysis here, the relevant measurement is the extent to which national welfare states conform to the various ideal typical welfare state regimes, and to what extent national welfare states are moving closer to each other, so-called ‘convergence’, or away from each other, ‘divergence’ (see also Chapter 10 by O’Connor in this volume). Another way of illustrating the diversity and analytical constructs applied here is the use of a truth table, displaying the combinations which are possible (Lazarsfeld, 1937). Table 9.1 shows this method for models of social citizenship, arising from simple yes-or-no dichotomies. Alternatively, a simple eight-cell table might be used (see Table 9.2) which, as Becker (1998) emphasized, has the advantage that the researcher can add a fourth variable by inserting the value in each cell. From the vantage point of concepts and ideal type analysis however, what the eightcell table shares with truth tables is the disadvantage of requiring aspects to be dichotomies.

204

Table 9.1


Truth table of social citizenship

Model New social-democratic Old social-democratic New labour Old labour New conservative Old conservative New liberal Old liberal

Table 9.2

Accessibility

Generosity

Obligations

–

Eight-cell table of social citizenship Accessible Generous

Not-generous

Not accessible Generous

Not-generous

Obligations No obligations

Here the interest is in substantive issues which are insufficiently captured by dichotomies, such as ‘yes’ or ‘no’. Such an approach does not allow the assessment of changes, both small and large, in relation to one another and to some analytical constructs. But how can the conformity of cases to ideal types, and to each other, be constructed and measured? The remainder of this chapter discusses the ways in which fuzzy-set theory may provide innovative and powerful answers to this question.

CONSTRUCTING FUZZY SETS ON CONCEPTS Fuzzy sets are not fuzzy in the sense of being imprecise or ambiguous. On the contrary, fuzzy sets should be designed to accurately reflect theoretical concepts and analytical constructs which have a precise meaning to those researchers using them. Fuzzy sets provide a way of operationalizing a concept into a 0-to-1 metric, from being ‘fully out’ to ‘fully in’ a set. This requires drawing a demarcation line between ‘A’ and ‘not-A’. In the analysis here, operationalizing implies the construction of sets that reflect accessibility, generosity and obligations. These, in turn, will – in different configurations – constitute different ideal types of welfare states.


205

Lines between different membership sets are drawn on the basis of substantive and theoretical knowledge. By having to draw a line or curve reflecting the particular concept under consideration, the researcher centres his or her focus on the concept rather than on the variables themselves. Focusing on the concept moves the analysis closer to the theoretical body that deals with concepts in the first place. A reference to ‘generous benefits’ is more informative than speaking of ‘benefits with a net replacement rate above X percentage’. Moreover using this term also helps to minimize measurement bias, that is, the gap between theory and reality. While ‘fully generous’ and ‘fully not-generous’ refer to extremes, many intermediary concepts link these two categories. Depending on the substance of the concept and the raw material, various fuzzy category intervals may be used (see Ragin, 2000). Here a nine-value fuzzy set is applied, where continuous fuzzy scores between 0 (fully out) and 1 (fully in) indicate partial membership in the following way: ● ● ● ● ● ● ●

scores from 0.83 to 0.99 is almost fully in; 0.67 to 0.82 is fairly in; 0.51 to 0.66 is more or less in; 0.5 is the cross-over point where the case is neither more in nor more out; 0.33 to 0.49 is more or less out; 0.17 to 0.32 is fairly out; 0.01 to 0.16 is almost fully out.

Using this nine-value fuzzy set throughout the chapter helps to translate interval fuzzy membership scores into verbal concepts or verbal qualifiers. For example, if a benefit has a fuzzy score of 0.75 the score is presented as a ‘fairly generous’ benefit, a fuzzy score of 0.60 translates to a ‘more or less generous’ benefit. Constructing fuzzy sets involves two steps: first to establish empirical indicators for the fuzzy set, and second to calibrate the fuzzy set. The following two subsections elaborate these steps. Empirical Indicators To reduce the gap between theory and reality, empirical indicators are needed which reflect the chosen concepts as closely as possible. The quest for useful empirical indicators should be guided by theories and substantive knowledge with reflections made explicit. Applying the example of social citizenship, three sets have been identified which reflect theoretically important concepts. The first set, accessibility of

206


unemployment benefits, is measured by an index based on scores for the scope of application and various eligibility criteria (e.g., work demands, definition of employment records, and membership requirements, if any). The set for generosity of unemployment benefits is measured by net replacement rates that express the ratio of benefits to former wages after taxation. This ‘generosity’ measure has become common in the literature. Here the net replacement rate for a single person with previous earnings at the level of the Average Production Worker (APW) has been used. This has two main caveats: first, net replacement rates calculated at other points in the income interval may suggest deviating values of generosity. Second, the existence of tax allowances and/or supplements for children may cause differences in net replacement rates of persons in single individual households and non-single households. Aggregate measures such as average net replacement rates for different income and family situations do not indicate how national systems work for any particular population group but simply conflate otherwise useful information. As most national unemployment insurance schemes are strongly individualized, and as unemployment is concentrated among groups with lower levels of education, the net replacement rate for a single APW as empirical indicator for benefit generosity seems justified here. The third set on obligations of unemployment benefit claimants can be measured in numerous ways. Here an index of negative sanctions, as stipulated in legal texts, has been applied. In other words, the measurement reflects negative sanctions that may be imposed if a person becomes unemployed voluntarily or because of misconduct, and on benefit claimants who refuse to accept a job offer or participation in an active labour market programme. Acknowledging that the implementation of sanctions may not always follow the letter of the law, legal stipulations give at least an important signal to both administrative authorities and claimants, and can therefore be seen as reflecting politicians’ positions on the issue of obligations. Calibration of Sets Having identified the best possible empirical evidence, how do the data reflect theoretical concepts? In practical terms, the best approach is, first, to establish when something is fully in and fully out of the set and, second, to fine-tune the set by describing how it looks in the range from fully out to fully in. This calibration of sets must be informed by theoretical and substantive knowledge since it affects the measurement of fuzzy membership scores. Any fuzzy-set analysis is only as good as its sets, making the infusion of knowledge into sets indispensable. The starting point for the accessibility index here is the assumption that people aged between 18 and the official retirement age should be able to


207

Table 9.3 Specification of empirical indicators and the translation of raw data into fuzzy membership scores and verbal labels 





90.0 82.0–90.0

90.0 79.3–89.9

85.0 69.5–85.0

1 0.84–0.99

72.0–81.9 60.0–71.9

67.7–79.2 55.6–67.6

54.9–69.4 41.3–54.8

0.68–0.83 0.51–0.67

59.0–59.9 47.0–58.9

54.5–55.5 42.4–54.4

40.2–41.2 27.6–40.1

0.50 0.34–0.49

37.0–46.9

30.8–42.3

16.0–27.5

0.18–0.33

28.9–36.9

20.1–30.7

5.4–15.9

0.01–0.17

28.9

20.0

5.4

0

Fuzzy Verbal labels membership scores Fully in the set Almost fully in the set Fairly in the set More or less in the set Cross-over point More or less out of the set Fairly out of the set Almost fully out of the set Fully out of the set

Notes:  to unemployment benefits is measured by an index taking into account personal scope of application, age groups and eligibility criteria.  of unemployment benefits is measured by net replacement rates for a single person with earnings at level of APW (%).  of claimants are measured by an index of negative sanctions imposed on claimants refusing to accept job and ALMP offers.

qualify for unemployment benefits after six months of employment within a 12 month period, taking into account activities other than ordinary paid work which might count towards eligibility, e.g. training and child caring. If qualification is possible under these conditions, the benefit system is deemed to be more easy than difficult to access (i.e. the membership score is greater than 0.5). If these conditions are insufficient for benefit qualification, the system is difficult to access (i.e. membership scores lower than 0.5). Table 9.3 shows the translation of raw data – index scores – into fuzzy membership scores and labels. The higher the index score, the easier the access to benefits. For the set on generosity, the first qualitative breakpoint occurs when the benefit is fully not-generous. Below this point variation is meaningless since distinguishing between degrees to which benefits exceed ‘fully not-generous’ does not make sense. The second qualitative breakpoint occurs when the benefit is fully generous. Above this point variation is meaningless because distinguishing between degrees to which benefits exceed ‘fully generous’

208


does not make sense. The third qualitative breakpoint is the cross-over point, where the benefit switches from being ‘more not-generous than generous’ to becoming ‘more generous than not-generous’. According to national consumption surveys (Hansen, 1998) persons cannot maintain any attained standards of living if their income is reduced by four-fifths, as they will soon have to rearrange their financial affairs dramatically. Hence, if the net replacement rate is below 20 per cent, we deem it fully not-generous. Having a job or participating in an active labour market policy programme involves costs for mobility and various other expenses. In most countries – for example, Denmark – workers are granted tax allowances to partially cover such costs and participants in active labour market policy programmes may earn something extra before their benefits are reduced. Both the earnings exemption and the tax allowances amount to approximately 10 per cent of the APW earnings in the Danish example. For this reason we label net replacement rates of 90 per cent and more as fully generous. Establishing when benefits are more generous than not is more difficult. We have put the point at 55.5 per cent. For the specific translation of net replacement rates into fuzzy scores and labels, see Table 9.3. The final fuzzy set on obligations concerns the severity of negative sanctions, measured by an empirical indicator of the number of weeks for which claimants may have their benefits suspended and the timing thereof. The earlier strict sanctions are imposed, the higher the score. The longer – and thus more severe – the sanctions, the higher the index score. Table 9.3 also shows the translation of fuzzy membership scores into nine verbal labels, ranging from ‘fully accessible’ to ‘fully not-accessible’. These labels are used for the analysis of the conformity of cases to concepts and ideal types. For example, if a benefit scores 70.2 in the set on generosity this translates as a ‘fairly generous’ benefit. Scoring Cases How can fuzzy sets be constructed? Basically, there are two options. The first is to separately investigate each dimension of the policy development; the second is to use formal set theory axioms to study configurations of sets. This section gives a brief example of the first option applied to empirical developments in Denmark. The subsequent section illustrates the potential of applying the second option. In 1990, Danish unemployment insurance benefits were almost fully accessible (see Table 9.4). By 1998, the benefits had become only more or less accessible, because the required work period preceding unemployment increased from 26 to 52 weeks. Benefit generosity fell in the same


209

Table 9.4 Fuzzy membership scores for Danish unemployment insurance benefits in ,  and , 1990–98

  

1990

1995

1998

0.98 0.71 0.22

0.74 0.63 0.73

0.53 0.60 0.94

period from fairly generous to more or less generous (see Table 9.4). This drop did not result from any direct cuts in benefit levels or in the benefit formulae but came about for two reasons: benefit indexation lagged behind wage developments and the introduction of a gross tax (a so-called ‘labour market contribution’) of 5 per cent in 1994, increased to 8 per cent in 1997. However obligations have seen the most dramatic degree of change. During the 1990s demands on wage and geographical and occupational mobility on the part of unemployed benefit claimants became stronger, accompanied by tougher negative sanctions for the rejection of jobs or training offers. While obligations were fairly lax in 1990, they became almost completely strong by 1998 (see Table 9.4). In other words, the marked development of obligations also led to a qualitative change from lax to strong obligations. In the 1980s benefits were easy to access in Denmark, with hardly any strings attached, leading some Danish observers at the time to describe the unemployment benefit system as a citizen wage (Langager, 1997). This description is no longer accurate. The tightening of eligibility criteria and the strengthening of obligations means that there is ‘no free lunch’ when it comes to claiming unemployment insurance benefits. Configuration of Fuzzy Sets into Ideal Types Esping-Andersen’s (1990) welfare state typology lives up to Weber’s definition of an ideal type as ‘formed by the one-sided accentuation of one or more points of view and by the synthesis of a great many diffuse, more or less present and occasionally absent concrete individual phenomena, which are arranged according to those one-sidedly emphasized viewpoints into a unified analytical construct’ (Weber, 1949: 147). Here, fuzzy-set theory is used for the study of ideal types. Although fuzzy sets and ideal types may be seen as opposites, they are not. Fuzzy-set theory

210


can be applied to the configurational view of crucial aspects and concepts combining in ideal types (see Ragin, 2000; Kvist, 2006). Logical operations in fuzzy-set theory allow the construction and measurement of alternative types in formal and precise manners. Basically, operations with fuzzy sets are generalizations of operations on crisp sets (see Zadeh, 1965; Ragin, 2000). Suppose case x has a membership value va in fuzzy set A for , a membership value vg in fuzzy set G for , and a membership value vo in fuzzy set O for . In the presentation of the analytical property space (Lazarsfeld, 1937; Becker, 1998), the new social-democratic model of social citizenship in relation to unemployed people can be expressed in fuzzy-set terms as the ideal typical location –  * *, (A*G*O*) – or, in plain English, as a model characterized by easily accessible, generous benefits with strong obligations imposed on claimants. Fuzzy set theory contains formal rules for dealing with set theoretic relationships like this one. In this case, we can make use of the intersection rule. The value of x in A*G*O is the minimum value of va, vg, and vo. This operation represents logical , denoted *, and is called the minimum principle in fuzzy-set theory. Applying the minimum principle to the membership scores in Table 9.4 suggests that the minimum value for Denmark in 1990 was 0.22. Put differently, due to lax obligation requirements Denmark was fairly out of the new social-democratic model in the early 1990s. As obligations were strengthened markedly in 1994, Denmark changed to belonging more or less (0.63) to the new social-democratic model by 1995. Due to tighter eligibility criteria making access to benefits more difficult in 1996, Denmark was barely a member of the new social-democratic model in 1998. Of course, cases’ membership in other models can be studied too. The old social-democratic welfare state model, for example, comprised easily accessible and generous benefits that were not subject to strong obligations. Here the complement rule can be applied, i.e. that the value of x in A is 1va, where means ‘not’. This operation finds the complement to A, and is called principle of negation in fuzzy-set theory. The value of x in A*G*O is the minimum value of va, vg, and 1vo. Again, looking at the scores in Table 9.4 suggests that Denmark operated a fairly old social-democratic welfare state model, i.e. equal to a score of 0.71, in 1990. However, eight years later it was almost fully out, with a score of only 0.06 (i.e. 10.94). Behind this fundamental change in the Danish model lies a labour market reform with a series of subsequent modifications, exacerbating certain traits (such as increased obligations) uncommon to the


211

old social-democratic model. Combining tougher obligations with stricter accessibility (see Table 9.4) moved the Danish case into the boundary regions of the new conservative (0.47) and new social-democratic welfare state models (0.53). Finally, let us accept, for the moment, Esping-Andersen’s (1990) argument that there is only one ‘liberal’ model, i.e. that a distinction between liberal and labour is not necessary. In this case the rule of union can be applied, e.g. the value of x in AGO is the maximum value of va, vg, and vo. This operation represents the logical , denoted , and is called the maximum principle. Applied to an analysis of the liberal model, it means that the value x in (A A)*G is given by the maximum of these two expressions. In plain English, it means that the fuzzy membership score of the liberal model is the highest score of the model characterized by either not-accessible, not-generous benefits or by accessible, not-generous benefits. Note that the aspect of obligations is left out entirely. The view of cases as configurations of aspects introduces the idea that a single difference in an aspect between two cases may constitute a difference in kind – a qualitative distinction. Moreover, the analytical property spaces or truth tables indicate that aspects should not be viewed as independent, separable variables, but rather as elements of configurations (Ragin, 2000). Of course, in principle it remains possible that not all eight feasible combinations have empirical validity or are of theoretical relevance. However, even when some of the ideal types are empirically irrelevant, listing them helps the researcher to get an overview of the subject (see Ragin, 1987; Becker, 1998; and Ragin, 2000 for set theory ways of reducing the property space).

IDEAL TYPE ANALYSIS Table 9.5 sets out fuzzy membership scores for seven countries in the eight possible welfare state ideal types for unemployment insurance. Using these qualitative distinctions, it can be analysed which ideal type a country comes close to and its degree of membership determined. Moreover, statements can be made about which ideal type the country is furthest away from. This analysis allows nuanced judgements on the (shifting) character of national welfare states. Table 9.5 shows that Denmark and Sweden have moved from belonging to an old social-democratic unemployment insurance model to belonging to a new social-democratic model. Moreover, these were not incremental shifts. Both Sweden and Denmark moved from being fairly out of the new socialdemocratic model to becoming fairly in, and more or less in respectively. Indeed, the greater emphasis on obligations in the two Nordic countries can

212

1990 1995 1999 1990 1995 1999 1990 1995 1999 1990 1995 1999 1990 1995 1999 1990 1995 1999 1990 1995 1999

Denmark

Sources:

UK

Germany

0.22 0.63 0.53 0.38 0.38 0.48 0.65 0.65 0.64 0.22 0.22 0.71 0.40 0.28 0.28 0.45 0.28 0.45 0.04 0.03 0.00

New socialdemocratic 0.71 0.27 0.06 0.62 0.62 0.52 0.25 0.25 0.25 0.78 0.77 0.19 0.41 0.28 0.08 0.34 0.28 0.34 0.04 0.03 0.00

Old socialdemocratic 0.22 0.37 0.40 0.38 0.38 0.43 0.35 0.35 0.35 0.04 0.19 0.29 0.25 0.25 0.28 0.42 0.25 0.45 0.48 0.48 0.51

New labour

Hansen (various years); NOSOSKO (various years); Kvist (2002).

Netherlands

Sweden

Norway

Finland

Year

Country

0.29 0.27 0.06 0.38 0.38 0.43 0.25 0.25 0.25 0.04 0.19 0.19 0.25 0.25 0.08 0.34 0.25 0.34 0.52 0.52 0.49

Old labour 0.02 0.26 0.47 0.34 0.29 0.48 0.22 0.22 0.36 0.08 0.22 0.23 0.40 0.40 0.72 0.55 0.40 0.54 0.04 0.03 0.00

New conservative 0.02 0.26 0.06 0.34 0.29 0.42 0.22 0.22 0.25 0.08 0.23 0.19 0.59 0.60 0.08 0.34 0.60 0.34 0.04 0.03 0.00

Old conservative 0.02 0.26 0.40 0.34 0.29 0.42 0.22 0.22 0.35 0.04 0.19 0.23 0.25 0.25 0.28 0.42 0.25 0.46 0.48 0.48 0.48

New liberal

Table 9.5 Fuzzy membership scores for seven European countries in unemployment insurance ideal types, 1990–99

0.02 0.26 0.06 0.34 0.29 0.42 0.22 0.22 0.25 0.04 0.19 0.19 0.25 0.25 0.08 0.34 0.25 0.34 0.48 0.48 0.48

Old liberal


213

be interpreted as having brought about a qualitative change in the unemployment insurance model. Furthermore, these two countries were almost fully out of many of the other ideal types during large parts of the 1990s. In other words, their unemployment insurance models are fairly distinctive most of the time. In contrast, Finland presents a more ambiguous case. Throughout the 1990s, Finland belonged to the old social-democratic ideal type, although to varying degrees. However, as the scores in Table 9.5 show, Finland is only more or less out of a number of other models. This can be interpreted as Finland having an unemployment insurance model which is less distinctive than the ones in the other two Nordic countries. The fourth Nordic country, Norway, had a fairly strong new social-democratic unemployment insurance model during the 1990s. Moreover, Norway was neither close to nor very distant from many other models. In short, the Nordic countries live up to the expectation that they operate social-democratic unemployment insurance models. However, with the exception of Finland, by the end of the 1990s they belonged to a version which could be labelled the new social-democratic model, stressing strong obligations. Due to welfare reforms, particularly in Denmark and Sweden, there has been some convergence toward the new social-democratic model in unemployment insurance across the Nordic countries. However there has been no convergence between the four Nordic and the three non-Nordic countries listed in Table 9.5. Although all seven countries intensified obligations on the part of the unemployed, the result was continued diversity, or what we have described elsewhere as ‘parallel trends, persistent diversity’ (Kautto and Kvist, 2002). The Netherlands in particular experienced a qualitative change from an old to a new conservative welfare state model. In contrast, developments in Germany and the UK have been less dramatic. Being a member of the new conservative model, Germany is not as distinct as the Netherlands, a fact which is underlined by an examination of its membership scores in other models. The scores for the UK indicate a distinctive case, i.e. that the country is fully out of both social-democratic and conservative models. In fact, the UK is situated in the border region between the labour and the liberal models. The UK seems to have stronger affinities across the Atlantic than with the other European countries.

CONCLUDING REMARKS The ideal type analysis has illustrated two advantages of using fuzzy-set theory for measurement purposes. First, fuzzy sets can be constructed in order to reflect the ideas of theoretical concepts, thereby directly tackling

214


the key concern of achieving measurement validity. Fuzzy-set theory demands a high degree of correspondence between concepts and fuzzy membership scores in sets established to reflect such concepts. Researchers must pay great attention to constructing analytical concepts, criteria for establishing qualitative breaking points, and the empirical evidence. Crucial decisions need to be made on the basis of theory, substantive knowledge and the availability and nature of data. In any case, such decisions should be sufficiently explicit to allow for scientific dialogue and replication of the analysis. Second, it has been demonstrated how axioms in fuzzy set theory allow the interrogation of the analytical property space (i.e. the inside of the cube), and the observation of how cases move around over time. Using comparative welfare state studies for illustrative purposes has indicated how concepts such as ‘resilience’ and ‘retrenchment’ are inadequate for capturing ongoing welfare reform. Instead, often the latter could be better characterized as ‘restructuring’ due to simultaneous change in several dimensions. Using fuzzy-set theory allows qualified statements on such changes and, in the case of unemployment insurance, the identification of a cross-country process of parallel trends but persistent diversity. In short, the chapter has aimed to illustrate how advanced fuzzy-set theory can be used for measurement purposes. If nothing else, applying the approach requires researchers to be more knowledgeable about theory and cases. In turn, such knowledge may greatly inform discussions of whether the glass is half-empty or whether it is more empty than full.

NOTE 1. I would like to thank Brian Gran, Olli Kangas and Charles Ragin for helpful comments on an earlier version of this chapter.

PART IV

Capturing the nature of welfare state change

10. Convergence in European welfare state analysis: convergence of what? Julia S. O’Connor INTRODUCTION Our understanding of institutional change and convergence is generally more intuitive than systematic. We think we know what we mean by convergence but in fact several alternative understandings are possible. Institutions and policies could become more alike, for instance, by becoming more like those already in existence in one country, or they could all change to some new configuration. (Kitschelt et al., 1999: 438) . . . policy convergence can be defined as any increase in the similarity between one or more of the characteristics of a certain policy (e.g. policy objectives, policy instruments, policy settings) across a given set of political jurisdictions over a given period of time. (Knill, 2005: 768) [Convergence is] the tendency of societies to grow more alike, to develop similarities in structures, processes and performances. (Kerr, 1983: 3)

These quotations point to two levels at which convergence is considered: institutional convergence and policy convergence. Studies of convergence often concentrate on the change in ‘performances,’ to use Kerr’s terminology, as reflected in structural and/or social indicators such as GDP per capita, social expenditure as a percentage of GDP or rates of unemployment but the basis on which convergence, or its absence, is asserted is often not sufficiently specified. This includes a failure to identify clearly the dependent variable, the scope of convergence and the relevant time period over which it is measured. With a view to assessing its usefulness in comparative welfare state analysis and the apparently contradictory claims relating to convergence or its absence, this chapter addresses three issues: first, it asks what are the essential elements of convergence, and how it might be measured. Second, it situates an examination of convergence in European Union welfare states within the context of a discussion of European integration which has the objective of political, economic and 217

218


social integration; it also focuses on Europeanization and harmonization as processes that differ from, but may contribute to, convergence. Third, it examines the evidence on convergence in the European Union 15 under three headings: convergence as decreased variation in measures of welfare effort, convergence through catch-up of laggards on leaders and convergence of ideas through change in the domestic frame of reference associated with the process of European integration. The chapter concludes with a discussion of the implications of the analysis for the effective measurement and examination of convergence in studies of welfare states in the context of changing global pressures.

CONVERGENCE: A CONCEPT WITH MULTIPLE MEANINGS The common element in the definitions of convergence outlined above is the concept of similarity. But similarity is an ambiguous and subjective concept without clear cut-off points. Overcoming this ambiguity calls for specification of ‘what’ is converging, the period of time over which change is measured and how it is to be measured. To this end I consider (1) societal level convergence; (2) policy convergence; and (3) measurement of convergence. Societal Level Convergence Macro-level convergence explanations are almost synonymous with the logic of industrialism view of social change originally formulated in 1960 by Kerr et al. as a response to the Marxist logic of capitalism analysis, which argued for convergence towards crisis because of the inherent contradictions of the capitalist mode of production (O’Connor, 1973). The essence of the logic of industrialism argument is that industrialization is associated with economic growth, changes in social organization, class and demographic structure and the emergence of new needs, such as a dependent population over retirement age. Social programmes are considered as a response to these needs and are facilitated by the additional resources available because of economic growth and technological change (Wilensky, 1975; 1976).1 The economic and technological determinism of the logic of industrialism and logic of capitalism approaches to social change have few proponents in the contemporary period. Yet, there is widespread acknowledgement that there are common pressures on national governments associated with globalization of trade, production and technological developments, trade openness and internationalization of finance, although the

Convergence in European welfare state analysis

219

positive and/or negative consequences of these and the ability and political commitment of national governments to resist them are debated. This resistance must take into account the role of supranational regional influences such as the European Union which may be seen as interacting with, exacerbating or filtering global pressures. Whereas the earlier arguments about convergence focused on addressing the development of welfare effort, as reflected in the percentage of GDP devoted to social welfare benefits and activities, the more recent variants have focused on whether or not there has been an impact of exogenous pressures on existing welfare state effort, in particular on whether a range of pressures associated with globalization have been associated with downward convergence in objectives and outcomes through retrenchment. However, at the regional level there may be different, or at least additional, dynamics operating. The European Union identifies economic and social convergence as a policy objective, for example. I consider this issue in the next section. Focusing on the global level some analysts have asked whether the exogenous pressures associated with capitalist development in the second part of the 20th century have been associated with institutional change and the convergence, or not, of all models of capitalism towards a neo-liberal model (Kitschelt et al., 1999). Based on variation in the degree to which scarce resources are controlled politically or by the market, Kitschelt et al. identify three models of the development of capitalism since the 1950s: the Liberal Market Convergence Model, the Mixed Economy Convergence Model and the Organized Capitalism Convergence Model.2 What is at issue here is not convergence or divergence but the path dependent contexts within which convergence pressures occur and their interaction with divergence pressures. The authors conclude that the three models demonstrate different types of convergence over three periods, 1950–73 (so-called ‘Golden Age of Postwar Capitalism’), 1973–82 (‘Initial Shocks and Crises’ associated with the oil crises) and 1982 onwards (‘Transformation of Advanced Capitalist Economies’). They examine pre-existing institutional arrangements, particularly those relating to producer groups, political parties and the capacity of the bureaucracy to influence and possibly constrain the effects of common exogenous challenges, and conclude that it is unlikely that there will be convergence to any unique model of capitalism. Of course this does not preclude convergence at sectoral level or on a range of dimensions. Kitschelt et al. (1999) argue for ‘a logic of “refracted divergence” in which some of the past patterns of diversity disappear, are replaced by new ones, reflecting institutionally mediated responses to the challenges posed by the new environment’ (ibid.: 443). In striving to balance economic growth, income inequality, political participation and government effectiveness within the context of contemporary capitalism, and the associated

220


institutional constraints, countries respond in partially path dependent ways and path dependence is greater where institutions are more embedded. Societal convergence may encompass policy convergence that ranges from relatively minor to extensive but the absence of societal convergence does not necessarily imply the absence of policy convergence. Furthermore, an exclusive macro-level focus may obscure processes leading to policy convergence (Bennett, 1991: 217; Busch and Jorgens, 2005: 862). Analysis of policy convergence raises a wide range of questions. Do findings of convergence or its absence refer only to one aspect of a policy area or policy or the entire area or policy; one dimension of a programme, such as a regulatory programme; one aspect of social protection or the entire programme or even the orientation of the overall social policy framework? And can convergence in some programmes co-exist with divergence in other programmes? This chapter focuses largely on dimensions of policy convergence with particular focus on policy effort as reflected in social protection expenditure processes. Policy Convergence In discussing policy convergence it is important to delineate its relationship with several other concepts which are sometimes used synonymously with it, sometimes as causal factors, and sometimes as mechanisms through which convergence is achieved. Policy transfer and policy diffusion are the most frequently occurring of these (Dolowitz and Marsh, 2000; Knill, 2005). Policy transfer and policy diffusion are processes of policy change describing how knowledge about policy characteristics is transferred or diffused. In contrast, convergence is the process effect. Policy convergence describes ‘the end result of a process of policy change over time towards some common point regardless of the causal process’ (Knill, 2005: 768). This process may be a policy transfer or policy diffusion although neither necessarily results in convergence (Radaelli, 2005). These concepts are often used interchangeably and seen to result from similar causal factors although some authors emphasize the uncoordinated and voluntary nature of diffusion. In contrast to these processes, the dependent variable in analysis of policy convergence is the change in the difference between units, for example policy goals, content, instruments, or expenditure levels, over a period of time. Measuring Convergence Overcoming the ambiguities associated with the concept of similarity underlying convergence necessitates clear specification not only of what is


221

converging but of the different types of policy development and of the time over which change is measured. Contrary to many of the earlier studies arguing for convergence, which based their conclusions on the demonstration of differences across countries at different levels of development at one point in time, contemporary analyses accept that convergence, as a temporal concept, must be demonstrated over time and across units. This means that ‘conclusions are drawn from the changes in difference over time’ (Seeliger, 1996: 289, emphasis in original). While the following illustrations use quantitative terminology, analysis does not have to be quantitative or use large numbers of cases; what is required is clear classification of the dependent variable which allows transparent measurement of change. The type of policy development depends on the direction of change in the difference () between entities at time one (t1) and time two (t2). Convergence is indicated if the difference in t2 is less than the difference in time t1 ( t1 t2; divergence is indicated if the difference is greater ( t1 t2).3 These analytical categories refer to relative policy development for a defined time period. The importance of specifying the time period has been demonstrated empirically in studies of welfare effort; for example, O’Connor (1988) and Bouget (2003) demonstrate periods of divergence within longer periods of convergence in social expenditure in OECD countries. Furthermore, the type of policy issue, the policy dimensions and the length of their typical policy cycle are all relevant for deciding on the appropriate time frame over which convergence should be measured. The two most widely identified types of convergence are decreasing variation over time, or sigma convergence ( ), and catch-up by laggards on leaders over time, that is beta convergence (). Sigma convergence is measured using the standard deviation or coefficient of variation; for example we may wish to measure the extent to which variation decreases over time in the European Union 15 or 25. Here we are talking about the variation across the entire EU 15 or 25 at different times. This must be supplemented by examination of the patterns for individual countries – to what extent individual countries are converging or diverging; whether some countries are contributing disproportionately to the overall pattern. A necessary condition for the existence of sigma convergence is catch-up by some units or beta convergence (). But the opposite is not necessary. We could find catch-up by some units without a decrease in the overall variation measured across the entire group. Catch-up convergence is measured by correlation or regression analysis demonstrating for example that social expenditure as a percentage of GDP at a certain time is negatively associated with its level at an earlier period and positively associated with exogenous variables influencing the growth of social expenditure. Improvement in the position

222


of the Cohesion Countries relative to the EU 15 level of expenditure or of the 2004 accession countries relative to EU 25 expenditure would exemplify catch-up convergence. This could be achieved by above average growth in social expenditure in the cohesion or accession countries and/or a slow down in its growth in other EU 15 or EU 25 countries. Although not widely used in welfare state analysis, mobility of units exemplifying a change in ranking over time may assist in specifying the type of change that is taking place for example within the European Union over time.4 These distinctions are important for interpreting results of convergence analysis and particularly in making comparisons across studies. Elements and Dimensions of Policy Convergence If we accept that convergence is an over time process effect, we specify the time period over which it is measured and identify the type of convergence being demonstrated, we are still left with the issue of the elements of policies being compared in policy convergence research (Bennett, 1991). Some analysts argue that the focus should be confined to policy output, on the grounds that the impact of intervening variables between output and outcome makes the identification of causative links very difficult in the case of outcomes (Holzinger and Knill, 2005: 776; see also Green-Pedersen, Chapter 2 in this volume). However, it is important to recognize that outcomes are increasingly the focus of convergence in the European Union reflected in the emphasis on Structural Indicators as an instrument to assess the progress towards meeting the Lisbon objectives (European Commission, 2006a; Eurostat, 2006b). Consequently, depending on circumstances, policy outputs, or outcomes as reflected in social indicators, may be the appropriate focus for the measurement of convergence. In line with this argument, policy convergence may refer to policy goals (for example, lower relative poverty rates, increased labour force participation or reduction of regional disparities in income); policy content (formal aspects such as legislation, regulations, administrative rules); policy instruments (regulatory, administrative or judicial tools); policy effort or orientation (as reflected in level and types of expenditure), or policy outcomes (poverty rates, labour force participation rates or regional income disparities). In summary, convergence or divergence is a process effect and must be demonstrated to have occurred over time and across units. It is the end result of quantitative or qualitative change in similarity or difference across units over time – these units may be societies, policies, elements of policies or expenditure. Its measurement is complex. A conclusion that convergence has occurred is indicated only if variation across units has decreased over a specified time period or if there has been catch-up by laggards on leaders


223

over time or if there has been a change in ranking over time. Specification of the type of convergence is theoretically and empirically important in analysing and explaining the nature of the change taking place. In addition, the elements and dimensions of policy must be precisely specified if legitimate comparisons are to be made. We now turn to the issue of convergence in the EU 15 welfare states with some reference to other OECD countries.

