The Institutional Logic of Welfare Attitudes: How Welfare Regimes Influence Public Support

THE INSTITUTIONAL LOGIC OF WELFARE ATTITUDES

This page intentionally left blank

The Institutional Logic of Welfare Attitudes How Welfare Regimes Influence Public Support

CHRISTIAN ALBREKT LARSEN Aalborg University, Denmark

© Christian Albrekt Larsen 2006 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, photocopying, recording or otherwise without the prior permission of the publisher. Christian Albrekt Larsen has asserted his moral right under the Copyright, Designs and Patents Act, 1988, to be identified as the author of this work. Published by Ashgate Publishing Limited Gower House Croft Road Aldershot Hampshire GU11 3HR England

Ashgate Publishing Company Suite 420 101 Cherry Street Burlington, VT 05401-4405 USA

Ashgate website: http://www.ashgate.com British Library Cataloguing in Publication Data Larsen, Christian Albrekt The institutional logic of welfare attitudes : how welfare regimes influence public support 1. Public welfare - Public opinion - Cross-cultural studies I. Title 361.6 Library of Congress Cataloging-in-Publication Data Larsen, Christian Albrekt. The institutional logic of welfare attitudes : how welfare regimes influence public support / by Christian Albrekt Larsen. p. cm. Includes bibliographical references and index. ISBN-13: 978-0-7546-4857-4 ISBN-10: 0-7546-4857-5 1. Public welfare--Public opinion--Cross-cultural studies. 2. Welfare recipients-Public opinion--Cross-cultural studies. 3. Welfare state--Cross-cultural studies. I. Title. HV51.L37 2006 361.6'5--dc22 2006013182 ISBN-10: 0 7546 4857 5 ISBN-13: 978-0-7546-4857-4 Printed and bound in Great Britain by Antony Rowe Ltd, Chippenham, Wiltshire.

Contents List of Figures and Box List of Tables Preface

vii ix xiii

1

Introduction

1

2

Welfare Regimes and the Attitudes of their Inhabitants

9

3

The Puzzle of Public Opinion

25

4

Welfare Regimes and Deservingness

45

5

Welfare Regimes and Perceived Causes of Poverty

65

6

Selectivism and Stigmatisation

99

7

Generosity and Stigmatisation

111

8

Regime-dependent Perceptions and Social Assistance

123

9

Deservingness and Welfare State Development

139

References Index

157 167

This page intentionally left blank

List of Figures and Box Figure 2.1.

Three mechanisms that link the macro-level phenomenon of welfare regimes with the micro-level phenomenon of public attitudes towards social policy

14

Box 3.1.

Introduction to the survey material used in the book

27

Figure 3.1.

Times high status occupations ought to earn more than a skilled factory worker. Western countries available in ISSP 1992 and ISSP 1999. Median

40

Times skilled factory worker ought to earn more than the unskilled factory worker. Western countries available in ISSP 1992 and ISSP 1999. Median

42

The theoretical link between welfare regime and cross-national differences in welfare attitudes

50

The main structure of welfare services and benefits in case of long-term unemployment in an ideal-type liberal and social democratic welfare regime

51

The connection between cross-national differences in perception of control of neediness found in Eurobarometer 1993/World Value Study 1990 and public support for welfare policy found in ISSP social inequality 1992

92

The distinction between the insurance track, the universal track, and the residual track at programme level

101

The connection (gamma-coefficients) between perceptions of job opportunities, level of generosity, work ethic, and public support for social assistance. Attitudes towards spending and the living standard that should be provided

128

Figure 3.2.

Figure 4.1.

Figure 4.2.

Figure 5.1.

Figure 6.1.

Figure 8.1.

viii

Figure 8.2.

Figure 8.3.

Figure 9.1

Figure 9.2

Figure 9.3

Figure 9.4

The Institutional Logic of Welfare Attitudes

Connection between perception of the share of immigrants on social assistance and relative and absolute support for social assistance in a social democratic regime. Denmark, 2004, gamma-coefficients

131

The connection between general perception of work ethic among claimants of social assistance, work ethic among immigrants on social assistance and absolute and relative support for social assistance. Denmark, 2004, gamma-coefficients

131

Illustration of reproduction of public support for welfare policy (left-hand side) and dynamic path-dependent public support for welfare policy (right-hand side)

151

Public support for welfare policy in Britain from 1983 to 2002. Share in favour of increased spending on health, education, and social benefits (solid line) and share in favour of increased spending on welfare benefits for the poor (dotted line). Per cent

152

Public support for welfare policy in Sweden from 1986 to 2002. Share that finds reduction in public sector a bad proposal (solid line) and share in favour of maintaining or increasing public spending on social assistance (dotted line). Per cent

153

Public support for welfare policy in Denmark from 1973 to 2001. Proportion of voters that 1) want to maintain or increase spending on social assistance (dotted line), 2) want to maintain social reforms at least at present level, and 3) who disagree that too many receive social benefits without needing them

154

List of Tables Table 2.1.

Selected features and effect of the three regimes

12

Table 3.1.

Legitimate income differences in Sweden, Germany and USA in 1992. Measured by mean, mean without 10 per cent extreme cases, median in relation to both unskilled and skilled factory workers

38

The effects from, respectively, selective and universal public policy on different dimensions of deservingness

52

Three regime dimensions that influence perceptions of the poor and unemployed and the theoretical position of the ideal-type countries of Sweden, West Germany and the USA

61

Table 4.1.

Table 4.2.

Table 5.1.

Share explaining poverty with ‘laziness and lack of will power’ in selected Western countries in the period from 1976 to 2001

Table 5.2.

Prediction of proportion answering in poverty caused by ‘laziness and lack of will power’ based on degree of selectivism, level of unemployment, social expenditures, and ethnic fractionalisation (OLS). World Value Study (1990)

74

Prediction of share answering ‘in control’ based on regime belonging, level of unemployment and social expenditures (OLS)

77

The connection between welfare regime, unemployment level, individual background variables, political orientation and perception of poor being in control of poverty in the World Value Study 1990–93, 16 Western countries

82

The connection between welfare regime, unemployment level, individual background variables, political orientation and perception of the poor being in control of poverty in the World Value Study 1990–93

86

68

Table 5.3.

Table 5.4.

Table 5.5.

x

Table 5.6.

Table 5.7.

Table 5.8.

Table 6.1.

Table 6.2.

Table 7.1.

Table 7.2.

Table 7.3.

Table 7.4.

Table 7.5.

The Institutional Logic of Welfare Attitudes

Cross-national differences in perception of control of neediness and public support for welfare state policy in countries where the ISSP social inequality module II was conducted. Shares, index and regression (OLS) between ‘laziness’ and dependent ISSP variables

90

Share explaining poverty with ‘laziness and lack of will power’ in Western Countries available in the second wave of the World Value Study

94

Correlations between perception of control of neediness and public support for welfare policy measured in the Eurobarometer 2001. Correlations at individual level within each country and overall correlation at individual level (gamma-coefficients and level of significance)

97

Share of population (18–) or share of families receiving various social benefits in four social democratic regimes in 1999/2000. Unemployment rate also included

103

Judgement of how often people tend to look down on recipients of different social benefits. Percentage answering ‘very often’ or ‘often’ and percentage difference (‘very often’ and ‘often’ minus ‘quite rarely’ and ‘very rarely’)

106

Experience of financial hardship among Danish and Finnish long-term unemployed

113

The experience of being ‘looked down on’ among long-term unemployed due to receiving unemployment benefits or social assistance. Denmark and Finland. Percentages

114

Personal feeling of being looked down on as a problem of being unemployed. Finland and Denmark. 10 point scale

115

Relationship between financial hardship and feeling of stigmatisation as a problem of being unemployed. Share between 6 and 10 on 10-point scale and average

116

Relationship between sex, age, number of children, health, type of social benefits, country and feeling of stigmatisation as a problem of unemployment

118

List of Tables

Table 8.1.

Table 8.2.

Table 8.3.

Table 8.4.

Table 8.5.

xi

One relative and two absolute measures of public support for social assistance. Denmark, 2004, per cent

125

Distributions of perception of job opportunities, level of generosity, and work ethic. Denmark, 2004, per cent

127

Attitudes towards immigrants’ entitlement to social support during unemployment in a social democratic regime. Denmark in 1993 and 2004, per cent

133

Attitudes towards immigrants’ entitlement to social assistance in a social democratic regime. Denmark 1998, 2000, and 2004, per cent

134

Causal model between perceptions of job opportunities, generosity, work ethic and support for social assistance controlled for political affiliation. Gamma-coefficients and explained variation

136

This page intentionally left blank

Preface

The tailback at Viale Alessandro Volta in Florence was a good place to speculate about the causal arguments within a book manuscript. While waiting to get to the European University Institute in Fiesole, the same beggar stood every morning outside my Scandinavian Volvo windscreen. Later I found out that her name was Diana. She was a gipsy from Romania who had come to Italy illegally with her husband and two children some years ago. Nevertheless, before I knew something about her it was the thoughts of the drivers in the tailback that interested me. Some rolled down the car window and gave Diana a euro and some kept the hands on the wheel. But what decided whether the drivers found Diana deserving or not deserving? According to which criteria was Diana judged? And could these deservingness criteria really be the solution to the scientific problem that occupied my mind? The answers turned out to be that we have good reasons to believe that all the drivers judged Diana according to very similar criteria. And whether Diana was perceived to fulfil these criteria was more than a matter of a driver’s mood, political orientation, economic resources etc. We can actually specify a number of mechanisms that explain how the judgement of deservingness is heavily influenced by the welfare institutions that are in place in different countries. Thereby, I became closer to explaining why my fellow Scandinavians and I are so much in favour of welfare policy; especially as, when it comes to welfare policy concerning the living conditions of the poor and unemployed, cross-national differences are very significant. I was also able to give an institutional explanation to the fact that Americans and others living in liberal welfare regimes are much more sceptical about welfare policy – in contrast to those scholars who explain the liberal reluctance through culture or racial divides. In the process of making these tailback speculations into this book I have been helped by a number of colleagues. Unfortunately, all of them cannot be mentioned. However, first of all I am grateful to Jørgen Goul Andersen who has given me constructive guidance throughout the process. It is also fair to say that in many respects I have tried to follow the basic assumptions, which can be deduced from Jørgen’s previous work within this field. Intellectually I am also grateful to another Dane, Gøsta Esping-Andersen, who synthesised the development in the Western welfare states in a way that have inspired a whole field of research. This book is a modest attempt to contribute with a brick to this theoretical building. Another great inspiration has been Wim van Oorschoot who during the process changed from being a remote scholar in the Netherlands to one of my colleagues at the Centre for Comparative Welfare Studies at Aalborg University. Wim gave me constructive and critical comments, especially in the later part of the process. Finally, I need to say

xiv

The Institutional Logic of Welfare Attitudes

thanks to Christoffer Green-Pedersen, Olli Kangas and Anette Borchorst. They read early versions of the manuscript and their comments contributed significantly to the present result. All that being said, there should be no doubt that I am responsible for all mistakes, interpretations and conclusions.

Christian Albrekt Larsen Centre for Comparative Welfare Studies, Department for Economics, Politics and Public Administration, Aalborg University, Denmark February 2006

Chapter 1

Introduction The development of large and generous welfare states is one of the most remarkable characteristics of Western countries in the period from the end of the Second World War to the mid 1970s. One can speak of a golden-age where the combination of capitalism and social protection almost came to represent the culmination of Western civilisation. Beneath this overall trend – in the 1970s, it was sometimes described as a necessary condition for advanced capitalism – one finds substantial variations between the Western countries. One of the most accepted explanations of this variation in the size and the principle of welfare states is the so-called power resource theory (Korpi 1983). The claim is that it very much depends on the strength of the labour movement and its ability to form coalitions with the rising middle class. However, the welfare optimism of the golden-age was interrupted by the oil crises in the 1970s and the following decades, with high levels of unemployment in most Western countries. The immediate result was reduced tax revenues and increased expenditures, which led to fiscal pressure on the welfare states. Combined with increased economic integration, which limited the possibilities to pursue expansive macro-economic policy, low productivity growth, and an increased number of old people, scholars came in the 1990s to speak of a period of permanent austerity (e.g. Pierson 2001b). In this present phase of welfare state retrenchment or restructuring, a number of scholars have argued, Pierson being the pioneer, that we are witnessing a so-called ‘new politics’ of the welfare state. This ‘new politics’ is centred around a conflict between a constituency in favour of the welfare state and a policy elite forced to take necessary decisions (1994, 1996, 2000, 2001b). The resistance to change is basically explained by the classic argument of concentrated costs (those who are to lose benefits, services or public employment) versus dispersed benefits (a balanced public budget). The end result is that politics under these conditions becomes an art of ‘blame avoidance’. There is, without doubt, a lot of truth in these ‘macro stories’ of welfare state development and this book will follow the comparative, and especially the institutional, approach taken in this literature. However, one of the major problems with these ‘macro stories’ is that they have rarely paid serious attention to the microlevel foundation of their causal reasoning. Basically, the power resource theory claims that the electorate is guided by (long-term) class-interest and the ‘new politics theory’ claims that the electorate is guided by (short-term) self-interest; but, as we shall see, the empirical evidence actually points in another direction. Respectively modelled around class-interest and self-interest, both theories also claim that the welfare state institutions in place influence the future preferences of the electorate.