CONVERGENCE IN EUROPEAN UNION WELFARE STATES In broad terms we can identify references to two types of convergence in the EU: on the one hand, references to upward economic and social convergence as an objective of European integration are pervasive in EU documents; on the other hand, downward convergence in the social policy area is identified as a probable negative unintended consequence of action on economic integration and more broadly as a response to the pressure of globalization forces. These facts make the EU a particularly interesting locus for analysis of convergence. Before discussing the evidence relating to types of convergence in EU welfare states it is necessary to outline what we mean by European integration and Europeanization and how they might contribute to convergence in social policy. European Integration and Europeanization European integration and Europeanization are frequently used interchangeably but they are not synonymous. In Radaelli’s words, Europeanization is ‘concerned with what happens once EU institutions are in place and produce their effects’; and it is distinct from the process leading to EU policy formation – it is ‘the reverberation of that policy in national arenas’ (Radaelli, 2003: 33–4). As with globalization and convergence, Europeanization is variously defined and often mistakenly used synonymously not only with European integration but with convergence and harmonization. Europeanization is not convergence – but convergence can be a consequence; it is not synonymous with harmonization – harmonization reduces diversity, Europeanization may or may not. Clearly Europeanization is a contested concept and is the subject of a voluminous literature. It is used in this chapter in the sense of change by and in Member States to meet the requirements and/or consequences of European integration. Suffice it to say here in relation to the European integration-Europeanization-convergence link that its examination

224


requires recognition of the mechanisms underlying European integration, namely, positive and negative integration (Scharpf, 1998) and framing mechanisms including the Open Method of Coordination (OMC). Positive integration refers to action taken by the Union to bring Member States’ policies and practices into conformity with a European institutional model, for example directives on environmental protection, health and safety, consumer protection and some aspects of social policy broadly conceived. Europeanization refers to the incorporation of these into the domestic framework. Negative integration refers to the removal of barriers to EU integration, for example barriers to competition to facilitate the single market. This may change the distribution of power and resources between actors in Member States. Negative integration is strengthened by European Court of Justice rulings ensuring Member State compliance with the removal of barriers to competition. These may impact on those social policy areas, such as health insurance, where a market dimension exists and where competition issues arise (Leibfried and Pierson, 2000). Privatization of other social and health areas will widen their range of influence and may constrain traditional social policy making capacity and increase pressure for convergence in strategies if not in principles. Rather than mandating particular changes, as in positive integration or altering the domestic situation to facilitate integration as in negative integration measures, framing integration alters the domestic frame of reference. Some analysts interpret this as making the domestic environment more supportive of future changes. For example, some recommendations and directives that can be transposed though relatively symbolic action may change the domestic frame of reference and the context within which future directives demanding greater change will be debated (Knill and Lehmkull, 1999). The concept of framing is used in a related but stronger and more encompassing sense by Liebert and colleagues in their studies of gender equality directives which they conclude have enhanced convergence without harmonization of EU Member States’ policy frameworks (Liebert, 2003).5 These two usages of the concept of framing indicate that it results from the working out of positive integration measures such as directives but increasingly in the social policy area it may also result from ‘facilitated coordination’ which is the term used by Bulmer and Radaelli (2005) to refer to the broad range of mechanisms of coordination where the EU institutions have relatively weak powers and where the concern is with the convergence of ideas. These include political declarations as at Council summits and/or soft law which are not legally enforceable but are directed to influencing policy practice. The OMC, which is based on soft law mechanisms such as guidelines, indicators, benchmarking and sharing of best practice and voluntary compliance, to which EU Member States have committed themselves, is an


225

increasingly important example of facilitated coordination or framing integration in the EU. This was initiated in 1997 in relation to employment and subsequently extended to other areas including poverty and social exclusion, pensions and health and social care. Already it has generated a huge literature and contradictory conclusions relating to its potential for the convergence of social policy goals and outcomes (Scharpf, 2002; Borrás and Jacobsson, 2004; Hemerijck and Berghman, 2004; O’Connor, 2005a; Zeitlin et al., 2005; Friedrich, 2006). Examination of the European integration-Europeanization-convergence link is subject to the same caveats relating to causality as any investigation of convergence with the added dimension of the European integrationEuropeanization-globalization link. Globalization is being used here to refer to the ‘transformation in the spatial organization of social relations and transactions – assessed in terms of their extensity, intensity, velocity and impact – generating transcontinental or interregional flows and networks of activity, interaction and the exercise of power’ (Held et al., 1999: 16). European integration and the resulting Europeanization can be seen as defensive strategy in the context of globalization and associated neoliberalism. This is not to deny the evidence of neo-liberalism in European Union policy making (Wincott, 2003). It recognizes the fact that globalization can be seen as a threat to the European social model and that Europeanization has been identified as a filter for globalization (Wallace, 2000: 381). Yet it is important to recognize that the demonstration of causal links between globalization and/or Europeanization, on the one hand, and downward pressure on welfare services and expenditure, on the other, cannot be unambiguously demonstrated (O’Connor, 2005b). Furthermore, the relative importance not only of European integration and globalization but also of domestic political factors may vary between countries, over time and depends on the type of pressure. Convergence in the EU 15 Welfare States In this subsection we discuss convergence as decreased variation by considering how it is measured in quantitative cross-national studies of welfare effort in the EU 15 and briefly consider the findings of comparative case studies and studies using qualitative policy analysis. This is followed by a discussion of evidence on the changing position of the Cohesion Countries relative to the EU 15 in terms of catch-up convergence. The section concludes with a discussion of evidence on frame convergence through directives on gender equality and the Open Method of Coordination which may contribute to convergence as reflected in decreased variation and catch-up.

226


Convergence as decreased variation The convergence of social protection objectives and policies was the subject of a recommendation of the European Council in 1992 (92/442/EEC).6 It is noteworthy that this was a compromise recommendation in place of a proposal for harmonization which has consistently been resisted in the social policy area by most Member States (Chassard, 2001). Convergence in social protection was also stressed in the 1994 White Paper on Social Policy, later papers on modernizing social protection and the 2000 Social Policy Agenda (European Commission, 1994; 1997; 1999; 2000). On the basis of EU social protection expenditure data, taxes and duties as a percentage of GDP, and expenditure on active compared to passive labour market measures, Greve (1996) argued that there was evidence of social policy convergence in the European Union from 1980 to 1993. Based on evidence of relatively greater percentage increases in social protection expenditure per capita at 1985 prices by Portugal, Spain, Greece and Italy than by the mature welfare states he concluded that this was largely a consequence of catch-up convergence (Greve, 1996: 363). What does the present evidence tell us and what kind of evidence would allow us to conclude that we are witnessing convergence? Several studies provide quantitative evidence relating to the period from 1980 to the late 1990s. Cornelisse and Goudswaard (2002), Wolf (2002), Bouget (2003) and Castles (2004) all base their analysis on data from the OECD but each argues for a different measure of social expenditure. The statistical measures of convergence used in all of the studies are relatively similar – standard deviation, coefficient of variation and analysis of the contribution of particular countries to the overall variation. Wolf and Castles respectively examine the evidence for catch-up convergence through regression analysis and correlation. All recognize the importance of measuring convergence/divergence as a process that must be measured over time.7 Despite the differences in the dependent variable the findings are relatively similar in giving some support for decreased variation in welfare effort in the EU 15, or the EU 15 minus one member depending on data availability, over the 1980 to 1998 period. The most detailed study of convergence in terms of its focus on EU issues is Bouget’s (2003) analysis of total public social expenditure and per capita social expenditure from 1980 to 1998 in 21 OECD countries and the EU 15 (minus Austria). He justifies these dependent variables on the grounds that the possibility for substitution between social benefits can be missed out when the focus is on particular social policies and the fact that macroeconomic pressure is a global one on national social expenditure. Using the standard deviation and coefficient of variation, he finds convergence in the period 1980–90, divergence in 1990–93 and convergence in the


227

sub-period 1993–98. He explains the divergence trend between 1990 and 1993 by the international economic recession when some countries were non-sensitive (Ireland and Luxembourg) and others (Sweden and Finland) responded with dramatic increases in social expenditure. In contrast to Greve’s conclusion relating to the 1980–93 period, Bouget’s analysis does not indicate a process of catch-up in social expenditure by the Southern European countries with the exception of Greece nor does it indicate a cutback by the original more generous welfare states. Unfortunately, Greve and Bouget are using different measures of welfare effort. Greve is using EU data on social protection while Bouget is using a broader measure of social expenditure from the OECD. These definitions of social expenditure differ significantly (see De Deken and Kittel, Chapter 5, this volume).8 Bouget’s overall conclusion is that the empirical data on social expenditure do not support the thesis of convergence due to European integration. Castles (2004), who also uses the OECD Social Expenditure database but a slightly different dependent variable to Bouget, reaches a similar conclusion relating to the absence of evidence of a European integration effect. Castles excludes spending on active labour market programmes because of incomplete data in the earlier years. While finding little evidence for a distinct European Social Model based on a range of measures of welfare expenditure, he does find evidence of convergence as measured by the coefficient of variation for the OECD, Europe as a whole and the EU 15 minus Luxembourg. But he finds no evidence of ‘downward harmonization’ of social spending. Wolf (2002) reaches a similar conclusion in his analysis of the European Union from 1980 to 1997. While finding no evidence for a race to the bottom he concludes that there is ‘a more general convergence trend towards an intermediate social expenditure share’ (Wolf, 2002: 8). Cornelisse and Goudswaard (2002) analyse social security expenditure as a percentage of GDP and replacement rates for unemployment benefits. Taking gross replacement rates of unemployment benefits in the EU 15, excluding Luxembourg, for 1979, 1989 and 1997 they find a consistently decreasing standard deviation and coefficient of variation. When they consider social security expenditure as a percentage of GDP over the 40 year period 1960–99 for the EU 15 (excluding Luxembourg), they find almost continuous relative convergence, as measured by the coefficient of variation, but absolute convergence, as measured by the standard deviation, only in the period after 1980.9 They argue that it is too early to conclude that the convergence trends within the EU are the result of European integration, although pointing out that intensified contacts and the possibility of demonstration effects of successful countries may promote convergence.

228


Considering a range of welfare state dimensions Kautto and Kvist (2002) analyse convergence in the EU 15 over the 1990 to 1997 period. Their primary focus is on trends in Nordic welfare states, Germany, the Netherlands and the UK. They argue for restructuring rather than retrenchment and while recognizing some weak evidence of convergence in some welfare state dimensions, including social protection expenditure, they conclude that Nordic welfare states illustrate ‘parallel trends’ to the other countries but ‘persistent diversity’ even in response to similar challenges. Updating their analysis of social protection expenditure as a percentage of GDP to the period from 1990 to 2003 the trend of very modest convergence in social protection expenditure is maintained (see second, third and fourth columns of Table 10.1). However, when we analyse the variation in social expenditure in Purchasing Power Standards (PPS)10 evidence of convergence is stronger (see second, third and fourth columns of Table 10.2). The significance of PPS is that it removes the distortions due to price level differences across countries so that we are comparing equivalent purchasing power in each country. Ideally, these measures should be considered in parallel with social outcome indicators. Unfortunately this is not possible due Table 10.1

Social protection expenditure (as a percentage of GDP)

EU15 Belgium Denmark Germany Greece Spain France Ireland Italy Luxembourg Netherlands Austria Portugal Finland Sweden UK Coefficient var.

1990

2000

2003

% GDP change 1990 to 2000

% GDP change 1990 to 2003

25.2 26.4 28.2 25.4 22.9 19.9 27.4 18.4 24.7 21.4 31.1 26.2 16.3 25.1 33.1 22.9

27.2 26.8 28.9 29.3 26.3 19.6 29.3 14.1 25.2 20.3 27.4 28.3 21.7 25.3 31.0 27.0

28.3 29.7 30.9 30.2 26.3 19.7 30.9 16.5 26.4 23.8 28.1 29.5 24.3 26.9 33.5 26.7

2.0 0.4 0.7 3.9 3.4 0.3 1.9 4.3 0.5 1.1 3.7 2.1 5.4 0.2 2.1 4.1

1.1 3.3 2.7 4.8 3.4 0.2 3.5 1.9 1.7 2.2 3.0 3.3 8.0 1.8 0.4 3.9

0.19

0.18

0.17

Source: http://epp.eurostat.ec.europa.eu/portal/page.

229


to the non-availability of a range of such indicators over a sufficient length of time to measure convergence. Considering, first, social protection expenditure as a per cent of GDP we find that the Netherlands, Ireland and Spain, the latter to a minuscule extent, are the only countries with decreased expenditure over the entire period from 1990 to 2003 (Table 10.1). It is noteworthy that the decrease in the Netherlands was in the context of expenditure that was second highest of the EU 15 in 1990 while expenditure in Ireland and Spain was among the lowest in the EU 15. Luxembourg and Sweden had decreases in expenditure in the 1990 to 2000 period. But this was not only reversed between 2000 and 2003, as it was in the other cut-back countries, it had slightly surpassed its former level in both countries by 2003 and Sweden continued to be the highest social protection spender in terms of GDP percentage. In contrast, social protection expenditure in PPS increased in all EU 15 countries between 1992 and 2003 (see column 5 of Table 10.2). The analysis of Table 10.2

Social protection expenditure in Purchasing Power Standards

EU15 Belgium Denmark Germany Greece Spain France Ireland Italy Luxembourg Netherlands Austria Portugal Finland Sweden UK Coefficient var.

1992

2000

2003

% change 1992 to 2003

4 510.0 4 913.4 5 403.9 4 992.7 2 342.7 2 859.9 5 075.8 2 616.0 4 472.8 5 849.5 5 597.5 5 172.4 2 011.5 5 184.7 6 063.2 4 298.1 0.30

6 192.6 6 195.0 7 313.9 6 580.5 3 764.0 3 632.4 6 696.0 3 571.9 5 624.2 8 788.2 6 583.1 7 144.9 3 512.7 5 750.4 7 334.0 6 000.3 0.27

6 926.2 7 476.2 8 115.0 7 086.5 4 567.1 4 186.0 7 433.9 4 813.6 6 023.5 10 904.6* 7 604.5 7 699.8 4 076.4 6 560.3 8 258.0 6 812.3 0.27

53.5 52.2 50.2 41.9 95.0 46.4 46.5 84.0 34.4 86.4 35.9 48.9 102.7 26.5 36.2 58.5

Note: * A large share of benefits is paid to persons living outside Luxembourg (mainly family allowances). Adjusted for this, expenditure is about 10 per cent lower than indicated above, that is about 9810 PPS. (Eurostat, ‘Social Protection in Europe’ Statistics in Focus Population and Social Conditions 14/2005: footnote 1). Source: http://epp.eurostat.ec.europa.eu/portal/page.

230


the contribution of individual countries points, with some exceptions, to a process of catch-up by the initially lower spending countries and some variation in ranking particularly in the middle of the distribution. Specifically, Portugal, Greece and Ireland all increased their social protection expenditure in PPS by 84 per cent or more over the 1992 to 2003 period compared to EU 15 average and median increases of 54 and 49 per cent, respectively. In contrast, both Spain and Italy had relatively low increases of 46 and 34 per cent. Luxembourg is the exception amongst the initially higher spending countries, as measured by social protection expenditure in PPS, with an increase of 86 per cent over the period bringing its social expenditure in PPS to the top of the EU 15 countries in 2003 from a position of third highest in 1990. As noted in Table 10.2 Luxembourg is atypical due to the share of benefits paid to persons living outside the country (mainly family allowances). Adjusting for this, its social expenditure in PPS is still highest in the EU 15 in 2003 (Eurostat, 2005b). More importantly, Luxembourg is also atypical because of its extreme outlier status in terms of GDP per capita. In 2003 its GDP per capita was 194 per cent of the EU average. The next highest in GDP per capita terms were Ireland at 123 and the Netherlands at 115 per cent (European Commission, 2005a). Like the quantitative studies of expenditure, recent comparative case studies of various aspects of policy give weak support for convergence in particular policy areas across two to four countries and illustrate the difficulties of reaching conclusions about convergence (Taylor-Gooby, 2003).11 These difficulties include the identification of the aspect of policy in relation to which convergence can be demonstrated – convergence of objectives that are not operationalized and remain at the symbolic level, what Hvinden (2003) identifies as ‘thin’ convergence, or convergence of operational goals and policy instruments. For example, Daguerre and Taylor-Gooby (2004) conclude that while France and the UK had relatively similar objectives in addressing challenges associated with labour market change in the 1990s, they display strong patterns of path dependency in responses and do not exhibit ‘strong tendencies to converge’. The importance of the time period over which convergence is measured has already been illustrated, for example by Bouget (2003) and in earlier studies by Castles (1982b) and O’Connor (1988) and is highlighted by Hinrichs and Kangas (2003) in their study of pension systems in Germany and Finland. They point out that change in pension systems that may eventually be characterized as a systemic change tends to occur incrementally making it difficult to identify it except in the long run. The time factor also influences the appropriateness of expenditure patterns as a yardstick for assessing change: specifically, current pension expenditure is the result of past decisions, often the quite distant past, and the result of present decisions


231

may not be evident for many years and sometimes even decades. Jaeger and Kvist (2003) make a related point in distinguishing between ‘crises’ and challenges, which create sufficient pressure to demand a policy response such as welfare state restriction and/or greater efficiency in use of resources, and ‘controversies’, such as demographic and labour market change, whose impact is long-term, not clear-cut and can only be established empirically. These studies not only support the emphasis on identifying the appropriate time frame within which convergence should be considered, they also point to the importance of taking different dimensions of expenditure into account in explaining over time patterns – the appropriate time frame is likely to vary across dimensions of expenditure and depending on the nature of the pressure on the welfare state. Hvinden (2003) finds no evidence for the occurrence, or likelihood, of convergence in redistributive policies relating to disability across four welfare regimes in Western Europe – the Nordic, Continental, Southern and Western, that is the liberal, ‘Social Europes’ (Ferrera, 1998) in the 1990s.12 He argues that the potential for convergence is greater in regulatory areas where there are more ‘vacant areas’ in the sense of less policy development. These are the areas in which EU action in relation to equal opportunities may be seen as more legitimated. They are also the areas in which the EU has been developing a disability strategy since the 1990s; this has been a rights based approach relating to equal opportunities, antidiscrimination and mainstreaming (CEU, 1996) that resulted in a directive establishing a general framework for equal treatment in employment and occupation in 2000 (CEU, 2000a, 2000b). This relates to the issue of convergence through framing that is discussed further on. In summary, quantitative measures of welfare effort over the EU 15 provide more evidence of convergence than do qualitative studies of particular dimensions of social policy in two to four countries. Several studies using quantitative measures of welfare effort as a percentage of GDP provide some evidence of decreased variation over time in the EU 15 but the strength of convergence is rather weak. When welfare effort is measured by social protection expenditure in PPS there is stronger evidence not only of decreased variation but of catch-up convergence. There is no evidence of a race to the bottom in terms of welfare effort in the EU 15 between 1990 and 2003 whether measured as a percentage of GDP or as social protection expenditure in PPS. We now turn to a more detailed analysis of catch-up convergence. Catch-up Convergence? Catch-up convergence is of particular relevance in the European Union in the context of the 2004 enlargement. The Structural Funds are one of the

232


key measures developed by the EU to address regional inequalities in income, employment, investment and growth.13 Structural Funds and national social expenditure are rarely linked in the analysis of convergence. Yet, if standard of living is the concern both dimensions of redistribution are relevant. When we examine both dimensions we find conflicting patterns. Despite acknowledging the persistence of significant disparities across the EU, the European Commission reports on economic and social cohesion provide evidence of convergence, in particular the catching up of the Objective 1 Regions in terms of GDP per capita throughout the 1989–93 and 1994–99 programme periods (European Commission, 1996; 2001a). The 2004 report points to the positive impact of structural intervention on the growth of the Cohesion Countries – Greece, Ireland, Portugal and Spain (European Commission, 2004). The catch-up, or not, of these countries since they joined the EU (or former EC) is of particular relevance in the context of the EU enlargement process. The 2004 accession countries and the applicant countries are, like the Cohesion Countries at their time of accession, considerably below the EU average in terms of GDP per capita and social development. Despite some evidence to the contrary, most academic studies of the impact of the Structural Funds support the Commission position and conclude that the allocation of regional funding to new EU members does assist the process of catch-up convergence as the additional financial resources allow for investment in crucial infrastructure such as transportation networks. This, in turn, has an impact on the speed at which economic development progresses. A key factor in influencing this is the national policy choices and context (see McClelland and O’Connor, 2006 for a review). Delhey (2001) analyses GDP per capita, social expenditure as a percentage of GDP in the Cohesion Countries and concludes that EU membership has had largely positive implications for new Member States in the 1970 to 1997 period. In the following analysis the time period is extended to 2003, the performance of the Cohesion Countries is considered relative to the EU 1514 and social expenditure per capita in purchasing power standards (PPS) is included in the analysis. The GDP per capita in PPS of the Cohesion Countries relative to the EU 15 for selected years from 1970 to 2004 is presented in Table 10.3. With the exception of Greece, these countries have improved their position, as measured by GDP per capita, relative to the EU average between their dates of accession and 2004 although the pattern and degree of catch-up has varied considerably. Ireland has been a member of the EU since 1973 when its GDP per capita in PPS was 60 per cent of the EU 15 average. Its pattern is one of modest catch-up to the mid 1990s – its GDP per capita in PPS was only 65 per cent of the EU 15 average

233


Table 10.3 GDP per capita as percentage of EU 15 100 selected years, 1970–2004

1970 1973 1980 1986 1990 1995 2000 2003 2004

Ireland EU15100

Greece EU15100

Portugal EU15100

Spain EU15100

60.3 68.2 65.3 76.0 89.2 115.1 123.4 124.6

71.8 70.8 79.7 62.6 65.8 64.2 65.1 73.5 75.4

55.2 57.9 60.5 54.4 66.2 68.6 73.4 70.8 69.9

74.9 77.0 74.3 71.8 77.6 79.5 84.3 90.8 90.5

Source: European Economy No. 70 (2000): Table 9 and updates from http://ec.europa.eu/ economy_finance/publications/european_economy/2005/ statannex0105_en.pdf.

in 1986 when Spain and Portugal joined the EU. From the mid 1990s catchup of GDP per capita was marked and is now second highest in the EU after Luxembourg. In contrast to Ireland, Portugal demonstrates relatively significant catch-up in its first 12 years of EU membership – its GDP per capita in PPS increased from 54 per cent of the EU average in 1986 to almost 74 per cent in 1999. Since then the pattern has been one of decline to 70 per cent in 2004. Spain’s catch-up in GDP per capita has been relatively steady throughout the period of its membership – from 72 to 91 per cent between 1986 and 2004. Greece as noted above is the exception to this pattern of catch-up. Its GDP per capita in PPS declined from almost 80 per cent of the EU average in 1981, when it became an EC member, to about 64 per cent in the late 1990s. By 2004 it had reached 75 per cent, which although lower than its relative position in 1981, indicates considerable progress from the late 1990s.15 While all the Cohesion Countries benefited from the Structural Funds, national policy choices relating to their use, the fruition of long-term choices, for example relating to education in Ireland, and other factors such as a favourable demographic structure contributed to the outcomes. The pattern or catch-up demonstrated in GDP per capita is reversed when social expenditure as a percentage of GDP is considered from 1993 to 2003 (Table 10.4).16 Ireland and Spain demonstrated the strongest catch-up in GDP per capita but their social expenditure as a percentage of GDP and relative to the EU 15 has declined consistently since the early 1990s. This is

234 – – 76 80 97 93

EU 15 100

http://epp.eurostat.ec.europa.eu/portal/page.

– – 22.0 22.7 26.3 26.3

% GDP

Source:

– – 57 – – 69

PPS as % of EU 15

13.2 23.6 20.8 19.9 14.1 16.5

– – 72 70 52 58

EU 15 100

Greece

1970 1985 1993 1995 2000 2003

% GDP

Ireland

– – 51 – – 66

PPS as % of EU 15 – 14.2 21.0 20.7 21.7 24.3

% GDP – – 72 73 80 86

EU 15 100

Portugal

– – 48 – – 59

PPS as % of EU 15 – 20.0 24.4 22.6 19.6 19.7

% GDP

– – 84 79 72 70

EU 15 100

Spain

– – 65 – – 60

PPS as % of EU15

– – 28.7 28.2 27.2 28.3

EU 15 Average expenditure % GDP

Table 10.4 Social protection expenditure as a percentage of GDP, relative to EU 15100 and PPS for selected years, 1970–2003


235

particularly evident in the case of Ireland whose social protection expenditure as a percentage of GDP was only 58 per cent of the EU average in 2003. This measure of social protection expenditure did increase significantly in the earlier years of Ireland’s membership up to the late 1980s. In contrast to Spain and Ireland, social expenditure as a percentage of GDP in Portugal and Greece demonstrates catch-up relative to the EU 15 average between accession and 2003, when Portugal had reached 86 per cent, and Greece 93 per cent, of the EU average. At one level the relatively poor performance of Ireland and Spain is not surprising since in the context of GDP growth maintenance of the real value of social expenditure is possible despite a constant or even declining percentage of GDP. When social protection expenditure is presented in PPS per capita, all of the Cohesion Countries with the exception of Spain have improved their position relative to the EU 15 over the period from 1993 to 2003 (Table 10.4, third column for each country). The improvement is most marked for Greece which increased it PPS expenditure per capita from 45 to 67 per cent of the EU 15 average; Portugal increased its PPS expenditure from 46 to 60 per cent of the average while Ireland improved from 59 to 71 per cent. Ireland’s relative position in terms of the EU average is the best of the Cohesion Countries in 2003 but the catch-up of Portugal and Greece between the early 1990s and 2003 was considerably greater; a review of their expenditures in PPS demonstrates that Portugal more than doubled and Greece almost doubled its social protection expenditure in real terms over this period (see column 5 of Table 10.2). Despite these improvements over the 1990 to 2003 period, the Cohesion Countries in 2003 still had the lowest social expenditure in PPS in the EU 15 and while Ireland is best in PPS terms it is noteworthy that this is in the context of a significantly higher GDP than the other Cohesion Countries and its expenditure in PPS is exceedingly low when considered relative to countries at the same level of GDP per capita. Ireland’s low social expenditure is in part due to its low old-age dependency – 17 per cent relative to an EU average of 24 per cent and an associated less than 4 per cent of GDP on old-age and survivors benefits compared to almost 13 per cent for the EU 15 in 2001 (calculated from Eurostat, 2005b). However, the exclusion of this category of expenditure still leaves Irish expenditure below the EU 15 average, higher than the other Cohesion Countries and Italy but considerably lower than those countries around the average EU 15 GDP per capita (O’Connor, 2003: 393). The summary of the outcomes in terms of GDP per capita and social protection expenditure for the Cohesion Countries up to 2003 presented in Table 10.5 reinforces the conclusions of earlier studies in demonstrating that convergence has to be viewed as a long-term process that may include periods of stagnation or reversal of progress in relative terms. But it also

236


Table 10.5 Cohesion Countries: catch-up relative to EU 15 accession to 2003 Country and year of accession

GDP per capita

Social protection expenditure as % GDP relative to EU 15100

Social protection expenditure in PPS 1993–2003

Ireland (1973)

Reached EU average 1997; 123% of EU average in 2003

Progress towards catch-up to late 1980s; reversal since early 1990s

Progress towards catch-up: from 59 to 71% of EU 15 average

Greece (1981)

Almost consistent reversal to 1996; catch-up thereafter – 75% of EU average in 2003

Strong progress towards catch-up

Strong progress towards catch-up: from 45 to 67% of EU 15 average

Portugal (1986)

Progress towards Strong progress catch-up to 1999; towards catch-up reversal 2000 to 2003 – 73% to 70% of EU average

Progress towards catch-up: from 46 to 60% of EU 15 average

Spain (1986)

Relatively steady progress towards catch-up; 91% of EU average in 2003

No progress towards catch-up

Consistent reversal

demonstrates that catch-up in one dimension of progress or in one measure of welfare effort does not guarantee catch-up in other dimensions or measures; in particular, growth in GDP per capita relative to the EU 15 average does not guarantee a similar level of progress in social expenditure as a percentage of GDP – this is particularly evident in relation to Ireland and Spain; furthermore, divergence in one measure of welfare effort – social protection as a percentage of GDP – may occur in the context of convergence in welfare effort as measured by PPS as is evident for Ireland because of its increased GDP. In contrast to Ireland and Spain, both Greece and Portugal demonstrate strong catch-up relative to the EU 15 in both social protection expenditure measures since accession to the EU despite only recent modest catch-up, following earlier decline, in GDP per capita for Greece and relatively modest catch-up in GDP for Portugal since its accession. Case studies indicate progress in some areas other than income maintenance, for example gender equality provisions in Ireland (O’Connor,


237

2003) and welfare services in Spain and Greece (Guillén and Matsaganis, 2000). The situation is relation to poverty is less encouraging. Matsaganis et al. (2003) examine anti-poverty policies in Greece, Italy, Portugal and Spain and conclude that while there were significant policy innovations in relation to poverty and social exclusion during the 1990s stimulated by EU initiatives, the safety net for people who are highly vulnerable economically continues to be frail when considered in the EU context. They point to particular constraints associated with the role of families and the relative weakness of the administrative capacity of the state in this area. In relation to Ireland, O’Connor (2003) demonstrates the persistence of high levels of relative income inequality despite improvement in ‘consistent poverty’, a measure weighted by indicators of material deprivation. Because of the short length of time over which data are available it is not possible to present a statistical assessment of the pattern of variation in poverty measures reflected by income after social transfer payments in the EU 15. Based on the share of persons with a disposable income below 60 per cent of the median equivalized disposable income after social transfers the evidence for the period 1999 to 2003 and 2004 indicates continuing high levels of relative poverty and social inequality in the Cohesion Countries. The evidence also indicates an overall increase in this measure of poverty in the EU 15 in 2004 at 17 per cent compared to 16 per cent in 1999, and it shows continuing diversity and the persistence of the highest levels of poverty in the EU 15 in the Cohesion Countries (Eurostat, 2006b).17 The development of Structural Indicators and the associated data together with the reporting on the National Action Plans on Poverty and Social Exclusion element of the OMC will over time enhance the possibility of examining the evidence for convergence of outcomes or its absence in this area as in others in the EU 25. In summary while we have strong evidence of catch-up as measured by GDP per capita in the Cohesion Countries relative to the EU 15, the evidence for social expenditure as a percentage of GDP presents a more mixed picture. Catch-up in this measure of social expenditure is evident only for Portugal and Greece, the countries with the weakest dynamic of economic catch-up.18 When we consider social protection in purchasing power standards we find catch-up for these countries and for Ireland. This indicates that while the share of social protection as a percentage of GDP may decline, as in Ireland, the relative welfare effort in purchasing power standards may be maintained or increased because of increased GDP. A somewhat similar pattern associated with rapidly increasing GDP is evident for Luxembourg with the significant difference that its welfare effort as percentage of GDP over the entire 1990 to 2003 period also increased. Despite catch-up convergence as measured by both measures of welfare effort in

238


Greece and Portugal and in social expenditure in PPS in Ireland, the Cohesion Countries were still amongst the countries with the lowest social protection expenditure as a percentage of GDP in the EU 15 in 2003 as in 1990 and had the lowest social protection expenditure in PPS in both years. Despite the evidence of varying catch-up in expenditure patterns and progress in some other areas there is no evidence of change in the parameters of existing institutional structures – a liberal welfare regime in Ireland and adherence to the Southern model in Greece, Spain and Portugal (Guillén and Matsaganis, 2000; O’Connor, 2003) – although we may find convergence in policy discourse and frame convergence in areas other than social protection. Convergence through Framing? The Structural Funds may contribute to catch-up convergence in particular in GDP per capita and may contribute to convergence in social expenditure but convergence in aspects of the broad social policy domain may also be the outcome of other processes in particular the related concepts of frame convergence and ‘cognitive Europeanization’. Understood as ‘the shaping and reshaping of the perceptions of, and attitudes towards, social problems and the way to tackle them’ (Radaelli, 2000), cognitive Europeanization is increasingly recognized as a significant influence on aspects of the social policy framework, for example, welfare services in Spain, Portugal and Greece (Guillén and Palier, 2004) and gender equality in Ireland (O’Connor, 2003). Cognitive Europeanization contributes to frame convergence. We have already referred to the fact that frame convergence may be associated with the transposition of directives into national policy – an example of positive integration – and/or the broader mechanisms of facilitated coordination through political declarations as at Council summits and/or through soft law such as OMC, which is increasingly important in the broad social policy area. This form of coordination is concerned with the convergence of ideas (Bulmer and Radaelli, 2005). In both instances we are concerned with the change in the domestic frame of reference associated with the process of European integration. But this change does not necessarily imply harmonization in the sense of identity of policy output in member states nor does it necessarily imply convergence. We use harmonization to refer to a process that may but does not necessarily lead to convergence.19 Furthermore, we may identify convergence which is unrelated to EU initiatives that aim at harmonizing Member State legislation. For example, in discussing the impact of Europeanization on social and economic gender rights in six EU countries with contrasting welfare and gender regimes, Liebert (2003) argues that while EU equality


239

norms have enhanced convergence they have not produced harmonization.20 The implementation of the gender equality directives between 1975 and 2000 varied considerably depending on ‘domestic frames of mind’ but by 1998 ‘all 15 Member States had implemented the equality acquis communautaire, though without jeopardising national distinctions’ (Liebert, 2003: 13). The term ‘harmonization’ was used in the original EEC Treaty (Article 117) and has not been removed in subsequent treaties but by 1992, when the European Council made a Recommendation ‘on the convergence of social protection objectives and policies’ (92/442/EEC), it was clear that harmonization did not have the support of Member States. This is reflected in increasing emphasis on ‘a positive and active conception of subsidiarity’ as the key to cooperation between the European Union, Member States and citizens (European Commission, 1994: 4; Hantrais, 1995: 199–202). The processes by which EU directives are implemented by Member States vary, that is, the implementation mechanisms are not harmonized. Consistent with the primacy of subsidiarity, identical policy processes are not the objectives of convergence in the European Union – the interest is in convergence of outcomes as reflected in key structural and social indicators. Despite the transposition of directives on gender equality by Member States the Report on Equality between Men and Women by the European Commission (2006b) indicates that there is considerable leeway for progress and variation across member states, not only the 2004 accession countries but the EU 15, in outcomes such as the gender gap in employment, unemployment and pay rates. As Rubery puts it there are ‘many different aspirations lying behind an apparent common commitment to pursue equal opportunities’ (Rubery, 2003: 4). Yet, while we are at the stage of convergence of ideas about key issues in gender equality rather then convergence of outcomes, we cannot assume that the latter will not occur. At a minimum the introduction of a framework of structural and social indicators will afford the opportunity to assess outcomes comparatively over time, including outcomes not only in employment and unemployment disaggregated by gender but also in the gender pay gap. The OMC is a means for spreading best practice and achieving convergence towards EU goals (Bulmer and Radaelli, 2005). In so doing it is a mechanism for reconciling the potentially conflicting demands of economic and social policy on the one hand and Member State and EU objectives on the other (Moreno and Palier, 2004). But ideational convergence is not automatically or inevitably associated with effective policy change. The monitoring of implementation of commitments made in the various National Action Plans and the assessment of progress as well as the collection of data on key structural and social indicators are important dimensions of the radical potential of the OMC. But unless they are coupled with

240


the mobilization of national stakeholders around the formulation and implementation of National Action Plans the potential will not be realized (O’Connor, 2005a).