2

The Institutional Logic of Welfare Attitudes

This line of reasoning helps to explain why the restructuring of welfare states (even in a situation where they face very similar external pressures) followed distinct paths. Many current discussions of welfare state development relate to the notion of path-dependency. But again we shall see that empirical studies have had difficulties in finding the linking mechanisms between the macro and micro levels, which are suggested by the power resource and the new politics theory. One solution to these anomalies is to question the relevance and the quality of the empirical studies we have. In 1998, Korpi & Palme – being the primary defenders of the power resource theory – argued that, for example, ‘…the empirical testing of the macro-micro-links among institutions and the formation of interest and coalitions provides a major challenge for social scientists, but comparative micro-data currently are lacking’ (Korpi & Palme 1998, p. 682). Another solution is to gather the empirical evidence we actually have, try to find a theoretical explanation of the anomalies, and test the new theories on existing and new data. It is this latter research strategy that will be followed in this book. Based on available cross-national surveys, a number of studies have looked in particular for a connection between the famous distinction between social democratic (Scandinavian), conservative (Continental European), and liberal (Anglo Saxon) welfare regimes (Esping-Andersen 1990, 2000) and public support for welfare policy. Our review of these studies, in Chapter 3, leads to three overall conclusions. First, the previous studies clearly show that we sometimes find substantial cross-regime differences in public support for welfare policy. Secondly, the studies that find the expected pattern of high support in social democratic regimes, medium support in conservative regimes, and low support in liberal regimes, primarily measure attitudes towards policies that concern the living conditions of the poor and unemployed. In the liberal regimes, this taking care of the (potentially) poor is basically what welfare policy is all about. For Americans, welfare policy only refers to need-tested programmes targeted at those at ‘the bottom’ of society. For Europeans, welfare policy refers to a much broader range of policies that also include benefits and services given to all citizens. Nevertheless, the point is that when it comes to retrenchment or restructuring of welfare policy in the narrow or American meaning we could actually expect the policy elite in social democratic regimes to be ‘blamed’ and the policy elite in liberal regimes to be ‘blessed’. Thirdly, the previous studies have not been able to find the mechanisms that link the macro- and micro-structure. The substantial cross-regime differences in support for welfare policy cannot simply be explained by differences in the attitudes of the middle class, the number of people who receive welfare benefits and services, or the number of public employees. Thus, when it comes to welfare attitudes, the previous studies leave us with a regime pattern at the aggregated level on the one hand and on the other hand demonstrate an inability to explain this pattern. With this research frontier as our point of departure, the aim of this book is to search for the mechanisms that link the macro-structures of welfare regimes to the micro-structures of public opinion towards welfare policy – call it a search for the missing link or the intervening variables.

Introduction

3

Even though we are sceptical about the micro foundation of the power resource theory, the welfare regime theory, and the new politics theory, we borrow many insights from this literature and basically end up explaining cross-national differences in attitudes with cross-national differences in welfare state institutions. By pursuing this institutional line of reasoning we question two other popular explanations of cross-national differences in welfare attitudes. The first popular explanation basically claims that low support (by comparative standards) for welfare policy primarily has to do with the presence of ethnic heterogeneity. The argument is primarily inferred from the American experience. The claim is that the presence of Indians and the import of slaves simply gave, and continues to give, fundamentally different preconditions for the welfare discussion in USA. This idea is widespread among American scholars and is supported by studies that show a strong correlation between attitudes towards the race issue and the welfare issue (e.g. Quadagno 1994; Gilens 2000). Moreover, this explanation gains increasing influence as recent European discussions about immigration sometimes also link-up with the welfare discussions. From their study of so-called ethnic fractionalization, Alesina & Glaeser (2004) infer that eventually increased ethnic heterogeneity will lower the Europeans’ passion for welfare policy. However, simply by turning the argument around, i.e. to claim that higher support for welfare policy in the European countries is caused by ethnic homogeneity, one becomes more uneasy about the argument. Naturally, the USA has had a large minority with a black skin but European history is also filled with clashes between different ethnic groups. The building of nation states had a homogenising effect on the one hand; a process that has not been given the same time in the colonies. However, this long process of turning inhabitants of a given area into national citizens also established new and persistent divides between the majority and ethnic minorities who wanted their own nation state or wanted to belong to another nation state (e.g. Bommes & Geddes 2000). Furthermore, Alesina & Glaeser (2004) and others may argue that the degree of ethnic homogeneity was crucial for establishing different kinds of welfare institutions, but still they need to take into account that once welfare institutions are established they have an impact on their own. The impact institutions have on public welfare attitudes is the topic of this book. And we will demonstrate that the institutional line of reasoning indeed helps to explain the pattern of low support in the USA and other liberal regimes, medium support in the conservative regimes, and high support in the social democratic regimes. The second popular explanation of cross-national differences, which we will question, is the ‘culture thesis’. It basically claims that lower support in the USA and other liberal regimes is caused by a ‘passion for freedom over inequality’. The argument is in line with Lipset’s thesis of an American exceptionalism (e.g. Lipset 1996). The other way around, however, the ‘culture thesis’ claims that the high support in the social democratic regimes is caused by a ‘passion for equality’ (Graubard 1986). This explanation is widespread among Europeans; probably because, from a European perspective, it reproduces the ‘nice’ idea about a socially responsible Europe and a socially irresponsible USA. Not being able to find the expected effects from class-

4

The Institutional Logic of Welfare Attitudes

interest and self-interest, academic scholars sometimes also end up using different degrees of ‘egalitarianism’ as a cultural and residual explanation of cross-national differences in support for welfare policy (e.g. Andress & Heien 2001). However, the argument contradicts one of the very first and strongest impressions Europeans got after crossing the Atlantic. Coming from France, Tocqueville described in detail, and was indeed somewhat worried about, the ‘passion for equality’ that prevailed in the USA. Naturally, it was first of all a call for equality of opportunities and naturally much has happened in Europe since then. However, the comparative studies that more positively try to measure justice beliefs or level of egalitarianism do not find a pattern that neatly coincides with the described regime pattern. In Chapter 3 we will actually show that, measured in terms of the perception of just wage differences, the most anti-egalitarian attitudes are found in the conservative welfare regimes and not in the USA or the other liberal welfare state regimes (as Tocqueville would predict if he were still alive). Therefore, we argue that, in terms of egalitarian values, what distinguishes Americans and others who live in liberal welfare state regimes is not a general anti-egalitarian attitude. Instead, we find a specific anti-egalitarian attitude towards ‘the bottom’, which can be given an institutional interpretation. The main contribution from this book is to develop a theoretical framework that enables us to explain how the institutional structure of the different welfare regimes influence public support for welfare policy. The theory is established by combining the welfare regime theory, which is presented and critically discussed in Chapter 2, with the theory of deservingness criteria (Cook 1979; De Swaan 1988; Will 1993; Oorschot 2000). For some reason these two lines of reasoning have lived rather separate lives. The basic idea of our synthesis is that the structures that characterise the different welfare regimes influence the way the public perceives the poor and the unemployed, which again influences the judgement of deservingness and thereby support for welfare policy. The theoretical framework is inspired by the classic thesis that means testing in contrast to universalism gives the public a negative perception of recipients (‘them’ versus ‘us’ logic). But the combination of welfare regime theory and deservingness theory enables us to specify this classic argument and open a broader theoretical perspective that includes the influence from the generosity of the welfare state, and the job opportunities provided by the (regime specific) labour market. Even though the ethnic fractionalisation thesis (Alesina & Glaeser 2004) is treated as a competing theory our theoretical framework also allows us to specify through which mechanisms the ethnic heterogeneity potentially can influence support for welfare policy. This theoretical framework is presented in Chapter 4. Our line of reasoning has a lot in common with normative institutionalism, which in political science often is associated with March & Olsen (1984, 1989), but actually goes back to Talcott Parson and the old classic sociologists (Peters 1999). The idea that the political preferences of individuals are not exogenous, as in rational choice theory, but highly influenced by the institutional structures is shared by this book. Therefore we are sceptical about the rational choice institutionalism and its assumptions about an ‘economic man’, which has dominated the ‘new politics theory’. However, we are also sceptical about the variants of normative

Introduction

5

institutionalism where the attitudes of the ‘sociological man’ in a deterministic way are linked to a national cultural history. In contrast, we suggest what we shall call a ‘political man perspective’ on the formation of public support for welfare policy. This implies a reflective individual, where attitudes towards welfare policy are not only guided by self-interest or abstract societal values and norms but also based on concrete perceptions of the reality; in this case the perceptions of the poor and unemployed. Such a position fits nicely with recent studies that have shown that attitudes towards concrete policy proposals are highly dependent on the framing of the political issues (e.g. Gamson & Modigliani 1987; Kinder & Sanders 1996; Nelson et al. 1997; Kangas 1997). As the dependent variable we will only discuss attitudes related to policies that concern the living condition of the poor and unemployed. Above, we called it ‘support for welfare policy’ in the narrow meaning or American meaning. Thereby this book primarily helps to explain why the living conditions of the poor and unemployed vary – and continue to vary – so tremendously across the Western capitalist countries. If, for example, we take the ideal type countries used by EspingAndersen and the extent to which the income distribution generated in the market is altered (measured as the degree of reduction in the so-called Gini-coefficient), we find a clear pattern with the United States at the one end, Germany somewhere in the middle and Sweden at the other end (e.g. Ferrarini & Nelson 2002). If we look at the share of the population that lives below the relative poverty line, we find the same pattern. In 2000, 17 per cent of Americans had an income below the relative poverty line (less than 50 per cent of the equivalent median income) compared with 8.3 per cent in Germany and 6.5 per cent in Sweden.1 Such findings support the old saying in comparative welfare state research that ‘welfare for the poor becomes poor welfare’. This book will show how this process can be partly explained by regime-dependent welfare attitudes. Naturally, the focus on welfare policy in the narrow sense puts limitations on the conclusions we can draw but we have a number of reasons for taking this focus. First, from previous studies we have indications of a regime effect when it comes to welfare policy in the narrow sense. Second, even though good comparative data on welfare attitudes are unavailable we do have some comparative data on attitudes towards welfare policy in the narrow sense. And actually some of these data are little explored. Third, the management of ‘bad risks’ is a crucial variable in EspingAndersen’s argument for the existence of three, and only three, distinct regimes (Esping-Andersen 2000, pp. 77–81). Fourth, welfare policy in the narrow sense is also the crucial variable in the American literature on the connection between race and public support which, as already mentioned, will be one of the main opponents or ‘discussion partners’ in this book.

1 The estimates of the proportion of citizens living below the poverty line are heavily influenced by the method of calculation. Here we have simply taken the overall figures provided on the webpage of the Luxembourg Income study.

6

The Institutional Logic of Welfare Attitudes

The empirical backbone of the book is presented in Chapters 5, 6, 7 and 8. In these chapters we will prove the soundness of the causal reasoning in our synthesis of the welfare regime theory and the deservingness literature. In Chapter 5, we analyse how the three regime dimensions deduced in Chapter 4 (job opportunities, degree of selectivism, and degree of generosity), and the differences in ethnic fractionalisation influence public explanations of causes of poverty. Based on the World Values Studies (WVS) and Eurobarometer surveys, we are able to analyse cross-regime data over a long period. Still one cannot totally avoid the so-called ‘small n-problem’ – having too few cases and too many independent variables – that haunt all comparative research. In our case, the problem is that even though we find the expected connection between welfare regimes and the perception of poor and unemployed, one cannot totally dismiss the possibility that the pattern could be caused by differences in egalitarianism or ethnic homogeneity rather than differences in welfare institutions. Therefore, Chapters 6 and 7 supplement the ‘traditional’ comparative analysis in Chapter 5 with two so-called focused comparisons. The task is to ‘isolate’ the institutional effects derived respectively from the degree of selectivism and degree of generosity. By comparing the Nordic countries, we hold the size of the countries, ethnic fractionalisation, social trust, political systems, cultural legacy etc. more or less constant and look at the effects of differences in welfare policy (which actually can be found across the Nordic countries). The data for these ‘most similar designs’ come from a Nordic module to the International Social Survey Program (ISSP) from 1999 and a large Danish and Finnish dataset. The latter dataset both includes a general population sample and a large sub-sample of long-term unemployed. These focused comparisons enable us to conclude with very high certainty that the institutional effects suggested by our theoretical framework are present. However, one of the main problems with ‘most similar designs’, i.e. everything other than the independent variable is held constant, is that one cannot assess the relative importance of the independent variable in play. In our case, one would for example like to know how strong the institutional effects are compared with the effectsderived from ethnic divides. Therefore, Chapter 8 applies our theoretical framework on a national sample that has been established exclusively for this purpose. At this individual level within one nation we do not have ‘real’ cross-national institutional variation. But the sample enables us to analyse, in one model, how public perceptions of (1) immigrants, (2) the poor and unemployed in general, (3) the generosity of the welfare state, and (4) job opportunities, affects public support for welfare policy. The analysis shows that the perceptions, which we argue in Chapter 4 are likely to be regime-dependent, are highly relevant in order to explain variations in individuals’ support for welfare policies. From that we infer that they may also be highly relevant in explaining cross-national variations in public support for welfare policies. Besides summarising the theoretical argument and the main empirical findings, Chapter 9 includes four discussions. First, we discuss the thesis’ implication for the welfare regime theory. We have only analysed public support for welfare policy in the narrow sense, and we admit that what is most needed is better comparative

Introduction

7

data. However, we still believe that the book challenges the micro-foundation of present comparative welfare state theory and succeeds in presenting an alternative explanation. Secondly, we discuss to what extent regime-dependent public support for welfare policy actually influences the pursued welfare policy. Thirdly, we discuss the notion of path-dependency in relation to public support for welfare policy. And finally we discuss the future public support for welfare policy in the light of the recent trends in welfare state development and immigration. It leads to the prediction that, due to the impact of welfare institutions, the European future might not be American.

This page intentionally left blank

Chapter 2

Welfare Regimes and the Attitudes of their Inhabitants This chapter introduces the main concepts of the regime theory and describes how various researchers have theorised the link between the characteristics of the three regimes and attitudes towards welfare policy. We will discuss the argument of class interest and coalitions, short-term interests and the structure of costs and benefits and reproduction of values and norms. Finally, we will deliver a critique of these arguments. The latter section partly anticipates our review of previous studies in Chapter 3, in which it clearly emerges that it has been rather difficult to substantiate the theoretical arguments by empirical findings. However, first of all it might be helpful to clarify what Esping-Andersen actually meant by a liberal, a conservative and a social democratic welfare state regime. The first part of the concept is rather easy to comprehend as it simply refers to the main political force within a given regime. A number of historical studies (e.g. Baldwin 1990) have shown that, especially if we analyse the yearly formative years, the pattern of political forces and coalitions is much more complex than suggested by these labels, but at an overall level they are understandable. The regime label has caused larger problems. Basically, the concept is used to broaden the traditional state-centred comparative welfare state research, where ‘institution’ narrowly refers to the programmatic structure of social policies within a given country. In EspingAndersen’s own words, the regime concept is used (1) because the concept of ‘welfare state is too narrowly associated with the conventional social-amelioration policies’, (2) because the ‘contemporary advanced nations cluster not only in terms of how their traditional social-welfare policies are constructed, but also in terms of how these influence employment and general social structure’, and (3) to ‘denote the fact that in the relation between state and economy a complex of legal and organisational features are systematically interwoven’ (Esping-Andersen 1990, p. 2). In his followup book from 1999, Esping-Andersen even abolished the state label and gave the differences in family structures a more prominent role, which led him to define a welfare regime ‘as the ways in which welfare production is allocated between state, market, and household’ (Esping-Andersen 2000, p. 73). Therefore Esping-Andersen also emphasises that public welfare and social policies not only limit the market, which in the 1990 book was summarised as the degree to which citizens were decommodified from the (labour) market, but also affect the degree to which families absorb social risk; i.e. one can speak of degrees of ‘familialism’ or ‘de-familisation’ (Esping-Andersen 2000, p. 51). Both in terms of de-commodification and de-

10

The Institutional Logic of Welfare Attitudes

familisation the Nordic countries are at one extreme. At the other extreme we find the Anglo-Saxon countries with the lowest degree of de-commodification and the Continental, especially Southern Continental European, countries with the lowest degree of de-familisation. Nevertheless, the main point remains the same: a larger number of institutions and social structures cluster into three groups, each with their own internal logic. Childcare is a classic example. In Scandinavian countries, childcare has become the responsibility of the state, which provides highly subsidised childcare. In AngloSaxon countries, childcare is typically bought on the market, and in Continental European countries many children are still nursed by the family. This makes a difference not only in terms of price, quality and equality of childcare, but obviously also in terms of labour market and family structure. The focus of our study is the possible existence of an internal logic between the three clusters of institutions and the attitudes of the electorate. Before we present the dominant theoretical arguments, let us introduce the main features of the three regimes. The main features of the three welfare regimes Above, the different regimes were used as static ‘pictures’ of the western societies, but at the heart of Esping-Andersen’s theory is actually a dynamic explanation of welfare state development in the 20th century. In this brief presentation, we will make a rough distinction between three periods; the period of welfare state creation from the late 19th century to the Second World War, the period of expansion from the end of the Second World War approximately to the second oil crisis in the mid1970s, and finally the period of restructuring from the second oil crisis until now. In the first period, we can roughly distinguish between the countries that introduced compulsory social insurances financed by contributions from employers and employees and the countries that introduced targeted benefits to the needy, financed by general taxes. Two quite different principles were in play here. In the former schemes, access to and level of social benefits followed the principle of merit; you were entitled if you had paid a contribution, and the level normally depended on the amount paid. With the legislation in the 1890s that protected workers against sickness, disability, unemployment and old-age, Germany was the frontrunner in establishing these schemes in Continental Europe, but was soon followed by France, the Netherlands, Belgium and later by the Southern European countries. In the latter schemes, social benefits followed the principles of need and citizenship; if you were a needy citizen you were entitled, and the level was not dependent on what you had paid in taxes. With the pension legislation from 1891, Denmark was the frontrunner in this latter group of countries that came to include the other Scandinavian countries, New Zealand, Australia and Great Britain. In the last part of the first period, this latter group of countries can be divided into the Scandinavian countries, which expanded the entitlement to the whole population, most clearly the pension schemes, and the

Welfare Regimes and the Attitudes of their Inhabitants

11

Anglo-Saxon countries that continued with targeted benefits to the needy. The US became the prime example of a residual welfare state. In the following period, the so-called golden age of the welfare state, the differences between Western states began to grow, not only in principle but also in the size of the public budget. If we take the countries closest to Esping-Andersen’s ideal types, then, in 1960, the differences in public expenditures in percentage of GDP were not very large, with Sweden at 31 per cent, Germany at 32 per cent and the USA with 27 per cent. This is naturally a very rough measure, and USA’s high military expenses are a part of the story. But still the development over time is interesting.1 In 1975, the public expenditures had increased to 49 per cent in Sweden and Germany, but only to 36 per cent in USA (Goul Andersen 1998, p. 116). At this historical stage, the basic differences between the welfare states of the different regimes had crystallised. As synthesised in Table 2.1, continental European countries, to some extent with the exception of the Mediterranean countries, had expanded the compulsory social insurances to the whole workforce, which generated a fairly high amount of risk sharing within these countries, or at least within occupational groups. The Anglo-Saxon countries established a rather residual welfare state targeted at the ‘needy’, while Scandinavian countries had entitled the whole population to core transfers such as pensions, and services such as healthcare. In contrast to the British Beveridge plan, the Scandinavian countries came to provide benefits and services well above a minimum level. In the last part of ‘Three worlds of welfare state capitalism’ and in the work from 1996, Esping-Andersen describes how these regime differences affect the way countries developed since the economic crises of the 1970s, i.e. the development in what we have called the phase of restructuring. Many Continental European countries followed a trajectory of labour force reduction, primarily through early retirement, in order to protect the ‘insider’ male breadwinner and his family. The Scandinavian countries followed a trajectory of service job creation within the public sector. It started a self-perpetuating process where demand for female labour power increased, and the new public services made it possible for women to leave the home. As emphasised in Esping-Andersen’s 1999 book, this led to a high degree of de-familisation. Finally, many Anglo-Saxon countries followed the neo-liberal route where employment in the private service sector was created by larger wage flexibility, which as in Scandinavia both increased the demand and the possibility for greater female participation on the labour market. Thus, according to Esping-Andersen, the period of restructuring did not lead to convergence, but rather strengthened the difference between regimes (Esping-Andersen 1996b, 2000). If, in 1996, we again take the rough measure of government expenditures in relation to GDP, the share increased to 65 per cent in Sweden and stabilised respectively at 50 per cent in Germany and 36 per cent in the USA (Goul Andersen 1998, p. 16). 1 Social expenditures are naturally a better measure, but they cannot be calculated so far back in history. Even the subtraction of military expenses is difficult in the old national accounts, which are the basis of the OECD figures.

12

Table 2.1. Selected features and effect of the three regimes Conservative welfare regime

Liberal welfare regime

Social Democratic welfare regime

Compulsory social insurance financed through social contributions by employer and employees.

Targeted transfers and services to the ‘needy’ financed through general taxes.

All citizens entitled to transfers and services financed through general taxes.

Create safety

Avoid poverty

Create safety and equality

Regulated with insider/outsider divide. A medium public service sector Traditional: bread winner model

Unregulated with large wage flexibility. A large private service sector. Semi-modern:

Regulated with a large public service sector.

Medium de-commoficiation

Low de-commodification

High de-commodification

Low de-familisation

Medium de-familisation

High de-familisation

West Germany, Italy Continental European countries included Mediterranean latecomers

USA Anglo-Saxon: USA, Canada, Australia, New Zealand, Switzerland and Britain

Sweden Nordic countries: Sweden, Denmark, Norway and the latecomer Finland. Sometimes the Netherlands.

Welfare state: - basic principle

Labour market: Family structure: Effect towards the market: Effect towards the family: Country closest to the ideal type: Clusters of countries:

1

1 Taken from Esping-Andersen (2000, p. 85). Source: Selected features synthesised from Esping-Andersen (1990, 2000, p. 85).

Modern: Both men and women on the labour market

The Institutional Logic of Welfare Attitudes

- basic aim of the welfare state

Welfare Regimes and the Attitudes of their Inhabitants

13

These overall differences have a large impact on how welfare is ‘produced’ for the ‘bad risks’, ‘low-income groups’ or call it the ‘potential poor’. In his defence for the three regimes, mentioned in the previous chapter, Esping-Andersen (2000, p. 74) argues that ‘the relative weight of need-based social assistance compared to rights programmes should constitute an excellent indicator of “liberalism”’ and based on this measure (which is the strongest predictor) and the share of private pension, he finds a clear cluster of USA, Canada, Australia, Ireland, New Zealand and Britain. In other words, the degree to which the welfare states are concentrated on producing welfare to the ‘bad risks’ that cannot be covered in the market is a crucial feature of the liberal regime. With Medicaid to the poor, Medicare to the old, AFDC to lone mothers (abolished and handed over to the states in 1996) and tax credits to lowincome families with children, the USA is the prime example. In social democratic regimes, the ‘bad risks’ are in general covered by the same universal schemes as the rest of the population. Some other welfare states also have universal elements, e.g. British pensions and healthcare, but Esping-Andersen argues that ‘where the Nordic countries truly stand out in comparison to other tendentially universalistic systems (like the British) is their deliberate attempt to marginalize the role of needs-based assistance’ (Esping-Andersen 2000, p. 78). That means that low-income groups are given individual social rights even within the system of social assistance, and through active labour market policy it has become a state responsibility to provide jobs. Finally, the ‘bad risks’ of the conservative regime, i.e. those who are not entitled through the compulsory social insurance on the labour market, are primarily to be covered by the family. ‘What unites Austria, Germany, Italy and Spain is the continued legal prescription that parents (or children) be responsible for their children (or parents) in case of need’ (Esping-Andersen 2000, p. 83). However, as families are not always able to produce this welfare, a number of ad hoc residual schemes can also be found in the conservative regime, but ‘its target is very different: liberal residualism means picking up bad risks left behind by market failure; conservative residualism, in contrast, is primarily a response to family failure’ (Esping-Andersen 2000, p. 83). This book claims that these different approaches to the ‘bad risks’ are connected to – and also influenced by – welfare attitudes produced within each welfare regime. As we shall see below, many scholars believe in such a connection between the characteristics of a given welfare regime at the macro-level and the attitudes towards public policy at the micro-level, which then – through the processes of the democratic political system – feed back to the macro-level and thereby help explain continuity and change over time. In Figure 2.1 we distinguish between three theoretical arguments or mechanisms used in previous studies to establish the link between macro and micro; respectively (1) the formation of class interests and coalitions, (2) short-term self-interest, and finally (3) values and norms concerning equality,

14

The Institutional Logic of Welfare Attitudes

justice and solidarity.2 The mechanisms are deducted from very different theoretical traditions and will be treated separately in the following three sections.

Figure 2.1. Three mechanisms that link the macro-level phenomenon of welfare regimes with the micro-level phenomenon of public attitudes towards social policy

2 Empirical studies typically cite Esping-Andersen for the thesis that different labour market trajectories in the restructuring phase will generate different cleavages. The increased income mobility of the neo-liberal route in the Anglo-Saxon countries may increase the traditional class conflict within former segmented groups; ‘as women and Blacks are becoming more fully integrated into the prevailing class structure, the likelihood is that class differences will crystallize sharply within the various minority groups’ (Esping-Andersen 1990, p. 228). The strong protection of the ‘insiders’ in Continental European countries could create a conflict between these male breadwinner and outsiders, i.e. women, young and unemployed. Finally, the tremendous expansion of public service jobs in the Nordic countries could increase the tension between employed in this new sector (typically women) and employed in the private sector (typically men), who have to ‘pay the bill’ (Esping-Andersen 1990, chapter 9; e.g. cited by Svallfors 1997; Andress & Heien 2001). However, it remains unclear whether an increased class conflict between women and between blacks in Anglo-Saxon countries, an insider– outsider conflict in Continental Europe and a public–private sector conflict in Scandinavia should increase or decrease the level of public support for welfare state policy. Furthermore, we should be aware that these theses were developed to forecast possible conflicts associated with new labour market trajectories in the restructuring phase and not as explanations of public support. Thus, it seems a bit odd to pick these theses if the idea is to explain crossnational variation in public support for welfare state policy.

Welfare Regimes and the Attitudes of their Inhabitants

15

Welfare regimes and the formation of class interests and coalitions Esping-Andersen is not very explicit about a link between welfare regimes and public attitudes towards the responsibility of the state, but in several places he clearly implies such a feedback process through the electorate. In his early writings on the Danish and Swedish Social Democracy, it was actually a central thesis that differences in welfare policy generated differences in public support. Differences in housing policy were, for example, believed to unite the Swedish labour movement and divide the Danish labour movement (Esping-Andersen 1980, pp. 456–548). Thus, the basic line of reasoning seems to be an institutional addition to the power resource theory, i.e. the strength of the labour movement also relies on feedback from the pursued policy. The power resource theory is typically associated with Walter Korpi (1983), Esping-Andersen’s old research colleague, and basically argues that the welfare state development is dependent on the strengths of the labour movement. In Esping-Andersen’s ‘The three worlds of welfare capitalism’ from 1990, the importance of institutions was given further attention. The attempt to link the three phases of early welfare state creation, welfare state expansion and welfare state restructuring is one of the major strengths of the presented argument. In methodological terms, the welfare institutions as an independent variable received serious attention. In order to understand the pattern of welfare state expansion and later also the pattern of welfare state restructuring, Esping-Andersen discusses how inherited welfare institutions through feedback processes – sometimes by accident and sometimes by design – tend to reproduce themselves. In his words, the ‘welfare states are key institutions in the structuring of class and the social order’ (EspingAndersen 1990, p. 55), and combined with the three regimes the idea is that ‘each case will produce its own unique fabric of social solidarity’. ‘One may cultivate hierarchy and status, another dualisms, and a third univeralism’ (Esping-Andersen 1990, p. 58). Thus, we have a stabilising or even self-perpetuating feedback process from welfare institutions at one point in time to welfare institutions at a later point time. As mentioned, the idea of path-dependent processes has become increasingly popular within welfare state research, but it covers a number of different feedback processes (Goul Andersen & Larsen 2002) that are often under-specified (Pierson 2000). The same holds for Esping-Andersen. Following the work of T. H. Marshall, who saw civil, political, and especially social rights connected to citizenship as a way to moderate the class conflicts generated on the market (Marshall 1950),3 EspingAndersen uses the terminology that the welfare state influences the stratification of society. But the further causal mechanism and the level of commitment to the old power resource theory (Korpi 1983) remains a bit blurred. Walter Korpi and his college Joakim Palme are more explicit as they ‘view institutions as intervening 3 It is also a classic thought within Marxism that the welfare state is a way to ‘hide’ conflicts and thereby legitimate the existing class society (Offe 1972; O’Connor 1973; Piven & Cloward 1971).