CONCLUSIONS The concept of convergence has enjoyed varying levels of prominence in studies of welfare effort and policy output but its analytical and theoretical value has been constrained by a failure to specify clearly its range of meanings and the dependent variables being examined. Convergence refers to the end result of processes of quantitative or qualitative change in similarity or difference across units over time – these units may be societies, policies, elements of policies, expenditure or outcomes. Analysis of convergence has much to contribute to the understanding of public and social policy change. To achieve this it must be part of well designed studies that specify clearly each of the following factors. What is converging and how is it to be measured: is it institutional structures or policies, if policies are we talking of policy objectives, outputs or outcomes? If policy outputs what dimension/s of policies, if outcomes what are the appropriate indicators? Is the time frame appropriate for the type of policy issue, dimension or outcome? What type of convergence is being examined – decreased variation, catch-up by laggards on leaders or change in ranking? Decisions on these questions provide guidelines on appropriate comparative frameworks for the identification of convergence or its absence although data limitations are likely to impose constraints. A further significant constraint in the analysis of convergence is the fact that it is widely used to refer to a process of change that is difficult to delineate, namely convergence of ideas. This is evident in the general welfare state literature with reference to economic, financial and cultural globalization and common challenges associated with demographic and labour market changes. The issue of ideational convergence is particularly evident in discussion of welfare states in the European Union in the context of European integration and the phenomena of ‘cognitive Europeanization’ and the framing of integration. Such framing alters the domestic frame of reference within which policy discussion and formulation take place and varies from the explicit transposition of EU directives into national legislation and policy to the increasingly important ‘facilitated coordination’ (Bulmer and Radaelli, 2005) through European Council declarations and soft law as reflected particularly in the OMC in various areas. To what extent these processes will be reflected in changes beyond ideational convergence, specifically, convergence in operational goals and policy instruments, is a


241

matter for systematic over time comparative analysis. More importantly the key issue is to what extent they will result in convergence of policy outcomes irrespective of the policy instruments chosen. The non-availability of social outcome indicators for a sufficiently lengthy period of time and range of countries means that such analysis is not possible at present. It is also one of the reasons why analysts interested in the identification of convergence or its absence across more than a few countries over an extended period of time rely on expenditure data on welfare effort (but see Scruggs, Chapter 7, this volume). The potential contribution of this kind of analysis will be greatly enhanced by the over time availability of the social outcome indicators now being collected for the EU 25. European integration and Europeanization are frequently identified as the key causes of convergence towards the top or the bottom in terms of social policy standards in the European Union. It has been argued in this chapter that a hypothesized Europeanization-convergence link must be unpacked to take into account the mechanism underlying European integration – positive integration, negative integration or framing integration through positive integration as in directives or facilitated coordination as in the open method of coordination. None of these mechanisms operates in a vacuum – their operation is facilitated and constrained by a range of policy and country specific factors. Most importantly, they are constrained by domestic political factors and policy choices. This emphasis on the importance of political and institutional factors echoes the challenges to the logic of industrialism and logic of capitalism arguments of the 1960s and 1970s relating to the welfare state. Those studies emphasized the importance of political and institutional factors in explaining persisting differences and in some instances growing diversity across welfare states. Similarly, those studies that directly addressed the issue of convergence or divergence and found evidence of convergence in relation to particular measure of welfare effort over particular periods, and its absence in other time periods and in other measures of welfare effort, stressed the significance of political and institutional variables (Castles, 1982b; Uusitalo, 1984; O’Connor, 1988; Montanari, 2001 re social rights). In this analysis we have concentrated on the European Union 15 and identified the possible influence of various dimensions of European integration. This is an incomplete exercise: the European integrationEuropeanization-convergence linkage must also include the overarching influence of globalization of economic and financial flows – as was pointed out, Europeanization may be seen as a conduit or filter for globalization influences on Member States. Furthermore, influence on policy frames is not confined to the European Union – the influence of the OECD for example in its economic reports (Armingeon, 2004) and probably more strongly in

242


relation to activation is likely to have influenced the employment policy frame in most EU countries and in addition was filtered through EU policy. Work on convergence stems from sources with diverse theoretical, ideological and disciplinary backgrounds. While this has been associated to date with different parallel streams of analysis, cross-fertilization could contribute to a vibrant research agenda. Such a development is not only essential for effective analysis of convergence but is particularly appropriate in this area that brings into consideration the impact of globalization, European integration and Europeanization on contemporary public and social policy.

NOTES 1.

2.

3. 4. 5.

6. 7. 8.

9.

The findings and methodological approach of these studies were challenged by studies starting from a focus on the explanation of diversity in welfare state development amongst countries at a broadly similar level of development that demonstrated the importance of class mobilization in trade unions and political parties (Korpi, 1983; Stephens, 1979). Since both sets of studies were cross-sectional they demonstrated difference not convergence or divergence (Castles, 1982a). Later studies did consider the issue of change over time and demonstrated the importance of the time period over which convergence was measured and the indicator of the welfare state used. Convergence was evident only in relation to social transfer payments aspects of social expenditure (see O’Connor, 1988 for a review). Political control aggregates direct state control of the economy, as reflected in state ownership and the extent to which factors other than profitability influence decisions; indirect state control as reflected in subsidies, taxes and regulations; welfare state interventions and labour market organization. Seeliger (1996: 289–96) uses synchronous development to describe identical differences between units at both times. This is identified by Heichel, Pape and Sommerer (2005) as gamma-convergence ( ). They also identify delta-convergence () to describe the minimizing of distance from an exemplary model, for example, as promoted by an international organization. Liebert and colleagues define Europeanization as ‘transnational processes conducive to shared frameworks, such that, as Helen Wallace puts it, “a European dimension becomes an embedded feature which frames politics and policy within the European states” ’ (Wallace, 2000: 370) (Liebert, 2003: 14). In the same year the Council passed recommendation 92/441/EEC concerning sufficient resources for social protection in social protection systems. Many of the studies in the 1970s purporting to demonstrate convergence were crosssectional, based on countries at vastly different levels of economic development (Cutright, 1967; Miller, 1976; Wilensky, 1975). The components of social expenditure included in the OECD measure are public and private mandatory programmes of cash benefits for old age, disability and family; sickness and survivors’ benefits; unemployment; public expenditure on health; housing; family services; services for elderly and disabled people; occupational injury and disease; active labour market programmes; and other contingencies. Social protection in the European Union includes the following benefits: old-age and survivors’; sickness/health care; disability; family; unemployment; housing and social exclusion. In contrast, the non-EU countries considered – Norway, Switzerland, the United States, Japan and Australia – demonstrate consistent divergence as reflected in increasing standard deviations since 1980. This conflicts with Castles’ (2004) conclusions, in which


10.

11. 12. 13.

14. 15. 16. 17.

18.

19. 20.

243

he identifies the convergence of a Northern European grouping composed of the Scandinavian and continental Western European countries, all members of the European Union (Austria, France, Germany, the Netherlands and Belgium) but also Switzerland (Castles, 2004: 80). As pointed out above the measures of welfare effort used by Cornelisse and Goudswaard and Castles are very different (see note 8). Purchasing Power Standards refer to an independent unit of any national currency that removes the distortions due to price level differences. The PPS values are derived by using Purchasing Power Parities (PPPs) as a weighted average of relative price ratios in respect of a homogeneous basket of goods and services, comparable and representative for each Member State (Eurostat, 2005a). Social Policy and Administration 37(6) in 2003 brings together several studies. The division is based on four characteristics of the countries’ social protection systems: eligibility, benefit formulae, financing regulations and organizational-managerial relations (Hvinden, 2003: 613). Article 158 of the Consolidated Version of the Treaty Establishing the European Community outlines its economic and social cohesion provisions in terms of promoting overall ‘harmonious development’ to be achieved by ‘reducing disparities between the levels of development of the various regions and the backwardness of the least favoured regions or islands, including rural areas’. According to the European Commission the reduction of disparities in their geographical dimension is taken to mean ‘convergence of basic incomes through GDP growth, of competitiveness and of employment’ (European Commission, 1996: 13–14). Delhey (2001) situates social expenditure relative to the EU 12, that is the EU 15 without Austria, Finland and Sweden in the 1970 to 1997 period and GDP relative to the EU 15 up to 1998. Delhey (2001: 222) argues that the Greek experience of EU membership reflects the economic risks of premature accession. The year 1993 is the first year for which figures relative to the EU 15100 are available. http://epp.eurostat.cec.eu.int/portal/page?_pageid1996,45323734&_dadportal&_ schemaPORTAL&screenwelcomeref&open/&productSTRIND_SOCOHE& depth2. This provides the social cohesion tables that are part of the broader Structural Indicators series. Significant catch-up in social expenditure is not always associated with lower GDP. Focusing on 21 OECD countries between 1980 and 1998, Castles (2004: 25) found the highest increase in total public social expenditure based on OECD data in Switzerland, a country with high GDP per capita. Its increase was 13 per cent; Greece was next highest with 11 per cent, then Sweden and Norway with 8 per cent and Portugal with a 7 per cent increase. The overall mean increase was 4 per cent. Harmonization is sometimes used synonymously with convergence, for example, Castles (2004: 74–5) speaks of upward and downward harmonization in the same vein as upward and downward convergence of social protection. The countries studied are France, Germany, Italy, the UK, Spain and Sweden.

11. (In)Dependence as dependent variable: conceptualizing and measuring ‘de-familization’ Sigrid Leitner and Stephan Lessenich INTRODUCTION Looking at the current wave of social policy reform in Western Europe, it might seem as if these were pretty good times for the ‘women-friendly welfare state’ feminist politics and research have been striving for over the last two decades or so. In the context of a broad reform process framed almost everywhere throughout the EU (15) in terms of ‘employability’ and ‘activation’, the advancement of female labour market participation ranks high – if not first – on the political agenda of social policy makers. Women’s well-being (or ‘well-fare’), understood as their belated labour market individualization, stands at the centre of a new, postindustrial welfare equilibrium. As Gøsta Esping-Andersen (2002a: 3), the spiritus rector of European welfare state restructuring, puts it, ‘[i]deological predilections aside, it should be evident to all that we cannot afford not to be egalitarians in the advanced economies of the twenty-first century,’ and gender equality, at least as far as access to the labour market is concerned, is absolutely central to the agenda of the new egalitarianism. In order to free women from family care responsibilities and to enable their labour market participation, the ‘de-familization’ of welfare production has been established as the golden rule of welfare reform. The locus classicus of the mainstream understanding of ‘de-familization’ (or, as he puts it, ‘de-familialization’) is Esping-Andersen’s The Social Foundations of Post-Industrial Economies (1999: 45–6, 50–1). The simple formula to be found there, equating ‘an active policy committed to lessening the caring burdens of the family’ (1999: 45) with (a) the promotion of labour market participation and (b) the enhancement of economic independence of (c) women, has become common wisdom in scientific and political discourses alike. In line with Esping-Andersen’s definition, ‘de-familization’ is unequivocally used to denote ‘policies . . . that maximize individuals’ command of 244

Conceptualizing and measuring ‘de-familization’

245

economic resources independently of familial or conjugal reciprocities’ (1999: 45) – ‘individuals’ always and evidently meaning ‘females’: ‘Hence, defamilialization would indicate the degree to which social policy (or perhaps markets) render women autonomous to become “commodified”, or to set up independent households, in the first place’ (1999: 51). It comes as no surprise, then, that Esping-Andersen’s widely shared vision of ‘a Viable New European Welfare Architecture’ (2002a: 5) is centered around ‘defamilization’ – even if the term does not figure explicitly in his policy outline written for the Belgian Presidency of the European Union. To put it in a nutshell, Esping-Andersen’s recommendation to policy makers in postindustrial Europe (2002a, b, c) is to do it the Nordic way (minus labour market segregation by sex), i.e. to stick to a combined policy strategy focused on child welfare and gender equality, the links of which ‘all boil down to women’s employment’ (2002c: 67). The ‘basic “women-friendly” policy package’ (2002c: 73) propagated by Esping-Andersen thus includes measures to counteract the crisis of the industrial male breadwinner by way of ‘female life course masculinization’ (2002c: 95), i.e.: de-familization. According to the author, this strategy represents a substantial positive-sum game, reconciling in a virtuous circle family welfare (in terms of reduced child poverty and enhanced psychological well-being of children) and ‘the common good’ (meaning, above all, rising fertility rates and increased educational standards) with women’s liberation from family responsibilities and economic dependence. In the year 2000, the European Council acted in line with this argumentation when recommending to raise female employment rates in all member countries from the then average of 51 per cent to over 60 per cent by 2010. In 2002, the European Commission as well as the European Council even extended the European Employment Strategy and formulated targets for the expansion of child care facilities: by 2010 ‘[c]hild care should be available to at least 90 per cent of children between three years old and the mandatory school age and to at least 33 per cent of children under three years of age’ (CEC, 2002: 20). The so-called ‘adult worker model’ (Lewis, 2001) builds the new foundation for economic and social policies promoted by the European Union, and recent legislation in many European countries makes an attempt to comply with the EU targets. Even in the hitherto highly ‘familialistic’, conservative welfare regimes of continental Europe the welfare state has been reinvented as an ‘independent variable’ for female independence. Or so it may seem, at first glance. In what follows, we will challenge the notion of ‘de-familization’ and its equation with female independence so common in both feminist and main(male)stream welfare state analysis. It will be shown that the concept of ‘de-familization’ has to be deconstructed analytically in order to account for its – at a second glance – highly ambivalent effects on the choices and economic self-reliance of women.

246


RECONSTRUCTING ‘DE-FAMILIZATION’: BETWEEN DEPENDENCY AND INDEPENDENCE In this section we will embed the concept of de-familization within the feminist tradition of welfare state research. It will be shown that women’s care work provided within the family was identified as one of the main sources of the gender division in modern social policy. As a consequence, feminists called for the de-familization of women through the expansion of public care services for children and older people. But the shifting of caring responsibilities from the family to the welfare state is only one way to alter women’s dependency status. We will, in the first step of our argument, highlight the remaining pieces of the picture of female economic independence. Early feminist research strongly distrusted the welfare state as such which was conceived as a male dominated system that reinforces patriarchy. The focus of analysis was on the ways in which social policies worked in order to suppress women and to uphold male dominance (Riedmüller, 1984; Abramovitz, 1988; Gerhard et al., 1988; Pateman, 1989). As a result, the welfare state was often rejected as a whole – at least by radical feminists. Another branch of feminist work asserted a shift from private to public patriarchy (Hernes, 1984): the emergence and expansion of the welfare state decreased women’s economic dependence from individual men but – at the same time – established a new form of female dependence with regard to the welfare state and its institutions. Since women were under-represented in the political arena of parties, parliaments and corporatist decision making, they had little power over the structures of the welfare system. Thus, women were clients, consumers and employees of the welfare state but not fully integrated citizens. Men were the agents and women the objects within the political system (Siim, 1987). Nevertheless, the social rights provided by the welfare state were seen to hold an emancipatory potential for women. The transformation from private to public dependence has not only improved women’s economic independence from the male breadwinner but also provides new resources for mobilization, protest and political influence (Fox Piven, 1984; Dahlerup, 1987). While this early feminist work concentrated on ‘the welfare state’ as a macrostructure (or even ‘superstructure’) of patriarchal domination, further research tried to identify and analyse more specifically the arrangements and mechanisms making for the gendered character of social rights in the context of differing national histories of welfare state development. In general, feminist research identified two dimensions of gender discrimination in the welfare state (cf. Orloff, 1993: 315). On the one hand, welfare states distinguish between employment (or wage labour) and unpaid labour. Whereas employment is covered by contributory social security


247

systems that predominantly benefit men, unpaid labour’s social rights are bound to social assistance programmes that serve a predominantly female clientele. This ‘two-tier’ system (Nelson, 1984) prolongs the inferior status of women. Their unpaid labour does not generate equal social rights since social assistance programmes are mostly means tested and subject to social control of the clients while social insurance benefits are received as earned rights due to contribution payments.1 On the other hand, the welfare state treats women as dependants by granting derived social rights for married women. This incorporation of women on the basis of their husbands’ contributions to the social security system institutionalizes and reproduces women’s dependency on the male breadwinner, which makes them especially vulnerable in cases of divorce (Lewis, 1992). The familization of women, i.e. their dominant role in the provision of care, was identified as the main hindrance to women’s labour market participation. According to feminist accounts, their way to economic independence was blocked due to the division of care work between women and the state2 as well as between women and men. Moreover, the welfare state refused to reward paid and unpaid work equally. The inferior treatment of women’s unpaid care work kept them in a status of economic dependency. They were either clients of stigmatizing social assistance programmes or dependent on a breadwinner. From these feminist deficit analyses emerged demands for a women-friendly welfare state that would promote the independence of women. We will shortly discuss the most prominent proposals for a women-friendly welfare state in what follows. Ann Orloff (1993: 319–22) claimed that welfare states should provide women with the capacity to form and maintain an autonomous household. According to her, this could be achieved by two different strategies: (1) providing payments for family care that secure economic independence of the carer and her dependants or (2) improving women’s access to paid work and unburdening the family from care work through the expansion of public services.3 There is an obvious tension between the two strategies, but they could also be combined. Although Orloff is holding up both options, she seems to have a preference for the second strategy, arguing that the history of payments for care shows that the benefits provided were usually very low and not comparable to wage earners’ benefits. Taking a historical stance she claims that, recurring conflicts of interest among women notwithstanding, most postwar women’s movements promoted women’s labour market participation instead of a ‘mothers’ (or carers’) wage’. Since the political decision for one or the other of these strategies should democratically reflect women’s real interests, history suggests – according to Orloff – a normative preference for the second strategy which feminist welfare state theory cannot ignore. In a later article, Orloff (1997) distinguishes between contemporary women’s

248


movements in the liberal countries and in Scandinavia that pursue women’s access to employment, and women’s movements in some continental European countries and the historical women’s movements that are (were) in favour of payments for care. Orloff argues that women might well have different ‘contextualised interests’, but that there still is the overall goal of ‘emancipation’ operating as a common strategic gender interest throughout different political contexts. Since the emancipatory effects of payments for care are disputed, access to employment seems to be the more legitimate strategy – and the more so in times of fiscal austerity and budget cuts that render the guarantee of real ‘choice’ an unrealistic policy option.4 In her postindustrial thought experiment, Nancy Fraser (1997: 42–65) examines the two strategies which she calls the Caregiver Parity model and the Universal Breadwinner model. Whereas the first focuses on state provision of care giver allowances, the second relies on employment enabling services. Fraser concludes that both (ideal) models are good at preventing women’s poverty and exploitation. As far as it goes, each of them would contribute to the independence of women. However, Fraser goes beyond the notion of independence. She develops five additional criteria which have to be met in order to achieve true gender equity: income equality, leisure-time equality, equality of respect, anti-marginalization and antiandrocentrism. Only in a Universal Caregiver model can all of these criteria be met. This would imply that all, women and men, combine breadwinning and care giving and that the institutions of employment and care are redesigned in a way that systematically allows for this combination. Working hours would have to be reduced and employment enabling services would be available for everybody. Informal care work would be publicly supported and would be provided inside and outside households by relatives, friends and ‘civil society’ at large. Also, people without kin based responsibilities would have to provide care. Another approach that radicalizes the focus on care (and payments for care) as a fundamental element of social citizenship is the one elaborated by Trudie Knijn and Monique Kremer (1997). Within their concept of inclusive citizenship, paid work and care are of equal importance since every citizen will – at least once in his or her life – either provide care or be in need of care. Knijn and Kremer systematically distinguish between the right to time for care and the right to receive care. Whereas the first enables informal care giving – without enforcing it – by direct financial compensation of care, the second guarantees access to qualitatively good professional provision of care. Quite interestingly for our context, Knijn and Kremer do not give normative priority to the ‘independence’ of citizens but rather stress their mutual (inter)dependence and the reciprocity of care and ‘care giving’, which for them is not only a matter of giving help, but also of


249

receiving (immaterial, emotional etc.) rewards. Their claim is to recognize the central importance of care giving and care receiving in the context of not (only) the political economy, but the sociology (i.e. the social order) of the welfare state. It should be clear from this review of selected (and exemplary) feminist visions of a women-friendly welfare state that ‘de-familization’ understood (as is most generally done) as the unburdening of ‘the family’ (read: women) from care responsibilities represents only one dimension within an allencompassing feminist agenda for restructuring the gendered welfare state. The individualization of social rights is just as much on this agenda as the financial compensation of family care and the redistribution of care work from women to men within and beyond the family. Thus, it seems only fair at this point to stress that, according to feminist theorizing on the welfare state, access to paid labour or, more specifically, to public services enabling and facilitating women’s labour market participation is not the one and only way to women’s independence and to gender equality in the welfare state.

DECONSTRUCTING ‘DE-FAMILIZATION’ AND ‘INDEPENDENCE’ Having said (and shown) this, we will now try to adapt the ‘mainstream’ conceptualization of ‘de-familization’ to the whole of theoretical knowledge as developed by feminist welfare state research. In the next step of our argument we will make plausible that on a conceptual level – and contrary to what policy makers and their political consultants across Europe are telling us and trying to make us believe – de-familization (a) is not automatically and unequivocally increasing the ‘independence’ of women (or men), that de-familization (b) includes not only an economic but also a social dimension, and that de-familization (c) comprises not only the care giver’s but also the care receiver’s perspective. Things are more complicated than they seem at first glance – and than mainstream concepts of ‘defamilization’ would like to have them. De-familization is a complex, multifaceted concept. At its core is the question of ‘who cares’. De-familization was first defined by Eithne McLaughlin and Caroline Glendinning (1994: 65) as follows: . . . de-familisation is constituted by those provisions and practices which vary the extent to which well-being is dependent on ‘our’ relation to the (patriarchal) family. . . . de-familisation is about the terms and conditions under which people engage in families, and the extent to which they can uphold an acceptable standard of living independently of (patriarchal) ‘family’ participation.

250


Thus, de-familization refers to the question of (in)dependence within familial care relationships. Since a care relationship comprises the care giver on the one side and the care receiver on the other, de-familization cannot be reduced analytically to the situation of the care giver. De-familization processes change the care relationship as a whole and therefore also have an impact on the care receiver. The first part of the quote given above allows us to think of de-familization in this double perspective: on the one hand, de-familization has to be judged from the perspective of persons receiving care. We must ask for their status of dependence on or independence from the family, i.e. care receivers might be dependent on care given by a member of their own family (or household) or on financial support by the family in order to meet care needs. Care receivers might also be given a right to receive care as suggested by Knijn and Kremer (1997) who refer to a guaranteed access to qualitatively good professional care.5 On the other hand, de-familization has to be judged from the perspective of care givers. We must ask for their dependence on or independence from other members of their family or household. They might be dependent on financial support by the family breadwinner in order to meet care obligations or they might be bound by the expectations of care receivers that their needs should be met by close family members. Thus, the status of the care giver might be strengthened by a right to care which enables and rewards care giving through financial compensation as well as by a right not to care (cf. Lewis, 1997) which allows the care giver to leave a care relationship, e.g. because professional care takes over care obligations.6 Drawing from this clarification of different perspectives of (in)dependence we argue that the dependency or independence of both care receivers and care givers falls into two analytical dimensions: the degree of social (de-)familization, on the one hand, and of economic (de-)familization, on the other. In what follows, we will elaborate further these two dimensions with regard to care relationships between parents and children.7 The social dimension of (de-)familization is about the social and emotional relationship between the care giver (parent) and the care receiver (child). Usually, the parent feels an obligation to care for the child while the child is young and heavily emotionally bound to the parent.8 From the parent’s perspective, however, social de-familization would mean the right not to care for his or her child. Although we do not assume that parents could (or even should) be completely de-familized in this sense, they might nevertheless be de-familized gradually. On the one hand, the provision of child care services will contribute to the social de-familization of the parent, especially if they are low-cost, high-quality and easily available. On the other hand, the parent can become socially de-familized if other persons of the family or household take over (part of the) child care. Most obviously,


251

this could be the case within a model of shared parenthood with both parents accepting care responsibilities or within an intergenerational care arrangement with the grandparents providing care for their grandchildren.9 It should be acknowledged, however, that the social de-familization of the parent enables him or her to follow other interests besides child care. Labour market participation could be – but does not automatically have to be – such an interest. From the child’s perspective, social de-familization means the right not to be cared for by the (i.e. one) parent exclusively. Children have a need for additional care relationships with other family members (e.g. the other parent or the grandparents) and also with other persons from outside the family or household in order to fully develop their cognitive, intellectual and emotional potentials. The child’s right to both parents and its access to child care services outside the family’s household is thus central for the child’s degree of social de-familization. It is of course an open question (to be answered differently in different cases and contexts) what degree of social de-familization fits a child’s need. An intercessor or ‘child advocate’ could strengthen the position of the child within the care relationship and make sure that the child’s needs are assessed and met.10 The economic dimension of (de-)familization (to be distinguished from the social relationship conceptualized so far) describes different variants of choice for the care giver and the care receiver. From the perspective of the care giver (parent) the decision to what extent the child will be cared for by him or her depends also – and not least – on the financial possibilities (or restrictions) of the parent. A situation in which the parent is financially dependent on other members of his/her family or household might enable the parent economically to care just as well as if the parent receives direct payments for child care. But only financial independence of the parent gives him/her a substantial right to care and thus provides what could be called a ‘real choice’. The economic de-familization of the parent therefore requires his or her financial independence which would be best granted by payments for child care that allow the parent to form and maintain an autonomous household (cf. Orloff, 1993). An effective right to care is guaranteed only by direct and individualized payments for care.11 From an analytical perspective focusing on the child’s (or the care receiver’s) well-being, economic (de-)familization concentrates on the question of who covers the costs of the child’s care needs. Children have a right to choose additional care relationships besides their relations to close family members. These will standardly be care services or educational services which will have to be paid for. The costs of the child’s right to choose such services could be borne either by the family (or household) or by the welfare state. But, to be sure, only a socialization of these costs through

252


public transfer payments can, from the child’s perspective, guarantee ‘real choice’ and the economic de-familization of the child. When seen against the backdrop of this thorough analytical ‘deconstruction’ of the concept, the restrictions and shortcomings of the mainstream debate on de-familization should become immediately obvious. The common use of the concept distorts, and in a sense ‘de-problematizes’, the real-world processes and constellations of a ‘de-familization’ of care. We can distinguish two main aspects with regard to the inadequacy and undercomplexity of mainstream conceptualizations. First and most fundamentally, when talking of the need to de-familize care, the talk usually refers to the care giver only, disregarding the person in need of care and the consequences that de-familization may entail for him or her. (If the situation of the needy part of the care relationship is taken into account at all, it is commonly presupposed that the act of de-familization has unambiguously positive effects on the well-being of those receiving care.) Correspondingly, the independence the care giver is meant to enjoy through de-familization is strictly defined in the sense of financial independence, basically ignoring the social and emotional aspects of the care giver’s ‘liberation’ from his or her care responsibilities. (Again, where such considerations are given at all, there is no discussion of possible trade-offs between both dimensions because it is generally assumed that the social and/or emotional independence of the care giver with regard to the person in need of care is a value in itself.) Thus, what we are currently facing in the political as well as in the scientific debate on ‘de-familization’ is a significant conceptual reductionism (Table 11.1).12 Building on this reductionism, the next simplification characterizing current debates is that de-familizing the care giver (understood unidimensionally as the unburdening of the parent from care responsibilities, i.e. our dimension of social de-familization) is supposed to lead quasiautomatically to his or her economic independence, while the familization of care work (understood correspondingly as the burdening of the parent with care responsibilities, i.e. our dimension of social familization) is systematically equated with the economic dependency of the care giver – the care giver, almost needless to say, explicitly or implicitly being supposed to be the child’s mother and not the father. In other words: mainstream theorizing on and politicizing for ‘de-familization’ is characterized not only by Table 11.1

Conceptual reductionism in the debate on ‘de-familization’

Economic independence Social independence

Care giver

Receiver of care

X –

– –

253


an exclusive focus on care givers, on the social (de-)familization of care givers and on the economic (in)dependence (via labour market participation) of care givers – read: women – but also by a remarkably low tolerance for ambiguity, i.e. for the ambivalences and trade-offs the road to ‘defamilization’ (or, more exactly, to this variety of de-familization) might be paved with. Even if we, as a first step, accept part of the reductionism of mainstream conceptualizations and concentrate on the perspective of the care givers, there still remain counterintuitive interrelations between the two dimensions of social and economic (de-)familization and the (in)dependence of care givers. These amours dangereux for the dominant de-familization discourse, which commonly remain unacknowledged, may be systematized as shown in Table 11.2. Stating our argument in the most general terms, it is important to understand that there is no given, predetermined correlation between ‘(de-) familization’ of care (in its different facets) and the economic or social independence of women as care givers (always supposing that the ‘normal’ care giver is female, which is (a) empirically true and (b) politically assumed). The four ‘intuitive’ (‘i’) and ‘counter-intuitive’ (‘c’) cases derived from Table 11.2 will be explained in the following by referring again to care relationships between parent and child: (1) (i) If parents are socially familized as care givers, because there is no possibility to share child care either within the family nor by using child care Table 11.2 Intuitive and counterintuitive relations between social/economic (de-)familization and economic/social (in)dependence from the care giver’s perspective If . . .

then . . .

Social familization of care givers Social de-familization of care givers If . . .

then . . .

Economic familization of care givers Economic de-familization of care givers Note: i intuitive, c counter-intuitive relation.

Economic dependence of care givers

Economic independence of care givers

1 (i) 2 (c)

1 (c) 2 (i)

Social dependence of care givers

Social independence of care givers

3 (i) 4 (c)

3 (c) 4 (i)

254


facilities, the intuition would be (and actually is) to count them as financially dependent. Given that they are not willing to neglect their children, no options to gain financial independence via labour market participation are available for socially familized parents. (c) But: Even if the parent is the only provider of child care so that he or she is totally socially familized, the parent might be (at least partially) economically independent, e.g. due to individualized payments for care like paid parental leave. The degree of the parent’s independence would then be determined by the character (most importantly: the amount) of the payments for care, which allow the parent to develop (at least to a certain degree) economic independence from other household members and their earned incomes. However, payments for care might be (a) very low in terms of the amount and/or the duration of the benefit or (b) not fully individualized, e.g. if the amount of the benefit depends on the household income. In these cases (and especially if both (a) and (b) apply) the economic independence of the carer from his or her family will not be extended sufficiently. The most widespread individualized payment for child care is paid parental leave. Interestingly, countries which provide poor rates of formal child care (for children under three years old) and thus aim at the social familization of parents tend to have (if at all) also poor individualized benefits for child care. Their parental leave benefits are mostly low flat-rate payments which do not allow for the financial independence of the parent, e.g. the monthly flat rate amounts to €436 in Austria and – depending on the household income and the duration of the leave – to a maximum of €300 or €450 in Germany. In Italy, the benefit amounts to 30 per cent of the former wage and lasts only for six months whereas Germany provides payments for a maximum of two years and Austria even for a maximum of three years (cf. Leitner, 2003: 371). Thus, there is little empirical evidence for our argument, but the example of Luxembourg shows that high social familization of parents might indeed go hand in hand with high financial independence: Since 1999, parents can take six months of full-time leave from employment with a monthly benefit of €1,693 attached to it or 12 months of half-time leave with a monthly benefit of €846 (SSA, 2004). Moreover, the most recent German proposal to reform parental leave benefits aims at introducing a benefit based on 67 per cent of the parent’s former wage. Starting from 2007, the so called Elterngeld is supposed to be paid for ten months and can be extended to 14 months if the other parent takes at least two months of the leave. Besides individualized payments for care, socially familized parents can be thought of being economically independent due to contingent reasons (personal wealth or the existence of other income flows) or because of – and this is an important point to keep in mind with regard to possible social


255

policy reforms in the future – a welfare state systematically geared towards guaranteeing a basic income to every citizen. (We will be coming back to this last argument in the concluding part of the chapter.) (2) (i) If parents are socially de-familized – either due to the availability of child care facilities or because other family members share care responsibilities – they are supposed to be enabled to gain income via employment and thus (may) become economically independent. This is EspingAndersen’s (and the de-familization mainstream’s) line of argumentation described in the introduction to this chapter. We should nevertheless not forget to point out that there are also countries with high rates of formal child care (even for very young children) which provide comparatively high individualized payments for child care and thus contribute to the financial independence of parents: e.g. in Sweden the parental leave benefit amounts to 75 per cent of former earnings and the Danish flat rate during parental leave is rather high (cf. Leitner, 2003: 371).13 (c) But: We should be aware of the fact that the social de-familization of the parent does not necessarily increase his or her economic independence. In spite of the new opportunity structure (in terms of mothers’ labour market participation) brought about by the social de-familization of child care, the former care giver might not succeed in the labour market and earn only insufficient incomes to make him/her (or his/her household) truly ‘independent’. Many employed mothers are still financially dependent on their husband’s income or on welfare payments. In a situation of high unemployment, the chances for mothers to (re-) enter the labour market are a priori restricted. In fact, women’s increased employment rates tend to be based on more ‘flexible’ jobs with low job security and low wages (Lewis and Giullari, 2005: 82). Those who are employed are often working part-time to reconcile work and family life. Although women, in general, have high part-time rates, mothers’ incidence of part-time work is even higher, especially if there are two or more children. The Netherlands has the highest rate of part-time workers among mothers: 80 per cent of mothers with two or more children work part-time. In Australia, Germany, Switzerland and the United Kingdom, the share amounts to more than 60 per cent. Like women in general, working mothers earn, on average, 16 per cent less than men per hour worked. Besides direct wage discrimination, this gender wage gap accounts for the occupational and sectoral as well as the vertical segregation of employment by gender. Due to shorter working hours and the gender wage gap, there is also a considerable monthly wage gap between men and women. In the Netherlands and the United Kingdom, the monthly gender wage gap is highest with women earning just over half of what men earn. But

256


even in Sweden, one of Esping-Andersen’s best-practice countries, women earn only 70 per cent of men’s monthly income (cf. OECD, 2002). Thus, the achievement of economic independence through labour market participation is at least dubious for mothers. Their chances seem to rest more on a gradual economic autonomy. Moreover, despite the fact that the availability of child care, i.e. the social de-familization of the parent, is one of the most important prerequisites for mothers’ employment, the sound reconciliation of work and child care requires more than the availability of child care services. Flexible time structures would be needed for both child care and working hours to make ends meet. After all, at least part of the constraint on mothers’ employment is the low availability of part-time employment. This is especially important for Greece, Ireland, Italy, Luxembourg, and Spain (Daly, 2000: 491). Apart from these structural constraints of the labour market, cultural constraints for the employment of mothers have to be taken into consideration. The social de-familization of the parent will not lead to his/her economic independence via labour market participation if the employment of mothers is culturally rejected. In Belgium, e.g., half of all children between two and a half years and five years of age were attending child care facilities at the beginning of the 20th century. During the two world wars the number rose to two-thirds of all children in this age group and climbed up to the 100 per cent mark during the postwar period until 1970. But, up to the mid 1970s, this social de-familization of the parent did not result in a growing female employment rate. In contrast, the female employment rate fell constantly from two-thirds to one-third between 1850 and 1960. Yet in 1970, we find that Belgian women left the labour market without return after their first child had been born (cf. Leitner, 2005). (3) (i) Parents who are economically familized – e.g. by the lack of individualized payments for care or by indirect payments for care or by insufficient market income – are usually seen as socially dependent which means that they are obliged to take over child care. (c) But: This does not necessarily have to be the case. Parents could be economically familized and socially independent – i.e. free from child care obligations – at the same time, if e.g. a parent becomes (long-term) unemployed, and thus dependent on the income of the other parent, but does not increase his/her engagement in child care since the child has free access to extra-familial child care facilities anyway. The social independence of the economically familized parent would be best supported by child care services which not only have the function of enabling the reconciliation of work and child care (by just ‘billeting’ children), but which also aim at providing a high-quality educational function for the child.14


257

(4) (i) If parents are economically de-familized by an income from labour market participation, they could increase their social independence (and commonly are supposed to do so) in so far as child care is taken over by other family members or by child care services. (c) But: Social independence of parents with regard to their children will never be (and arguably should not be) reached completely. The economic defamilization of the parent could even strengthen his/her social dependence, if it comes through individualized payments for care. Quite interestingly, in Finland, parents were given the option to choose between the right to a daycare place for their children and a Child Home Care Allowance (CHCA) for those who did not take up a day-care place. This led to a high percentage of parents opting for the payment for care instead of the day-care place: in 1998, 45 per cent of children under three were cared for through CHCA whereas 24 per cent were covered by day care (Kröger et al., 2003: 40–3). Thus, the majority of parents opted for economic de-familization and social dependence. A lack of child care services and of inner-familial care sharing will also serve to uphold a high level of social dependence of the economically defamilized parent. Especially for working women (as opposed to men, at least in many or even most cases), de-familization in the sense of social independence from child care (and from familial as well as social child care expectations) can be said to be a chimera. What should have become clear at this point in our chapter is that ‘defamilization’ is a complex and, in terms of its potential for ‘autonomization’, highly ambivalent concept. To measure the degree of ‘dependency’ or ‘independence’ of care givers, then, it is of crucial importance to be clear about the dimensions of de-familization as well as about the facets of (female) independence that are under consideration. Under these circumstances, it should be uncontroversial at least analytically – if not politically – that different dimensions and degrees of female (in)dependence would have to be distinguished with regard to both the familization and de-familization of care givers in advanced welfare states. Moving now – if only briefly – on to the perspective of the care receiver, i.e. the child to be cared for by his or her parents, the ‘de-familization game’ becomes even more complicated. In the dominant discourse as reviewed above, de-familization is almost self-evidently supposed to be in the interest not only of women (because of it paving their way into the labour market) and the society at large (because of the mobilization of the hitherto untapped female labour reserve) but also of children. Children (or rather: many of them) are said to profit from being de-familized because of the improvement of their life chances brought about by detaching their education at least partially from their families (and thus relieving them

258


from their families’ educational failure; cf. Esping-Andersen, 2002b). But if the child’s perspective on de-familization really was to be taken seriously, things would look quite different – and rather more complex. Again, as in the case of the care giver’s perspective, there would have to be distinguished a social from an economic dimension of de-familization. The child as the person to be cared for has social needs, on the one hand, and economic requirements, on the other. The child’s social needs may be familized, meaning that his or her care is taken over by one (or both) of the parents, or they may be de-familized, care being then delivered by other institutions (private or public services). Similarly, the child’s economic needs may be met by his/her family, i.e. the costs of child raising may be internalized into the parents’ household and thus familized; or these economic needs may be effectively de-familized in the sense of externalizing, i.e. socializing the cost of children (by way of public transfers to households with children covering part or the whole of their child related expenses). As opposed to the case of the parent as the care giver, the child as care receiver cannot be supposed – the younger, the less – to be in a situation to make real choices between the familization or de-familization of his/her economic and (even less so) social needs; real-world children do not opt for social or economic dependence upon or independence from their families. But to do justice to children as a constitutive part of the child–parent care relationship, we should take into consideration the ‘would-be’ choices of children with regard to them being (de-)familized – choices (children as) real actors would make. In this sense, it is not too adventurous to assume that indeed economic de-familization is always in the child’s interest as care receiver. With regard to social de-familization, however, things are less clear, and it would seem to be in the child’s interest to really have, as far as possible, a real choice between being cared for by one of his or her parents, by both of them, by other family members (wherever applicable) or by extra-familial care givers – or even by a combination of two or more of these options. Just like in the case of a women-friendly welfare state, a ‘children-friendly welfare state’ would, above all, enable choices – and take children seriously as an essential and equitable part of the care relationship.

CONCLUSION Having said all this, we may keep our concluding section short and restrict it to two final remarks, an analytical and a political one. With respect to the analytical remark, we would like to finish our chapter by paraphrasing Esping-Andersen (as quoted in the introduction): ideological predilections aside, it should be evident to all that we cannot afford


259

not to deconstruct and differentiate the concept of ‘de-familization’ in advanced welfare state research of the 21st century. Taking ‘de-familization’ (analytically) seriously means taking into account both sides of the care coin: care givers and care receivers, parents and children alike. Care is ‘inherently relational’ (Lewis and Giullari, 2005: 94) – which means that not only women, but women, children and even men have to be analysed with regard to their needs, wants and interests concerning the familization or defamilization of care. Women’s, children’s and men’s autonomy and interdependence must be accounted for by an analytically sound concept of de-familization. Thus, the mainstream conceptualization focusing on liberating women from care burdens in order to guide them into (alleged) economic independence via labour market participation is only one possible variety of de-familization – and, to be sure, one that acknowledges gender equality, care work and children’s rights only ‘in a particular, partial and instrumental way’ (Lewis and Giullari, 2005: 78). In this context, a political remark may be permitted as well. If what is at stake is a welfare state that gives ‘a real choice’ – not only to women, but to children and, quite unconventionally, even to men – concerning the issue of care, its ‘familization’ or ‘de-familization’, we would have to head for a basic-income welfare state which would indeed leave the shaping of social relations of care open to the autonomous and interdependent decisions of its (family) citizens (van Parijs, 1995). Choice is socially ‘embedded’ and ‘genuine choice’ or ‘real freedom to choose’ in respect of the balance of paid and unpaid work at the level of the household (and hence gender equality) will involve not only a rebalancing of paid work between men and women, but a complicated rebalancing of unpaid work between the market, state and men and women. (Lewis and Giullari, 2005: 78)

If we include children as fully-fledged members of the care/work relationship in this picture of a world of ‘real choices’, we get an impression of what de-familization is – or should be – about: the collective well-being of women, men and children being part of a social collectivity of care givers and care receivers. Instead of subjecting care and care relations to productivist policy designs, we should wonder about how social policies could do justice to the complex arrangement of autonomy and interdependence – and of time, money, and love – which ‘care’ is all about.

NOTES 1.

This argument has been developed within the US context and might seem exaggerated when applied to the Scandinavian world. However, although Scandinavian-style universal

260

2.

3. 4. 5. 6.

7.

8. 9.

10.

11. 12.

13. 14.

Capturing the nature of welfare state change benefits are less disciminatory since they are not means tested, their flat-rate character makes them equally inferior to social security benefits. The expansion of public care services distinguished the Scandinavian model from other welfare states and was acknowledged (in principle) as a women-friendly policy by feminist social policy research though Scandinavian feminists also identified a trade-off between economic independence and dependence on the state for public services (Borchorst and Siim, 1987: 138). Or as Jane Lewis (1997: 173–4) puts it: (1) the right to do unpaid work and not to engage in paid work or (2) the right to do paid work and not to engage in unpaid work. Ilona Ostner (2004) discusses and criticizes this ‘ideological’ narrowing of the feminist debate which according to her eventually came to dominate the editorial policies of the main journal of feminist welfare state research, Social Politics. Obviously, family care could not be received by right, although if we think of the care relationship between parents and children, the care obligation of the parent is very strong, and sometimes even enforced by the law. Elisabeth Hammer and August Österle (2003: 41f.) propose – from a care giver perspective – to measure de-familization by (a) the freedom of choice to provide care due to payments for care and (b) the freedom of choice not to provide care which for them depends on the availability and accessibility of social services as well as on the care giver’s access to the labour market. Care relationships between (non-demented) adult persons differ from parent–child relationships in many respects. It is beyond the scope of this chapter to refer to these differences in a comprehensive way, but we will at least give some references when they seem appropriate. Similarly, in old-age care the care giver (usually) feels an obligation to care whereas the care receiver (usually) expects the family to provide care. Similarly, in old-age care the care giver could be socially de-familized by care services for the elderly or if care giving is shared with other family members. Due to the fact that in many cases the care giver is the partner of the care receiver, the idea of care sharing between partners is much more widespread with regard to child care than in terms of elderly care. In elderly care the social de-familization of care receivers would mean to enable them to become independent of the availability and willingness of family members to provide care. Easy access to social care services would guarantee this kind of choice. An intercessor would also be a necessary prerequisite for the social de-familization of demented care receivers. Payments that are adressed to the care receiver – as it is often the case in old-age care – will not (or only in a very indirect way) contribute to the economic de-familization of the care giver. This critique refers to mainstream social policy research. In contrast, feminist research comprises a much more comprehensive discourse on care (for an overview see Leira and Saraceno, 2005). From early on, care giving was analysed as a ‘labour of love’ (Finch and Groves, 1983) with a highly complex social dimension of interpersonal relationships of love and obligation. The interconnectedness of ‘caring for’ and ‘caring about’ (cf. Tronto, 1993; Sevenhuijsen, 1998) as well as the mutuality of the care relationship were described as a special ‘rationality of caring’ (Wærness, 1987). Moreover, care policies as well as policies to bring carers into paid employment have been a central field of analysis (e.g. Ungerson, 1997; Lewis, 2002; Gornick and Meyers, 2003). It should be noted, though, that services and payments are probably not used simultaneously. Parents would rather consume parental leave benefits immediately after the birth of the child and make use of care services later on. The sharing of care work within the family might be a less legitimate claim if one parent is not in the labour market.

12. Pension reform: beyond path dependency? Sven Jochem Change is eternal. Nothing ever changes. Both clichés are true. (Immanuel M. Wallerstein, 1974: 3)

INTRODUCTION1 It hardly seems necessary to emphasize that over the past two decades mature welfare states have changed in more than one respect. How can we assess these changes? Do these developments reflect dynamic adaptations of welfare programmes which, in the end, reinforce the basic welfare state institutions that had been founded several decades ago? In other words: are contemporary changes path dependent? Or do these changes imply programmatic and institutional innovations which alter the fundamental logic of welfare states? Contemporary welfare state research focuses on the question whether historical welfare paths are stable or undergo profound changes, which would imply that they have transcended the boundaries of historical paths. In this respect, the concept of path dependency is widely used, especially in comparative pension policy research. This chapter discusses basic dimensions of the path dependency concept in comparative welfare state research (cf. Thelen, 1999; 2004; Mahoney, 2000; Schwartz, 2000; Pierson, 2004; Streeck and Thelen, 2005). Different approaches to measuring processes of path dependency are introduced, discussed (second section) and confronted with results of case studies on recent pension developments in Denmark, Germany, the Netherlands, and Sweden (third section). Developments in mature pension systems are chosen because they serve as the locus classicus for path dependency arguments (Myles and Pierson, 2001). As a conclusion in the fourth section, I argue that path dependency is indeed a valuable concept which helps to understand contemporary welfare dynamics. The core problem of this approach, however, is the concise measurement of welfare state paths as well as the specific assessment of crucial thresholds which mark path departures. 261

262


PATH CONSTRUCTION AND DIMENSIONS OF CHANGE The path dependency concept was originally used to circumscribe the effects of economic and political institutions on economic and political dynamics of change. Within the broad field of contemporary neo-institutionalism, the concept was particularly discussed in the historical branch of the research community. Thus, the impact of historical patterns and specific political ‘rules of the game’ (Immergut, 1992) on welfare policies was analysed. Political institutions introduced earlier in history were assumed to have specific effects on policy making processes at later points in time, even if the political basis which led to the introduction of these institutions years ago was no longer in place. In other words, the historical inertia of institutions has an impact on future welfare developments. One might argue that this perspective, applied to welfare state dynamics, corresponds with the famous ‘freezing-hypothesis’ in party system research, formulated decades ago by Lipset and Rokkan (1967). In the literature dealing with this issue most attention is currently paid to the question of how institutions determine political behaviour. As Beyer (2005) shows, at least seven mechanisms might explain why a particular path chosen in the past is often followed in the future. Historical ‘sequencing’, the ‘layering’ of institutions, or ‘positive returns’ built into institutional configurations, to mention but three of such mechanisms, shed light on the difficulty of changing historically shaped paths. However, as Beyer (2005), Ebbinghaus (2005), Pierson (2004) or Streeck and Thelen (2005) emphasize, the same mechanisms which stabilize developmental paths may also be conducive to path departures if, for example, the political environment changes or political actors change their preferences. Beyond ‘simple’ notions of path dependency, which claim that there are some insurmountable lock-in effects blocking path deviant developments, there are also more ‘enlighted’ versions of the concept which aim to explain institutional path stability as well as path departure (for an early critique of this ambition cf. Schwartz, 2000). In the context of this volume, it is not the causal effects of political institutions, nor the usage of the path dependency concept as an independent variable, which this chapter concentrates on. Some of the theoretical aspects of path dependency will be discussed, but excursively in the case studies that follow. The main focus here is the use of the concept as a ‘dependent variable’, or a metaphor for the dependent variable, indicating the degree of welfare state change or continuity. Within comparative welfare state research the path dependency thesis is attractive to those commentators who do not regard the numerous changes during the past decades as having fundamentally altered the logic of modern welfare states. Nor have such

Pension reform: beyond path dependency?

263

changes diminished cross-national differences. Analysing national welfare states as an entity, contemporary research is dominated by three ‘holy trinities’. Assessed from a macro or ‘bird’s eye’ view (Siegel, 2002), the familiar concepts of different paths include the three ‘Regimes of Welfare Capitalism’ (Esping-Andersen, 1990; 1999), the different ‘Families of Nations’ (Castles, 1993; 2004) and the three ‘Varieties of Capitalism’ (Hall and Soskice, 2001; Hall and Gingerich, 2004). All these macrotypological approaches rest on a great number of different concepts, variables and data, quantitative as well as qualitative in nature, which were put together to describe the overall dynamics of different welfare paths. In accordance with these ‘holy trinities’ recent research has provided empirical evidence supporting the notion of path dependent dynamics. In the words of Gøsta Esping-Andersen: ‘In sum, within the advanced industrial democracies the contemporary politics of the welfare state is a politics of the status quo’ (Esping-Andersen, 1996b: 266–7; cf. also Esping-Andersen, 1999). Using public spending data, Francis G. Castles (2004) too provides evidence in favour of path dependent developments. Using his ‘family of nations’ concept as a means of grouping advanced OECD democracies, he concludes: ‘What appears to have been occurring during this period [i.e. 1980 until 1998] was that aggregate spending levels within expenditure share types and families of nations were becoming internally more homogeneous, without the spending levels of the types themselves necessarily becoming markedly more similar to each other.’ Indeed, Castles argues that incremental reforms seem to even stabilize historical paths and continue to foster cross-country differences. In those cases in which the political foundations of specific welfare institutions withered away, ‘their welfare state effects are likely to live long after them’ (Castles, 2004: 176). As a conclusion he predicts: ‘The diversity of welfare state regimes and families of nations is with us for the long-term’ (Castles, 2004: 174). In a similar vein, Peter A. Hall and Daniel W. Gingerich conclude their empirical analysis of the dynamics in different varieties of capitalism in favour of a path dependent explanation: ‘On balance, we read these figures as an indication that institutional practices did not converge dramatically across political economies during the 1980s and 1990s [. . .] The absence of wholesale convergence in the face of the substantial economic pressures experienced during the 1980s and 1990s suggests that the distinctions drawn by the varieties of capitalism literature between different types of political economies are likely to be relevant for some time to come’ (Hall and Gingerich, 2004: 35). I do not intend to explore these macro paths and their contemporary dynamics in any detail at this point (cf. Scruggs, Chapter 7, this volume). It will suffice to note that these assessments are based mostly on quantitative

264


data covering the aggregated output side of complex welfare state processes. Esping-Andersen (1999) combines quantitative data with some qualitative data, but he too concentrates to a great extent on the output side of the policy process. Hence, these perspectives on welfare state paths cover only a limited range of possible path characteristics. Possibly exceeding the methodological potential of this ‘bird’s eye’ perspective, particularly programmatic profiles of different welfare states are not covered in any detail. In what follows, I shall disentangle the ‘dependent variable’ and the path dependency concept to a certain extent and focus on pension paths and their dynamics during the past two decades. Due to the complex nature of the path dependency concept I follow Jørgen Goul Andersen (2005) who argues that the measurement of this concept should – at least – cover three dimensions: the (1) basic political process, (2) the basic programmatic rules and (3) the main outcome of welfare systems. Before discussing some possible measurements of each of these dimensions, however, it needs to be emphasized that at the start of empirical research which addresses the question of path dependency, major political institutions are regarded as crucial. In other words, formal institutions such as corporatism, proportional representation, or bicameralism, influence the policy making process in the welfare state. From this perspective, political institutions figure as independent variables whereas the policy output is the dependent variable. Paul Pierson has suggested that we transcend this separation by arguing that programmatic rules of welfare policies should themselves be interpreted as institutions. ‘Leaving aside the formal institutions typically explored by sociologists, the institutions that impinge on the modern citizen most directly and intensively as she goes about her daily life are in fact public policies, not the formal political institutions that have preoccupied political scientists’ (Pierson, 2004: 165, emphasis in original). This broadening of the institutional concept has the merit of covering exactly those policy rules that determine the ‘rules of the game’ in welfare policies to a great extent. The programmatic logic of PAYG pension schemes or the programmatic logic of labour market policies (cf. Clasen and Clegg, Chapter 8, this volume) differ across countries and, as we can assume, imply different rooms for manoeuvre for policy makers. But by accepting this broader perspective of the institutional path dependency concept, the contours of the welfare state become even more blurred. Especially the analytical separation between the dependent and the independent variable becomes a knotty problem in empirical research. Additionally, while formal political institutions are most of the time ‘sticky’, we can assume that welfare policies are not as stable as basic political institutions, such as constitutional structures. Hence, Pierson’s broadening of the path dependency concept further challenges valid measurements of welfare state changes as well as their explanations.


265

The path dependency argument for mature pension systems is discussed controversially in the literature. Assessing programmatic changes within this policy field some authors argue that most recent developments scarcely transcend historical paths (cf. Myles and Pierson, 2001). Others challenge path dependency arguments. As Karl Hinrichs (2000) states, the pension ‘elephants’ seem to be ‘on the move’ (cf. also Hinrichs and Kangas, 2003). And according to Martin Hering’s analysis of European pension schemes, due to the impact of the European monetary regime, a process of convergence is actually taking place within the European Union (Hering, 2005). Therefore, the current scientific dispute addresses the crucial question whether observable empirical changes are ‘minor’ or ‘big’, ‘not important’ for the logic of the system or, quite the contrary, changing the ‘fundamental rules’ of pension systems – and should therefore be seen as path breaking. How can we conceptually disentangle pension paths in a way that enables us to measure pension policy characteristics and their dynamics over time? First, the political process is crucial. Pension policies are naturally focused on long time horizons. This helps to explain that policy making in some European countries has been based not on political competition but on consensual reform practices. Policy consensus may entail broad issue-coalitions or a corporatist integration of interests into the framing of, and decision on, policy reforms. Therefore, the first measure of the welfare path should be a categorization of the political process in a dynamic perspective. Second, the policy profile of pension policies may be disentangled through different perspectives on different pension pillars and tiers (cf. Figure 12.1). Representing the ‘sector’ of a specific pension scheme the first pillar stands for the public sector, the second for the occupational sector, and the third for the private sector. This perspective makes use of the distinction between the state and the market sphere which is a prominent basic categorical differentiation in welfare research. Furthermore, the term ‘tier’ will reflect the type (or logic) of pension benefit. It matters whether benefits are targeted, minimum, flat-rate, earnings related or defined contribution. In this perspective, an additional distinction between mandatory and voluntary pension schemes becomes essential. Disentangling complex pension schemes further, it is possible to differentiate between various schemes for different occupational groups. Due to a lack of space such an undertaking is not possible in this chapter (but see the contributions to Anderson et al., 2006). In Figure 12.1 these perspectives, as well as some examples from European pension systems, are displayed. Applying this analytical perspective it becomes possible to localise programmatic changes. Hence, we know where to look. But this does not imply that we also know exactly how to measure the scope, extent or the quality of change. To address this problem, I propose a combined strategy. First,

266

Capturing the nature of welfare state change First pillar

Second pillar

Third tier (e.g. Swedish individually funded accounts)

Civil servants

Farmers

Self-employed

First tier Basic pension (e.g. Irish flat-rate pensions) Employees

Voluntary private pension (e.g. life insurance)

Government subsidized occupational pension (e.g. Danish tax deductible occupational pension schemes)

Government subsidized private pension (e.g. German Riester-Rente)

Mandatory occupational pension (e.g. Swiss second pillar or de facto mandatory Dutch occupational pensions)

Mandatory private pension (e.g. planned Portuguese Plafonamento)

Civil servants

Farmers

Self-employed

Employees

Second tier Earnings related part of pensions (e.g. French supplementary occupational pension schemes)

Voluntary occupational pension (e.g. British contributions above £3,600 per year)

Third pillar

Means tested part (e.g. Swedish guarantee pension) Social assistance (e.g. German social assistance substitutes for minimum pension)

Source: Anderson and Immergut (2006, Figure 1.1).

Figure 12.1

A classification of European pension systems

we can measure programmatic profiles for each pillar of a pension scheme, using the full range of information available, i.e. spending and financing data, programmatic rules, and changes to the programmatic profile of a specific pillar. Second, the outcome of pension schemes is to be assessed in this broader perspective, too. The replacement rate of the first pillar, for example, may be measured as well as the overall performance of pension schemes to avoid poverty in old age. Such an approach leads to two further challenges. First, how is it possible to measure programmatic rules? As Clasen and Clegg (Chapter 8, this


267

volume) argue, we can assess programmatic tendencies in a qualitative way. For pension policies I follow Hinrichs and Kangas (2003) who argue that the logic of the pillar mix should be the basis for the assessment of change. However, the crucial question in assessing the dynamics of change is: how can crucial thresholds be defined? When is a change still to be classified as path dependent, and in which cases do the dynamics transcend the historical path and result in path deviations? Here I propose a two-step procedure. As a first step, the overall programmatic contours of four European countries are illustratively sketched for the early 1980s. As a second step, the dynamics are described and measured. As a process, path departure is defined as a dynamic of change where the basic logic and the basic mix of the different pension pillars have changed (here since the beginning of the 1980s). Based on this definition, the introduction of a new pillar is to be seen as step towards path departure. Additionally, a process of path departure can be assumed if significant changes in the programmatic perspective occur. A significant change in the political process alone may not automatically imply new pension rules (albeit changing political constellations may provide the leeway for programmatic changes in the future). Similarly, the output of pension systems is determined at present by old programmatic rules. From an output perspective it seems necessary to measure the current output performance as well as taking account of output forecasts for the near future. The political processes as well as the outcome of pension schemes provide additional information, but the decisive measurement is the programmatic profile of national pension schemes. In sum, the measurement of institutional path dependency or path departure is a complex endeavour. As a means to disentangle the path dependency concept it seems useful to measure change or continuity in the political process, the programmatic profile, and selected outcome dimensions. However, the programmatic dimension especially is difficult to measure. As will be shown below, the perspective of pillars and tiers, as well as the use of qualitative and quantitative data, might provide a first step towards measuring pension system dynamics across several countries over time.

PENSION DYNAMICS – BEYOND PATH DEPENDENCY? At the beginning of the 1980s, the pension systems of Denmark, the Netherlands, Sweden, and Germany were clustering into two families with distinctive historical legacies. In the first, Germany represented the protoBismarckian country which had pioneered and established a public PAYG

268


system for industrial workers in the 1880s. Subsequently, this pension system was extended, particularly remarkably after World War II, in respect of coverage, benefit levels, and range of entitlements. In contrast, Denmark, Sweden, and the Netherlands are all members of the family of Beveridge countries, in which pension policies developed out of the ‘poor-law’ tradition. During the 1960s, the Beveridge countries split up, as Sweden successfully implemented a mandatory public complementary pension scheme (ATP) (Hinrichs, 2000). However, the Swedish pension system still differed from the German one, as the dominant first pillar had been supplemented by a multitude of occupational pension schemes since the 1960s. Before the introduction of the ATP system, labour market parties already negotiated occupational pension schemes. In the early 1980s the coverage for white collar employees and blue collar workers differed by between 10 per cent and 40 per cent (Palme and Svensson, 1997: 20). In Denmark and the Netherlands, in contrast, a similar shift towards the Bismarckian model did not occur. In both countries, complementary pension schemes were more or less absent until the early 1980s (in Denmark), or occupational pension schemes served as a functional equivalent for those

First pillar

Second pillar

Third pillar

Third tier

Voluntary occupational pension

Voluntary private pension

Second tier Earnings related part of pensions

Government subsidized occupational pension

Government subsidized private pension

Germany Sweden

Sweden (low/medium coverage)

First tier Basic pension

Mandatory occupational pension

Sweden Denmark Netherlands

Quasi-mandatory: Netherlands

Mandatory private pension

Social assistance

Figure 12.2

Pension characteristics in the early 1980s


269

pensioners who looked back on lengthy employment careers. Regulated by the state but negotiated by labour market parties, an early dispersion of occupational pension schemes was observable in the Netherlands. Applying the conceptual map of pension policy changes outlined in the previous section, until the early 1980s the German as well as the Swedish pension path relied on the first pillar and the first and second tier. In Sweden, the second pillar had already been growing in size and importance since the 1960s. In contrast to Germany it should be noted that in Sweden a minimum protection level was established through the national basic pension. Unlike this first group, Denmark relied on the first pillar and the first tier, whereas the Netherlands relied on the first pillar and a quasi-mandatory second-pillar, i.e. a mixture of mandatory occupational and voluntary occupational schemes (cf. Figure 12.2). Voluntary private pension schemes are not covered in this overview. What happened after the early 1980s? For which countries are path departing pension dynamics observable? And what were the political reasons for continuity and change? These questions are addressed in the following case studies. Denmark – Incremental Path Departure Introduced in 1956, the universal pension scheme was the cornerstone of the Danish pension system until the early 1990s. In 1964, the government and the Danish parliament agreed to expand this pillar towards a generous, flat-rate, tax financed universal pension scheme (the reform came into force in 1970). The upgrading of the universal pension scheme was partially the consequence of blocked ambitions which had been pursued by major parts of the Danish Social Democrats (SD) who had preferred the path chosen by the Swedish ‘brother’. They aimed at layering the basic pension scheme with a generous and mandatory income related state pension scheme. Because of the weakness of the SD in parliament, and because of internal splits in the trade union movement, this reform ambition was watered down. Instead, a marginal ATP scheme was introduced which by no means guaranteed income maintenance after the age of 67, the then official retirement age (which was reduced to the age of 65 only in 2002) (cf. GreenPedersen, 2006). This political defeat of social-democratic pension plans in Denmark did not only cause a significant divergence from the development of the Swedish pension system. This political stalemate also left the question of how to regulate an income related pension scheme unsettled. Danish civil servants, and increasingly since the 1960s, white collar employees in the public sector, had privileged access to funded occupational pension plans.

270


Since 1956, the government had provided marginal subsidies for private and occupational pension schemes. While in 1960 approximately 20 per cent of all employees were covered by occupational pension schemes (mainly in the public sector), the coverage increased only slightly until the early 1980s (to approximately 35 per cent of all Danish wage earners). As a consequence, the majority of the working age population was not covered by occupational pensions and relied on the (relatively generous) basic pension scheme (Green-Pedersen, 2006). The labour movement criticized this situation as low-wage earners especially faced old-age poverty. During the 1980s, the SD as well as trade unions pushed the bourgeois government to make occupational pension schemes mandatory for all wage earners. After the national elections in 1988, the bourgeois coalition lost its tacit support in parliament and had to rely on the right-wing populist Progress Party – or the SD. As a consequence and to calm down tensions in labour relations (cf. Jochem, 2003), the executive revitalized tripartite concertation talks. The broad pension commission (Arbejdsmarkedpensionsudvalget) favoured mandatory occupational pension schemes but the bourgeois coalition was internally split on this issue. While employers and employees in the public sector continued to expand occupational pension schemes through self-governance, the trade unions in the private sector hesitantly waited for legislation and focused exclusively on wage issues. The government finally invited the largest party in opposition, the SD, to negotiations. As Green-Pedersen argues (2006), the executive was willing to accept a legislative solution in exchange for commitments on the part of the SD on tax and other reform issues. However, aiming to strengthen its profile as a tough opposition party, the SD eventually undermined negotiations. When negotiations failed, the Metalworkers Union changed its strategy and opted for occupational pension schemes regulated via wage bargaining, which were, for the first time, negotiated in 1991. This reform process – together with the adjustment in 1993/94, which made national pensions more means tested – paved the way for an impressively rapid spread of occupational pension plans. Currently up to 93 per cent of all Danish employees are covered by occupational pension plans (Green-Pedersen, 2006). As a result, a universal ‘ “Rolls Royce” version of flat rate universalism’ (Goul Andersen, 2002a: 133) is now combined with occupational income related and funded pension schemes. Given the broad coverage of wage negotiations, occupational pensions have become ‘quasi-mandatory’ (Green-Pedersen, 2006). For the first pillar, the government is the responsible regulatory body, overseeing the expansion of means testing in 1993/94. For the second pillar, trade unions and employers’ associations


271

negotiate and regulate occupational pension issues without state interference. Hence, the development of the Danish pension scheme led to a multipillar scheme that went beyond the old pillar structure and logic. A parliamentary stalemate, inefficient tripartism, and the politics of nondecisions opened up the way for social partners to transcend their role to move beyond wage bargaining. This ‘silent revolution’ (Goul Andersen, 2005: 7) may be interpreted as a path departure which is not observable in public pension spending data (which remained rather stable). Hence it clearly shows that expenditure figures are to be interpreted very cautiously and are not useful as an indicator of change (or continuity) for this type of outcome. Additionally, new actors entered the policy field, new pension schemes expanded rapidly without direct interference of the executive – a paradigmatic case of incremental path departure. Germany – Incremental Path Layering Germany is one of the most cited examples of reform gridlock and path dependent welfare state change. Despite numerous reform steps, it is argued that path dependent and incremental change was prevalent until the end of the 1990s (Alber, 2000; Jochem, 2001; Schmidt, 2005). The last encompassing pension reform before German unification was passed in parliament in 1989 and, with most parts becoming effective in 1992 or later, contributed to a continuation of the historical path (Hinrichs, 1998) despite far reaching cost containment measures (Schulze and Jochem, 2006). Until the mid 1990s Norbert Blüm, the Minister for Labour and Social Affairs then in office, repeated that the future of German pensions would be safeguarded, despite the financial burden of German unification, and high and increasing unemployment, as well as ongoing demographic change. Consequently, in the process of unification the West German pension system was transferred to the former GDR, based on a generous benefit calculation mechanism for most pensioners in the East. From the mid 1990s onwards, the centre-right government started to opt for new policy instruments in the first pillar. However, these innovations (e.g. the introduction of a ‘demographic factor’ influencing the uprating of pension levels) were postponed and finally abolished by the red-green government after 1998. Furthermore, the Grand Coalition in pension policies between the ruling Christian Democrats and the Social Democrats in Opposition was eroded further during this time. While there were some new policy ideas swirling around in the public debate, during the last few years of the centre-right coalition no structural innovations were implemented. By and large, the Kohl government restricted its reform efforts to cost containment measures.