16

The Institutional Logic of Welfare Attitudes

variables that reflect conflicts of interest among different interest groups and affect the definitions of interests and coalition formation among citizens, which in turn have consequences for the size of budgets available for redistribution and the final degree of redistribution achieved’ (Korpi & Palme 1998, p. 682). Scholars committed to the power resource theory have especially focused on the potential for a coalition between the working class and the middle class. The ‘Designed positive’ feedback process has probably been the most pronounced among socialist policy makers as the generation of solidarity among workers has always been a fundamental dilemma for the socialist movement. Thus, the success of socialism in Scandinavia can, according to Esping-Andersen, partly be explained by a low degree of ‘ghetto-socialism’ (i.e. radical groups in which the movement itself became the welfare provider) and the ability to form a coalition with the small farmers who also argued for social benefits provided by the state and based on citizenship – at such a level that the middle class was covered (see also Baldwin 1990). The results became the Scandinavian universal welfare state that crystallised in the 1950s and 1960s. According to Esping-Andersen, the strength of the social democratic welfare state ‘was the consolidation of a vast popular majority wedded to its defence. “Middle-class” universalism has protected the welfare state against backlash sentiments’ (1990, p. 69). In contrast, the decision to stick to the principle of need in Anglo-Saxon countries made such a coalition between the working class and middle class unlikely. In the ideal type description of the liberal regime, all but the ‘needy’ must cover themselves through private savings or private insurance companies. However, in reality, a number of occupational insurance schemes were introduced through individual or collective bargaining. The result of this residual welfare state is, according to Esping-Andersen, ‘a curious mix of individual selfresponsibility and dualisms: one group at the bottom primarily reliant on stigmatising relief; one group in the middle predominantly the clients of social insurance: and, finally, one privileged group capable of deriving its main welfare from the market’ (Esping-Andersen 1990, p. 65). The point emphasised by scholars like Esping-Andersen and Korpi & Palme (see also Rothstein 1994, p. 179) is that the ability to form coalitions around the long-term class interests are dependent on previous choices and institutions. According to the power resource theory (with an institutional addition), we should expect the working class to hold very positive attitudes towards the welfare state, whereas especially the attitudes of the middle class will differ in a social democratic and liberal welfare regime. How class interests and coalition possibilities are influenced in the idealtype conservative regime is more blurred. This is one of the weak points in the power resource theory, which has been highlighted by, for example, Kersbergen (1995). Bismarck’s introduction of the first compulsory insurance schemes in the 1880s is often interpreted as a successful attempt to dampen the socialist ideas of the potential labour movement. Whether due to such strategic political decisions or due to the influence from religion, it is a fact that the Christ Democrats became the major force in the conservative welfare regime, and consequently many workers did not join the Social Democratic movement. If this picture of consensus in the

Welfare Regimes and the Attitudes of their Inhabitants

17

conservative welfare state is valid, we should expect the class division to be less tense. But whether this consensus should increase or decrease support for welfare policies that concern the poor and unemployed is difficult to infer directly. Without further theoretical discussion, the previous empirical studies expect the conservative regimes to fall in between. In Chapter 4 we will make a theoretical argument for this ‘in-between’ expectation, but it will not be deduced from class interests. Welfare regimes and short-term interests – the rational choice argument One can also formulate the process of institutional feedback as centred on the shortterm interest of individuals. This line of reasoning is very pronounced in the current literature on welfare state restructuring. According to Pierson (2001), the (new) politics of welfare states in economic lean times can be described as an exercise in ‘blame avoidance’ where the tension between an elite pursuing necessary reforms and a reluctant constituency forms a major conflict line in modern politics. The situation can be understood as a result of concentrated costs versus dispersed benefits. Those who would lose benefits/services or receive benefits/services at lower levels will experience a concentrated cost, whereas a slightly lower tax rate is a dispersed benefit (Wilson 1980). Thus, to help explain the possibility of retrenchment in Anglo Saxon countries such as the USA and Great Britain, the tendency towards more moderate reforms in Continental European and simple cost containment in Scandinavian countries, Pierson points to a variation in what he calls the welfare state clienteles. Measured as recipients of pension, disability pension, unemployment benefits and social assistance plus public service sector employees, the welfare state clienteles make up 57 per cent of the electorate in Sweden, 51 per cent in Germany and only 33 per cent in the USA (Pierson 2001, p. 413). In his 1999 book, Esping-Andersen seems to agree with Pierson as he argues that ‘the Nordic (and European) welfare states are so much more enshrined than is, say, the American. Leftists, righties, and the median voter all uphold the edifice because all benefit from it directly and personally’ (Esping-Andersen 2000, p. 147). In the period of economic growth and welfare state expansion, the argument can be turned around. The larger the group that is entitled to public transfers and services (and the more workers in the public welfare service sector), the stronger ‘concentrated’ political pressure for expanding the size of the budget (Kristensen 1980). Naturally, someone has to pay the bill, but following this logic the political influence of this issue will be limited because the costs are dispersed. A number of the previous studies reviewed in the next chapters also apply this theoretical reasoning, which is typically operationalised as an expected connection between recipients of benefits and positive attitudes towards the welfare state. We will not anticipate the findings, but simply reveal that the difference in welfare attitude between those who receive and those who (mainly) contribute is far from clear-cut.

18

The Institutional Logic of Welfare Attitudes

Welfare regimes and the (re-)production of values and norms The third mechanism, the reproduction of norms and values concerning equality, justice and solidarity, builds on the idea of homo sociologicus. In this perspective, the important forces behind attitudes towards social policy are neither class interest nor self-interest, but rather values and norms that are typically formed in youth. Thus, the attitudes towards social policy reflect the norms and values that individuals in a given society try to fulfil. In our opinion, there is no doubt that values and norm influence attitudes towards social policy, but the challenge is to explain how these norms and values are connected to welfare regimes. As described by Kelley & Evans (1993: note 10), many studies, including their own, use ‘culture’ as a residual or not investigated explanation. The same logic is applied by, for example, Andress & Heien (2001) who, after control for a number of variables, label the remaining difference between countries the effect of ‘dominant welfare state ideologies’. Sometimes it may be correct to refer to ‘culture’ or ‘dominant welfare state ideologies’, but it should at least be specified which cultural characteristic we think of, and how they can influence cross-national differences in attitude. To the latter request, we can imagine two ways in which ‘culture’ can influence the relationship between regime and public attitude. Culture could be the ‘major force’ that influences both the way a welfare regime was created and the attitudes of present populations, i.e. it is not the regime that influences welfare attitudes, but rather the culture in which the regime is embedded. Such a perception comes close the idea that the USA and Americans, both in terms of institutions and opinions, are embedded in a culture of liberalism. This has been argued in a number of publications by Lipset (e.g. ‘American Exceptionalism’ 1996). It has also been claimed that Nordic people share a special ‘passion for equality’ (Graubard 1986), which potentially explains both the existence of the social democratic welfare state regime and the attitudes of the citizens. However, if we take this ‘easy way out’ and explain the link between regimes and opinion with a culture of respectively Liberalism, Conservatism and Socialism, it seems that we are also obliged to answer the big question of how such a culture comes about and how it is reproduced. Esping-Andersen is not very implicit about such effects, but appears to recognise the influence from religion. Following the work of Weber and his successors, one could suggest that the regimes were ‘products’ of Calvinist Protestantism in Anglo-Saxon countries, Catholicism in Continental Europe and Lutheranism in Scandinavia. However, even if religion were a major force when the initial welfare schemes were implemented, we still need to specify how this culture keeps reproducing public attitudes towards social policy in the 20th and the 21st centuries. A belief in deterministic reproduction of culture through socialisation mechanisms is a possible solution, but that position seems more than hard to defend. Nevertheless, the main problem with the ‘master force thesis’ is that it is almost impossible to falsify. The other line of reasoning is that the welfare state regime at a given time, i.e. the institutional complex of welfare state, labour market and family, through socialisation

Welfare Regimes and the Attitudes of their Inhabitants

19

mechanisms, produce a ‘culture’ or ‘dominant welfare state ideology’ that either is in favour of, or against, welfare state policy. Thus, in this line of reasoning, ‘culture’ is an intervening variable between welfare state regimes and welfare attitudes (as implied in Figure 2.1). Feedback based on such theoretical assumptions also seems to be used by Esping-Andersen as, ‘… each case will produce its own unique fabric of social solidarity’ (Esping-Andersen 1990, p 58). In his 1999 book, Esping-Andersen a bit playfully replaces the homo economicus of traditional economic theory with three new species: homo liberalismus believes in ‘noble self-reliance, … and may be generous [but believes that] kindness for others is a personal affair, not something dictated from above. His ethics tell him that a free lunch is amoral … [and] prefers a welfare regime where those who can play the market do so, whereas those who cannot must merit charity’. Homo familius on the other hand thinks that ‘self-interest is amoral’ and that ‘a person will find his equilibrium when he puts himself at the service of his family … He is a satisficer, not a maximizer, because what really counts is stability and security’. The male variant is the male breadwinner who ‘wants a welfare regime that tames the market and exalts close-knit solidarity’. Finally, homo socialdemocraticus, is ‘inclined to believe that he will do better when everybody does better. Doing good is not an act of charity and can, indeed, be coolly calculative … [he/she] is fully convinced that the more we invest in the public good, the better it will become. And this will trickle down to all, himself especially, in the form of a good life’ (Esping-Andersen 2000, pp. 171–72). From these speculations one can deduce a number of norms and values we should look for – our second request to a specification of the culture explanation – but it remains unclear whether these new human species created the welfare state regimes or are the product of the welfare state regimes. Within the social justice literature, some studies have more thoroughly measured cross-national differences in egalitarian values and norms, and we will present some of the major findings in the next chapter. Within this tradition, we find a thesis of feedback or reproduction, as Homans is often cited for the argument that ‘the rule of distributive justice is a statement of what ought to be, and what people say ought to be is determined, in the long run, and with some lag, by what they find in fact to be the case’ (Homans 1974). Following the idea of de-commodification, belief in the justice of market inequalities might be the crucial cultural characteristic. This refers to the classic right–left ideological conflict. The right wing typically argues that the income distribution produced on the market is just because it reflects the motivation, initiative and effort of the individual. By contrast, the left wing argues that the income distribution produced on the market is unjust because there is no real equality of opportunities; because wages do not reflect the real value of the labour force; or because capitalism as an economic system produces a number of dysfunctional outcomes, e.g. unemployment. If attitudes towards social policy are a simple reflection of this left–right dimension, we should expect the acceptance of market inequalities to be low in Scandinavian, medium in Continental Europe and high in Anglo-Saxon countries. Whether this is the case will be discussed in our review of previous studies.

20

The Institutional Logic of Welfare Attitudes

Some objections and the call for a political man Esping-Andersen’s (1990) ‘The Three Worlds of Welfare Capitalism’ is already a classic and is often used as the point of departure in introductions to comparative welfare state research. The strength of the book is the capability to synthesise a large number of country characteristics, historical developments, and theoretical arguments into one overall framework. Thereby, the major weakness of the book is almost also given; namely the lack of theoretical clarity and consistency. Nevertheless, most of the academic debate has focused on the question of the numbers of welfare state regimes. Castles & Mitchell (1993) and Castles (1996) have argued in favour of a fourth Australian or Antipodean welfare regime, basically (1) because the income ceiling for eligibility of the need-tested benefits was drawn at middle incomes (not at poverty lines), and (2) because the so-called wage arbitration system was a successful social democratic strategy to produce welfare through regulation of the labour market. Esping-Andersen acknowledges that an Antipodean model might have provided a ‘package’ of welfare guaranties in the 1960s and 1970s that was more social democratic than the other liberal state regimes, but he persuasively argues (Esping-Andersen 2000, 89) that the passage of time (like the Thatcher period in Britain) has pushed Australia and certainly New Zealand closer to the ideal type liberal regime. Others have called for a fourth Mediterranean model, but much of the difference between North Continental Europe and South Continental Europe seems to be caused by differences in economic development and not in the basic characteristics. Ferrera (1996) argues that, in terms of social assistance, the South European countries actually have a distinct practice, but as argued by EspingAndersen one social scheme is only a very small part of what is meant by the label welfare regime. The same holds for Leibfried (1992), who argues that the Italian welfare schemes have served the purpose of political clientelism, and for others who find deviant cases on the program level. These differences do exist, and they are important for our understanding of the given welfare state (and in our analysis we will have the descriptions of deviant cases in mind) but – in our opinion – it simply does not add up to a persuasive argument for a distinct regime. And, quite frankly, the discussion of the number of regimes and the endless list of deviations from the three ideal types has not been very fruitful. At least is seems safe to conclude that a better account of cross-national differences in welfare states, labour markets and family structures has not been developed. Instead, the contribution of this book is to clarify, contest and reformulate the theoretical line of reasoning in the welfare regime theory in order to identity what could be the missing link between the three regimes and the citizens’ attitudes towards welfare. The above distinction between three theoretically different linking mechanisms found in previous studies is already a step in the right direction; at least it gives us a better basis for discussing why the empirical findings in previous studies often go against our expectation (see the next chapter). First of all, the idea that welfare attitudes are an outcome of long-term class interests can be questioned by the fact that the stability of the relationship between voters and the elite in recent

Welfare Regimes and the Attitudes of their Inhabitants

21

decades has decreased tremendously. Furthermore, one can question the very idea that the long-term self-interest of the middle class should be different in each regime. It is a pivotal thesis in the power resource theory that the level of redistribution is higher in the social democratic welfare state regime (or call it an all-encompassing welfare design) because it provides the well-off with a vested interest in the maintenance of the welfare schemes, and therefore provides the ‘bad risks’ with a strong ally in their lobbying for better benefits and services (Korpi 1983; Korpi & Palme 1998). But following Tullock’s (1984) criticism one can basically argue ‘that redistribution is redistribution’, i.e. if the programmatic structure of the social democratic welfare schemes causes more redistribution, it is not in the class interest of the well-off (see also Overbye 1996). Therefore Tullock argues that it is the lesser visibility of the redistribution that enables political elites in the social democratic regimes to create more redistribution than they expect the majority to accept. The level of strategic thinking in the policy elite can be discussed (e.g. Albrekt Larsen & Goul Andersen 2004 on the Danish case), but basically we find the visibility argument valid. Furthermore, we show in Chapter 4 that the visibility of net-winners and net-losers in the different welfare state regimes is important in order to understand the public judgement of deservingness. So far, the main point is that it is not so self-evident that substantial cross-national differences in welfare attitudes should be caused by differences in the interests of the middle class; especially not when we speak of welfare policy in the narrow or American sense. If the general public opinion is less influenced by long-term class interest, it may seem ‘natural’ to pay more attention to the influence of the short-term selfinterest. The academic shift from the power resource discussion in the 1980s to the new politics discussion in the 1990s can actually be understood in those terms. The shift fits nicely with the general intellectual criticism of today’s citizens and the idea that the argument of concentrated costs versus dispersed benefits actually enable us to understand a number of present events. One of the most illustrative areas is the political struggle over pension reforms. Especially in conservative welfare regimes, we have seen a mobilisation around the concentrated costs of pensioners. Why it has been so legitimate for pensioners to organise around their self-interest can be explained by the fact that they fulfil almost all relevant deservingness criteria, which we will discuss further in Chapter 4. At this point, we will argue that we also have situations where groups with vested interests end up accepting reforms. This was for instance the case in Denmark in the mid-1980s, where the Danish bourgeois government in the wake of the economic crises brutally lowered the income for most people (named the potato cure), but got public acceptance of the measure and even managed to get re-elected shortly after. Actually, the overall impression is that the Western welfare states despite concentrated costs exhibit ‘declining resistance to change’ (Kersbergen 2000), are ‘restructured’ or ‘recalibrated’ (Ferrera & Rhodes 2000; Sharpf & Schmidt 2000; Pierson 2001); even when it comes to pension schemes, Hinrichs (2000) argues that ‘elephants (are) on the move’. Thus, framed around a political discourse of necessity (Schmidt 2002) and the common good, it seems possible to replace ‘economic man’ with a more ‘political man’.