272


After some fumbling of the red-green government on welfare policies, in 2001 a pension reform changed the contours of the German pension scheme (Trampusch, 2005b). As Winfried Schmähl put it, the 2001 reforms marked a ‘paradigm shift’ in German pension policies (Schmähl, 2002; see also Hering, 2003; Hinrichs, 2004; Rüb and Lamping, 2005). The replacement rate was significantly reduced. The joint (employer and employee) contribution rate was capped at 20 per cent of gross wages. In order to compensate for the traditional principle that the first pillar should secure the standard of living of pensioners (Lebensstandardsicherung), the government subsidized private pensions (Riester-Rente), targeted especially at the lowincome brackets. After political negotiations and the intervention of trade unions, the opportunity for occupational pensions (Entgeltumwandlung) was included in the system of tax incentives, thereby giving the labour market parties new responsibilities (Schmähl, 2004). The reform complemented or layered the old pension system, which will in fact change the public-private mix of future pensioner generations. This paved the way (as was the case in Denmark) for new actors – private insurance companies – to enter the inner core of pension policy making. Because of far reaching tax incentives it also gave the Ministry of Finance an important veto power in this policy domain (Rüb and Lamping, 2005). In 2004 this ‘direction setting law’ (Rüb and Lamping, 2005: 2) was complemented by two reforms which openly rejected the idea that pensions provided by the first pillar could secure a standard of living which used to be the norm. Under the heading of a ‘sustainability factor’ the replacement rate was further reduced in the first policy package. Its construction enables the government to adjust pension upratings in an ad hoc fashion if the combined contribution rate of employers and employees threatens to rise above the fixed limit of 20 per cent. Additionally, the second reform (following a verdict of the Constitutional Court) will make future pension benefits taxable, which implies a further reduction of the net replacement rate. Undoubtedly, many parts of the old system still exist and for most retired Germans the bulk of pension income will continue to be provided through the first pillar pension in the near future. A successful financial consolidation of the public pension schemes has not occurred. On the contrary, public pension spending has been steadily increasing since the 1990s. The reduction of benefit levels will be noticeable after 2010 at the earliest. Nevertheless, the logic of the German pension system has changed. The role of the first pillar has been reduced and will be further diminished in the future. Until today, the incentives of the Riester-Rente are only used scantily, especially by low-income earners. However, under the Grand Coalition, in 2005 the government announced that mandatory private or occupational pension schemes were not necessary and, hence, law making not on the


273

political agenda, because coverage rates for both private and occupational voluntary schemes were increasing (Rentenbericht, 2005). In order to dampen future expenditure growth for the first pillar, the Grand Coalition decided to increase the official retirement age to 67 years. This change will be gradually implemented between 2007 and 2029. Incrementally, the pension system, the pillar mix and the logic of the whole pension system will change due to these and more recent reform measures. The future replacement rate of the first pillar will be significantly lower than at present and a consensus is observable in the political discourse as all political parties emphasize the increasing importance of occupational and private supplementary pensions. The prototypical Bismarckian German pension system has thus started to change its path due to institutional layering with supplementary and voluntary pension schemes. The Netherlands – Path Dependent Policy Blockage The Dutch pension system illustrates the power of policy inertia and path dependency because of reform blockages. In the Netherlands, a complex multi-pillar pension system with a universal basic pension scheme emerged after World War II (adopted in 1956), which is complemented by occupational pensions. The universal basic pension is financed through contributions (in 2005 the contribution rate stood at 17.9 per cent of gross wages). Every Dutch resident over 65 (with 50 years of residence)2 benefits from the basic pension (AOW). This first pillar scheme is complemented by funded earnings related occupational pensions. As early as 1908, the Dutch state regulated occupational pensions for the first time and 750 different occupational pension funds existed by 1938. Today, approximately 90 per cent of all wage earners are covered by occupational pension schemes. De jure, the Dutch government has the power to declare occupational pensions as mandatory for specific economic sectors. The government provides far reaching tax incentives in order to stimulate coverage of these pension schemes. In 2003, these tax deductions amounted to 2.1 per cent of GDP (Anderson, 2006). Since the implementation of the AOW, the contribution rate to the basic pension scheme increased significantly from 6.75 per cent of gross wages in 1957 to 17.9 per cent in 2001 (Anderson, 2006). One prominent measure used frequently by governments since the 1980s has been to freeze the universal pension benefit. For three subsequent years (1993, 1994, 1995) the indexation of pension benefits was suspended (Visser and Hemerijck, 1997). However, as most political observers of the 1994 general elections commented, the then major party in government – the Christian Democratic CDA – had to pay a high price for these cost containment measures. Two

274


parties representing the interests of pensioners were able to gain votes and entered parliament for the first time. And as a consequence of a disastrous electoral defeat, for the first time since 1918 the CDA was not part of the ruling coalition government. Instead, the Social Democratic Party (PvDA) together with two liberal parties (VVD, D66) formed a ‘purple coalition’ in 1994. The new government blocked the cost saving measures which the previous government had suggested and tried to introduce new calculation rules for supplements in the AOW. However, after fierce opposition, the coalition withdrew this reform initiative. Since the 1994 national elections, cost containment measures in the AOW seem to be politically taboo. The PvDA started to discuss the future role of the basic pension and publicly deliberated whether well-off pensioners should contribute to the financing of the scheme (which would in effect imply means testing) and to set up a supplementary pension fund to cushion the impact of the future increase of the number of pensioners as a result of demographic change. After fierce political debates, the government failed to reach its goals. In effect, an AOW fund was established, but the government feeds this fund with taxes and through contributions. Well-off pensioners were not obliged to co-finance this fund. Hence, Dutch reform politics could not structurally alter the basic pension scheme, but policy blockage implied the introduction of a functional ‘lifebelt’ for the AOW for the time when the baby boomers approach retirement age. A discussion of the politics and policies of pension reform in the Netherlands cannot exclude a major area of reform: the disability pension scheme (WAO) as the ‘jewel in the crown’ of the social security system (Kuipers, 2004: 150). Introduced in 1967, the Dutch disability scheme serves as a major route from work to welfare. Since the economic crises of the 1970s and 1980s, a steady inflow of participants has induced huge financial burdens. The Grand Coalition first introduced cost containment measures in the WAO in 1991 (tightening access criteria, reducing benefit replacement rates and other measures). This caused the largest public protest march since World War II (Hemerijck, 2003: 249). In 1992, the focus of intervention changed. Now administering the system, the social partners were accused of not following political guidelines. As the head of a parliamentary commission stated: while the historical legacy has led to corporatist administrations, policy making can not suddenly abolish them, however, ‘an attempt must be made to break their power in another way’ (Buurmeijer, cited in Kuipers, 2004: 177). The work of this commission opened the way for a shift in the governance structure of the WAO. This way of making a political intervention enabled the partial privatization of the sickness insurance as well as a


275

de-corporatization of Dutch welfare administration. Televised hearings in the commission convinced public opinion that the administration was to blame for the Dutch ‘disability crisis’, enabling politicians to forge ahead with this reform despite resistance from the trade unions. But it is necessary to emphasize that this institutional reform neither altered the programmatic instruments of the WAO nor did it result in a decline of the dependency ratio. The Dutch history of pension policies during the past two decades shows how policy inertia shapes future developments. The early dynamics of the second pillar have influenced the historical dynamic until today. Beyond short-sighted cost containment measures, the members of the executive could not alter the fundamental profile and principal logic of the pension system. Especially in the field of early retirement, various governments could not push through institutional innovations. They all failed (to a great extent) because the pension issue was highly prioritized in public debate. Since the elections in 1994, no major party has openly suggested far-reaching structural reforms in the first pillar. Blocked pension policies imply dynamics in line with historical institutions, which were only layered by a functional ‘lifebelt’, the AOW fund. It is necessary to note that the Dutch pension system already had a diversified pillar mix in the early 1980s. Nevertheless, the dynamics during the past decades reflect policy stalemate and the status quo regarding the logic of the whole pension system. Sweden – Big-Bang Reform with Path Departure? In the comparative welfare state literature Sweden is usually presented as an example of a country which introduced major reform of a mature pension system. It was an informal Grand Coalition, consisting of the four bourgeois parties in government and the major opposition party, the Social Democratic Party (SAP), which changed the fabric of pension rules in one major policy package. The political circumstances as well as the one-shot political reform process both stand out in a comparative perspective. Sweden was an ‘early bird’ in pension scheme innovation, implementing a basic pension as early as 1914, which was updated and renamed in 1935 (folkpension). Pension benefits were meagre, and in the early 1950s the mean benefit level was only equal to approximately 30 per cent of the average industrial wage (Anderson and Immergut, 2006). Similarly to the Danish situation, employees in the public sector were members of an earnings related occupational pension scheme, and trade unions in the private sector forcefully demanded an earnings related public pension scheme for the whole working population. After several years of policy deliberations and despite highly insecure political circumstances (Immergut and Jochem,

276


2006), the national supplementary pension scheme (ATP), as ‘jewel in the [Swedish] crown’ (Lundberg, 2003), passed parliament. Rather unusually for a contribution based public pension scheme the ATP was partially funded.3 Even more importantly, the ATP became a national symbol for social-democratic Sweden (Svensson, 1994; Lundberg, 2003). Hence, due to historical developments, by the early 1980s the Swedish pension pillar mix was to a great extent diversified. The big pension reform that was agreed upon in 1994 was gradually implemented until 1998 under minority governments led by the Social Democratic Party. Several reform aspects contribute to the conclusion that this pension reform seems to be a ‘radical overhaul of the existing system’ (Anderson and Immergut, 2005: 20). First, the universal flat-rate basic pension was replaced by a new ‘guarantee pension’. Pension levels have been indexed in line with inflation since 2003, hence also marking a major shift away from wage or wage and price related pension adjustments. Second, the old ATP defined benefit scheme was replaced by two defined contribution schemes. The new public ‘income pension’ is supplemented by a ‘premium pension’ which is a funded individual account scheme administered by a state agency. Into this ‘premium reserve’ 2.5 percentage points of gross wages (out of a total contribution rate of 18.5 per cent) can be invested in different funds. Alternatively the wage earner may decide to let the state administer the money. Third, the new income pension benefits are indexed in line with economic growth and life expectancy. Furthermore, earnings related benefits are now based on lifetime earnings instead of the best 15 out of 30 years. Finally, while in the ATP system pension contributions were paid entirely by employers, the new system is jointly financed by employers and employees, each paying half of the contribution rate. Indisputably, the pension reform of 1994–98 changed the dominant pension rules with one major reform step. But from our perspective the logic of the whole system mainly changed the first pillar, while other pillars were not directly targeted by the reform. As far-reaching as these changes might be, the programmatic changes ‘only’ altered the functioning of the first and second tier within the first pillar. The single privatization measure was the introduction of the ‘premium pension’, through which employees have the freedom of choice how to invest the money. It is noteworthy that in 2006 benefit increases in the new system were higher than they would have been under the old ATP rules (Settergren, 2005). Additionally, occupational pensions spread further across the economy. Today, occupational pensions in Sweden are quasi-mandatory (Anderson and Immergut, 2006). In sum, the programmatic rules within the first pillar changed to a great extent; the pillar mix, however, remained more or less stable. Therefore, the Swedish case is in our conception of path dependency a borderline case.


277

CONCLUSION Strictly speaking, every policy change is path dependent, as every change in the policy profile responds to problems built into the old system to a certain extent (Goul Andersen, 2005). Hence, pension reform always involves programmatic change and programmatic continuity. Contemporary welfare state research is confronted with the challenge of assessing whether observable changes are undermining historical paths, or have at least the potential to do so in the long run. From a ‘bird’s eye’ view, concepts such as welfare regimes, families of nations, or varieties of capitalism (Esping-Andersen, 1999; Castles, 2004; Hall and Gingerich, 2004) may still offer valuable analytical frameworks for the shape of welfare paths over time. From a ‘micro’ perspective that focuses on changes within existing pension systems, however, we can observe changes which, in some cases, transcend the historical pension scheme logic. In Germany and Denmark, and with some limitations even in Sweden, political reforms and self-steering processes of labour market parties have paved the way for profound new programmatic logics in pension policies. As Hering (2005) and Bonoli (2003b) argue, there is indeed evidence that current pension systems start to converge ‘towards the multipillar end of the spectrum’ (Bonoli, 2003b: 414). Therefore, the usage and usefulness of the path dependency concept depend on the perspective we take. The holistic perspective may provide evidence for dynamic continuity and the robustness of crucial differences between several countries, while a more detailed perspective, which also takes programmatic dynamics into account, may provide evidence for path departures. The ‘dependent variable problem’ may not be solved with a ‘silver bullet’. Different perspectives are not only legitimate but necessary. We should bear in mind that our arguments concerning historical patterns crucially depend on our analytical perspective and categories. This chapter has dealt with the measurement of path dependency in four European pension systems. The proposed analytical perspective covers the political process, the programmatic profile as well as the outcome of pension policy. Three observations can be made. First, analysing pension spending patterns, the overall stability is impressive (cf. Table 12.1). Despite some small fluctuations – and the German trend of increasing spending – the overall assessment reveals continuity. If we analyse spending projections, which are based on OECD calculations and updated by Reimann (2005), the results are different. The most severe increases are forecast for the Netherlands and Germany, whereas the anticipated spending increase in Denmark is rather moderate. According to these estimates, the Swedish pension system will face one of the lowest growth dynamics until the year 2050. If the most recent German reforms (between 2001 and 2004) were included the anticipated

278

Table 12.1


Pension dynamics – spending and politics

Dimension Pension spending (1995–2003) (% GDP) Projection of pension expenditure, 2000 until 2050 (increase in percentage points, pension spending in relation to GDP) Pension politics

Denmark

Germany

The Netherlands

Sweden

11.7→11.1

11.1→12.0

9.5→9.2

12.0→12.2

4.1

5.0 (3.3)*

6.2

2.6

Consensus Politics of End of Party about basic commissions, corporatist competition pension, selfnew policy administration, leading to governance of actors and party broad occupational fragile policy competition in consensus, new pensions consensus a volatile party policy actors system

Note: * in brackets: the most recent pension reforms (2001, 2004) are taken into account. Sources: Eurostat (2006a: 77), Reimann (2005), and the case studies above.

spending increase would be only 3.3 percentage points (instead of 5.0) by the year 2050, approaching the group of the best performers (Reimann, 2005: 5). Second, the governance mode changed notably in Denmark as well as in Germany and in Sweden, where the (partial) funding of pension schemes opened access for new policy actors in policy making. In the Dutch case, dismantling the corporatist welfare administration was a notable change. However, the consequence of this development for future programmatic dynamics remains to be seen. Third, as the case studies showed, important changes to the programmatic profile and the mix of pension pillars occurred in Denmark and Germany. The Swedish case stands for an all-encompassing reform in a single step, but it seems questionable whether path departure is really as noticeable as it is considered to be in several assessments of only the Swedish example. The Dutch case provides evidence for policy stalemate and a stable logic of pension policies (cf. Figure 12.3). In all four countries programmatic innovations were influenced by policy inertia. Hence, to focus on the historical paths provides useful insights into

279

Pension reform: beyond path dependency? First pillar

Second pillar

Third pillar

Third tier Sweden

Voluntary occupational pension

Voluntary private pension

Second tier Earnings related part of pensions

Government subsidized occupational pension

Government subsidized private pension Germany

Germany (considerable retrenchment) Sweden

Germany (fragile expansion)

First tier Basic pension

Mandatory occupational pension

Sweden Denmark The Netherlands

Quasi-mandatory: The Netherlands Denmark Sweden

Mandatory private pension

Social assistance

Note: Bold: Programmatic innovations of the pillar mix for each country.

Figure 12.3

Pension dynamics – programmatic profiles

the dynamics of pension systems over time. Most innovations reported here layered already existing institutions. Both old and new schemes when combined, however, induced a new logic of the Danish and German pension systems. The most impressive case in this regard is the change of the German pension system. In the Federal Republic the pillar mix was expanded stepwise towards subsidized occupational and private pensions. The Danish case is an example of political stalemate and – as a consequence of the encompassing nature of labour relations – of an impressive growth of quasi-mandatory occupational pensions negotiated outside the parliamentary arena. The Swedish development is ambivalent in this respect. Some basic programmatic rules in the first pillar have changed. Noteworthy is the introduction of a third tier in the first pillar. The expansion of occupational pensions, however, had already been initiated in the 1970s. Strictly speaking therefore, the latest pension reform did not generate a major change within the pillar mix, but mainly recalibrated the rules within the first pillar. Finally, the Dutch pension system already represented an example of a diversified pension system in the early 1980s and

280


indicates merely gradual change since then. From this perspective, cost containment in the first pillar is the most urgent political question, which has not yet been solved. Up to this point, changes in different dimensions have been reported, but how are they measured? Certainly, the assessment offered in this chapter relies on qualitative comparative analyses which are difficult to translate into quantitative data. But the perspective applied opens the way to focus on programmatic innovation. Changes in this dimension are made explicit and are traceable. Because the path dependency concept is in danger of becoming a ‘catch-all’ phrase (Goul Andersen, 2002a: 131), the approach adopted in this chapter provides a clear perspective on pension dynamics. It defines thresholds of path departures and enables qualitative assessment. The approach applied in this contribution enables us to assess degrees of change and their implication for the pillar mix in mature welfare states. Its application underlines Immanuel Wallerstein’s statement that ‘change is eternal. Nothing ever changes. Both clichés are true’ (Wallerstein, 1974: 3).

NOTES 1. I am indebted to Jochen Clasen, Olli Kangas, Isabell Schulze and Nico A. Siegel for their support and critical remarks while preparing this chapter. For helpful comments on an earlier version of this chapter I thank the participants in the project workshop at Stirling in May 2005 as well as the participants in the workshop ‘Comparative Analysis of Welfare Reform: The Dependent Variable Problem’ at the ESPAnet Conference in Fribourg in September 2005. I especially thank Kara Ballarin for her very much appreciated improvement of the English text. The usual disclaimer applies. 2. For fewer years of residence, 2 per cent is subtracted from the pension benefit for each missing year. 3. The AP funds cushioned the transition costs during the reform. Up to 2004, the AP funds transferred approximately €38 billion to the government budget in order to compensate for the costs of transforming the pension system (Andersen and Immergut, 2006).

References Abramovitz, Mimi (1988), Regulating the Lives of Women: Social Welfare Policy from Colonial Times to the Present, Boston, MA: South End Press. Abramovitz, Moses (1984), ‘Catching up, forging ahead, and falling behind’, Journal of Economic History, 46(2), 385–406. Adcock, Robert and David Collier (2001), ‘Measurement validity: a shared standard for qualitative and quantitative research, American Political Science Review, 95(3), 529–46. Adema, Willem (1997), ‘What do countries really spend on social policies? A comparative note’, OECD Economic Studies, (28), 153–67. Adema, Willem (1999), ‘Net social expenditures’, OECD labour market and social policy occasional papers no. 39, Paris. Adema, Willem (2001), ‘Net social expenditures’, 2nd edn, OECD labour market and social policy occasional papers no. 52, Paris. Adema, Willem and Marcel Einerhand (1998), ‘The growing role of private social benefits’, OECD labour market and social policy occasional papers no. 32, Paris. Adema, Willem and Maxime Ladaique (2005), ‘Net social expenditure, 2005 edition. More comprehensive measures of social support’, OECD social, employment and migration working papers, no. 29, Paris. Adema, Willem, et al. (1996), ‘Net public social expenditures’, OECD labour market and social policy occasional papers no. 19, Paris. Alber, Jens (1987), Vom Armenhaus zum Wohlfahrtsstaat: Analysen zur Entwicklung der Sozialversicherung in Westeuropa, Frankfurt am Main: Campus. Alber, Jens (1988), ‘Continuities and changes in the idea of the welfare state’, Politics & Society, 16(4), 451–68. Alber, Jens (1996), ‘Selectivity, universalism, and the politics of welfare retrenchment in Germany and the United States’, manuscript, University of Constance. Alber, Jens (2000), ‘Der deutsche Sozialstaat in der Ära Kohl: Diagnosen und Daten’, in Stephan Leibfried and Uwe Wagschal (eds), Der deutsche Sozialstaat: Bilanzen-Reformen-Perspektiven, Frankfurt am Main and New York: Campus, pp. 235–75.

281

282

References

Alestalo, Matti and Hannu Uusitalo (1992), ‘Social expenditure: a decompositional approach’, in Jon Eivind Kolberg (ed.), The Study of Welfare State Regimes, New York: M.E. Sharpe, pp. 37–68. Allan, James and Lyle Scruggs (2004), ‘Political partisanship and welfare state reform in advanced industrial societies’, American Journal of Political Science, 48(3), 496–512. Amenta, Edwin (1993), ‘The state of the art in welfare state research on social spending efforts in capitalist democracies since 1960’, American Journal of Sociology, (99), 750–63. Amenta, Edwin (2003), ‘What we know about the development of social policy: comparative and historical research in comparative and historical perspective’, in James Mahoney and Dietrich Rueschemeyer (eds), Comparative Historical Analysis in the Social Sciences, Cambridge: Cambridge University Press, pp. 91–130. Anderson, Karen (2001), ‘The politics of retrenchment in a social democratic welfare state: reform of Swedish pensions and unemployment insurance’, Comparative Political Studies, 34(9), 1063–91. Anderson, Karen (2006), ‘The Netherlands: political competition and neocorporatist accommodation’, in Karen Anderson, Ellen M. Immergut and Isabelle Schulze (eds), The Oxford Handbook of West European Pension Politics: Political Institutions and Policy Outcomes in Comparative Perspective, Oxford: Oxford University Press. Anderson, Karen and Ellen M. Immergut (2005), ‘The politics of pension reform in Western Europe’, paper prepared for the annual meeting of the American Political Science Association, Washington, DC, 31 August – 4 September. Anderson, Karen and Ellen M. Immergut (2006), ‘Sweden: after social democratic hegemony’, in Karen Anderson, Ellen M. Immergut and Isabelle Schulze (eds), The Oxford Handbook of West European Pension Politics: Political Institutions and Policy Outcomes in Comparative Perspective, Oxford: Oxford University Press. Anderson, Karen, Ellen M. Immergut and Isabelle Schulze (eds) (2006), The Oxford Handbook of West European Pension Politics: Political Institutions and Policy Outcomes in Comparative Perspective, Oxford: Oxford University Press. Armingeon, Klaus (2004), ‘OECD and national welfare state development’, in Klaus Armingeon and Michelle Beyeler (eds), The OECD and European Welfare States, Cheltenham, UK and Northampton, MA, USA: Edward Elgar, pp. 226–42. Armingeon, Klaus and Giuliano Bonoli (eds) (2006), The Politics of Postindustrial Welfare States: Adapting Postwar Social Policies to New Social Risks, London: Routledge.

References

283

Armingeon, Klaus and Nathalie Giger (2006), ‘The electoral consequences of welfare state retrenchment in OECD nations’, paper presented at the International Conference of the Council for European Studies, Chicago, IL, 29 March–2 April. Armingeon, Klaus, Michelle Beyeler and Harmen Binnema (2001), ‘The changing politics of the welfare state – a comparative analysis of social security expenditures in 22 OECD countries, 1960–1998’, presentation to the ECPR Conference, Canterbury, UK, 6-8 September. Audier, Florence et al. (1998), ‘Le RMI comme mode particulier d’indemnisation du chômage’, in Philippe Mehaut and Philippe Mosse (eds), Les politiques sociales categorielles: Tome 2 – fondements, portée et limites, Paris, L’Harmattan, pp. 124–43. Bäckman, Olof (1992), ‘Sjukfrånvaro och arbetslöshet – Samband eller skensamband?’, Sociologisk forskning, 29, 38–49. Bambra, Clare (2005), ‘Cash versus services: worlds of welfare and the decommodification of cash benefits and health services,’ European Journal of Social Policy, 34(2), 195–213. Barbier, Jean-Claude (2004), ‘Systems of social protection in Europe: two contrasted paths to activation, and maybe a third’, in Jens Lind, Hans Knudsen and Henning Jørgensen (eds), Labour and Employment Regulation in Europe, Brussels: Peter Lang, pp. 233–54. Barbier, Jean-Claude and Wolfgang Ludwig-Mayerhofer (2004), ‘The many worlds of activation’, European Societies, 6(4), 423–36. Barr, Nicholas (2001), ‘The truth about pension reform’, Finance & Development, 38(3). Beck, Nathaniel and Jonathan N. Katz (1995), ‘What to do (and not to do) with time-series-cross-section data in comparative politics’, American Political Science Review, 89(3), 634–47. Becker, Howard (1998), Tricks of the Trade: How to Think about Your Research While You’re Doing It, Chicago: Chicago University Press. Bennett, Collin J. (1991), ‘Review article: what is policy convergence and what causes it?’, British Journal of Political Science, 21, 215–33. Beyer, Jürgen (2005), ‘Pfadabhängigkeit ist nicht gleich Pfadabhängigkeit! Wider den implizierten Konservatismus eines gängigen Konzepts’, Zeitschrift für Soziologie, 34(1), 5–21. Bleses, Peter and Martin Seeleib-Kaiser (2004), The Dual Transformation of the German Welfare State, London: Palgrave. Bolderson, Helen and Deborah Mabbett (1996), ‘Cost containment in complex social security systems: the limitations of targeting’, International Social Security Review, 96(1), 3–12.

284

References

Bollen, Kenneth A., Barbara Entwistle and Arthur S. Alderson (1993), ‘Macrocomparative research methods’, Annual Review of Sociology, 19, 321–51. Bonoli, Giuliano (2000), The Politics of Pension Reform. Institutions and Policy Change in Western Europe, Cambridge: Cambridge University Press. Bonoli, Giuliano (2003a), ‘Social policy through labour markets. Understanding national differences in the provision of economic security to wage earners’, Comparative Political Studies, 36, 983–1006. Bonoli, Giuliano (2003b), ‘Two worlds of pension reform in Western Europe’, Comparative Politics, 399–416. Bonoli, Giuliano (2005), ‘The politics of the new social policies. Providing coverage against new social risks in mature welfare states’, Policy and Politics, 33(3), 431–49. Bonoli, Giuliano (2006), ‘New social risks and the politics of postindustrial social policies’, in Klaus Armingeon and Giuliano Bonoli (eds), The Politics of Postindustrial Welfare States: Adapting Postwar Social Policies to New Social Risks, London: Routledge, pp. 3–26. Bonoli, Giuliano, and André Mach (2000), ‘Switzerland: adjustment politics within institutional constraints’, in Fritz W. Scharpf and Vivian Schmidt (eds), Welfare and Work in the Open Economy, Oxford: Oxford University Press. Bonoli, Giuliano, Vic George and Peter Taylor-Gooby (2000), European Welfare State Futures: Towards a Theory of Retrenchment, Cambridge: Cambridge University Press. Borchorst, Anette and Birte Siim (1987), ‘Women and the advanced welfare state – a new kind of patriarchal power?’, in Anne Showstack Sassoon (ed.), Women and the State: The Shifting Boundaries of Public and Private, London: Hutchinson, pp. 128–57. Borrás, Susanne and Kerstin K. Jacobsson (2004), ‘The open method of co-ordination and new governance patterns in the EU’, Journal of European Public Policy, 11(2), 185–208. Bouget, Denis (2003), ‘Convergence in the social welfare system in Europe: from goal to reality’, Social Policy and Administration, 37(6), 674–93. Brady, Henry E. and David Collier (eds) (2004), Rethinking Social Inquiry: Diverse Tools, Shared Standards, Lanham, MD: Rowman & Littlefield. Brandl, Bernd and Franz Traxler (2005), ‘Industrial relations, social pacts and welfare expenditures: a cross-national comparision, British Journal of Industrial Relations, 43(4), 635–58. Briggs, Asa (2000), ‘The welfare state in historical perspective’ (originally published 1961), in Robert E. Goodin and Deborah Mitchell (eds), The Foundations of the Welfare State Vol. III, Chelthenham, UK and Northampton, MA, USA: Edward Elgar, pp. 221–58.

References

285

Buhr, Petra (2003), ‘Wege aus der Armut durch Wege in eine neue Armutspolitik?’, in Angelika Gohr and Martin Seeleib-Kaiser (eds), Sozial- und Wirtschaftspolitik unter Rot-Grün, Wiesbaden: Westdeutscher Verlag, pp. 147–66. Bulmer, Simon J. and Claudio M. Radaelli (2005), ‘The Europeanisation of national policy’, in Christiane Lequene and Simon Bulmer (eds), The Member States of the European Union, Oxford: Oxford University Press. Busch, Per-Olof and Helge Jörgens (2005), ‘The international sources of policy convergence: explaining the spread of environmental policy innovations’, Journal of European Public Policy, 12(5), 860–84. Cameron, David (1978), ‘The expansion of the public economy: a comparative analysis’, American Political Science Review, 72, 1243–61. Carroll, Eero (1999), Emergence and Structuring of Social Insurance Institutions: Comparative Studies on Social Policy and Unemployment Insurance, Stockholm: Swedish Institute for Social Research. Castles, Francis G. (ed.) (1982a), The Impact of Parties, London: Sage. Castles, Francis G. (1982b), ‘The impact of parties on public expenditure’, in Francis G. Castles (ed.), The Impact of Parties. Politics and Policies in Democratic Capitalist States, London: Sage, pp. 21–96. Castles, Francis G. (1985), The Working Class and Welfare: Reflections on the Political Development of the Welfare State in Australia and New Zealand, 1890–1980, Sydney: Allen & Unwin. Castles, Francis G. (1989), ‘Social protection by other means: Australia’s strategy of coping with external vulnerability’, in Francis G. Castles (ed.), The Comparative History of Public Policy, Cambridge: Polity Press, pp. 16–55. Castles, Francis G. (ed.) (1993), Families of Nations: Patterns of Public Policy in Western Democracies, Aldershot: Dartmouth. Castles, Francis, G. (1994), ‘Is expenditure enough? On the nature of the dependent variable in comparative public policy analysis’, Journal of Commonwealth and Comparative Politics, 32(3), 349–63. Castles, Francis, G. (1998), Comparative Public Policy: Patterns of PostWar Transformation, Cheltenham, UK and Lyme, USA: Edward Elgar. Castles, Francis G. (2001), ‘On the political economy of recent public sector development’, Journal of European Social Policy, 11(3), 195–212. Castles, Francis G. (2002), ‘Developing new measures of welfare state change and reform’, European Journal of Political Research, 41, 613–41. Castles, Francis G. (2004), The Future of the Welfare State: Crisis Myths and Crisis Realities, Oxford: Oxford University Press. Castles, Francis G. and Deborah Mitchell (1992), ‘Identifying welfare state regimes: the links between politics, instruments and outcomes’, Governance, 5(1), 1–26.

286

References

Castles, Francis G. and Deborah Mitchell (1993), ‘Three worlds of welfare capitalism or four?’, in Frank Castles (ed.), Families of Nations, Brookfield, VT: Dartmouth. Castles, Francis G. and Herbert Obinger (2006), ‘Towards more comprehensive measures of social support: adding in the impact of taxes and private spending or netting out the impact of politics on redistribution?’, ZeS working paper 04/2006, University of Bremen, Zentrum für Sozialpolitik. CEC (Commission of the European Communities) (2002), ‘Communication from the Commission to the Spring European Council in Barcelona. The Lisbon Strategy – making change happen. COM (2002) 14 final’, Brussels. CEU (1996), Resolution of the Council and the Representatives of the Governments of the Member States meeting within the Council of 20 December 1996 on equality of opportunities for people with disabilities, Official Journal of the European Communities C 012, 13/01/1997. CEU (2000a), ‘Council Directive 2000/78/EC of 27 November 2000 establishing a general framework for equal treatment in employment and occupation’, Official Journal of the European Communities L303, 02/12/2000:0016. CEU (2000b), ‘Council Decision 2000/750/EC of 27 November 2000 establishing action programme to combat discrimination (2001–2006)’, Official Journal of the European Communities L303: 23–8. Chassard, Yves (2001), ‘European integration and social protection: from the Spaak Report to the Open Method of Coordination’, in D.G. Mayes, Jos Berghman and Robert Salais (eds), Social Exclusion and European Policy, Cheltenham, UK and Northampton, MA, USA: Edward Elgar, pp. 277–305. Clasen, Jochen (1994), Paying the Jobless. A Comparison of Unemployment Benefit Policies in Great Britian and Germany, Aldershot: Avebury. Clasen, Jochen (2005), Reforming European Welfare States: Germany and the UK Compared, Oxford: Oxford University Press. Clasen Jochen, and Wim Van Oorschot (2002), ‘Changing principles in European social security’, European Journal of Social Security, 4(2), 89–115. Clasen, Jochen and Daniel Clegg (2003), ‘Unemployment protection and labour market reform in France and Great Britain in the 1990s: solidarity versus activation?’, Journal of Social Policy, 32, 361–81. Clasen, Jochen and Daniel Clegg (2006a), ‘New labour market risks and the revision of unemployment protection systems in Europe’, in Klaus Armingeon and Giuliano Bonoli (eds), The Politics of the Post-Industrial Welfare State, London: Routledge, pp. 192–210.

References

287

Clasen, Jochen and Daniel Clegg (2006b), ‘Beyond activation: reforming European unemployment protection systems in post-industrial labour markets’, European Societies, 8(4), 527–53. Clayton, Richard and Jonas Pontusson (1998), ‘Welfare state retrenchment revisited – entitlement cuts, public sector restructuring, and inegalitarian trends in advanced capitalist societies’, World Politics, 51(1), 67–98. Cornelisse, Peter A. and Kees P. Goudswaard (2002), ‘On convergence of social protection systems in the European Union’, International Social Security Review, 55(3), 3–17. Cox, Robert H. (2001), ‘The social construction of an imperative: why welfare reform happened in Denmark and the Netherlands, but not in Germany’, World Politics, 53(3), 463–98. Crepaz, Markus (1998), ‘Inclusion versus exclusion: political institutions and welfare expenditures’, Comparative Politics, 31(1), 61–80. Cutright, Philip (1965), ‘Political structure, economic development, and national social programs’, American Journal of Sociology, 70, 550–7. Cutright, Philip (1967), ‘Inequality: a cross national analysis’, American Sociological Review, 32, 562–78. Daguerre, Anne and Bruno Palier (2006), ‘France: from social treatment to activation of the unemployed’, in Jochen Clasen, Maurizio Ferrera and Martin Rhodes (eds), Welfare States and the Challenge of Unemployment: Reform Policies and Institutions in the European Union, London: Routledge. Daguerre, Anne and Peter Taylor-Gooby (2004), ‘Neglecting Europe: explaining the predominance of American ideas in New Labour’s welfare policies since 1997’, Journal of European Social Policy, 14(1), 25–39. Dahlerup, Drude (1987), ‘Confusing concepts – confusing reality: a theoretical discussion of the patriarchal state’, in Anne Showstack Sassoon (ed.), Women and the State: the Shifting Boundaries of Public and Private, London: Hutchinson, pp. 93–127. Daly, Mary (2000), ‘A fine balance: women’s labor market participation in international comparison’, in Fritz W. Scharpf and Vivien A. Schmidt (eds), Welfare and Work in the Open Economy. Volume II: Diverse Responses to Common Challenges, Oxford: Oxford University Press, pp. 467–510. Daniel, Christine (2000), ‘L’indemnisation du chômage depuis 1974: d’une logique d’intégration à une logique de segmentation’, Revue Française des Affaires Sociales, 54(3–4), 29–46. Daniel, C. and C. Tuchszirer (1999), L’Etat face aux chômeurs: l’indemnisation du chômage de 1889 à nos jours, Paris: Flammarion. Danmark Statistik (2005), Statistical Yearbook 2005, Copenhagen.

288

References

de Marchi, Scott (2006), Computational and Mathematical Modelling in the Social Sciences, Cambridge: Cambridge University Press. Delhey, Jan (2001), ‘The prospects of catching up for new EU members: lessons for the accession countries to the European Union from previous enlargements’, Social Indicators Research, 56, 205–31. DeViney, Stanley (1983), ‘Characteristics of the state and the expansion of public social expenditure’, Comparative Social Research, 6, 151–73. Dolowitz, David P. and David Marsh (2000), ‘Learning from abroad: the role of policy transfer in contemporary policy making’, Governance, 13(1), 5–24. Dufour, Pascal, Gérard Boismenu and Alain Noel (2003), L’aide au conditionnel: la contrepartie dans les mesures envers les personnes sans emploi en Europe et en Amérique du Nord, Brussels: P.I.E.-Peter Lang. Eardley, Tony, Jonathan Bradshaw, John Ditch, Ian Gough and Peter Whiteford (1995), Social Assistance in OECD Countries: Volume 1, Synthesis Report, London: HMSO. Ebbinghaus, Bernhard (1998), ‘European labor relations and welfare-state regimes: a comparative analysis of their “elective affinities”, paper presented at the International Conference of Europeanists, Baltimore, MD. Ebbinghaus, Bernhard (2005), ‘Can path dependence explain institutional change? Two approaches applied to welfare state reform’, Max Planck Institute for the Study of Societies, discussion paper 05/2, Cologne. Ebbinghaus, Bernhard and Philip Manow (eds) (2001), Comparing Welfare Capitalism: Social Policy and Political Economy in Europe, Japan and the US, London: Routledge. Esping-Andersen, Gøsta (1985), Politics against Markets: The Social Democratic Road to Power, Princeton, NJ: Princeton University Press. Esping-Andersen, Gøsta (1990), The Three Worlds of Welfare Capitalism, Cambridge: Polity Press. Esping-Andersen, Gøsta (1996a), ‘Welfare states without work: the impasse of labour shedding and familialism in continental European social policy’, in Gøsta Esping-Andersen (ed.), Welfare States in Transition. National Adaptations in Global Economies, London: Sage, pp. 66–87. Esping-Andersen, Gøsta (1996b), ‘Positive sum solutions in a world of trade-offs?’, in Gøsta Esping-Andersen (ed.), Welfare States in Transition. National Adaptations in Global Economics, London: Sage, pp. 256–67. Esping-Andersen, Gøsta (1999), The Social Foundations of Post-Industrial Economies, Oxford: Oxford University Press. Esping-Andersen, Gøsta (2002a), ‘Towards the good society, once again?’, in Gøsta Esping-Andersen et al. (eds), Why We Need a New Welfare State, Oxford: Oxford University Press, pp. 1–25.