22

The Institutional Logic of Welfare Attitudes

Normally ‘sociological man’ is used as the contrast to ‘economic man’, but both traditions actually share a rather mechanic view of attitude formation; at least if we stick to the causal reasoning suggested in Figure 2.1. Internalised values and norms of equality and solidarity often coincide with support for welfare policy, but it is not always a stable relationship. Gilens (2000, p. 2), for example, argues that the lack of public support for welfare policy in the US is not merely a reflection of individualism, as most Americans actually believe ‘that government can and should play a central role in providing both the means for individuals to better themselves and a cushion for times when individuals’ own efforts are not enough’. Further, the next chapter will show that besides the methodological problems of the culture argument, it is actually not that easy to find cross-national differences in values and norms that can explain the gross-national differences in support for welfare policy. Thus, at the overall level we argue that the previous theoretical links between the macro-level of welfare state regimes and the micro-level of attitudes towards public policy suggest a too mechanic formation of attitudes towards welfare policy; (1) within the power resource theory, the individual is believed to follow the interest of her/his class, (2) within the theory of concentrated versus dispersed benefits and cost, the individual is believed to follow her/his short-term interest, (3) and within the theory of sociological man, the individual is believed to reproduce the norms and values inherited through socialisation. In contrast, we suggest a political man theory, where the individual’s attitudes towards public policy are open to discussions of the common good, justice, necessity, suitability in relation to the experienced reality. In doing so, we follow political science researchers like Bo Rothstein (1994) and Jørgen Goul Andersen (forthcoming), and the old idea that attitudes towards public policy are partly a result of a public debate. As the Western societies get richer and richer and the educational level keeps rising, one could further argue that such a reflective constituency becomes more and more likely. Actually, this idea of a context-dependent and moveable political man is also consistent with the fact that welfare attitudes captured in surveys are very dependent on the framing of the questions (e.g. Goodin 1986; Kangas 1997). If welfare attitudes were determined strictly by class interest, self-interest or internalised values and norms, we should expect greater stability, e.g. less context dependence. For example, Kangas reports that confronted with the statement ‘there should be a work tax with which the employed would support the unemployed’, only 28 per cent were in favour. Confronted with the statement ‘Unemployment causes great difficulties to families. Therefore there should be a work tax with which the employed would support the unemployed’, the share in favour increased from 28 per cent to 40 per cent (Kangas 1997, p. 487). Thus, specifying a special perception of reality, namely that unemployment causes great difficulties for families, forces the political man to take a new stand on the question. Instead of worrying about this inconsistency and question the methodological foundation of survey research,4 we take these 4 ‘Sceptical scholars’ (e.g. DeMaio 1984; Overbye & Eia 1995) would argue that survey responses are basically a matter of positive self-representation, i.e. the answers serve

Welfare Regimes and the Attitudes of their Inhabitants

23

findings as an argument for being more concerned about the perceptions of reality produced by welfare regimes than the effects on class interest, self-interest and basic egalitarian values. In order to develop causal reasoning in this field, we find Aalberg’s (2003) distinction between ideals, perceptions, and policy attitudes very relevant. Ideals refer to abstract norms, values and principles and rather lasting ideas of what is desirable. Perceptions refer to more specific observations of reality that are more likely to change and are not necessarily normative. Finally, policy attitudes refer to normative beliefs about specific situations and policies. In Chapter 4, we argue that the regime effect on policy attitudes towards welfare policy is primarily caused by the regime effects on public perceptions of the poor and unemployed.

to present the respondent as a reasonable, good and attractive person for himself and others. Such a ‘social desirability bias’ is definitely present in many answers, but it does not help us explain the substantial cross-national differences in welfare attitudes.

This page intentionally left blank

Chapter 3

The Puzzle of Public Opinion The number of previous studies that investigate the link between welfare state regimes and public support for welfare policy is rather limited. This is partly due to a lack of interest in a theoretical micro-foundation, but the primary reason is a lack of cross-national comparative data. As mentioned, Korpi & Palme (1998, p. 682) argue that the ‘the empirical testing of the macro-micro-links among institutions and the formation of interest and coalitions provides a major challenge for social scientists, but comparative micro-data currently are lacking’. It is true that we do not have a cross-national survey module that deals exclusively with attitudes towards welfare policy, but the more general surveys in ISSP (International Social Survey Programme), World Values Studies, ESS (European Social Survey) and Eurobarometer offer relevant items. In particular, the ISSP data have been used to test the connection between welfare regimes and support for welfare policy. Social science rarely produces clear-cut results, and scholars devoted to empirical research will gladly admit that the real world is a fuzzy place. Nevertheless, the previous analyses of cross-national differences in welfare attitudes conducted on the available data not only deliver blurred results, but, as mentioned, they also deliver puzzling results that challenge a number of the theoretical arguments suggested in the previous chapter. Based on a review of these studies, this chapter has three major arguments. First, is has been difficult to find a clear-cut pattern between welfare state regimes and welfare attitudes, but if we take the items that (primarily) measure support for welfare policies that concern the poor and unemployed, a regime effect actually seems to present. Secondly, the mechanisms of class interest and selfinterest have been very difficult to find; both in studies that measure welfare policy in the narrow and broader sense. Thirdly, this leaves us with the problematic ‘culture explanation’, but actually the few comparative studies of egalitarianism do not find a regime pattern. By means of simple descriptive analyses of the little-explored ISSP data on public attitudes on the ‘just salary’, we show that in terms of difference between skilled workers and elite occupations, the pattern does not coincide with the regime pattern of public support. Measured this way the conservative welfare regime is – as expected – most anti-egalitarian. However, measured in terms of the just difference between unskilled and skilled workers, we do find a pattern that coincides with the pattern of public support for welfare policy. The findings that underpin these three arguments are presented in this chapter. But first of all we have to address the question of validity of cross-national survey findings.

26

The Institutional Logic of Welfare Attitudes

The validity of cross-national survey findings There are (as always) many problems with the survey data used in previous studies as well as in this book. One problem in cross-national studies is that the same question may not mean the same thing in different countries. The ISSP has set up a procedure where a steering committee of national experts discusses wording and translation, which reduces – but definitely does not eliminate – such problems. As an example, let us take the (relatively good) question from the ISSP role of government module: ‘On the whole do you think it should be or not be the government’s responsibility to reduce income differences between the rich and the poor?’ Within a Scandinavian and Anglo-Saxon context, it may be obvious what is meant by ‘the government’, but for Continental European respondents it may be unclear whether compulsory insurances – the major part of welfare policy in these countries, typically run by the labour market organisations – belong to the government or not. Respondents in countries with federal structures have another problem: Are we talking about the federal level or the national level? Furthermore, the terms ‘poor’ and ‘rich’ might refer to different groups. In an Anglo-Saxon context, the concept of the poor is well established (intense public discussions use the poverty category), whereas the concept is less familiar to the European respondents. Finally, even though the question is posed in absolute terms (‘should it be or not be the government’s …’), we cannot rule out the possibility that respondents think in relative terms, i.e. should the government do more or less. If this is the case, we have the problem that respondents living in countries with a high degree of redistribution reach a saturation point and therefore answer more negatively than they ought to from a comparative perspective. The opposite effect may be present in societies with very little redistribution. Here respondents might answer more positively. The problems with different meanings in different countries are increased by the fact that we only have a small number of relevant items that measure attitudes towards welfare policy, i.e. problems with one or two indicators become much more serious. Furthermore, these items vary from one survey to another, which makes comparisons difficult. Often, scholars simply take the items available within a given survey and merge them into a one-dimensional index of ‘welfare attitudes’, and end up measuring different dimensions of welfare attitudes and therefore find different results. We know from a number of studies that public support varies tremendously across programmes. Coughlin – being the pioneer in this field – concluded that the public is most in favour of redistribution to old-people, followed by the sick and disabled, needy families with children, the unemployed and finally people on social assistance (Coughlin 1980; see also Goul Andersen 1982; Petterson 1995). Therefore, our review pays close attention to the items behind the dependent variables, which actually lead to the interesting finding that it is, especially in terms of welfare policy, in the narrow sense that attitudes are regime-dependent. In addition to validity problems, there is a reliability problem caused low response rates in some countries (for a further introduction to the data used in this book see Box 3.1).

The Puzzle of Public Opinion

27

These methodological problems have pushed the research field in different directions. One direction is to use more sophisticated statistical methods, e.g. structural equation models, on the available data. In particular, the use of confirmatory factor analyses within these studies is promising, as it enables us formally to test whether the indexes measure the same thing in each country (e.g. Gelissen 2002). Another direction is to collect new cross-national data as the small numbers of items available in previous surveys put hard restrictions on even the most advanced statistical models. A third direction is to accept that the puzzling findings might not only be due to methodological problems. We follow the latter line of reasoning and argue that the previous studies actually deliver findings that highlight the need for a new theoretical link between the macro-level and the micro-level. Box 3.1.

Introduction to the survey material used in the book

World Values Survey (primarily used in Chapter 5) The World Values Survey (WVS) may be described as a series of intercultural studies based on interviews (primarily face to face) with representative population samples (above 18 years old) from some 80 countries in the latest wave. The programme is directed by Professor Ronald Ingelhart. The first wave was conducted in 1990–93, the second wave in 1995–97, and the third wave in 1999–2001. We primarily use the first wave. The second wave includes a smaller number of the countries we are interested in, and the response categories to the question we analyse are problematic (see Chapter 5). The third wave does not ask the question (expect in the European countries). In the first wave we end up with 16 rich Western countries of interest, where the data collection was conducted by professional research institutes. We hereby avoid many of the problems related to a real random sampling procedure and low response rates, which haunt the surveys from (economic) less developed countries in the WVS. It should also be mentioned that the World Values Survey historically started as an addition to the European Values Survey. Further documentation can been found on ICPSR’s webpage.

Eurobarometer (primarily used in Chapter 5) The Eurobarometer was established in 1973. Each survey consists of approximately 1000 primarily face-to-face interviews (with persons above 18) per member state (except Germany: 2000, Luxembourg: 600, United Kingdom 1300 including 300 in Northern Ireland). We use four Eurobarometer surveys, which all posed the same question. The four cross-cuts are from 1976 (number 5), 1989 (number 31A), 1993 (number 40) and 2001 (number 56.1). The surveys were conducted by professional research institutes based on representative samples. Furthermore, we have used the standard weights provided in the data set, except for 1976 where no weight was provided. For further information see the codebooks provided by Eurobarometer.

The International Social Survey Programme (primarily used in Chapter 6) The International Social Survey Programme undertakes annual surveys in – at the moment (2004) – 39 countries. The samples are probability-based nationwide samples containing grown-ups (above 18). The sample size is well above 1000 (with

28

The Institutional Logic of Welfare Attitudes few exceptions). The topics change from year to year with a view to replication approximately every five years. In the following chapters we will mainly use the ISSP module on social inequality from 1992 and 1999. In 1992, 18 countries participated, 10 of them of interest to us. In 1999, 27 countries participated, 12 of them of interest to us. Even in these (economically) developed countries, the response rate is sometimes not too good, i.e. below 60 per cent. Among the countries we use from the 1992 module, Great Britain and West Germany have low response rates, and the response rates from Austria and Italy are not available. Documentation for the 1999 study is not yet available (2004). The codebook for the 1992 module is available on the GESIS webpage. In Chapter 6, we use additional questions to ISSP’s 1999 survey in Denmark, Norway and Sweden. These data are not available from the international data archives, but were accessed through national data archives and then merged into one file. As mentioned, the codebook is not yet available, but the response rates are believed to be around 60 per cent (the level in Denmark, Norway and Sweden in the 1998 sample).

The Nordic Citizenship, Marginalisation and Unemployment Data (used in Chapter 7) In Chapter 7 we also use two samples of long-term unemployed from Denmark and Finland. The Danish sample is based on a gross sample of 1626 persons, randomly selected from administrative registers in autumn 1998 among recipients with a minimum of 3 months of previous unemployment. 1034 interviews were obtained. This a decent response rate at 64 per cent. However, as there were six months between sampling and interviewing, only 456 respondents were still unemployed when the interview took place. These 456 persons constitute a representative sample of persons with an unemployment period of at least nine months prior to interviewing. The Finnish sample is drawn from the Ministry of Labour’s register among those who were registered as unemployed job seekers in the last quarter of 1999. Another criterion was that of an unemployment period of at least 92 days. The maximum duration of unemployment was not fixed. Among the 1995 drawn, the response rate was 40.4 per cent, which is quite low. However, a closer analysis of the drop-out does not show a heavily biased sample (Saurama, 2001). The survey material is part of the project ‘Unemployment, marginalisation and citizenship: The Danish welfare state in a comparative perspective’ and a Nordic project on ‘Unemployment, early retirement and citizenship: Marginalisation and integration in the Nordic countries’. For a full project description and detailed data information, see the CCWS homepage (www.ccws.dk) (direct: www. socsci.auc.dk/ccws/projects/BChangingLabourMarkets/b1.htm).