References

289

Esping-Andersen, Gøsta (2002b), ‘A child-centred social investment strategy’, in Gøsta Esping-Andersen et al. (eds), Why We Need a New Welfare State, Oxford: Oxford University Press, pp. 26–67. Esping-Andersen, Gøsta (2002c), ‘A new gender contract’, in Gøsta Esping-Andersen et al. (eds), Why We Need a New Welfare State, Oxford: Oxford University Press, pp. 68–95. Esping-Andersen, Gøsta and Marino Regini (eds) (2000), Why Deregulate Labour Markets?, Oxford: Oxford University Press. Esping-Andersen, Gøsta, Duncan Gallie, Anton Hemerijck and John Myles (eds) (2002), Why We Need a New Welfare State, Oxford: Oxford University Press. European Commission (1993), Old Age Pension Replacement Ratios: Relation Between Pensions and Income at the Moment of Retirement, Brussels: Office for Official Publications of the European Communities. European Commission (1994), A White Paper – European Social Policy. A Way Forward for the Union, OOPEC (COM (94) 333 final 27 July 1994). European Commission (1996), The Impact of Structural Policies on Economic and Social Cohesion in the Union 1989–1999: A First Assessment Presented by Country, regional development studies, Brussels: Office for Official Publications of the European Communities. European Commission (1997), ‘Modernising and improving social protection in the European Union’, communication from the Commission, Brussels. European Commission (1999), ‘A concerted strategy for modernising social protection’, communication from the Commission, Brussels. European Commission (2000), ‘Social policy agenda’, COM(2000) 379 final, communication from the Commission, Brussels. European Commission (2001a), Second Report on Economic and Social Cohesion, Brussels: Office for Official Publications of the European Communities. European Commission (2001b), ‘Structural indicators’, Com(2001) 619 final, communication from the Commission, Brussels. European Commission (2004), A New Partnership for Cohesion Convergence Competitiveness Cooperation, Third Report on Economic and Social Cohesion, Brussels: Office for Official Publications of the European Communities. European Commission (2005a), Statistical Annex of European Economy Spring 2005, Brussels: Office for Official Publications of the European Communities. European Commission (2005b), ‘VAT rates applied to the member states of the European Community, DOC/1636/2005’, accessed at http://europa. eu.int/comm/taxation_customs/resources/documents/taxation/vat/how_ vat_works/rates/vat_rates_2005_en.pdf.

290

References

European Commission (2006a), ‘Time to move up a gear: the new partnership for growth and jobs’, communication from the European Commission to the Spring European Council, accessed at http://ec. europa.eu/growthandjobs/annual-report_en.htm. European Commission (2006b), Report from the Commission to the Council, the European Parliament, the European Economic and Social Committee and the Committee of the Regions on Equality between Men and Women, Brussels: Office for Official Publications of the European Communities. Eurostat (1995), Basic Statistics of the European Communities 1994, Brussels: Eurostat. Eurostat (1996), ESSPROS Manual, Luxembourg: Office for Official Publications of the European Communities. Eurostat (2005), Social Protection in the European Union, Statistics in Focus: Population and Social Conditions 14/2005, Brussels: Office for Official Publications of the European Communities. Eurostat (2006a), European Social Statistics: Social Protection Expenditure and Receipts: Data 1995–2003, Luxembourg: Office for Official Publications of the European Communities. Eurostat (2006b), Structural indicator social cohesion tables, accessed at www.epp.eurostat.cec.eu.int/portal/page?_pageid1996,45323734&_da dportal&_schemaPORTAL&screenwelcomeref&open/&produc tSTRIND_SOCOHE&depth2. Ferrarini, Tommy (2006), States, Families and Labour Markets, Cheltenham, UK and Northampton, MA, USA: Edward Elgar. Ferrera, Maurizio (1996), ‘Modèles de solidarité, divergences, convergences: perspectives pour l’Europe’, Swiss Political Science Review, 2(1), 55–72. Ferrera, Maurizio (1998), ‘The four “Social Europes”: between universalism and selectivity’, in Martin Rhodes and Yves Meny (eds), The Future of European Welfare, Basingstoke: Macmillan. Ferrera, Maurizio and Martin Rhodes (eds) (2000a), Recasting European Welfare States, London: Frank Cass. Finch, Janet and Dulcie Groves (eds) (1983), A Labour of Love: Women, Work and Caring, London: Routledge & Kegan Paul. Finn, Dan (1998), ‘Labour’s “New Deal” for the unemployed and the stricter benefit regime’, in Edward Brunsdon, Hartley Dean, and Roberta Woods (eds), Social Policy Review 10, London: Social Policy Association, pp. 105–22. Flora, Peter (ed.) (1986), Growth to Limits: The Western European Welfare States since World War II, Volume 1, Sweden, Norway, Finland, Denmark, New York: de Gruyter.

References

291

Flora, Peter (1987a), Growth to Limits. The Western European Welfare States Since World War II, Volume 4, Appendix (Synopses, Bibliographies, Tables), Berlin and New York: de Gruyter. Flora, Peter (1987b), State, Economy and Society in Western Europe 1815–1975, Volume 1, London: Macmillan. Flora, Peter and Jens Alber (1982), ‘Modernization, democratization and the development of welfare states in Western Europe’, in Peter Flora and Arnold J. Heidenheimer (eds), The Development of Welfare States in Europe and North America, London: Transaction Publishers, pp. 17–34. Flückiger, Y. and J. Cordero (1995), ‘Analyse économique des différentes propositions de réforme du financement des assurances sociales’, Geneva: University of Geneva. Fox Piven, Frances (1984), ‘Women and the state: ideology, power and the welfare state’, Socialist Review, 14(2), 13–22. Fraser, Nancy (1997), Justice Interruptus: Critical Reflections on the ‘Postsocialist’ Condition, New York and London: Routledge. Freyssinet, Jacques (2002), ‘La réforme de l’indemnisation du chômage en France (mars 2000–juillet 2001)’, Institute de Recherches Economiques et Sociales Working Paper 02.02. Friedrich, David (2006), ‘Policy process, governance and democracy in the EU: the case of the Open Method of Coordination on social inclusion in Germany’, Policy & Politics, 34(2), 367–83. Gal, John (1997), ‘Categorical benefits in welfare states: findings from Britain and Israel’, London School of Economics/STICERED Welfare State Programme working paper WSP/132. Galasso, Vincenzo and Paola Profeta (2004), ‘Lessons for an ageing society: the political sustainability of social security systems’, Economic Policy, 38, 63–115. Garrett, Geoffrey (1998), Partisan Politics in the Global Economy, Cambridge: Cambridge University Press. Garrett, Geoffrey and Deborah Mitchell (2001), ‘Globalization, government spending and taxation in the OECD’, European Journal of Political Research, 39(2), 145–77. Gerhard, Ute, Alice Schwarzer and Vera Slupik (eds) (1988), Auf Kosten der Frauen. Frauenrechte im Sozialstaat, Weinheim and Basel: Beltz. Gilbert, Neil (2002), The Transformation of the Welfare State: The Silent Surrender of Public Responsibility, New York: Oxford University Press. Gilbert, Neil and Ailee Moon (1988), ‘Analyzing welfare effort: an appraisal of comparative methods’, Journal of Policy Analysis and Management, 7, 326–40. Goodin, Robert (2000), ‘Crumbling pillars: social security futures’, The Political Quarterly, 71(2), 144–50.

292

References

Goodin, Robert and Martin Rein (2001), ‘Regimes on pillars: alternative welfare state logics and dynamics’, Public Administration, 79(4), 769–801. Goodin, Robert, Bruce Headey, Ruud Muffels and Henk-Jan Dirven (1999), The Real Worlds of Welfare Capitalism, Cambridge: Cambridge University Press. Gornick, Janet and Marcia C. Meyers (2003), Families that Work: Policies for Reconciling Parenthood and Employment, New York: Russell Sage. Gough, Ian, Jonathan Bradshaw, Tony Eardley and Peter Whiteford (1997), ‘Social assistance in OECD countries,’ Journal of European Social Policy, 7(1), 17–43. Goul Andersen, Jørgen (2002a), ‘Change without challenge? Welfare states, social construction of challenge and dynamics of path dependency’, in Jochen Clasen (ed.), What Future for Social Security? Debates and Reforms in National and Cross-national Perspective, Bristol: Policy Press, pp. 121–38. Goul Andersen, Jørgen (2002b), ‘Work and citizenship: unemployment policies in Denmark, 1980–2000’, in Jørgen Goul Andersen and Per Jensen (eds), Changing Labour Markets, Welfare Policies and Citizenship, Bristol: Policy Press, pp. 59–81. Goul Andersen, Jørgen (2005), ‘Ambiguity of welfare state change: institutions, output and outcomes’, paper presented at the EPSAnet Conference, Fribourg, 22-24 September, 2005. Green-Pedersen, Christoffer (2000), ‘How politics still matters: retrenchment of old-age pensions, unemployment benefits, and disability pensions/early retirement benefits in Denmark and in the Netherlands from 1982 to 1998’, doctoral thesis for the Department of Political Science, University of Åarhus, Åarhus. Green-Pedersen, Christoffer (2002), The Politics of Justification: Party Competition and Welfare State Retrenchment in Denmark and the Netherlands from 1982 to 1998, Amsterdam: Amsterdam University Press. Green-Pedersen, Christoffer (2004), ‘The dependent variable problem within the study of welfare state retrenchment: defining the problem and looking for solutions’, Journal of Comparative Policy Analysis, 6(1), 3–14. Green-Pedersen, Christoffer (2006), ‘A pension system according to the World Bank: development of the Danish pension system from 1980 to 2002’, in Karen Anderson, Ellen M. Immergut and Isabelle Schulze (eds), The Oxford Handbook of West European Pension Politics: Political Institutions and Policy Outcomes in Comparative Perspective, Oxford: Oxford University Press.

References

293

Green-Pedersen, Christoffer and Markus Haverland (2002), ‘The new politics of the welfare state and the new scholarship of the welfare state’, Journal of European Social Policy, 12(1), 243–51. Green-Pedersen, Christoffer and John Wilkerson (2006), ‘How agenda setting attributes shape politics: basic dilemmas, problem attention and health politics developments in Denmark and the US’, Journal of European Public Policy, 13(7), 1039–52. Greve, Bent (1996), ‘Indications of social policy convergence in Europe’, Social Policy and Administration, 30(4), 348–67. Guillén, Ana M. and Manos Matsaganis (2000), ‘Testing the “social dumping” hypothesis in Southern Europe: welfare policies in Greece and Spain during the last 20 years’, Journal of European Social Policy, 10, 120–45. Guillén, Ana M. and Bruno Palier (2004), ‘Introduction: does Europe matter? Accession to EU and social policy developments in recent and new member states’, Journal of European Social Policy, 14(3), 203–10. Gujarati, Damodar (2003), Basic Econometrics, 4th edn, New York: McGraw-Hill. Hacker, Jacob (2002), The Divided Welfare State, Cambridge: Cambridge University Press. Hacker, Jacob (2004), ‘Privatizing risk without privatizing the welfare state: the hidden politics of social policy retrenchment in the United States’, American Political Science Review, 98(2), 243–60. Hall, Peter A. (2003), ‘Aligning ontology and methodology in comparative research’, in James Mahoney and Dietrich Rueschemeyer (eds), Comparative Historical Analysis in the Social Sciences, Cambridge: Cambridge University Press, pp. 373–404. Hall, Peter A. and Daniel W. Gingerich (2004), ‘Varieties of capitalism and institutional complementarities in the macroeconomy: an empirical analysis’, Max Planck Institute for the Study of Societies, discussion paper 04/5, Cologne. Hall, Peter A. and David Soskice (eds) (2001), Varieties of Capitalism: the Institutional Foundations of Comparative Advantage, Oxford: Oxford University Press. Hammer, Elisabeth and August Österle (2003), ‘Welfare state policy and informal long-term care giving in Austria: old gender divisions and new stratification processes among women’, Journal of Social Policy, 32(1), 37–53. Hansen, Finn Kenneth (1998), ‘Forbrugerstyrelsens Familiebudget’, Råd & Resultater, 8, 13–20. Hansen, Hans (2002), Elements of Social Security, 9th edn, Copenhagen: Danish National Institute of Social Research.

294

References

Hansen, Hans (various years), Elements of Social Security in Six European Countries, Copenhagen: Danish National Institute of Social Research. Hantrais, Linda (1995), Social Policy in the European Union, London: Macmillan. Haverland, Marcus (2001), ‘Another Dutch miracle? Explaining Dutch and German pension trajectories’, Journal of European Social Policy, 11(4), 308–23. Heichel, Stephan, Jessica Pape and Thomas Sommerer (2005), ‘Is there convergence in convergence research? An overview of empirical studies on policy convergence’, Journal of European Public Policy, 12(5), 817–40. Held, David A., Anthony G. McGrew, David Goldblatt and Jonathan Perraton (1999), Global Transformations: Politics, Economics and Culture, Stanford, CA: Stanford University Press. Hemerijck, Anton C. (2003), ‘A paradoxical miracle: the politics of coalition government and social concertation in Dutch welfare reform’, in Sven Jochem and Nico A. Siegel (eds), Konzertierung, Verhandlungsdemokratie und Reformpolitik im Wohlfahrtsstaat: Das Modell Deutschland im Vergleich, Opladen: Leske and Budrich, pp. 232–70. Hemerijck, Anton C. and Jos Berghman (2004), ‘The European social partimony: deepening social Europe through legitimate diversity’, in Theo Sakellaropoulos, Jos Berghman, Anton Hemerijck, A. Stergiou and Y. Stevens (eds), Connecting Welfare Diversity Within the European Social Model, Antwerp: Intersentia, pp. 9–54. Hering, Martin (2003), ‘The politics of institutional path-departure: a revised analytical framework for the reform of welfare states’, Mannheimer Zentrum für Europäische Sozialforschung working paper, 65/2003, Mannheim. Hering, Martin (2005), ‘The politics of structural pension reform in Western Europe: does the EU matter?’, paper presented at the annual meeting of the American Political Science Association, Washington DC, 1-4 September. Hernes, Helga Maria (1984), ‘Women and the welfare state: the transition from private to public dependence’, in Harriet Holter (ed.), Patriarchy in a Welfare Society, Oslo: Universitetsforlaget, pp. 26–45. Hicks, Alexander (1994), ‘Introduction to pooling’, in Alexander Hicks and Thomas Janoski (eds), The Comparative Political Economy of the Welfare State, New York: Cambridge University Press, pp. 169–88. Hicks, Alexander (1999), Social Democracy and Welfare Capitalism: A Century of Income Security Politics, Ithaca, NY: Cornell University Press. Hicks, Alexander and Joya Misra (1993), ‘Political resources and the growth of welfare in affluent capitalist democracies, 1960–1982’, American Journal of Sociology, 99, 668–710.

References

295

Hicks, Alexander and Duane Swank (1992), ‘Politics, institutions, and welfare spending in industrialized democracies, 1960–1971’, American Political Science Review, 86, 658–74. Hinrichs, Karl (1998), ‘Reforming the public pension scheme in Germany: the end of traditional consensus?, Zentrum für Sozialpolitik working paper, 11/98, Bremen. Hinrichs, Karl (2000), ‘Elephants on the move: patterns of public pension reform in OECD countries’, European Review, 8(3), 353–78. Hinrichs, Karl (2001), ‘Elephants on the Move: patterns of public pension reform in OECD countries’, in Stephan Leibfried (ed.), Welfare State Futures, Cambridge: Cambridge University Press, pp. 77–102. Hinrichs, Karl (2004), ‘Alterssicherungspolitik in Deutschland: Zwischen Kontinuität und Paradigmenwechsel’, in Petra Stykow and Jürgen Beyer (eds), Gesellschaft mit beschränkter Hoffnung: Reformfähigkeit und die Möglichkeit rationaler Politik, Wiesbaden: VS, pp. 266–86. Hinrichs, Karl and Olli Kangas (2003), ‘When is a change big enough to be a system shift? Small system-shifting changes in German and Finnish pension policies’, Social Policy and Administration, 37(6), 573–91. Hobson, Barbara, Jane Lewis and Birte Siim (eds) (2002), Contested Concepts in Gender and Social Politics, Cheltenham, UK and Northampton, MA, USA: Edward Elgar. Holzinger, Katharina and Christoph Knill (2005) ‘Causes and conditions of cross-national policy convergence’, Journal of European Public Policy, 12(5), 775–96. Howard, Christopher (1997), The Hidden Welfare State: Tax Expenditures and Social Policy in the United States, Princeton, NJ: Princeton University Press. Hsiao, Cheng (1990), Analysing of Panel Data, Cambridge: Cambridge University Press, accessed at http://ec.europa.eu/economy_finance/ publications/european_economy/2005/statannex0105_en.pdf, and at http: //gstudynet.com/gwcsg/publications/ops/papers/wolf.pdf. Hu, Jingjing, Moira Nelson and John Stephens (2006), ‘Social democracy: decommodification or activiation’, paper presented at the 15th International Conference of the Council of European Studies, Chicago, 29 March–2 April. Huber, Evelyn and John D. Stephens (1993), ‘Political parties and public pensions: a quantitative analysis’, Acta Sociologica, 36, 309–25. Huber, Evelyn, Charles Ragin and John Stephens (1993), ‘Social democracy, Christian democracy: constitutional structure and the welfare state’, American Journal of Sociology, 99(3), 711–49.

296

References

Huber, Evelyn and John D. Stephens (2000), ‘Partisan governance, women’s employment, and the social democratic service state’, Amercian Sociological Review, 65, 323–42. Huber, Evelyn and John D. Stephens (2001), Development and Crisis of the Welfare States: Parties and Policies in Global Markets, Chicago: Chicago University Press. Hvinden, Bjørn (2003), ‘The uncertain convergence of disability policies in Western Europe’, Social Policy and Administration, 37(6), 609–24. Hvinden, Bjørn, Jon Kvist and Hannu Uusitalo (eds), Nordic Welfare States in the European Context, London and New York: Routledge, pp. 198–224. Immergut, Ellen M. (1992), Health Politics: Interests and Institutions in Western Europe, Cambridge: Cambridge University Press. Immergut, Ellen M. and Sven Jochem (2006), ‘The political frame for negotiated capitalism: electoral reform and reactions to crisis in Japan and Sweden’, Governance, 19(1), 99–133. International Labour Organisation (ILO) (various years), The Cost of Social Security, Geneva: ILO. Iversen, Torben (1998), ‘Wage bargaining, hard money and economic performance: theory and evidence for organized market economies’, British Journal of Political Science, 28(1), 31–61. Iversen, Torben and Anne Wren (1998), ‘Equality, employment, and budgetary restraint: the trilemma of the service economy’, World Politics, 50, 507–46. Jaeger, Mads M. and Jon Kvist (2003), ‘Pressures on state welfare in postindustrial societies: is more or less better?’, Social Policy and Administration, 37(6), 555–72. Jochem, Sven (2001), ‘Reformpolitik im deutschen Sozialversicherungsstaat’, in Manfred G. Schmidt (ed.), Wohlfahrtsstaatliche Politik: Institutionen, politischer Prozess und Leistungsprofil, Opladen: Leske & Budrich, pp. 193–226. Jochem, Sven (2003), ‘Nordic corporatism and welfare state reforms: Denmark and Sweden compared’, in Frans van Waarden and Gerhard Lehmbruch (eds), Renegotiating the Welfare State. Flexible Adjustment through Corporatist Concertation, London and New York: Routledge, pp. 114–41. Kangas, Olli (1991), The Politics of Social Rights: Studies on the Dimension of Sickness Insurance in OECD Countries, Stockholm: Swedish Institute for Social Research. Kangas, Olli and Joakim Palme (2000), ‘Does social policy matter? Poverty cycles in OECD countries’, International Journal of Health Services, 30, 335–52.

References

297

Kautto, Mikko and Jon Kvist (2002), ‘Parallel trends, persistent diversity: Nordic welfare states in the European and global context’, Global Social Policy, 2(2), 189–208. Kemmerling, Achim and Oliver Bruttel (2006), ‘ “New politics” in German labour market policy? The implications of the recent Hartz reforms for the German welfare state’, West European Politics, 29(1), 90–112. Kenworthy, Lane (1999), ‘Do social-welfare policies reduce poverty? A cross-national assessment’, Social Forces, 77(3), 119–39. Kenworthy, Lane (2007), Jobs with Equality, New York: Russell Sage. Kerr, Clark (1983), The Future of Industrial Societies: Convergence or Continuing Diversity?, Cambridge, MA: Harvard University Press. Kerr, Clark, John T. Dunlap, Frederick H. Harbison and Charles A. Myers (1960), Industrialism and Industrial Man, Cambridge, MA: Harvard University Press. Kerschen, Nicole (2005), ‘Contracts and commitments in French unemployment legislation’, in Els Sol and Mies Westerveld (eds), Contractualism in Employment Services: A New Form of Welfare State Governance, The Hague: Kluwer Law, pp. 307–23. Kiddal, Nina and Stein Kuhnle (2005), ‘The principle of universalism: tracing a key idea in the Scandinavian welfare model’, in Nina Kiddal and Stein Kuhnle (eds), Normative Foundations of the Welfare State: The Nordic Experience, London: Routledge. King, Gary, Robert O. Keohane and Sidney Verba (1994), Designing Social Inquiry, Princeton, NJ: Princeton University Press. Kitschelt, Herbert (2001), ‘Partisan competition and welfare state retrenchment: when do politicians choose unpopular policies?’, in Paul Pierson (ed.), The New Politics of the Welfare State, Oxford: Oxford University Press, pp. 265–302. Kitschelt, Herbert et al. (1999), ‘Convergence and divergence in advanced capitalist democracies’, in Herbert Kitschelt, Peter Lange, Gary Marks and John D. Stephens (eds), Continuity and Change in Contemporary Capitalism, Cambridge: Cambridge University Press, pp. 427–60. Kittel, Bernhard (1999), ‘Sense and sensitivity in pooled analysis of political data’, European Journal of Political Research, 35, 225–53. Kittel, Bernhard (2004), ‘A crazy methodology? On the limits of macroquantitative social science research’, mimeo, University of Amsterdam. Kittel, Bernhard and Herbert Obinger (2003), ‘Political parties, institutions, and the dynamics of social expenditure in times of austerity’, Journal of European Public Policy, 10(1), 20–45. Kittel, Bernhard and Hannes Winner (2005), ‘How reliable is pooled analysis in political economy?’, European Journal of Political Research, 44(2), pp. 269–93.

298

References

Klingemann, Hans-Dieter, Richard I. Hofferbert and Ian Budge (1994), Parties, Policies, and Democracy, Boulder, CO: Westview Press. Knijn, Trudie and Monique Kremer (1997), ‘Gender and the caring dimension of welfare states: toward inclusive citizenship’, Social Politics, 4(3), 328–61. Knijn, Trudie and Ilona Ostner (2002), ‘Commodification and decommodification’, in Barbara Hobson, Jane Lewis and Birte Siim (eds), Contested Concepts in Gender and Social Politics, Cheltenham, UK and Northampton, MA, USA: Edward Elgar, pp. 141–69. Knill, Christoph (2005), ‘Introduction: Cross-national policy convergence: concepts, approaches and explanatory factors’, Journal of European Public Policy, 12(5), 764–74. Knill, Christoph and Dirk Lehmkuhl (1999), ‘How Europe matters: different mechanisms of Europeanization’, in European Integration Online Papers, 3(7), accessed at http://eiop.or.at/eiop/texte/1999-007.htm Korpi, Walter (1983), The Democratic Class Struggle, London: Routledge & Keegan Paul. Korpi, Walter (1989), ‘Power, politics, and state autonomy in the development of social citizenship: social rights during sickness in eighteen OECD countries since 1930’, American Sociological Review, 54, 309–28. Korpi, Walter (1991), ‘Political and economic explanations for unemployment: a cross-national and long-term analysis’, British Journal of Political Science, 21, 315–48. Korpi, Walter and Joakim Palme (2003), ‘New politics and class politics in the context of austerity and globalization: welfare state regress in 18 countries, 1975–95’, American Political Science Review, 97(3), 425–46. Kröger, Teppo, Anneli Anttonen and Jorma Sipilä (2003), ‘Social care in Finland: stronger and weaker forms of universalism’, in Anneli Anttonen, John Baldock and Jorma Sipilä (eds), The Young, the Old and the State: Social Care Systems in Five Industrial Nations, Cheltenham, UK and Northampton, MA, USA: Edward Elgar, pp. 25–54. Kuipers, Susanne Liesbeth (2004), Cast in Concete? The Institutional Dynamics of Belgium and Dutch Social Policy Reform, Delft: Eburm. Kuivalainen, Susan (2004), ‘A comparative study on last resort social assistance schemes in six european countries’, States (Finnish national statistics) research report 146, Helsinki. Kvist, Jon (1997), ‘Retrenchment or restructuring – the emergence of a multitiered welfare state in Denmark’, in Jochen Clasen (ed.), Social Insurance in Europe, Bristol: Policy Press, pp. 14–39. Kvist, Jon (1999), ‘Welfare reform in the Nordic countries in the 1990s: using fuzzy-set theory to assess conformity to ideal types’, Journal of European Social Policy, 9(3), 231–52.

References

299

Kvist, Jon (2002), ‘Activating welfare states: how social policies can promote employment’, in Jochen Clasen (ed.), What Future for Social Security? Debates and Reforms in National and Cross-National Perspectives, Bristol: Policy Press, pp. 197–210. Kvist, Jon (2006), ‘Diversity, ideal types and fuzzy sets in comparative welfare state research’, in Benoit Rihoux and Heike Grimm (eds), Innovative Comparative Methods for Policy Analysis: Beyond the Quantitative–Qualitative Divide, New York: Springer, pp. 167–84. Langager, Klaus (1997), Indsatsen overfor de forsikrede ledige, Copenhagen: Danish National Institute of Social Research. Lazarsfeld, Paul (1937), ‘Some remarks on typological procedures in social theory’, Zeitschrift für Sozialforschung, 6, 119–39. Leibfried, Stephan and Paul Pierson (2000), ‘Social policy left to courts and markets?’, in Helen Wallace and W. Wallace (eds), Policy Making in the European Union, 4th edn, Oxford: Oxford University Press, pp. 267–92. Leira, Arnlaug and Chiara Saraceno (2005), ‘Care: actors, relationships and contexts’, in Barbara Hobson, Jane Lewis and Birte Siim (eds), Contested Concepts in Gender and Social Politics, Cheltenham, UK and Northampton, MA, USA: Edward Elgar, pp. 55–83. Leitner, Sigrid (2003), ‘Varieties of familialism: the caring function of the family in comparative perspective’, European Societies, 5(4), 353–75. Leitner, Sigrid (2005), ‘Conservative familialism reconsidered: the case of Belgium’, Acta Politica, 40(4), 419–39. Leitner, Sigrid and Stephan Lessenich (2003), ‘Assessing welfare state change: the German social insurance state between reciprocity and solidarity’, Journal of Public Policy, 23(3), 324–47. Leitner, Sigrid, Ilona Ostner and Margit Schratzenstaller (2004), ‘Was kommt nach dem Ernährermodell? Sozialpolitik zwischen Re-Kommodifizierung und Re-Familisierung’, in Sigrid Leitner, Ilona Ostner and Margit Schratzenstaller (eds), Wohlfahrtsstaat und Geschlechterverhältnis im Umbruch, Jahrbuch für Europa- und Nordamerika-Studien, Wiesbaden: VS, pp. 9–27. Lewis, Jane (1992), ‘Gender and the development of welfare regimes’, Journal of European Social Policy, 2(3), 159–73. Lewis, Jane (1997), ‘Gender and welfare regimes: further thoughts’, Social Politics, 4(2), 160–77. Lewis, Jane (2001), ‘The decline of the male breadwinner model: the implications for work and care’, Social Politics, 8(2), 152–70. Lewis, Jane (2002), ‘Gender and welfare state change’, European Societies, 4(4), 331–57. Lewis, Jane and Susanna Giullari (2005), ‘The adult worker model family, gender equality and care: the search for new policy principles and the

300

References

possibilities and problems of a capabilities approach’, Economy and Society, 34(1), 76–104. Lieberson, Stanley (1991), ‘Small n’s and big conclusions: an examination of the reasoning in comparative studies based on a small number of cases’, Social Forces, 70 (December), 307–20. Liebert, Ulrike (2003), ‘Between diversity and equality: analysing Europeanisation’, in Ulrike Liebert (ed.), Gendering Europeanisation, Brussels: P.I.E.- Peter Lang, pp. 11–46. Lijphart, Arend (1999), Patterns of Democracy: Government Forms and Performance in Thirty-Six Countries, New Haven, CT: Yale University Press. Lindbom, Anders (2005), ‘Swedish welfare state retrenchment in the 1990s: a new politics of welfare’, paper for European Consortium for Political Research Joint Secession of Workshops, Granada, 14–19 April. Lindert, Peter (2004), Growing Public, Volume 2, Cambridge: Cambridge University Press. Lipset, Seymour M. and Stein Rokkan (1967), ‘Cleavage structure, party systems, and voter alignments: an introduction’, in Seymour M. Lipset and Stein Rokkan (eds), Party Systems and Voter Alignments: Cross National Perspectives, New York: Free Press, pp. 1–64. Lødemel, Ivar and Heather Trickey (eds) (2001), An Offer You Can’t Refuse: Workfare in International Perspective, Bristol: Policy Press. Lundberg, Urban (2003), Juvelen i kronan: Socialdemokraterna och den allmännna (sic) pensionen, Stockholm: Hjalmarson & Högberg. Mabbet, Deborah (1995), Trade, Employment and Welfare: a Comparative Study of Trade and Labour Market Policies in Sweden and New Zealand, 1880–1980, Oxford: Clarendon Press. Mach, André (1999), ‘ “Small European states in world markets” revisited: the questioning of compensation policies in the light of the Swiss case’, Berlin: Wissenschaftszentrum für Sozialforschung. Mackenroth, Gerhard (1957), Sozialpolitik und Sozialreform, Tübingen: Mohr. Mahoney, James (2000), ‘Path dependence in historical sociology’, Theory and Society, 29, 507–48. Majone, Giandomenico (1994), ‘The rise of the regulatory state in Europe’, West European Politics, 17, 77–101. Mares, Isabela (1996), ‘Firms and the welfare state: the emergence of new forms of non-employment’, Berlin: Wissenschaftszentrum fuer Sozialforschung. Marshall, Thomas H. (1950), Citizenship and Social Class and Other Essays, Cambridge: Cambridge University Press.

References

301

Marshall, Thomas H. (1981), The Right to Welfare and Other Essays, London: Heinemann. Martin, Andrew and George Ross (eds) (2004), Euros and Europeans: Monetary Integration and the European Model of Society, Cambridge: Cambridge University Press. Matsaganis, Manos, Maurizio Ferrera, Luis Capucha and Luis Moreno (2003), ‘Mending nets in the South: anti-poverty policies in Greece, Italy, Portugal and Spain’, Social Policy and Administration, 37, 639–55. McClelland, Ian and Julia O’Connor (2006), ‘Structural Funds and convergence in the European Union’, University of Ulster. McLaughlin, Eithne and Caroline Glendinning (1994), ‘Paying for care in Europe: is there a feminist approach?’, in Linda Hantrais and Steen Mangen (eds), Family Policy and the Welfare of Women: Cross-National Research Papers. Third Series: Concepts and Contexts in International Comparisons, Loughborough: Loughborough University of Technology, pp. 52–69. Messner, Steven and Richard Rosenfeld (1997), ‘Political restraint of the market and levels of criminal homicide: a cross-national application of institutional-anomie theory’, Social Forces, 75, 1393–1416. Micklewright, John (1994), ‘The analysis of pooled cross-sectional data: early school leaving’, in Angela Dale and Richard B. Davies (eds), Analyzing Social and Political Change, London: Sage, pp. 78–97. Miller, L.S. (1976), ‘The structural determinants of welfare effort: a critique and a contribution’, Social Service Review, 50, 57–79. Mishra, Ramesh (1990), The Welfare State in Capitalist Society – Policies of Retrenchment and Maintenance in Europe, North America and Australia, New York: Harvester Wheatsheaf. Mitchell, Deborah, Ann Harding and Fred Gruen (1994), ‘Targeting welfare’, The Economic Record, 70(210), 315–40. Mitchell, Neil (1996), ‘Theoretical and empirical issues in the comparative measurement of union power and corporatism’, British Journal of Political Science, 26(3), 419–28. Montanari, Ingalill (2001), ‘Modernization, globalization and the welfare state: a comparative analysis of old and new convergence of social insurance since 1930’, British Journal of Sociology, 52, 469–94. Montanari, Ingalill, Kenneth Nelson and Joakim Palme (2006), ‘Convergence pressures and responses: recent social insurance development in modern welfare states’, mimeo, Swedish Institute for Social Research, Stockholm University. Moreno, Luis and Bruno Palier (2004), ‘The Europeanization of welfare. Paradigm shifts and social policy reforms’, paper presented at the

302

References

Welfare Reform and Management of Societal Change Conference, 23-25 April, Berlin. Myles, John (1984), Old Age in the Welfare State, Boston, MA: Little, Brown. Myles, John and Paul Pierson (2001), ‘The comparative political economy of pension reform’, in Paul Pierson (ed.), The New Politics of the Welfare State, Oxford: Oxford University Press, pp. 305–33. Naumann, Ingela (2005), ‘Child care and feminism in West Germany and Sweden in the 1960s and 1970s’, Journal of European Social Policy, 15(1), 47–63. Nelson, Barbara (1984), ‘Women’s poverty and women’s citizenship: some political consequences of economic marginality’, Signs, 10, 209–32. Nelson, Kenneth (2003), Fighting Poverty: Comparative Studies on Social Insurance, Means-tested Benefits and Income Redistribution, Stockholm: Swedish Institute for Social Research. Nickell, Stephen (2003), ‘Labour market institutions and unemployment in OECD countries’, Center for Economic Studies – Institute for Economic Research (CESifo) DICE report, 2, 13–26. Nordic Statistical Office (1995), Yearbook of Nordic Statistics, Copenhagen. Nordic Social Statistical Committee (NOSOSCO) (various years), Social Security in the Nordic Countries, Copenhagen: NOSOSCO. O’Connor, James (1973), The Fiscal Crisis of the State, New York: St. Martin’s Press. O’Connor, Julia S. (1988), ‘Convergence or divergence? Change in welfare effort in OECD countries 1960–1980’, European Journal of Political Research, 16, 277–99. O’Connor, Julia S. (2003), ‘Welfare state development in the context of European integration and economic convergence: situating Ireland within the European Union context’, Policy and Politics, 31, 223–40. O’Connor, Julia S. (2005a), ‘Employment-anchored social policy, gender equality and the Open Method of Coordination in the European Union’, European Societies, 7, 27–52. O’Connor, Julia S. (2005b), ‘Convergence in welfare state analysis: conceptual and measurement issues’, paper presented to the Network for European Social Policy Analysis conference, University of Fribourg, 22-24 September. O’Connor, Julia S. and Robert J. Brym (1988), ‘Public welfare expenditure in OECD countries: towards a reconciliation of inconsistent findings’, British Journal of Sociology, 39, 47–68. Obinger, Herbert and Francis G. Castles (2006), ‘Towards more comprehensive measures of social support: adding in the impact of taxes and

References

303

private spending or netting out the impact of politics on redistribution?’, ZeS-Arbeitspapier, 4, University of Bremen. Organisation for Economic Co-operation and Development (OECD) (1985), ‘Social Expenditure 1960–1990. Problems of Growth and Control, Paris: OECD. OECD (1987), Financing and Delivering Health Care, Paris: OECD. OECD (1988), Reforming Public Pensions, Paris: OECD. OECD (1990), Employment Outlook, July, Paris: OECD. OECD (1994), OECD Jobs Study: Facts, Analysis and Strategy, Paris: OECD. OECD (1998), Employment Outlook, June, Paris: OECD. OECD (2001), 1980–1998: 20 Years of Social Expenditures: The OECD Database, online explanation, accessed at www.oecd.org/dataoecd/3/63/ 2084281.pdf. OECD (2002), ‘Women at work: who are they and how are they faring?’, in Employment Outlook, July, Paris: OECD, 61–122. OECD (2003), Employment Outlook, Paris: OECD. OECD (2004), Social Expenditure Database (SOCX), Paris: OECD. OECD (2005a), Social Expenditure Database, 1980–2001, Paris: OECD. OECD (2005b), Benefits and Wages 2004, Paris: OECD, accessed at www.oecd.org/document/0/0,2340,en_2649_34633_34053248_1_1_1_1, 00.html. OECD (2005c), Pensions at a Glance: Public Policies across OECD Countries, Paris: OECD. OECD (various years), Taxing Wages, Paris: OECD. Orloff, Ann Shola (1993), ‘Gender and the social rights of citizenship: the comparative analysis of gender relations and welfare states’, American Sociological Review, 58, 303–28. Orloff, Ann Shola (1997), ‘Comment on Jane Lewis’s “Gender and welfare regimes: further thoughts” ’, Social Politics, 4(2), 188–202. Orszag, Peter, and Joseph Stiglitz (1999), ‘Rethinking pension reform: ten myths about social security’, paper presented to the New Ideas About Social Security conference, The World Bank, Washington, DC, 14-15 September. Ostner, Ilona (2003) ‘ “Individualisation” – the origins of the concept and its impact on German social policies’, Social Policy and Society, 3(1), 47–56. Ostner, Ilona (2004), ‘Gender and welfare revisited’, in Sigrid Leitner, Ilona Ostner and Margit Schratzenstaller (eds), Wohlfahrtsstaat und Geschlechterverhältnis im Umbruch: Was kommt nach dem Ernährermodell?, Wiesbaden: Verlag für Sozialwissenschaften, pp. 44–61. Palier, Bruno (2002a), ‘Beyond retrenchment: four problems in current welfare state research and one suggestion how to overcome them’, in

304

References

Jochen Clasen (ed.), What Future for Social Security? Debates and Reforms in National and Cross-national Perspective, Bristol: Policy Press, pp. 105–20. Palier, Bruno (2002b), Gouverner la sécurité sociale: les réformes du système français de protection sociale depuis 1945, Paris: Presses Universitaires de France. Palme, Joakim (1990), Pension Rights in Welfare Capitalism: The Development of Old-Age Pensions in 18 OECD Countries 1930 to 1985, Stockholm: Swedish Institute for Social Research. Palme, Mårten and Ingemar Svensson (1997), ‘Social security, occupational pensions and retirement in Sweden’, Stockholm University Department of Economics working paper series in economics and finance no. 184. Pampel, Fred C. and Rosemary Gartner (1995), ‘Age structure, sociopolitical institutions and national homicide rates’, European Sociological Review, 11, 243–60. Pampel, Fred C. and John B. Williamson (1985), ‘Age structure, politics and cross-national patterns of public pension expenditures’, American Sociological Review, 50, 782–99. Pampel, Fred C. and John B. Williamson (1989), Age, Class, Politics, and the Welfare State, New York: Cambridge University Press. Pampel, Fred C. and John B. Williamson (2001), ‘Age patterns of suicide and homicide mortality rates in high income countries’, Social Forces, 80(1), 251–82. Pateman, Carole (1989), The Disorder of Women: Democracy, Feminism and Political Theory, Stanford, CA: Stanford University Press. Pierson, Paul (1994), Dismantling the Welfare State? Reagan, Thatcher and the Politics of Retrenchment, Cambridge: Cambridge University Press. Pierson, Paul (1996), ‘The new politics of the welfare state’, World Politics, 48(2), 143–79. Pierson, Paul (1998), ‘Irresistible forces, immovable objects: post-industrial welfare states confront permanent austerity’, Journal of European Public Policy, 5, 539–60. Pierson, Paul (2001), ‘Coping with permanent austerity: welfare state restructuring in affluent democracies’, in Paul Pierson (ed.), The New Politics of the Welfare State, Oxford: Oxford University Press, pp. 410–56. Pierson, Paul (ed.) (2001), The New Politics of the Welfare State, Oxford: Oxford University Press. Pierson, Paul (2003), ‘Big, slow-moving, and . . . invisible: macrosocial processes in the study of comparative politics’, in James Mahoney and Dietrich Rueschemeyer (eds), Comparative Historical Analysis in the Social Sciences, Cambridge: Cambridge University Press, pp. 177–207.

References

305

Pierson, Paul (2004), Politics in Time: History, Institutions, and Social Analysis, Princeton, NJ and Oxford: Princeton University Press. Plümper, Thomas, Vera Troeger and Philip Manow (2005), ‘Panel data analysis in comparative politics: linking method and theory’, European Journal of Political Research, 44(2), 327–54. Polanyi, Karl (1957), The Great Transformation, Boston, MA: Beacon Press. Pontusson, Jonas (2006), Inequality and Prosperity: Social America versus Liberal Europe, Ithaca, NY: Cornell University Press. Powell, Martin (2004), ‘In search of the dependent variable: welfare change in Europe’, paper presented to the COST A15 conference, 21-22 May, Nantes. Pryor, Frederic L. (1968), Public Expenditures in Communist and Capitalist Nations, London: Allen & Unwin. Radaelli, Claudio M. (2000), ‘Whither Europeanization? Concept stretching and substantive change’, paper presented to the Political Studies Association Annual Conference, London, April. Radaelli, Claudio M. (2003), ‘The Europeanization of public policy’, in Kevin Featherstone and Claudio M. Radaelli (eds), The Politics of Europeanization, Oxford: Oxford University Press, pp. 27–56. Radaelli, Claudio M. (2005), ‘Diffusion without convergence: how political context shapes the adoption of regulatory impact assessment’, Journal of European Public Policy, 12(5), 775–96. Radcliff, Benjamin (2001), ‘Politics, markets and life satisfaction: the political economy of human happiness’, American Political Science Review, 95(4), 939–52. Ragin, Charles (1987), The Comparative Method, Berkeley, CA and Los Angeles: University of California Press. Ragin, Charles (2000), Fuzzy Set Social Science, Chicago: Chicago University Press. Reimann, Axel (2005), ‘Die Alterssicherung in Deutschland aus internationaler Sicht – Ergebnisse aktueller Studien’, Presseseminar Deutsche Rentenversicherung 29 and 30 November, accessed at www.deutscherentenversicherung-bund.de/lang_de/nn_20776/de/Inhalt/Presse/ Presseseminare/presseseminare_drv_bund/30_11_05_reimann_pdf, propertypublicationFile.pdf/30_11_05_reimann_pdf). Rentenbericht (2005), ‘Bericht der Bundesregierung über die gesetzliche Rentenversicherung, insbesondere über die Entwicklung der Einnahmen und Ausgaben, der Nachhaltigkeitsrücklage sowie des jeweils erforderlichen Beitragssatzes in den künftigen 15 Kalenderjahren (Rentenversicherungsbericht 2005) und Gutachten des Sozialbeirats zum Rentenversicherungsbericht 2005 und zum Alterssicherungsbericht

306

References

2005’, BT-Drucksache, 16/905, accessed at http://dip.bundestag.de/btd/ 16/009/1600905.pdf. Riedmüller, Barbara (1984), ‘Frauen haben keine Rechte: zur Stellung der Frau im System sozialer Sicherheit’, in Ilona Kickbusch and Barbara Riedmüller (eds), Die armen Frauen. Frauen und Sozialpolitik, Frankfurt am Main: Suhrkamp, pp. 46–72. Room, Graham (1979), The Sociology of Welfare: Social Policy, Stratification and Political Order, London: Martin Robertson. Rose, Richard (1990), ‘Inheritance before choice in public policy’, Journal of Theoretical Politics, 2, 263–91. Rosenfeld, Rachel and Elizabeth Birkelund (1995), ‘Women’s part-time work: a Cross-national comparison’, European Sociological Review, 11, 111–34. Ross, Fiona (1997), ‘Cutting public expenditures in advanced industrial democracies: the importance of avoiding blame’, Governance, 10(2), 175–200. Ross, Fiona (2000a), ‘Interest and choice in the not quite so new politics of welfare’, West European Politics, 23(2), 11–34. Ross, Fiona (2000b), ‘Beyond left and right: the new partisan politics of welfare’, Governance, 13(2), 155–83. Rüb, Friedbert W. and Wolfram Lamping (2005), ‘The politics of reforming a “Holy Grail”: paradigm shifts, experimental law-making, and the politics of German pension reform’, paper presented to the Network for European Social Policy conference, Fribourg, 22-24 September. Rubery, Jill (2003), ‘Gender mainstreaming and gender quality in the EU: the impact of the EU employment strategy’, European Union Center of Excellence, University of Wisconsin, accessed at http://eucenter.wisc. edu/omc/Summer 03Conf/Rubery_mainstreaming.pdf. Rueda, David and Jonas Pontusson (2000), ‘Wage inequality and varieties of capitalism’, World Politics, 52, 350–83. Saari, Juho (2001), ‘Reforming social policy: a study on institutional change in Finland during the 1990s’, Turku: University of Turku, Department of Social Policy. Samek Lodovici, Manuela (2000), ‘The dynamics of labour market reform in European countries’, in Gøsta Esping-Andersen and Marino Regini (eds), Why Deregulate Labour Markets?, Oxford: Oxford University Press, pp. 30–65. Saunders, Peter and Friedrich Klau (1985), ‘The role of the public sector’, OECD Economic Studies, 4 (Spring), special issue. Sayrs, Lois (1989), Pooled Time Series Analysis: London, Sage. Scarborough, Elanor (2000), ‘West European welfare states: the old politics of retrenchment’, European Journal of Political Research, 38(1), 225–59.

References

307

Scharpf, Fritz W. (2002), ‘The European Social Model: coping with the challenges of diversity’, Journal of Common Market Studies, 40(4), 645–70. Scherer, Peter (2001), Age of Withdrawal from the Labour Force in OECD Countries, Paris: OECD. Schmähl, Winfried (2002), ‘The “2001 pension reform” in Germany – a paradigm shift and its effects’, ZeS Working paper, 11/02, Bremen: Centre for Social Policy Research. Schmähl, Winfried (2004), ‘Paradigm shift in German pension policy: measures aiming at a new public-private mix and their effects’, in Martin Rein and Winfried Schmähl (eds), Rethinking the Welfare State: the Political Economy of Pension Reform, Cheltenham, UK and Northampton, MA, USA: Edward Elgar, pp. 153–204. Schmidt, Manfred G. (1982), Wohlfahrtsstaatliche Politik unter bürgerlichen und sozial-demokratischen Regierungen: ein internationaler Vergleich, Frankfurt am Main: Campus. Schmidt, Manfred G. (1996), ‘When parties matter: a review of the possibilities and limits of partisan influence on public policy’, European Journal of Political Research, 30(3), 155–83. Schmidt, Manfred G. (1998), Sozialpolitik in Deutschland: historischer und internationaler Vergleich, Opladen: Leske & Budrich. Schmidt, Manfred G. (2005), ‘Germany: still on the middle way?’, paper presented to the workshop ‘Das Modell Deutschland‚ Revisited: Bestandsaufnahmen, Analysen, Perspektiven’, Humboldt-Universität Berlin, 10-11 November. Schmidt, Vivien A. (2002), ‘Does discourse matter in the politics of welfare state adjustment?’, Comparative Political Studies, 35(2), 168–93. Schulze, Isabelle and Sven Jochem (2006), ‘The development of pension policies in Germany’, in Karen Anderson, Ellen M. Immergut and Isabelle Schulze (eds), The Oxford Handbook of West European Pension Politics: Political Institutions and Policy Outcomes in Comparative Perspective, Oxford: Oxford University Press. Schwartz, Herman (2000), ‘Down the wrong path: path dependence, markets, and the abuse of evolutionary metaphors’, unpublished manuscript. Schwartz, Herman (2003), ‘Globalisation/welfare: what’s the preposition? And, or, versus, with?’, in C. Bochel, N. Ellison and M. Powell (eds), Social Policy Review 15, Bristol: Policy Press. Scruggs, Lyle (2005), ‘Redistributive consequences of welfare state entitlements’, paper presented to the workshop on Democracy, Inequality and Representation: Europe in Comparative Perspective, Maxwell European Union Center, Syracuse University.

308

References

Scruggs, Lyle and James Allan (2006a), ‘Welfare state decommodification in eighteen OECD countries: a replication and revision’, Journal of European Social Policy, 16(1), 55–72. Scruggs, Lyle and James Allan (2006b), ‘The material consequences of welfare states: benefit generosity and absolute poverty in 16 OECD countries’, Comparative Political Studies, 39, 880–904. Seeliger, Robert (1996), ‘Conceptualizing and researching policy convergence’, Policy Studies Journal, 24(2), 287–306. Settergren, Ole (2005), ‘Kritiserat pensionssystem slår ATP för femte året i rad’, Dagens Nyheter, October. Sevenhuijsen, Selma (1998), Citizenship and the Ethics of Care: Feminist Considerations on Justice, London: Routledge. Shalev, Michael (forthcoming), ‘Limits and alternatives to multiple regressions in comparative political economy’, Comparative Social Research. Siegel, Nico A. (2001), ‘Jenseits der Expansion? Sozialpolitik in westlichen Demokratien, 1975–1995’, in Manfred G. Schmidt (ed.), Wohlfahrsstaatliche Politik, Institutionen, Prozesse, Leistungsprofil, Opladen: Leske & Budrich, pp. 54–89. Siegel, Nico A. (2002), Baustelle Sozialpolitik: Konsolidierung und Rückbau im internationalen Vergleich, Frankfurt am Main: Campus. Siegel, Nico A. (2003), ‘Worlds of comparative welfare state analysis: some suggestions for potential bridge-over strategies’; paper presented to the Network for European Social Policy Annual Conference, Copenhagen, 13-15 November. Siegel, Nico A. (2004), ‘EMU and the German welfare capitalism’, in Andrew Martin and George Ross (eds), Euros and Europeans: Monetary Integration and the European Model of Society, Cambridge: Cambridge University Press, pp. 103–26. Siegel, Nico A. (2005) ‘When (only) money matters: the pros and cons of expenditure analysis’, paper presented to the Network for European Social Policy conference, Fribourg, 11–13 September. Siegel, Nico A. and Sven Jochem (2004), ‘Staat und Markt im internationalen Vergleich: empirische Mosaiksteine einer facettenreichen Arbeitsbeziehung’, in Roland Czada and Reinhard Zintl (eds), Politik und Markt, Wiesbaden: Verlag für Sozialwissenschaften, pp. 351–88. Siim, Birte (1987), ‘The Scandinavian welfare states – towards sexual equality or a new kind of male domination?’, Acta Sociologica, 30(3/4), 255–70. Sjöberg, Ola (2000b), ‘Unemployment and unemployment benefits in the OECD 1960–1990: an empirical test of neo-classical economic theory’, Work, Employment, and Society, 14, 51–76. Social and Cultural Planning Office of the Netherlands (SCP) (2004), Public Sector Performance, The Hague: SCP.

References

309

Stata (2005), Longitudinal/Panel Data: Release 9, College Station, PA: Stata Press. Stephens, John D. (1979), The Transition from Capitalism to Socialism, London: Macmillan. Streeck, Wolfgang and Kathleen Thelen (eds) (2005), Beyond Continuity: Institutional Change in Advanced Political Economies, Oxford: Oxford University Press. Svensson, Torsten (1994), ‘Socialdemokratins dominans: en studie av den svenska socialdemokratins partistrategi’, Uppsala: Acta Universitatis Upaliensis. Swank, Duane (2002), Diminished Democracy? Global Capital, Political Institutions, and Policy Change in Developed Welfare States, Cambridge: Cambridge University Press. Taylor-Gooby, Peter (2002), ‘The silver age of the welfare state: perspectives on resilience’, Journal of Social Policy, 31(4), 597–621. Taylor-Gooby, Peter (2003), ‘Open markets versus welfare citizenship: conflicting approaches to policy convergence in Europe’, Social Policy and Administration, 37(6), 539–54. Taylor-Gooby, Peter (ed.) (2004), New Risk, New Welfare: The Transformation of the European Welfare State, Oxford: Oxford University Press. Thelen, Kathleen (1999), ‘Historical institutionalism in comparative politics’, Annual Review of Political Science, 2, 369–404. Thelen, Kathleen (2004), How Institutions Evolve: The Political Economy of Skills in Germany, Britain, the United States and Japan, New York: Cambridge University Press. Timonen, Virpi (2003), Restructuring the Welfare State, Cheltenham, UK and Northampton, MA, USA: Edward Elgar. Titmuss, Richard (1958), Essays on the Welfare State, London: Allen & Unwin. Titmuss, Richard (1963), ‘The social division of welfare: some reflections on the search for equity’, in R. Titmuss (ed.), Essays on the Welfare State, London: Allen & Unwin, pp. 34–55. Titmuss, Richard (1974), Social Policy, London: Allen and Unwin. Torfing, Jakob (1999), ‘Workfare with welfare: recent reforms of the Danish welfare state’, Journal of European Social Policy, 9(1), 6–28. Trampusch, Christine (2005a), ‘Institutional resettlement: the case of early retirement in Germany’, in Wolfgang Streeck and Kathleen Thelen (eds), Beyond Continuity: Institutional Change in Advanced Political Economies, Oxford: Oxford University Press, pp. 203–28. Trampusch, Christine (2005b), ‘Sequenzorientierte Policy-Analyse: Warum die Rentenreform von Walter Riester nicht an Reformblockaden

310

References

scheiterte’, Max Planck Institute for the Study of Societies working paper, 05/3, Cologne. Tronto, Joan C. (1993), Moral Boundaries: a Political Argument for an Ethic of Care, New York: Routledge. Tsebelis, George (2002), Veto Players: How Political Institutions Work, Princeton, NJ: Princeton University Press. Ungerson, Claire (1997), ‘Social politics and the commodification of care’, Social Politics, 4(3), 362–81. US Social Security Administration (SSA) (2004), Social Security Programs throughout the World, accessed at www.ssa.gov/policy/docs/progdesc/ ssptw/2004-2005/europe/luxembourg.html. US Department of Health and Human Services (USDOHHS) (2005), Social Security Programs Throughout the World, Washington, DC: USDOHHS. Uusitalo, Hanno (1984), ‘Comparative research on the determinants of the welfare state: the state of the art’, European Journal of Political Research, 12, 403–42. van Kersbergen, Kees (1995), Social Capitalism, London: Routledge. van Parijs, Philippe (1995), Real Freedom for All: What (if Anything) Can Justify Capitalism?, Oxford: Clarendon Press. Vielle, Pascale, Philippe Pochet and Isabelle Cassiers (eds) (2005), L’Etat social actif: vers un changement de paradigme?, Brussels: P.I.E-Peter Lang. Vis, Barbara (2007), ‘States of welfare or states of workfare? A fuzzy-set ideal type analysis of major welfare state restructuring in sixteen advanced capitalist democracies, 1985–2002’, Policy and Politics, 35(1), 105–22. Visser, Jelle and Anton Hemerijck (1997), A Dutch Miracle: Job Growth, Welfare Reform and Corporatism in the Netherlands, Amsterdam: Amsterdam University Press. Wærness, Kari (1987), ‘On the rationality of caring’, in Anne Showstack Sassoon (ed.), Women and the State: the Shifting Boundaries of Public and Private, London: Hutchinson, pp. 207–34. Wallace, Helen (2000), ‘Europeanisation and globalisation: complementary or contradictory trends’, New Political Economy, 5(3), 369–81. Wallerstein, Immanuel M. (1974), The Modern World System I: Capitalist Agriculture and the Origins of the European World Economy in the Sixteenth Century, San Diego, CA: Academic Press. Weaver, Kent R. (1986), ‘The politics of blame avoidance’, Journal of Public Policy, 6, 371–98. Weber, Max (1904), ‘Objectivity in social science and social policy’, in Max Weber, The Methodology of the Social Sciences, reprinted 1949, Glencoe, IL: Free Press, pp. 49–112.

References

311

Wennemo, Irene (1994), Sharing the Costs of Children: Studies on the Development of Family Support in the OECD Countries, Stockholm: Swedish Institute for Social Research. Western, Bruce (1996), ‘Vague theory and model uncertainty in macrosociology’, Sociological Methodology, 26, 165–92. Whiteford, Peter (1995), ‘The use of replacement rates in international comparisons of benefit systems’, Social Policy Research Centre working paper, no. 54, accessed at www.sprc.unsw.edu.au/dp/dp 054.pdf. Whiteford, Peter and S. Kennedy (1995), Incomes and Living Standards of Older People, London: UK Department of Social Security. Whiteside, Noel and Robert Salais (1998), ‘Comparing welfare states: social protection and industrial politics in Britain and France’, Journal of European Social Policy, 8, 139–54. Wilensky, Harold L. (1975), The Welfare State and Equality, Berkeley, CA: University of California Press. Wilensky, Harold L. (1976), The ‘New Corporatism’, Centralization, and the Welfare State, London: Sage. Wilensky, Harold L. (2002), Rich Democracies: Political Economy, Public Policy, and Performance, Berkeley, CA: University of California Press. Wilensky, Harold L. and Charles N. Lebeaux (1958), Industrial Society and Social Welfare: The Impact of Industrialization on the Supply and Organization of Social Welfare in the United States, New York: Russell Sage. Wilkerson, John (2003), ‘The political economy of health in the United States’, Annual Review of Political Science, 6, 327–43. Wincott, Daniel (2003), ‘The idea of the European Social Model: limits and paradoxes of Europeanization’, in Kevin Featherstone and Claudio M. Radaelli (eds), The Politics of Europeanization, Oxford: Oxford University Press, pp. 279–301. Wolf, Holger (2002), ‘Globalization and the convergence of social expenditure in the European Union’, GW Center for the Study of Globalization occasional paper series. Zadeh, Lotfi (1965) ‘Fuzzy sets’, Information Control, 8, 338–53. Zeitlin, Jonathan, Philippe Pochet and L. Magnusson (2005), The Open Method of Coordination in Action: the European Employment and Social Inclusion Strategies, Brussels: P.I.E.-Peter Lang.

Index accessibility, welfare see welfare accessibility Adcock, Robert 201 Adema, Willem 75, 76, 81, 83, 127, 137, 139 adult worker model 245 see also male breadwinners aged population 107, 110, 115–16, 120–22, 124, 125, 137 see also population ageing agricultural policy 32 Alber, Jens 47, 93, 149, 200, 271 Allan, James 15, 163 analytic inferences 47–9, 69–70 Andersen Jørgen Goul see Goul Andersen, Jørgen Anderson, Karen M. 15, 17, 273, 276 assistance based benefits see meanstested benefits Australia 26–7, 32, 153–4, 156, 158, 159, 160, 255–6 Austria parental leave benefits 254 replacement rates 144, 145, 146, 147, 148 in social rights, structural needs and social expenditure comparative study 112 welfare generosity 156, 158, 159, 160 Average Production Worker (APW) 143–55, 162, 206 background concepts 201 Bambra, Clare 140 Becker, Howard 203, 210, 211 behavioural conditions see conduct, conditions of; obligations Belgium pension expenditure 84, 85 replacement rates 144, 145, 146, 147, 148, 157 social insurance coverage 150, 152

in social rights, structural needs and social expenditure comparative study 118 welfare generosity 156, 158, 159, 160 working women 256 benefits see child benefit expenditure; means-tested benefits; parental leave benefits; pensions; sickness benefits; unemployment benefits Bennett, Collin J. 220, 222 beta convergence 221–3, 226, 230, 231–6 Beyer, Jürgen 262 bias, in statistical analysis 163–4 Bollen, Kenneth A. 198 Bonoli, Giuliano 5, 6–7, 27, 31, 32, 33, 37–8, 277 Bouget, Denis 221, 226–7, 230 Brady, Henry E. 19 breadwinners 27, 34, 36, 245, 247, 248, 250 Briggs, Asa 25 Bulmer, Simon J. 224, 238, 239, 240 Canada 144, 145, 146, 147, 148, 150, 152 capital markets 102 capitalism 218, 219–20, 263 capitalist societies see industrial societies care givers 244, 245, 247, 248–9, 250–51, 252, 253–7, 258, 259 care receivers 250, 251–2, 253, 257–8, 259 Caregiver Parity model 248 case studies 19, 44, 135, 199 Castles, Francis G. 15, 19, 26–7, 31, 32, 46, 51, 52, 58, 72, 73, 74, 168–9, 199, 226, 227, 230, 263 catch-up convergence 221–3, 226, 230, 231–6 313

314

Index

category, conditions of 172–3, 175, 177, 178, 179, 180–82, 183, 184, 187–97 centre-right governments 13, 271–4 child benefit expenditure 107, 109–10, 122, 123, 124 child care 245, 246, 250, 251–2, 253–4 child care services 245, 246, 250, 251, 255, 256–7 children 107, 110, 115, 122, 123, 138 see also care givers; care receivers choice, in de-familization 258, 259 circumstance, conditions of 173–4, 175, 177, 178, 179, 180, 182, 183–4, 187–97 civil servant pensions 152–3, 269 Clasen, Jochen 177, 178, 179, 184, 186 Clayton, Richard 44, 48, 51, 52, 198 Clegg, Daniel 184, 186 coefficient of variation 221, 226–7 cognitive Europeanization 238 Cohesion Countries 222, 231–6 see also Greece; Ireland; Portugal; Spain collective agreements 28, 76, 77–8, 102 collective bargaining 27, 28, 29, 77–8 collectively negotiated agreements (extension) 28, 29, 76, 102 Collier, David 19, 201 commodification 34, 245 see also decommodification comparative empirical analyses measurement problems 198–9 paradigm shifts 166 time and space 134–6 and welfare dependency structure 136–7 see also cross-national comparisons; pooled time cross-sectional methods; social rights, structural needs and social expenditure comparative study Comparative Welfare Entitlements Dataset (CWED) comparative indices of welfare generosity 155–61 described 140–43 replacement rate definition 143 sickness benefit replacement rate 141, 144–5

social insurance eligibility conditions 149–55 social pension replacement rates 141, 146–7 standard pension replacement rates 141–2, 147–9 unemployment replacement rates 141, 143–4 competition 32–3 complexity 61–2, 124 concepts 47, 199, 201 see also fuzzy sets conceptual validity 133, 135 conditionality conditions of category 172–3, 175, 177, 178, 179, 180–82, 183, 184, 187–97 conditions of circumstance 173–4, 175, 177, 178, 179, 180, 182, 183–4, 187–97 conditions of conduct 174–5, 177–8, 179, 180, 181, 182, 183, 184, 187–97 criticism 185–6 levels and levers 175 and retrenchment 184 and risk management 172 and social citizenship 172, 175, 184–5 social rights and responsibilities 171–2 and unemployment provision Denmark 181–2, 183, 184, 185, 195–7 France 180–81, 183–4, 185, 192–4 Germany 178–80, 183–4, 185, 189–91 UK 177–8, 183, 184, 185, 187–8 conduct, conditions of 174–5, 177–8, 179, 180, 181, 182, 183, 184, 187–97 see also obligations conservative welfare states/welfare regimes ideal types 200, 204, 212, 213 modelling change 201–2 pension take-up rates 152 replacement rates 144, 145, 146, 147, 148, 149, 158

Index unemployment benefit coverage 152 welfare generosity 158, 161 continental Europe 27, 28, 29, 31, 36, 37, 149, 158, 231 see also individual countries contractual private schemes 77 convergence defined 217, 218–20 measurement 220–22, 227–31, 232–5, 236, 237–8 see also divergence; policy convergence; societal-level convergence convergence in European Union catch-up convergence and Cohesion Countries 222, 226–31 as decreased variation in EU-15 226–31 and European integration and Europeanization 223–5, 238, 241 framing 238–40 Cornelisse, Peter A. 226, 227 correlation 221, 226 cost containment policies 44, 45, 48, 50, 52–3, 271, 273, 274, 280 coverage, unemployment benefits and sickness benefits 149–52 cross-national comparisons analytic theories 69–70 data quality and measurement error 62–7 GDP and social expenditure 19, 36–7, 45, 52, 137 generalizability 69 indicators 69 measurement problems 46 multidimensional research 54–5, 68 reliability 69 social expenditure data 19, 48, 50–52 social expenditure of Ireland compared to Sweden 137–9 SOCX vs. ESSPROS calculation methods 74 and theories of welfare reform 134–5 validity 69 see also comparative empirical analyses Cutright, Philip 106

315

Daguerre, Anne 180, 230 datasets see Comparative Welfare Entitlements Dataset (CWED); ESSPROS (European System of Integrated Social Protection Statistics); Flora datasets; historical data; ILO datasets; OECD datasets; social expenditure data; SOCX (Social Expenditure Database); statistical datasets De Deken, Johan 19, 62, 122, 227 de-familization care giver’s perspective 244, 245, 247, 248–9, 250–51, 252, 253–7, 258, 259 care receiver’s perspective 250, 251–2, 253, 257–8, 259 and child care services 245, 246, 250, 251, 255, 256–7 conceptual reductionism 252–3 defined 244–5, 249–50 economic perspective 251–2, 253, 254–7, 258 feminist perspective 246–9 and individualized payments for care 254 and labour market participation 244, 245, 246–7, 248, 252, 255–7 social perspective 249–50, 252, 253–5, 256–7, 258 decommodification 24, 34, 36, 170–71 see also commodification decommodification indices 140, 143, 155–61, 170–71, 173 decreased variation convergence 221, 226–31 deindustrialization 34–5 Delhey, Jan 232–3 demographic factors and comparative social expenditure 136–7 and social expenditure modelling using varying definitions of dependent variable 92, 93–4, 95, 97–8 and social expenditure of Ireland compared to Sweden 138 and time sensitivity in pooled time cross-sectional methods 58

316

Index

see also aged population; breadwinners; children; dependent population; life expectancy; long-term sickness; male breadwinners; single parent households; single person replacement rates; spouses; two-earner households; unemployment; women Denmark parental leave benefits 255 pension expenditure 84–6, 126–7, 277, 278 pension path dependency 268–71, 277, 278–9 replacement rates 144, 145, 146, 148 retrenchment 184 social insurance coverage 149, 150, 152 in social rights, structural needs and social expenditure comparative study 115 SOCS vs. ESSPROS social expenditure data 75 tax system 127 unemployment benefit conditionality 181–2, 183, 184, 185, 195–7 unemployment benefit model configured by fuzzy sets 208–9, 210–11, 212, 213 dependent population 136–7, 138 see also aged population; children; unemployment dependent variable problem and analytic inferences 47–9 and concept specification 47 defining 4, 14, 15 operational definitions of retrenchment 16, 18–22 and ‘politics matter’ argument 14–15 qualitative vs. macro-quantitative research 44–6 theoretical definitions of retrenchment 16–18 dependent variables in industrial vs. postindustrial age 39 (see also industrial societies; postindustrial age)

in quantitative vs. qualitative research 133–4 and research questions 39 total social expenditure 86–91 see also path dependency; social expenditure data; social expenditure modelling using varying definitions of dependent variable; social rights descriptive inferences 47, 69 disability pensions 274–5 dismissal, protection against 27, 29 divergence 219, 221, 222, 226, 227 see also convergence diversity see ideal types duration of benefits 154 Earned Income Tax Credit (US) 137 economic conditions 45–6, 54, 56–7 economic crises 52, 82 economic dependence and independence of women 244–5, 246–9 see also de-familization economic growth 137, 138, 218 see also GDP economic models, and measurement of retrenchment 21 economic perspective, of defamilization 251–2, 253, 254–7, 258 economic recessions 45, 56–7, 61, 227 efficiency, in health care expenditure 20 Einerhand, Marcel 76 eligibility conditions conditions of circumstance 173–4, 175, 177, 178, 179, 180, 182 duration of benefits 154 funding ratio 155 pensions 152–5 and pooled time cross-sectional methods 56 qualifying conditions 154–5 and retrenchment 16 sickness benefits 149–52, 154 in social rights, structural needs and social expenditure comparative study 110 and unemployment 20

Index unemployment benefits 149–52, 154, 209 waiting days 154 see also circumstance, conditions of employment see breadwinners; defamilization; labour laws; labour markets; low-skilled workers; nonemployment; occupational welfare; part-time employment; trade unions; two-earner households; unemployment; wages employment protection laws 27, 29 entitlement in conditionality 173–4, 175 conditions of circumstance 173–4, 175, 177, 178, 179, 180, 182 and cost containment policies 53 and pooled time cross-sectional methods 56 and retrenchment 16 in social rights, structural needs and social expenditure comparative study 110 and unemployment 20 see also circumstance, conditions of Esping-Andersen, Gøsta 5, 18, 22, 30, 36, 49, 50, 54, 137, 139, 143, 155–61, 170, 173, 198, 200, 209, 211, 244–5, 255, 256, 258–9, 263, 264 ESSPROS (European System of Integrated Social Protection Statistics) contractual private schemes 77 gross social expenditure 81 pension expenditure classification 78, 84–6 sickness benefit classification 76–7 vs. SOCX 74–5, 105 vs. SOCX in pension expenditure modelling 96–9 vs. SOCX in total social expenditure modelling 94–6 total social expenditure 88–9, 90 EU-15 222, 226–31 EU-25 222 Europe see Cohesion Countries; Continental Europe; European labour laws; European Union;