The national sample (Chapter 8) Finally, Chapter 8 is based on an additional module to the Danish ISSP 2004 sample. As with the other Danish ISSP sample, it is a nationwide sample that includes 2000 adults (above 18). The fieldwork was conducted by The Danish National Institute for Social Research, which is the leading institute for Danish research surveys. The response rate was 58 per cent and there does not seem to be any substantial bias in the sample. The questionnaire is available through the author (but will later, in the normal ISSP procedure, be stored at the GESIS).

The Puzzle of Public Opinion

29

Welfare regimes and cross-national differences in public attitudes towards welfare policy Most studies connecting welfare state regimes and cross-national survey results use an index of the ISSP items that measure preferences for a range of governmental responsibility, i.e. they look for gross-national variation in the level of welfare state support. Based on a rather simple and rigid understanding of the regime theory, the previous studies expect (without much discussion) support for welfare policy to be highest in the social democratic regimes, lowest in the liberal regimes and somewhere in-between in the conservative regimes. Walter Korpi did defend a rather one-dimensional understanding of welfare states in the 1980s, and Esping-Andersen did measure the welfare state regimes on a single dimension of de-commodification in 1990. However, at least in the latter case, the operationalisation in the empirical studies does not earn enough credit to the theoretical debt of the regime theory (for a similar criticism see Bonoli 2000). Nevertheless, if we follow the arguments presented in the previous chapter, there are reasons to believe that differences in class interest and coalitions, differences in self-interest and differences in egalitarian values could produce large support in the social democratic regime, medium support in the conservative regime and low support in the liberal regime. The items on the principal responsibility of the government are both found in the ISSP ‘role of government’ module and in the ISSP ‘social inequality’ module. The studies that do not find the expected regime pattern are primarily based on the role of the government module. Using seven items (‘provide job for every one’, ‘provide healthcare’, ‘provide decent standard of living for old’, ‘provide decent standard of living for unemployed’, ‘reduce income differences’, ‘provide financial help to college students from low income families’, ‘provide decent housing’) from 20 countries, Arts & Gelissen (2001) distinguish between three groups of countries; (1) the immature welfare states of Spain, Ireland and Italy, (2) the pure social democratic welfare state of Norway and Sweden plus France and Great Britain, and (3) the liberal regimes of the USA, Australia, Canada, New Zealand plus Japan, Philippines and West Germany. Statistically it is an excellent study, but logically we see the problem with measuring the dependent variable on a one-dimensional scale (or call it a latent variable). Both logically and based on Couglin’s and others’ empirical findings, ‘provide decent standard of living for the old’ and ‘reduce income differences between the rich and the poor’ are very different things.1 1 The exact wording of the question was: On the whole, do you think it should be or should not be the government’s responsibility to: a) provide a job for everyone who wants one? b) provide healthcare for the sick? c) provide decent standard of living for the old? d) provide decent standard of living for the unemployed? e) reduce income differences between the rich and the poor? f) give financial help to university students from low-income families? g) provide decent housing for those who can’t afford it?

30

The Institutional Logic of Welfare Attitudes

Bean & Papadakis (1998), using the role of government module from 1985 and 1990 from six countries, are better at keeping the items apart, but basically reach the same rather blurred pattern. On the item ‘provide a job for everyone’, the expected pattern is found, USA followed by Australia, Britain, West Germany, Italy and Norway; on the item ‘decent standard of living for unemployed’, USA and Australia could be distinguished from the others, whereas no real differences were found on the items ‘provide healthcare for the sick’ and ‘provide decent standard of living for the old’. Furthermore, Bean & Papadakis, more or less as the only ones in the field, used the items of attitudes towards government spending which, however, also failed to create the expected regime differences. Finally Gelissen (2000) has analysed the EU countries where he measured attitudes towards the extensiveness (‘social security is a major achievement…’, ‘Government must continue to provide everybody a broad range of social security benefit’, ‘the right to claim assistance from the authorities should be guaranteed ’) and attitude towards the intensity (‘Social security is too costly ...’, ‘should only provide a limited number of essential benefits ...’, ‘government should provide healthcare to everyone’). He concludes that ‘no evidence was found for the thesis of there being a relationship between the type of welfare state, as defined by Esping-Andersen, and levels of popular support for the welfare state’ (Gelissen 2000, p. 298).2 Gelissen (2002, chapters 5 and 6) has also made separate analyses of the health and pension area. In the field of health care he finds some effect from living in social democratic regimes whereas ‘welfare state regimes only marginally affected the level of support for government responsibility in the field of old-age pensions’ (Gelissen 2002, p. 201). However, a number of other studies primarily based on the ISSP social inequality module actually find substantial cross-country differences that, to some extent, follow the pattern suggested by Esping-Andersen. By means of a structural equation measurement model based on three items (‘provide job for all’, ‘provide everyone a guaranteed basic income’, ‘reduce income differences’) and four countries from the ISSP 1992 module, Andress & Heien (2001) confirm the expected ranging with USA 2 The exact wording of the questions was: a) Social security is a major achievement of modern society. The government should make sure that nobody is left deprived when unemployed, poor, ill or disabled. b) The government must continue to provide everyone with a broad range of social security benefits even if this means increasing taxes and contributions. c) Some people cannot earn enough income to live on, because of disability, old age etc. To what extent do you agree or disagree that the right to claim assistance from the authorities should be guaranteed? d) Social security is too costly for society. Benefits should be reduced and contributions should be lowered. e) The government should provide everyone with only a limited number of essential benefits (such as healthcare and minimum income) and encourage people to provide for themselves in other respects. f) The government should provide everyone only with essential services such as care for serious diseases, and encourage people to provide for themselves in other respects.

The Puzzle of Public Opinion

31

at the bottom, West Germany in the middle and Norway at the top (East Germany as a former communist country above Norway). Studying six countries, Heien & Hofäcker (1999) using more or less the same items as dependent variables confirm the position of Norway, the USA, West Germany and East Germany, but find Italy and Britain to be deviant cases (both too much in favour of intervention). Analysing eight countries using a simple index for the same three items, Svallfors (1997) distinguishes between Norway, Sweden, Germany, Austria on the one hand and Australia, New Zealand, Canada, the USA on the other hand. Using cluster analysis on these three items, but including the whole range of ISSP countries, Evans (1995) finds three clusters: (1) the former communist countries plus Norway and Italy, 2) West Germany, Britain, Sweden, Netherlands, Austria, and (3) Australia, Canada, New Zealand, Switzerland and the USA.3 Thus, there is a clear divide between USA, Canada, New Zealand, Australia and other western countries where the internal pattern is a bit blurred. However, if we limit our cases to Norway and West Germany, we find a division between the social democratic and the conservative regimes. Furthermore, we can use a number of ad hoc explanations to explain the deviant cases of Sweden, Britain and Italy. The relatively low support in Sweden could be caused by the fact that Swedes during the 1990s experienced the worst economic crisis since the Second World War. The relatively high support in Britain could be a public reaction to the harsh policy pursued by Thatcher during the 1980s and the uncertainty about whether Britain really belongs to the liberal regime or is what Esping-Andersen labels a ‘stalled social democratization’ (Esping-Andersen 2000, p. 87). Finally, the relatively high support in Italy could be caused by its status as a latecomer, and we may even have to operate with a distinct Mediterranean regime (even though we took the sceptical position in the previous chapter). The latter group of confirmatory studies combined with the ad hoc arguments for the deviant cases leads to the conclusion that regime-dependent attitudes might be present at least on one dimension, namely the dimension captured by the items ‘provide job for all’, ‘provide basic income’ and ‘redistribution between those people with high incomes and those with low incomes’. The common denominator of these items is, in our opinion, the focus on public policy targeted at the poor and unemployed of working age. Admittedly, this cannot be directly inferred from the items, but we will substantiate this interpretation later.

3 The exact wording of the questions was: ‘Please show how much you agree or disagree with each statement…’ a) ‘The government should provide a job for everyone who wants one’ b) ‘The government should provide everyone with a guaranteed basic income’ c) ‘It is the responsibility of the government to reduce the differences in income between people with high incomes and those with low incomes’

32

The Institutional Logic of Welfare Attitudes

Welfare state regimes, class interest and self-interest. The puzzle of the intervening variables One thing is to find a match between the welfare regimes and support for welfare policy, another to demonstrate the mechanisms that lead to these cross-national differences. Following Andress & Hein (2001) we can distinguish between two effects. Firstly, a higher level of public support could be caused be the relative size of the group of ‘welfare supporters’ within different countries. In all countries, the dependants of welfare state transfers/services and public employees are believed to support the welfare state (if we take the self-interest mechanism), and as the size of these groups varies substantially across countries, this could explain differences in the aggregated level of support. To use the statistical terms, this compositional effect would also be influenced by the level of unemployment and demography (as especially the old are more dependent on the welfare state). Secondly, higher levels of public support could be the results of ‘the same groups’ holding different attitudes due to the strength of (left-wing) political mobilisation and the character of the institutional settings in each regime. Following the power resource theory, especially the attitudes of the middle class should be crucial for this so-called interaction effect. In the social democratic regime, the (strongly mobilised) workers have an alliance with the middle class around the welfare state; in the liberal regimes, the middle class joins the upper class so the (weakly mobilised) working class is on its own. In the conservative regimes, the theoretical expectations from the power resource theory are a bit blurred. From the discussion about Bismarck’s strategic attempt to prevent class-conflicts by means of compulsory social insurances, we could expect class to be less important for welfare attitudes in the conservative regimes. Another thesis within the empirical studies is that women, who in general support welfare policy, might be more positive in social democratic regimes as they benefit tremendously from the institutional setting that provides them with both jobs and possibilities to leave the home (Svallfors 1997), i.e. women hold different attitudes within different institutional settings. All studies presented in the previous section also looked for differences in individual support for welfare state policy, but the task of explaining country differences by such individual level variables often remains unsolved or blurred. As to the group of studies that do not find the expected cross-country differences, one could say that there is not much to explain. Anyway, we will summarise some of the findings. Arts & Gelissen (2001) do not compare differences within countries, but summarise the analyses in a multilevel regression model. The inclusion of individual level variables (education, employment status, household income, sex and age) does not reduce the regime effect much, i.e. the compositional effect seems to be modest. So the difference found between liberal regimes, on the one side, and social democratic regimes and Mediterranean countries, on the other side, remains stable. Gelissen (2000) uses the same method and finds more or less the same result; strong and significant cross-country variation in spite of the introduction of individual level variables. Bean & Papadakis (1998) analyse the influence of class in different

The Puzzle of Public Opinion

33

countries more thoroughly and find a rather strong relationship between positive attitudes towards the unions, left party identification and their ‘welfare attitudes index’. This supports the power resource theory. However, the actual occupational position of individuals (the objective class variable) had very little impact on their dependent variable – which probably has to do with the fact that subjective class variables are in the regressions (for a similar critique see Svallfors 1999, p. 97). Nevertheless, they did not find the middle class effect by comparing Norway and the USA. Actually the difference between the working class and the middle class was rather similar in the two countries. Finally, they conclude that ‘attempts to link support for the welfare state with transfer classes once again failed to bring out any consistent pattern among different regimes’ (Bean & Papadakis 1998, p. 231). As for the group of studies that, to a larger extent, find the expected regime pattern, i.e. studies that use the items ‘provide job to all’, ‘guarantee basic income’ and ‘redistribution’, we have higher expectations of significant compositional and interaction effect. By means of structural equation models on the most narrow range of countries – Norway, the USA, West and East Germany – Andress & Heien explicitly try to distinguish between these two effects. Controlled for differences in the mean of the independent variables, the differences between the countries are not altered remarkably, which again leads to the conclusion that the compositional effect cannot explain the cross-national differences in support for welfare policy. Allowing the independent variable to vary both in terms of mean and strength within countries, they found the differences between countries reduced, which leads to the conclusion that interaction effects are important. However, for strange reasons they did not include a class variable. They do have a measure of (subjective) position on a scale from ‘top’ to ‘bottom’ of society, but they do not discuss the crucial middle class effect at all. Finally, Andress & Heien (2001, p. 352) conclude that ‘antiegalitarianism, support for the need principle, education and income emerge as the individual-level attitude predictors with the strongest cross-country variation’, but they do not discuss which theoretical arguments these findings at the individual level support. Anyway, the tables (Andress & Heien 2001, p. 350) show that the subjective ‘top-bottom-scale’ has more or less the same strength in all four countries, which runs against the expectation of the class mobilisation thesis.4 Other studies using the same items discuss the class interest and self-interest thesis more thoroughly, but arrive at a sceptical position. In his eight-country study, Svallfors finds the unemployed to be more in favour of redistribution (as do Gelissen 2000; Andress & Heien 2001), but according to Svallfors (1997, p. 294), ‘the most striking results is the great similarity we find in cleavages structures’ so ‘the various arguments that different regime types tend to create different cleavage structures in the population receives very little support’. The expected difference between 4 They also found support for the theses ‘such as the class cleavage in the United States, the market cleavage in West Germany [and] the gender cleavage in Norway’ (Andress & Heien 2001, p. 352), but as argued in a previous chapter (see Footnote 2), it remains unclear how these conflict lines are supposed to influence the level of public support.