317

Europeanization; Latin Europe; Nordic Europe; Northern Europe; Southern Europe; Western Europe; individual countries European Commission 222, 226, 232, 239, 245 European Council 226, 239, 245 European integration 223–5, 227, 238, 241, 242 European labour laws 27–8 European Union convergence (see convergence in European Union) de-familization 244, 245 European integration and Europeanization 223–5, 227, 238, 241, 242 gender equality 224, 236–7, 238–9 harmonization 223, 238–9 National Action Plans 237, 239–40 Open Method of Coordination (OMC) 224–5, 237, 238, 239–40 Structural Funds 231–2, 233, 238 Structural Indicators 222, 237 see also Cohesion Countries; EU-15; EU-25 Europeanization 223–5, 238, 241, 242 Eurostat 74–5 see also ESSPROS (European System of Integrated Social Protection Statistics) extension 28, 29, 76, 102 families see de-familization; familization ‘families of nations’ 27, 31, 263 familization 247 family allowance expenditure see child benefit expenditure family benefit replacement rates 141, 144, 145, 149, 157–8 family pension replacement rate 141, 142, 147–8 family unemployment replacement rate 141, 144 feminist perspective, de-familization 246–9 Ferrara, Maurizio 169, 231 Finland child care services 257

318

Index

convergence 227, 228, 229 economic recession and social expenditure 57 replacement rates 144, 145, 146, 147, 148 social insurance coverage 149, 150, 152 in social rights, structural needs and social expenditure comparative study 112, 115, 118 unemployment 118, 124 unemployment benefit model configured by fuzzy sets 212, 213 welfare generosity 156, 158, 159, 160 first order differences 64–5 fiscal welfare 126, 127 Flora, Peter 111, 115, 149, 200 Flora datasets 111, 112 framing 224–5, 238–40 France replacement rates 144, 145, 146, 147, 148 social insurance based pension systems 30 in social rights, structural needs and social expenditure comparative study 115 unemployment benefit conditionality 180–81, 183–4, 185, 192–4 welfare generosity 156, 158, 159, 160 Fraser, Nancy 248 ‘functional equivalents’ 26–7 funding, pension systems 82 funding ratio, social insurance 155 fuzzy sets 204–5 fuzzy sets applied to unemployment benefits calibration of sets 206–8 configuration into ideal types 209–11 empirical indicators 205–6 ideal type analysis 211–13 scoring cases 208–9 GDP and cross-national comparison of social expenditure 19, 36–7, 45, 52, 137 and data quality and measurement error 63, 65–6, 108

economic recession and social expenditure growth 57 social expenditure and convergence 219, 227, 232–5 in social expenditure modelling using varying definitions of dependent variable 92, 93, 94–5, 97, 98 social expenditure of Ireland compared to Sweden 138 social protection expenditure and convergence 226, 227, 228, 229–30, 235–6, 237–8 in SOCX vs. ESSPROS social expenditure calculation methods 75, 77, 78 gender see male breadwinners; women gender equality 224, 236–7, 238–9 see also de-familization generosity, see welfare generosity Germany convergence 228, 229 parental leave benefits 254 pension expenditure 84, 85, 86, 277–8 pension fund classification in ESSPROS 78 pension fund classification in SOCX 78, 80 pension path dependency 267–8, 269, 271–3, 277, 278–9 replacement rates 144, 145, 146, 147, 148 sickness benefit classification in ESSPROS 77, 78 sickness benefit classification in SOCX 76, 77, 78, 80 social insurance based pension systems 30 in social rights, structural needs and social expenditure comparative study 112, 115 SOCS vs. ESSPROS social expenditure data 75 time sensitivity and social expenditure data measurement error 66–7 unemployment benefit conditionality 178–80, 183–4, 185, 189–91

Index unemployment benefit model configured by fuzzy sets 212, 213 variable impact chains 61 welfare generosity 156, 158, 159, 160–61 working mothers 255–6 Gilbert, Neil 168, 198 Gingerich, Daniel W. 263 Giullari, Susanna 255, 259 Glendinning, Caroline 249 globalization 218–19, 225, 241, 242 Goudswaard, Kees P. 226, 227 Goul Andersen, Jørgen 264, 270, 271, 277, 280 Greece convergence 226, 227, 228, 229, 230, 232, 233, 234, 235, 236, 237–8 part-time employment 256 Green-Pedersen, Christoffer 14, 21, 270 Greve, Bent 226, 227 gross social expenditure 81 Guillén, Ana M. 237, 238 Hacker, Jacob 15, 17–18, 20 Hall, Peter A. 263 harmonization 223, 238–9 health care expenditure 20, 52 Held, David A. 225 Hering, Martin 265, 272, 277 Hinrichs, Karl 10, 167, 265, 267, 268, 271, 272 historical data 135–6, 137 Huber, Evelyn 14, 19, 47, 51, 53, 61, 126 Hvinden, Bjørn 230, 231 hypothesis testing 46–7, 50 ideal types 199, 200–204, 209–13 see also conservative welfare states; labour welfare states; liberal welfare states; social-democratic welfare states ILO datasets 111, 112, 115, 120, 122, 126 in-depth analysis see case studies income distribution 30 income inequality 30, 237 income maintenance 126–7

319

income transfer programmes 28, 29, 30–31 independent variable, welfare state as 5 indicators 51–2, 69 indices decommodification 140, 143, 155–61, 170–71, 173 institutional traits 22 social rights 114–15 individualized payments for care 254 industrial relations 80 see also collective agreements; collective bargaining; extension; labour laws; trade unions industrial societies 25–33, 218 inequality 17–18, 20 see also gender equality; income inequality inferences 47–9, 61–2, 66, 69, 163–4 institutional aspects, welfare reform 18, 22 institutional convergence, defined 217 institutional frameworks, in social expenditure modelling 92, 93–4, 95, 97, 98 institutional welfare states 200 Ireland convergence 228, 229, 230, 232–5, 236–8 data quality in SOCX dataset 63 gender equality 236–7 part-time employment 256 replacement rates 144, 145, 146, 148 social expenditure compared to Sweden 137–9 in social rights, structural needs and social expenditure comparative study 118 Italy convergence 226, 228, 229, 230 parental leave benefits 254 part-time employment 256 replacement rates 144, 145, 146, 147, 148 in social rights, structural needs and social expenditure comparative study 118 welfare generosity 156, 158, 159, 160

320

Index

Jaeger, Mads M. 231 Japan 144, 145, 146, 147, 148, 156, 158, 159, 160 Jochem, Sven 19, 50, 126, 271 Kangas, Olli 10, 118, 167, 230, 267 Kautto, Mikko 213, 228 Kerr, Clark 217, 218 King, Gary 19 Kitschelt, Herbert 15, 21, 47, 94, 217, 219 Kittel, Bernhard 15, 48, 55, 61, 62, 72–3, 92, 93, 95, 122, 227 Klau, Friedrich 116 Knijn, Trudie 34, 248–9, 250 Knill, Christoph 217, 220, 222, 224 Korpi, Walter 47, 48, 109, 124, 135, 139, 149, 163 Kremer, Monique 248–9, 250 Kvist, Jon 182, 203, 213, 228, 231 labour law/labour laws 27–8, 29, 31, 36 labour markets adult worker model 245 and de-familization 244, 245, 246–7, 248, 252, 255–7 and deindustrialization 34–5 regulation 26–31, 36–7 labour movements 270 labour welfare states 201, 204, 211, 212, 213 Ladaique, Maxime 81, 137, 139 Lamping, Wolfram 272 Latin Europe 27, 28, 29, 31, 36 see also individual countries Lazarsfeld, Paul 203, 210 Lebeaux, Charles N. 53, 200 left parties/left-wing parties 48, 93, 94, 95, 97, 98, 124, 125, 126 see also labour welfare states; socialdemocratic governments; socialdemocratic welfare states legislation 27–8, 29, 31, 36, 115, 126, 187–97, 272 Lehmkuhl, Dirk 224 Leitner, Sigrid 170, 255 Lessenich, Stephan 170 Lewis, Jane 34, 245, 247, 250, 255, 259 liberal welfare states/welfare regimes ideal types 200, 204, 211, 212, 213

replacement rates 144–5, 146, 147, 148–9 sickness benefit coverage 152 welfare generosity 127, 161 Liebert, Ulrike 224, 238–9 life expectancy 17, 111, 115–16, 117, 118, 119, 120, 121, 125 Lindbom, Anders 15, 17, 21 long-term sickness 120 long-term unemployment 34, 116, 118, 124 low-skilled workers 35 Luxembourg 227, 228, 229, 230, 237–8, 254, 256 Mach, André 32 macro-level convergence 218–20 macro-level research 44–5, 135, 198–9, 263–4 male breadwinners 27, 34, 36, 245, 247 mandating 79–80 mandatory occupational pensions 127, 265, 266, 268, 270, 272–3, 276 mandatory pensions 79, 80, 127, 265, 266, 268, 270, 272–3, 276 mandatory private expenditure 74, 75, 76, 77, 78, 79, 80, 86–7, 88, 89, 90, 91, 127 mandatory private pensions 79, 80, 127, 265, 266, 268, 272–3 market forces 25–7 Marshall, Thomas H. 26, 108–9, 200 Matsaganis, Manos 237, 238 McLaughlin, Eithne 249 means-tested benefits 16, 17, 173–4, 177, 270, 274 measurement case studies 19, 44, 135, 199 and concepts 199 convergence 220–22, 227–31, 232–5, 236, 237–8 errors 65–6, 108, 122, 124, 126, 163–4 macro-level analysis 44–5, 135, 198–9, 263, 264 outcome 20–21, 55 output 21–2 pension expenditure 82–6 problems in comparative empirical analyses 198–9

Index of retrenchment 21 welfare generosity 162–4 measurement validity 199 minimum wages 27–8, 29 Mishra, Ramesh 200 mothers, de-familization see defamilization multidimensional research 54–5, 68 National Action Plans 237, 239–40 negative integration 224 neo-institutionalism 262 neo-liberalism 219, 225 net replacement rates 206 net social expenditure 81 Netherlands convergence 228, 229, 230 occupational pensions 126–7 pension expenditure 82–3, 84, 85, 277, 278 pension fund classification in ESSPROS 78 pension fund classification in SOCX 75–6, 78, 79, 80 pension path dependency 268, 269, 273–5, 278–80 replacement rates 144, 145, 146, 147, 148 sickness benefit classification in ESSPROS 77, 78 sickness benefit classification in SOCX 76–7, 78, 79, 80 in social rights, structural needs and social expenditure comparative study 115, 118 SOCS vs. ESSPROS social expenditure data 75 unemployment benefit model configured by fuzzy sets 212, 213 welfare generosity 156, 158, 159, 160 working women 255–6 new conservative welfare state model 204, 212, 213 new labour welfare state model 204, 212 new liberal welfare state model 204, 212 new politics of welfare state 4, 43–4, 47, 48, 68

321

new social-democratic welfare state model 204, 210, 211, 212, 213 new social risk policies 35–8 new social risks 4, 5–6, 34–8 New Zealand protectionism 32–3 replacement rates 144, 145, 146, 147, 148 retrenchment 157 welfare generosity 156, 157, 158, 159, 160 non-employment 138 Nordic Europe convergence 228, 231 income transfer programmes 29, 31 labour market regulation 27, 28, 29, 31 pension systems 36, 37 social expenditure on new social risks 37, 38 in social rights, structural needs and social expenditure comparative study 112, 118 welfare generosity 156, 158, 159, 160 see also Scandinavian welfare model; individual countries Northern Europe 34, 36, 37, 52–3 see also individual countries Norway 112, 115, 144, 145, 146, 147, 148, 212, 213 Obinger, Herbert 48, 55, 72–3, 92, 93, 95 obligations 202–3, 204, 206, 207, 208, 209, 211, 213 see also conduct, conditions of occupational pensions 79, 80, 82, 126–7, 152–3, 265, 266, 268, 269–71, 272–3, 275, 276, 279 occupational welfare 126–7 O’Connor, Julia S. 218, 221, 225, 230, 235, 236, 237, 238, 240 OECD countries 36, 37, 38, 44, 51, 52–3 see also social rights, structural needs and social expenditure comparative study OECD datasets 111, 112, 114–15, 118, 120, 126, 140, 226–7

322

Index

see also SOCX (Social Expenditure Database) old-age pensions old-age pension expenditure 75–6, 78, 79, 80, 84–6, 109–10, 120–22, 124, 137 old conservative welfare state model 204, 212 old labour welfare state model 204, 212 old liberal welfare state model 204, 212 old politics 47, 48, 68 old social-democratic welfare state model 204, 210–11, 212, 213 old social risks/old social risk policies 25–33, 36–7, 38 OLS regression analysis 57, 61, 94–9, 113–14, 117, 119, 121, 123 Orloff, Ann S. 246, 247–8, 251 outcome measures 20–21, 55 see also replacement rates; social expenditure data outcome perspective 16–18, 201, 222 output measures 21–2 output perspective 16, 17, 201, 222, 264, 267 Palier, Bruno 169–70, 180, 239 Palme, Joakim 47, 48, 120, 163 Pampel, Fred C. 107 paradigm shifts 166, 168, 184–5 parental leave benefits 254 parents, defamilization see defamilization part-time employment 255, 256 partisanship 20, 94–5, 99, 125–6, 163 path dependency 219–20, 230, 262–7 path dependency of old-age pensions Denmark 268–71, 277, 278–9 Germany 267–8, 269, 271–3, 277, 278–9 Netherlands 268, 269, 273–5, 278–80 pension pillars and tiers 265, 266, 267–76, 279–80 and pension policies 265, 267, 278–80 and programmatic changes 265–6, 279 programmatic rules 266–7, 278, 279

Sweden 268, 269, 275–6, 277, 278–9 PAYG (Pay-As-You-Go) pension systems 82, 102, 285–6 pension expenditure 36–7, 52, 78, 82–6, 96–9, 107 pension replacement rates 141–2, 146–9, 154 pension rights 107, 122, 127 pensions civil servants 152–3, 269 design 82 eligibility conditions 152–5 generosity 36, 37, 44, 156 and new social risks 36 PAYG (Pay-As-You-Go) pension systems 82, 102, 285–6 pillars and tiers 79, 80, 82, 126–7, 265, 266, 267–76, 279–80 reform 15, 17, 20, 21 social insurance based, and income inequality 30 take-up rates 152–4 and tax relief 83–4, 272, 273 welfare generosity 36, 37, 44, 127 see also civil servant pensions; disability pensions; mandatory pensions; occupational pensions; old-age pension expenditure; path dependency of old-age pensions; private pensions; public pensions; quasi-mandatory pensions; social pensions; survivors’ pensions; voluntary pensions Pierson, Paul 4, 10, 15, 16, 18, 43, 47, 69, 92, 93, 200, 262, 264 Polanyi, Karl 26, 33 policy convergence 217, 220, 222–3, 225–31 policy diffusion 220 policy transfer 220 political institutions, and path dependency 262, 264 politics and output measures of retrenchment 21 and path dependency of old-age pensions 265 and pension rights 107 and responsibilities 171

Index and social expenditure 72–3, 92, 93–4, 95, 96, 98, 99, 124–6 and social rights 51, 107, 124–6 and time sensitivity in pooled time cross-sectional methods 58–9 and variable impact chains 59–61 and welfare efforts 54, 59–61 and welfare generosity 54, 127, 139 see also partisanship; social policies ‘politics matters’ argument 13–15, 55–6, 107 Pontusson, Jonas 44, 48, 51, 52 pooled time cross-sectional methods 14–15, 48, 55–6, 57–9, 61–2 see also social rights, structural needs and social expenditure comparative study population ageing 37–8, 62 see also aged population Portugal 226, 228, 229, 230, 232, 233, 234, 235, 236, 237–8 positive integration 224 postindustrial age 33–8 poverty 225, 237, 270 private health care 20 private pensions 79, 127, 265, 266, 268, 269, 270, 272–3, 276, 279 private social expenditure 74, 75–6, 77, 78, 79–80 privatization 16 programmatic changes 265–6, 278 programmatic rules 264, 266–7, 277, 278, 279 protectionism 32–3 public pensions 79, 80, 82, 127, 265, 266, 267–9, 270, 271–2, 273–4, 275–6, 279–80 public social expenditure 74, 75–6, 78–9, 80, 86–8, 89, 90, 91, 127 Purchasing Power Standards (PPS) 228–30, 232–3, 234, 235, 236, 238 qualifying conditions 154–5 qualifying period 154 qualitative change in benefits 201–2 qualitative research 16, 19, 44, 47, 54, 133–4, 135, 264, 265 quantitative change in benefits 201–2 quantitative research 44–5, 54–5, 133, 134, 135, 198–9, 263–4

323

see also measurement; OLS analysis; regression analysis; statistical datasets; statistical methods quasi-mandatory pensions 269, 276, 279 Radaelli, Claudio M. 220, 223, 224, 238, 239, 240 Ragin, Charles 199, 205, 209, 210, 211 redistributive social insurance 30 reductionism 252–3 refracted divergence 219 regression analysis/OLS regression 46, 49, 61–2, 163–4, 221, 226 regulation 26–31, 27, 36, 54 reliability 16, 19, 69, 72, 135, 199 replacement rates advantages and disadvantages as retrenchment indicator 20–21 definitions 143 in social rights, structural needs and social expenditure comparative study 116, 117, 118, 119, 124 as social rights indicator 170 and tax system 127 and welfare state commitments 140 see also family benefit replacement rates; net replacement rates; pension replacement rates; sickness benefit replacement rates; single person replacement rates; unemployment benefit replacement rates replication 46–7 research questions 39 residency requirements 149 residual welfare states 200 responsibilities 171, 172 restructuring 17, 21, 228 retirement age 36, 37, 273 retrenchment and conditionality 184 Denmark 184 in history of welfare state 200 Ireland 137–9 New Zealand 157 operational definitions 16, 18–22 output perspective 17 and ‘politic matters’ argument 13–15 qualitative research 44, 47

324

Index

quantitative research 47 as research focus 167–8 vs. restructuring 17 and social expenditure 168 Sweden 137–9, 157 Switzerland 155–6 theoretical definitions 16–18 UK 184 right-wing parties 48, 270 see also centre-right governments; conservative welfare states risk management 172, 173–4, 175, 184 risks, social see social risks Rüb, Friedbert W. 272 Rubery, Jill 239 Saari, Juho 57 Saunders, Peter 116 Scandinavia see Nordic Europe Scandinavian welfare model 115 Schmähl, Winfried 272 Schmidt, Manfred G. 93, 271 Scruggs, Lyle 15, 163 selectivity 18 sensitivity analysis 94 services sector 34–6 sickness absence 118, 119, 120, 124 sickness benefit replacement rate 141, 144–5 sickness benefits 76-7, 78, 79, 80, 127, 149–52, 154, 156, 157 sickness insurance expenditure 109–10, 118–20, 124 Siegel, Nico A. 15, 19–20, 44, 48, 50, 51, 52, 55, 94, 161, 170, 263 sigma convergence 221, 226–31 single parent households 162 single person replacement rates 141, 144, 147, 149, 158 small-n research see case studies social assistance benefits, see meanstested benefits social change 205–6, 218 social citizenship 170–71, 172, 184–5, 201–4, 205–6, 248–9 see also fuzzy sets applied to unemployment benefits Social Citizenship Indicator Program (SCIP) 108–10, 111–12, 140, 202

social-democratic governments 13, 14, 48, 127, 269, 270, 271, 276 social-democratic welfare states ideal types 200, 204, 210–11, 212, 213 modelling change 201, 202 replacement rates 144, 145, 146, 147, 148 welfare generosity 161 social expenditure commitments, statistical data sets 139–42 and convergence 219, 226–7 and cost containment policies 44, 45, 48, 50, 52–3, 271, 273, 274, 280 definitions 74, 227 and economic conditions 45–6 and economic crises 52 and economic recessions 45, 56–7, 61 and GDP and convergence 219, 227, 232–5 and GDP in cross-national comparisons 19, 36–7, 45, 52, 137 as indicator of welfare effort 168–9 and industrial relations centralization 80 Ireland compared to Sweden 137–9 and net costs of social policies 53 and pension expenditure 36–7, 84, 97–9 and political commitment 139 political determinant ambiguities in using SOCX (Social Expenditure Database) 72–3 and ‘politics matters’ vs. social rights 107–8 in social rights, structural needs and social expenditure comparative study 111–12, 115 and social risk 36-8, 168, 169 structure 168–9 and tax system 81, 127, 168 trends 44–6 and welfare generosity 51–2 see also child benefit expenditure; gross social expenditure; health care expenditure; mandatory private expenditure; net social

Index expenditure; old-age pension expenditure; pension expenditure; public social expenditure; sickness insurance expenditure; social expenditure modelling using varying definitions of dependent variable; social protection expenditure; social rights, structural needs and social expenditure comparative study; total social expenditure; unemployment benefit expenditure; voluntary private expenditure social expenditure data and complexity 54–5 critiques 49, 50 cross-national comparisons 19, 48, 50–52, 136–7 and GDP 19, 36–7, 45, 52, 137 implicit constant causality assumption 61–2 myths and practices 50–52 as outcome measure 55 problems 19–20, 22, 136–7 quality and measurement error 62–7 and tax state 51 ‘time-lag’ problem 20, 169 social expenditure modelling using varying definitions of dependent variable conclusions 99–102 introduction 91–4 pension expenditure SOCX vs. ESSPROS 96–9 total social expenditure SOCX vs. ESSPROS 94–6 social insurance 30, 149–55 see also child benefit expenditure; means-tested benefits; parental leave benefits; pensions; sickness benefits; social expenditure; social protection; unemployment benefits social need and conditions of circumstance 173–4, 178 confounded with social rights 107 and economic recessions 61

325

and implicit constant causality assumption in social expenditure analysis 62 indicators 51–2 and pooled time cross-sectional methods 56–7 residual vs. institutional welfare state 200 and welfare efforts 53–4 see also social rights, structural needs and social expenditure comparative study; structural needs social pension replacement rates 141, 146–7 social pensions 146 social perspective, of de-familization 249–50, 252, 253–5, 256–7, 258 social policies labour markets regulation 26–31, 36 and male breadwinners 34, 36 net costs 53 for new social risks 35–8 and path dependency of old-age pensions 265, 267 protectionism 32–3 and regulation 54 and replacement rates 21 and structural change 168 social protection 169–70 see also total social protection social protection expenditure 226, 227, 228, 229–30, 235–6 see also social expenditure social rights alternative types 162 concept 201 and conditionality 171–2 confounded with social need 107 definitions 108–9 and household composition 162 and implicit constant causality assumption in social expenditure analysis 61–2 and income distribution 162 indices 114–15 legislated 115 modelling 201–2 see also welfare accessibility; welfare generosity

326

Index

and politics 51, 107 and social citizenship 202 vs. social expenditure 107–8 and social risks 171 and tax system 127 and variable impact chains 59–61 as welfare effort indicator 170–71 and welfare efforts 54 see also social rights, structural needs and social expenditure comparative study social rights, structural needs and social expenditure comparative study data 108–12 discussion 122–7 methods 112–14 results 114–22 pension expenditure 120–22 sickness insurance expenditure 118–20 total social expenditure 114–16, 117 unemployment insurance expenditure 116–18 social risks 4, 168, 169–70, 171, 172–4 see also new social risk policies; new social risks; old social risk policies; old social risks social welfare, defined 126 societal-level convergence 218–19, 238–40, 241 SOCX (Social Expenditure Database) convergence in European Union 227 data quality and measurement error 62–7, 81 vs. ESSPROS 74–5, 105 vs. ESSPROS in pension expenditure modelling 96–9 vs. ESSPROS in total expenditure modelling 94–6 gross social expenditure 81 mandatory private expenditure 74, 75, 76, 77, 78, 79, 80, 86–7, 88, 89, 90, 91 pension expenditure classification 75–6, 78, 79, 80, 84, 85 political determinants of social expenditure ambiguities 72–3

public expenditure 74, 75–6, 78–9, 80, 86–8, 89, 90, 91 reliability 72 sickness benefit classification 76–7, 78, 79, 80 and social expenditure in Finland 57 total social expenditure 86–9, 90, 91 total social protection 75, 77, 78, 88, 89, 902 voluntary private expenditure 74, 75–6, 77, 78, 79, 80, 86, 87–9, 90, 91 Southern Europe 36, 37, 38, 226, 231 see also individual countries Spain 226, 228, 229, 230, 232, 233–5, 236, 237, 238, 256 spatial dimension, in comparative empirical analyses 135–6 see also cross-national comparisons; pooled time cross-sectional methods spending caps, in social expenditure 53 spouses 153, 158 standard deviation 221, 226–7 standard of living 208, 232, 272 standardized welfare efforts 51 statistical datasets 62–7, 86–91, 108, 122, 124, 126, 139–42 see also ESSPROS (European System of Integrated Social Protection Statistics); Flora datasets; ILO datasets; OECD datasets; Social Citizenship Indicator Program (SCIP); SOCX (Social Expenditure Database) statistical methods 14–15, 19, 21, 163–4 see also coefficient of variation; correlation; OLS analysis; regression analysis; standard deviation Stephens, John D. 19, 47, 51, 53, 58, 61, 126 structural change 168 Structural Funds 231–2, 233, 238 Structural Indicators, European Union 222, 237

Index structural needs 110–11 see also social need; social rights, structural needs and social expenditure comparative study survivors’ pensions 84–6 Swank, Duane 19 Sweden convergence 227, 228, 229 demands of women 34 parental leave benefits 255 pension expenditure 277, 278 pension path dependency 268, 269, 275–6, 277, 278–9 pension reform 15, 17, 21 replacement rates 127, 144, 145, 146, 148 retrenchment 21, 157 social expenditure compared to Ireland 137–9 social insurance coverage 149, 151, 152 in social rights, structural needs and social expenditure comparative study 112, 115, 118 tax system 127 time sensitivity and social expenditure data measurement error 66 unemployment 118, 124 unemployment benefit model configured by fuzzy sets 208–9, 210–11, 212, 213 welfare generosity 156, 157, 158, 159, 160 working women 256 Switzerland low-skilled workers 35 pension system 30 protectionism 32–3 replacement rates 144, 145, 146, 147, 148 retrenchment 155–6 social expenditure 37 social insurance coverage 150, 152 in social rights, structural needs and social expenditure comparative study 112 welfare generosity 156, 158, 159, 160 working mothers 255–6

327

take-up rates, pensions 152–4 tax credits 35, 137 tax rates 138–9 tax relief 83–4, 272, 273 taxation 21, 51, 81, 127, 137, 168 Taylor-Gooby, Peter 4, 44, 200, 230 theoretical validity 133, 134 thin convergence 230 time factors 83–4, 135–6, 137, 221–3, 230–31 see also cross-national comparisons; duration of benefits; pooled time cross-sectional methods; waiting days time-lags 20, 21, 55–6, 92–4, 137, 169 time sensitivity 57–9, 64–7 time series analysis 62–7 see also pooled time cross-sectional methods; social rights, structural needs and social expenditure comparative study Titmuss, Richard 126, 200 total social expenditure 86–91, 94–6, 101, 114–16 total social protection 75, 77, 78, 88, 89, 90, 91 trade unions 28, 35, 102, 149, 152, 270–71, 272, 275–6 transfer payments 137 Tsebelis, George 59–61 two-earner households 162 UK convergence 228, 229 cost containment policies 53 labour market regulation 27, 28, 29 low-skilled workers 35 pension system 30, 37 retrenchment 184 unemployment benefit 137, 177–8, 183, 184, 185, 187–8, 212, 213 welfare state and market forces 26 Working Families Tax Credit 137 working women 255–6 unemployment and deindustrialization 34, 35 effect of changes on unemployment benefits 20 and employment protection laws 27, 29

328

Index

and social expenditure 56–7, 92, 94, 95, 97, 98, 124 in social rights, structural needs and social expenditure comparative study 110, 115, 116–18, 124–5 and time sensitivity in pooled time cross-sectional methods 57–8 and unemployment benefit expenditure 137 see also labour markets; nonemployment unemployment benefit expenditure 19–20, 52, 109–10, 116–18, 124, 137 unemployment benefit generosity 156, 157, 206, 207, 208–9, 210, 211 unemployment benefit replacement rates 137, 141, 143–4, 157, 206, 227 unemployment benefits conditionality 175–82 eligibility conditions 149–52, 154 fuzzy sets see fuzzy sets applied to unemployment benefits generosity 156, 157, 206, 207, 208–9, 210, 211 and labour market regulation 30 and replacement rates 137, 206, 227 and time sensitivity in pooled time cross-sectional methods 57–8 welfare state commitments 140 unemployment rate 110, 115, 116, 117, 118, 124–5 Universal Breadwinner model 248 universalism and conditionality 174 and institutional welfare reform 18 pensions 269, 270, 273, 275–6 and residual vs. institutional welfare state 200 and social insurance coverage 149, 152, 153–4 US cost containment policies 53 Earned Income Tax Credit 137 health care expenditure 20 labour market regulation 27, 28 low-skilled workers 35 outcome and output perspectives of welfare state 15, 18

pension take-up rates 152–3 replacement rates 144, 145, 146, 147, 148–9 social insurance coverage 150, 152 welfare generosity 156, 158, 159, 160 validity 16, 19, 69, 133–4, 199 variable impact chains, and political decision making 59–61 varieties of capitalism, and path dependency 263 veto player theory 59–61 voluntary occupational pensions 265, 266, 269, 273 voluntary pensions 79, 80, 127, 265, 266, 268, 269–70, 273, 279 voluntary private pensions 79, 80, 265, 266, 268, 269–70, 273, 279 voluntary private social expenditure 74, 75–6, 77, 78, 79-80, 86, 87–9, 90, 91, 126–7 wages 21, 27–8, 29, 35, 127, 247, 248–9, 255–6 see also parental leave benefits waiting days 154 Weber, Max 203, 209 welfare, types 126 welfare accessibility 201, 202, 203, 204, 205–7, 208, 209, 210, 211 welfare benefits see child benefit expenditure; means-tested benefits; parental leave benefits; pensions; sickness benefits; unemployment benefits welfare-capitalism 26, 33–8, 263 welfare efforts 53–4, 59–61 welfare expenditure see social expenditure; social protection expenditure welfare generosity and cost containment policies 44 decommodification indices 140, 143, 155–61 dimensions 201 fuzzy sets applied to unemployment benefits 206–7 measurement techniques 162–4 and partisanship 163 pensions 36, 37, 44, 127

Index and politics 15, 54, 127, 139 and pooled time cross-sectional methods 56 and residual vs. institutional welfare state 200 in social citizenship modelling 201, 202, 203, 204 and social expenditure 51–2 and social policies 170 in social rights, structural needs and social expenditure comparative study 112, 116, 122 in social rights modelling 201, 202 unemployment benefits 156, 157, 206, 207, 208–9, 210, 211 and variable impact chains 59–61 and welfare efforts 54 welfare state definitions 6–7, 16, 24, 25–6 ‘functional equivalents’ 26–7 ideal types see ideal types as independent variable 5 new politics 43–4

329

output perspective 15, 16, 17 outcome perspective 15, 16–17 paradigm shifts 166, 168, 184–5 types 17 welfare state restructuring see restructuring welfare state retrenchment see retrenchment welfare-to-need ratios 51, 52 Western Europe 27, 28–31, 34, 52–3, 231 see also individual countries Wilensky, Harold L. 53–4, 106, 200, 218 Williamson, John B. 107 Winner, Hannes 15, 61 Wolf, Holger 226, 227 women 34, 244–5, 246–9, 255–6 see also de-familization; gender equality work accident insurance 109–10 Working Families Tax Credit (UK) 137

Investigating Welfare State Change: The Dependant Variable Problem in Comparative Analysis

Welfare State Transformations: Comparative Perspectives

Welfare in the Kantian State

Welfare State Change: Towards a Third Way?

The Hidden Welfare State

The Hidden Welfare State

Revisiting the welfare state

Renegotiating the Welfare State

The Welfare State Reader

Practical analysis in one variable

Practical Analysis in One Variable

Practical Analysis in One Variable

Practical Analysis in One Variable

Practical Analysis in One Variable

Practical Analysis in One Variable

Contradictions of the Welfare State

Democracy and the Welfare State

Children, Welfare and the State

Globalization and the Welfare State

Is the Welfare State Justified?

Children, Welfare and the State

State Variable Methods in Automatic Control

European Welfare States: Comparative Perspectives

European Welfare States: Comparative Perspectives

Worlds of Welfare, Worlds of Consent?: Public Opinion on the Welfare State (International Comparative Social Studies) (International Comparative Social Studies)

The White Welfare State: The Racialization of U.S. Welfare Policy

Social Theory, Social Change and Social Work (State of Welfare)

Class Formation, Civil Society and the State: A Comparative Analysis

Citizenship and Welfare State Reform in Europe

Class Formation, Civil Society and the State: A Comparative Analysis

Welfare, Exclusion & Political Agency (The State of Welfare)

Investigating Welfare State Change: The Dependant Variable Problem in Comparative Analysis

Welfare State Transformations: Comparative Perspectives

Welfare in the Kantian State

Welfare State Change: Towards a Third Way?

The Hidden Welfare State

The Hidden Welfare State

Revisiting the welfare state

Renegotiating the Welfare State

The Welfare State Reader

Practical analysis in one variable

Practical Analysis in One Variable

Practical Analysis in One Variable

Practical Analysis in One Variable

Practical Analysis in One Variable

Practical Analysis in One Variable

Contradictions of the Welfare State

Democracy and the Welfare State

Children, Welfare and the State

Globalization and the Welfare State

Is the Welfare State Justified?

Children, Welfare and the State

State Variable Methods in Automatic Control

European Welfare States: Comparative Perspectives

European Welfare States: Comparative Perspectives

Worlds of Welfare, Worlds of Consent?: Public Opinion on the Welfare State (International Comparative Social Studies) (International Comparative Social Studies)

The White Welfare State: The Racialization of U.S. Welfare Policy

Social Theory, Social Change and Social Work (State of Welfare)

Class Formation, Civil Society and the State: A Comparative Analysis

Citizenship and Welfare State Reform in Europe

Class Formation, Civil Society and the State: A Comparative Analysis

Welfare, Exclusion & Political Agency (The State of Welfare)

Recommend Documents