34

The Institutional Logic of Welfare Attitudes

public and private was also hard to find. Svallfors concludes that ‘suggestions about private–public sector employment constituting a new important fault line also receives precious little support. In most countries attitudinal differences between private- and public sector employees are simply non-existent’ (Svallfors 1997, p. 294). Evans, who analysed the USA, Britain, Australia, Switzerland, West Germany, Austria, Italy and the Netherlands, also concludes that ‘in most cases the countries are so similar in their patterns of association, especially those between class and welfare attitudes, that uniformity is the most sensible interpretation’ (Evans 1995, p. 202). If class mobilisation and coalition building is the mechanism behind crossnational differences in public support for welfare policy, we should expect the exact opposite conclusion. So far, the review of previous studies leads us to conclude that, measured in certain ways, the expected country pattern is found, but the mechanisms of class interest and self-interest – believed to cause these differences in aggregated support for welfare state policy – seem difficult if not impossible to find.5 It brings examples from the introduction courses to statistics to mind. As in the case of the almost perfect correlation between decline in arctic foxes in Finland and divorces in France, we have a situation where we are unable to explain the causal mechanism. If the regime-specific pattern is not caused by differences in class interest and self-interest, we have to search for another link between the macro-level and the micro-level; at least when we speak about welfare attitudes that concern ‘providing a job to all’, ‘guaranteeing basic income’, and ‘redistributing from rich to poor’. Our theoretical proposal for an alternative link is presented in the next chapter, but first we need to review the comparative studies of values and norms, i.e. the third link discussed in the previous chapter. Welfare regimes and cross-national differences in values and norms The above-reviewed studies that discuss the inability to explain the (sometimes) substantial cross-national differences in support for welfare policy often turn to the notion of ‘culture’ or ‘dominant welfare state ideology’. We have already argued that reference to ‘culture’ as a residual explanation for cross-national differences in support for welfare policy is problematic, or at least not ambitious enough. To substantiate the argument we need to explain which kinds of cultural characteristics, i.e. values and norms, we are looking for, and how they are believed to be reproduced. The empirical studies that come closest to fulfilling such criteria are analyses of justice belief; again primarily based on the ISSP Social Inequality module. We will limit ourselves to a short presentation of these studies of ‘what is’ and leave the vast 5 For a very interesting country study see Goul Andersen. Based on Danish data, he shows that even if we delimit the analyses to the group of private employees without own or family experience of unemployment, there is no majority against the welfare policy in general. He also shows that support for increased public expenditures in a given area is not higher among employees working in this area (Goul Andersen, forthcoming).

The Puzzle of Public Opinion

35

theoretical literature of ‘what ought to be’. Furthermore, we concentrate on crossnational survey studies and for the moment leave the number of mostly American survey studies (for an overview see Scherer 1992) and experimental studies (for a good overview see Miller 1992) behind. Having reduced the social justice tradition in such a cruel manner, we end up with a very limited number of studies. The standard question of ‘income differences being too high’ is found in the ISSP social inequality module, the ISSP role of government module and in a number of Eurobarometer surveys. This item could tap to what degree the populations have faith in a just income distribution provided on the market, but the perception of excessive income differences naturally depends on the actual differences in disposable income. And we know that the income differences vary substantially in Western societies. Therefore, it is uncertain what we should conclude, e.g. from the finding that not even a difference between Americans and Scandinavians can be found on this item (Evans 1995, p. 195). In the latest ISSP inequality module (1999), around 65 per cent of the Americans agree that income differences are too high, compared with around 70 per cent Norwegians and Swedes and 59 per cent Danes. Based on Eurobarometer surveys, Inglehart even found that Danes showed the least support among 11 European countries for a ‘greater effort to reduce income inequality’, which he goes on to explain in the following way: ‘… in countries like Denmark [the classic welfare state policies] have largely solved the problems they are capable of solving most readily – and have thereby reduced the demand for more of the same’ (Inglehart 1997, p. 257). His point is that measured in this way, the belief in equality provided on the market is influenced by the actual income differences produced on the market and the degree of redistribution conducted by the state.6 One could also argue that the question of excessive income differences basically refers to a left–right ideological conflict that does not capture the multidimensional nature of the welfare regimes. Another relevant justice belief could be the weighting of merit, need and equality, which has been studied in several experiments. The principle of need corresponds with the basic principle in the liberal welfare state regime, the principle of merit with the basic principle in the conservative welfare state regime, and the principle of equality with the basic principle in the social democratic welfare state regime. Using the European Value Study from 1999, Arts & Gelissen (2001) try to operationalise the concepts of equality (‘a just society should eliminate large inequalities in income among citizens’), need (‘a just society should guaranty that basic need are met for all…’) and merit (‘a just society should recognize people on their merits’), but the differences were rather small and inconclusive.7 6 Still it is a mystery that we do not find a majority in any of the countries that considers the present income distribution just. For a good analysis of the impact of framing when it comes to equality see Kangas (1997). 7 Using a multi-level regression analysis Arts & Gelissen actually find the social democratic countries (defined in this study as Denmark, Sweden, The Netherlands and Austria) to be less in favour of the principle of equality, and Mediterranean countries (Greece, Portugal, Spain, Italy) to be more in favour of this principle than Britain and Ireland (used as a liberal reference category). West Germany could not be distinguished from Britain and

36

The Institutional Logic of Welfare Attitudes

However, the ISSP inequality module also includes items that measure what people perceive as a just wage for different occupations. This is probably the best measure of egalitarianism as the professions are relatively clearly specified, i.e. the respondents think of more or less the same thing in each country (e.g. a skilled factory worker), and it is more or less independent of the actual income distribution. Unfortunately, these items have not been explored very much, so after reporting the main findings from the few studies, we will conduct some basic descriptive analyses ourselves. First of all Kelley & Evans (1993) have conducted a good study, which is based on nine nations within the ISSP module on inequality from 1987. Their main finding is that there is little cross-national variation in perception of the appropriate wage of low status jobs (they use unskilled factory worker, skilled factory worker, bus driver and farm worker), while there is much more variation in the perception of the appropriate wage for elite occupations (they use doctor, cabinet minister and chairman of a large national firm). When the salary that the low status jobs ought to be paid, seen in relation to the actual income of unskilled workers, is seen in relation to the salary that the high status jobs ought to be paid (again seen in relation to the actual income of unskilled workers), they find a range with Americans at one end and Poles at the other. Americans thought that the high status group should earn almost four times that of unskilled workers. Americans are followed by Austrians, Germans, Dutch, Swiss, Britons, Australians, Hungarians and then finally the Poles. The Poles think that the high occupations should earn below twice as much as unskilled workers (Kelley & Evans 1993, p. 98). So far, the liberal regime closest to Esping-Andersen’s ideal type is at the one end, but it is clear that the study does not find a cluster of liberal regimes. Actually the Germans – living in the country closest to the ideal-type conservative regime – accept larger income differences than Britons and Australians. Furthermore, Kelley & Evans arrive at a quite different ranking when they control for compositional effects concerning sex, age, income, class, party affiliation and previous communist country. After this control procedure, Austrian respondents indicate the highest differences in the ‘right’ wages. Austria is followed by Germany, The Netherlands, Switzerland, the USA, Britain, Australia, Hungary and Poland (Kelley & Evans 1993, p. 114). This leads to the conclusion that ‘American political exceptionalism is less evident once structural differences between the United States and Western Europe are taken into account’ (Kelley & Evans 1993, p. 114). Their

Ireland. But again the present degree of redistribution could influence the results, even though the question – in contrast to the question of income differences being too high – is asked in principle terms (‘a just society should ...’). On the item of need, social democratic countries favoured applying this need principle significantly less than the liberal countries, whereas the conservative and Mediterranean countries could not be distinguished. Finally it was shown that the social democratic, conservative and Mediterranean countries were all less in favour of the principle of equity than the liberal reference category. Thus, on these items we can distinguish the liberals (Brits and Irish) from other Europeans, but it is clear that a given welfare regime does not directly coincide with a basic justice principle of a given population.

The Puzzle of Public Opinion

37

control procedure is somewhat problematic8 but, as we shall see below, a similar ranking can be reached by simple descriptive statistics. As the Nordic countries had not joined the ISSP in 1987, none of the social democratic regimes were included. A more regime-confirming result is found by Svallfors (1997), who by means of the second ISSP module on inequality from eight countries found a remarkable difference between Sweden and Norway at one end and the USA at the other. By simple descriptive measure it was shown, for example, that while Norwegians and Swedes think that a chairman of a large national company should earn less than 2.5 times as much as an unskilled worker, Americans think that he should earn more than 11 times as much (see the first line in Table 3.1). After the Scandinavian extremes come New Zealand, Australia, Canada, Austria and Germany – the latter being the closest to the USA. Furthermore, in contrast to Kelley & Evans’ study (see Footnote 13), the pattern is not altered by Svallfors’ statistical model that includes the USA (reference category), Germany, Australia and Norway. Thus, in terms of legitimate income difference, Svallfors does find the social democratic regimes at the one end, which speaks in favour of the ‘passion for equality thesis’, and the USA at the other, which speaks in favour of the ‘American exceptionalism thesis’. Using the third ISSP social inequality module (1999) and analysing Sweden, Great Britain, West Germany and USA, Svallfors confirms that Sweden is at one extreme and the USA is at the other. The differences have reduced slightly as the Swedes have become a bit less egalitarian: a chairman should now earn 3.2 times more than an unskilled worker, and the Americans have become a bit more egalitarian: a chairman should now only earn 10.5 times that of an unskilled worker (Svallfors 2004, p. 83).9 The differences are still very substantial, which speaks against Gilens (2000), who claimed (cited in the previous chapter) that the lack of support for welfare policy in USA was not a matter of Americans having different egalitarian values.

8 Their method is to ‘apply the regression equation predicting the legitimate income of high-status occupations in [e.g.] Hungary to the entire U.S sample [reference] and predict what each American would think if he or she followed the Hungarian pattern as characterised by the Hungarian regression equations’. This is especially problematic because they control for party affiliation. In practice this is done by turning American left-wing supporters, i.e. those affiliated with the Democrats, into European left-wing supporters, i.e. those affiliated with social democratic and communist parties. In the German case, it is for example calculated what an American affiliated with the Democrats would perceive as a just wage difference if he (besides having the same characteristics in terms of income, occupation, sex and age) supported either SPD (the social democrats), DKP (the German communist party), or Die Grüne (a new red and green party with a strong legacy from socialism). In the Hungarian and Polish cases, this American had even be exposed to four decades of communist rule, so no wonder his American exceptionalism is less evident. 9 Furthermore, Svallfors shows that the class differences concerning income inequality are modest and rather similar across countries. This speaks against the power resource theory – at least if we expect political (left-wing) mobilisation primarily to influence certain classes and not the whole society – and speaks in favour of the difficult culture argument.

The Institutional Logic of Welfare Attitudes

38

Table 3.1.

Legitimate income differences in Sweden, Germany and USA in 1992. Measured by mean, mean without 10 per cent extreme cases, median in relation to both unskilled and skilled factory workers1 Sweden

Germany (W)

USA

Mean:

2.39

7.11

11.14

Standard deviation:

1.57

8.00

16.40

Mean with 10 pct. extreme cases excluded:

1.95

4.75

6.47

Median:

1.92

4.29

5.33

Median in relation to skilled factory worker:

1.56

3.33

3.14

Mean

1.96

4.32

5.46

Standard deviation

0.80

3.16

5.25

Mean with 10 pct. extreme cases excluded:

1.76

3.48

4.14

Median

1.74

3.38

3.87

1.64

2.95

3.00

Times more a chairman of a large national company should earn than unskilled factory worker:

Income distribution index (Svallfors, 1997)2:

Median only in relation to skilled factory worker

1 We have followed Svallfors’ calculations. Those that would not have given the exact figures, but were categorised as ‘1,000,000 or more’ were given the value 1,000,000. 2 Difference between unskilled factory worker, farm worker, skilled factory workers and chairman of a large national company, cabinet minister and doctor. 3 Question wording: Next, what do you think people in these jobs ought to be paid – how much do you think they should earn each year before taxes, regardless of what they actually get? (Please write in how much they should earn each year before tax).

The Puzzle of Public Opinion

39

However, Svallfors’ analyses (1995, 1997, 2004) have a severe methodological problem because he calculates the mean of what each respondent in a given country thinks the differences should be between different occupational groups. Thus, if a small number of respondents think that the difference should be extremely large, these ‘out layers’ have a large impact on the calculated mean (Kelley & Evans overcome this problem by taking the logarithm of the wages). In Table 3.1, we illustrate this point by expanding the calculations in Svallfors (1997) for Sweden, the USA, and West Germany. If we take the most extreme case, that of a chairman of a large national company, a closer analysis of the data reveals that a small number of respondents in favour of extremely high differences between an unskilled worker and a chairman influences the calculated means heavily in the USA, somewhat in West Germany and little in Sweden. The different amounts indicated by these extreme respondents cause the large differences in the standard deviations in the second row in Table 3.1. If we exclude the 10 per cent that indicate the highest differences, the means for the just wage difference between an unskilled worker and a chairman are reduced significantly in the USA, from 11.14 to 6.47, substantially in West Germany, from 7.11 to 4.75 in West Germany, and slightly in Sweden, from 2.39 to 1.95. This result is more or less repeated if we instead take the median, i.e. look at the income judgement of the respondents in America, Sweden and West Germany, where one half of the population thinks that the difference should be minor and the other half thinks the difference should be higher; the figures are respectively 5.33 for USA, 4.29 for West Germany, and 1.92 for Sweden. However, we know that, in particular, the salary of unskilled workers differs between Europe and USA, so in contrast to Kelley & Evans and Svallfors it might be more valid to use the skilled factory worker as a reference point (as done in most economic literature, often labelled ‘average production worker’). If we compare with the skilled industrial worker and take the median, it actually turns out that ‘the average man’ in Germany thinks that a chairman should earn more (3.33 times) than ‘the average man’ in the USA thinks (3.14 times). Still the Swedish ‘average man’ is more egalitarian (1.56 times), but the differences have decreased. Finally, these results apply not only to the chairman case. We get the same results when we take Svallfors’ income distribution index, which expresses the ratio between the three top occupations (like Kelley & Evans uses the chairman of a large national company, cabinet minister and doctor) and the three bottom occupations (he uses unskilled factory worker, skilled factory worker and farm worker). The median is respectively 3.87 for the USA, 3.38 for West Germany, and 1.74 for Sweden and can be further reduced if we only take the skilled industrial worker as the reference point (see Table 3.1). As Svallfors (1997) used data from the first ISSP module (1987) on the Austrian case, his choice of top occupations was restricted to those also analysed by Kelley & Evans. However, in the 1992 module, two new top professions were added, which enables us to make a better measure of the perceived just wage for elite occupations.10 10 Except in Sweden where the occupations mentioned in the ISSP 1987 survey simple was replicated. The explanation is that the Swedish survey was conducted prior, namely in

40

The Institutional Logic of Welfare Attitudes

In the new high status group we exclude the doctor occupation, which in fact is not a top position, and add ‘owner of a large factory’ and ‘judge in the highest court’ respectively to the two remaining high status occupations, i.e. ‘cabinet minister’ and ‘chairman for a large national company’. For each respondent we calculate what he/she thinks these four high status jobs in average should earn. This figure is then divided by what he/she thinks a skilled industrial worker should earn. By taking the median of these ratios in each country, we get the country ranking presented in Figure 3.1. If we start with the results from 1992, it is clear that we do not find a pattern where the low support for welfare policy in the liberal regime, the medium support for welfare policy in the conservative regimes, and the high support in the social

Figure 3.1. Times high status occupations ought to earn more than a skilled factory worker. Western countries available in ISSP 1992 and ISSP 1999. Median 1991, to the ISSP 1992 survey.

The Puzzle of Public Opinion

41

democratic regimes coincide neatly with differences in egalitarian values. Actually, the ideal type conservative countries, Germany (W) with 3.44 and Italy with 3.38, turn out to be the countries where ‘the average man’ indicates that the difference between the high status jobs and a skilled factory worker should be highest. This finding supports the idea that the conservative regime in Esping-Andersen’s words ‘cultivates hierarchy and status’ (cited above). At the other end we find the representative from the social democratic regimes, namely Norway (1.81), which clearly is distinct from the other countries. In between we find Great Britain (3.38), the USA (3.15), Australia (2.81), New Zealand (2.61) and Canada (2.58). Thus, rather than being exceptional in terms of (anti) egalitarianism, ‘the average man’ in America holds attitudes close to ‘the average men’ in the other liberal regimes (Great Britain is actually above the USA) which, in general, are more egalitarian than ‘the average men’ in the ideal-type conservative welfare state regimes. Using the third ISSP module on social inequality from 1999, we lose one country (Italy), but can include six new Western countries, which gives us a total of 13 countries. First of all, it is reassuring that for the countries surveyed both in 1992 and in 1999, the results are, in general, similar (Germany 3.44 and 3.43; Great Britain 3.30 and 3.33; USA 3.15 and 3.13; Australia 2.81 and 3.00; Canada 2.58 and 2.75; Norway 1.81 and 2.13). The biggest change is seen in New Zealand, from 2.61 to 3.14, whereby ‘the average man’ in New Zealand just becomes a bit more antiegalitarian than the American ‘average man’. However, two of the six new countries disturb the nice ranking we had in the 1992 figures with conservative regimes in top, liberal regimes in the middle and social democratic regimes at the bottom. With 3.2, the Austrian ‘average man’ is a bit more egalitarian than the British ‘average man’, but more surprising is the very egalitarian Spanish ‘average man’ (2.33), who holds views of income differences above those of the Danish ‘average man’ (2.33). The French (4.08) and Portuguese (3.33) ‘average men’ come in at the top, which confirms the perception of status-preserving conservative regimes. Believing that the four high status jobs should be paid 4.08 times more than a skilled factory worker, the French ‘average man’ is actually well above the other conservative ‘average men’. The overall conclusion is that the pattern of low support for welfare policy in the liberal regimes, medium support in the conservative regimes and high support in the social democratic regimes is not replicated. The Nordic countries are more egalitarian, but they do not seem so distinct from the liberal regimes, e.g. if we compare the Danish (2.50) and Canadian (2.75) ‘average man’. Measured on the just difference between a skilled factory worker and the elite occupations, the social democratic regimes seem much more distinct from the conservative regimes. However, if we look at the (just) wage difference between a skilled factory worker and an unskilled factory worker, we see a picture that coincides with the expected regime pattern, see Figure 3.2. Using the same technique, the figures illustrate that, in 1992, the American ‘average man’ is at the one end, thinking that a skilled worker should have twice the salary of an unskilled worker, and the Norwegian and Swedish ‘average men’ are at the other end, thinking that the difference should only be 1.33. In the countries available in the 1992 module, only Australia ‘disturbs’ the pattern

42

The Institutional Logic of Welfare Attitudes

with the liberal regimes at the top (USA, 2.0; Great Britain, 1.8; Canada, 1.79; New Zealand, 1.6), the conservative regimes in the middle (Germany, 1.6; Italy, 1.5) and social democratic regimes at the bottom (Sweden, 1.33; Norway, 1.33). Furthermore, if we accept Homans’ idea of a connection between what ‘is’ and what ‘ought to be’, the ad hoc explanation for the Australian case could be the special wage arbitration system, which during the 1960s and 1970s especially in Australia (less so in New Zealand) secured a very compressed wage structure. As discussed in the previous chapter, it was the effect from this wage arbitration system that made Castle argue for a fourth Antipodean regime. Looking at the Western countries covered by the ISSP 1999 module, we see once again that Australia – by being too egalitarian in terms of the just wage difference between skilled and unskilled factory workers – disturbs the regime pattern. The Portuguese ‘average man’ also disturbs the regime pattern, but this time by being too

Figure 3.2. Times skilled factory worker ought to earn more than the unskilled factory worker. Western countries available in ISSP 1992 and ISSP 1999. Median

The Puzzle of Public Opinion

43

anti-egalitarian. But otherwise we again find the pattern of liberal welfare regimes at the top (USA, 1.50; Canada, 1.50; Great Britain, 1.40; New Zealand, 1.38), the conservative welfare regimes in the middle (Germany W, 1.36; Austria, 1.33; France, 1.25; Spain 1.25), and the social democratic regimes at the bottom (Norway, 1.21; Denmark 1.20; Sweden, 1.18). If Homans is right about a connection between what ‘is’ and what ‘ought to be’, these figures could be explained by the actual differences between unskilled workers and skilled workers found in each regimes, cf. the discussion of regime specific labour market trajectories discussed in the previous chapter. However, at this point the overall conclusion is that if we are to explain the regime pattern by support for welfare policy with ‘culture’, we cannot refer to cross-national differences in general egalitarian values. Instead we should speak of differences in egalitarian values that concern the relationship between the skilled industrial worker – or to stretch it a bit, call it the well-adjusted working class – and the unskilled industrial worker – or again to stretch it a bit, call it the ‘bottom’, ‘potential poor’ or ‘poor’. Furthermore, we do not seem to find a distinct American attitude towards ‘the bottom’ of society, which seems to be implied by Gilens (2000) and Alesina & Glaeser (2004), who try to explain American exceptionalism with the extreme racial divides. In Figure 3.2, USA is at the top, but the anti-egalitarian views about ‘the bottom’ seem to be part of a wider phenomenon connected to the liberal welfare state regimes. Furthermore, we do not find a distinct European egalitarian view of ‘the bottom’, which also seems to be implied by Alesina & Glaser (2004). Instead we find a divide between the conservative welfare state regimes and the social democratic regimes. If this is the right interpretation of the figures, the challenge is to find the ‘regime mechanisms’ that explain how these different ‘cultures towards the bottom’ are produced and reproduced. However, following the discussion in the previous chapter, we would rather speak of the regime mechanisms that produce the perception of the poor and unemployed, which influence ‘the political man’s’ judgement of deservingness. Theorising these mechanisms is the task for the next chapter, but let us first sum up the findings. Previous studies and our research agenda Whether these previous attempts to link Esping-Andersen’s typologies with support for welfare state policy measured in surveys is a success or failure naturally depends on one’s expectations. If we expected a clear pattern of three ‘families’ of public opinion, we would definitely be disappointed. Bean & Papadakis (1998) seem to be among the most sceptical as they found no cross-national difference in attitude towards core areas like government responsibility for healthcare and old age pension. Such findings could support the old functional theory of welfare states being the result of a ‘need’ felt by citizens living in capitalist societies (e.g. Wilensky 1975) and suggest that survey studies of public opinion are of limited relevance

44

The Institutional Logic of Welfare Attitudes

for understanding cross-national differences and reproduction of cross-national differences in pursued welfare policy. However, a number of studies have confirmed the expected pattern, especially if the analysis is limited to the countries that come closest to Esping-Andersen’s ideal types. These studies used the three items ‘provide job for all’, ‘guarantee basic income’ and ‘redistribute income’ as dependent variables. We believe these findings indicate that a regime pattern is present when it comes to welfare attitudes towards the poor and unemployed. Therefore, it seems promising to theorise the impact welfare regimes have on support for welfare policy in the narrow sense. Nevertheless, based on the previous studies one could also object to this more limited thesis of a connection between regimes and welfare attitudes because the expected causes of such regime patterns were hard to find. The group with concentrated benefits and costs, i.e. those receiving benefits/service and those working in the public service sector, do not score much higher on the attitude indexes, which question the idea of feedback processes centred on short-term self-interest. Furthermore, the attitudes of different classes did not seem to differ much from country to country, which questions the thesis that left-wing mobilisation and coalition building with the middle class is the major explanation for cross-regime differences in support of welfare policy. This basically leaves us with the problematic culture explanation. Previous studies have not done much to specify this culture explanation. Furthermore, turning to the few comparative studies of egalitarian values, we saw that it is actually problematic to assign a general anti-egalitarian attitude to liberal regimes, a medium egalitarian attitude to conservative regimes, and a high egalitarian attitude to social democratic regimes. If we take our best comparative measure for egalitarian attitudes, namely what different occupations ought to earn, we argue that, in general, the conservative regimes are the least egalitarian. However, looking at the cross-national difference in the just income differences between skilled and unskilled workers, we actually find the expected regime pattern. With few exceptions, the ‘average men’ of the liberal regimes indicated the largest differences, the ‘average men’ of the social democratic regimes the smallest difference, and the ‘average men’ of the conservative regimes were somewhere in between. Thus, if we are to explain the regime pattern in support of welfare policy with culture, it is a regime-dependent ‘culture towards the bottom’ that has to be specified.

Chapter 4

Welfare Regimes and Deservingness In this chapter, we theorise how different welfare regimes can influence the perception of the poor and unemployed, i.e. those who will benefits from the sort of public policy captured by the questions of guaranteed minimum income, jobs for all, and redistribution to low income groups. The thesis is that the degree to which the poor and unemployed fulfil a number of so-called deservingness criteria could be the missing link between welfare regimes and the cross-national differences in attitudes described in Chapter 3. Previous studies do yield evidence on the thesis that the public perception of poverty and unemployment influence – or at least are connected to – public attitudes towards social inequality and social spending. However, in order to establish the ‘missing link’, we need to explain how welfare regimes influence public perceptions of poverty and unemployment. This chapter aims to do so by combining the literature on deservingness criteria with the welfare regime theory. Only a few studies have analysed perceptions of the poor and unemployed, and they have primarily been conducted within (what could be called) the Anglo-Saxon poverty tradition. In an American survey study from 1969 (Feagin 1972), the public was asked, for example, ‘As you know, there has been a lot of interest recently in the extent to which poverty exists in the USA and in how poor people can be assisted. The following questions are concerned with possible causes of poverty. Listed below are some reasons some people give to explain why there are poor people in this country’ (Feather 1974, p. 203). The public was then given 11 different reasons that were classified into individualistic reasons, e.g. lack of thrift, lack of effort, loose morals and drunkenness; structural reasons, e.g. external social and economic sources such as lack of education and low wages; and finally fatalistic reasons, e.g. illness and bad luck. The study was later replicated in Australia (Feather 1974), and both studies found that the public put strong emphasis on individual causes. Feagin found that among the 11 reasons suggested in the questionnaire, ‘lack of thrift and proper money management’ (58 per cent found it very important) and ‘lack of effort by the poor themselves’ (55 per cent found it very important) were the two most important causes according to the American public. Later American surveys confirmed these early findings (Alston & Dean 1972; Hochschild 1981; Kluegel & Smith 1986). These empirical studies were largely descriptive and disconnected from the comparative welfare state research (to some extent except Kluegel et al. 1995). To put it very boldly, the studies were ‘institution-blind’ because they were all conducted in liberal welfare state regimes. If the results are given a broader interpretation, scholars typically referred to the impact of American liberalism and exceptionalism, which we (as argued) find problematic and not ambitious enough.

46

The Institutional Logic of Welfare Attitudes

By contrast we will link this mainly Anglo-Saxon literature on public poverty explanations and deservingness with the comparative welfare state regime theory. The chapter is divided into five sections. The first section presents five deservingness criteria, which we believe are very important yardsticks for the ‘political man’s’ support or lack of support for welfare policy. The following sections discuss how three different regime characteristics generate regime-dependent perceptions of poor and unemployed. The regime characteristics are degree of selective welfare policy (the next section), level of generosity (the section after), and structure of the labour market (the fourth section). The fifth section summarises the three dimensions and the causal reasoning and sketches the analytical strategy in the following empirical chapters. Before we start this theoretical journey, we will briefly discuss the terms ‘poor’ and ‘unemployed’ as the boundaries of these groups are by no means clearcut. Furthermore, one could add that we primarily speak of able-bodied poor. It refers to persons of working age who are able to work (i.e. excluding children and pensioners), and who, compared with an average citizen, have a low income. Thus, we think of a smaller group than ‘the poor’ in general (even though we think most respondents actually have this middle-aged able-bodied group in mind when they answer questions about ‘the poor’). Furthermore we intentionally do not replicate the formal definition of poverty lines, e.g. citizens with a disposable income below half of an average disposable income within a given country. Basically, we want to capture the group that respondents think would benefit from a welfare state that ‘provides jobs to all’, ‘guarantees a basic income’, and ‘redistributes from those with high incomes to those with low incomes’. It is a well-recognised problem that it is difficult to define the boundaries of ‘the poor’ (Lee et al. 1990; Lee et al. 1992; Wilson 1996), which in comparative research makes it easier to use the concept ‘unemployed’. In Furnham’s (1982) replication of Feagin’s and Feather’s studies, he asked about causes of unemployment in Britain and not about causes of poverty. Without further explanation, the ‘unemployed’ excludes children, pensioners, persons not able to take a job, and the group does not have to have an income as low as half of an ordinary income. Defined by the boundaries of existing social security schemes, we then primarily speak of persons who receive unemployment benefits or social assistance. One could also label this group the potential poor, as their living conditions very much depend on the welfare institutions in place. However, we cannot only use the term ‘unemployed’ as, especially the liberal regimes, have groups of working poor who indeed would benefit from redistribution to ‘those with low income’. Thus, the best operationalisation of the potential poor is probably to define it as the two bottom deciles in the income distribution. However, such questions about the group of potential poor is not available in the existing comparative data and it is obvious that is difficult to make respondents think of ‘the two bottom deciles in the income distribution’. Therefore we have had to stick to the somewhat blurred concepts of ‘poor’ and ‘unemployed in the data material used in the following chapters. When possible, we will use the concept ‘unemployed’ as it

Welfare Regimes and Deservingness

47

is familiar to respondents in both a European and an American context.1 Naturally, perceptions of what constitutes unemployed also vary among respondents, but the term has more clear references to a person of working-age who is physically and mentally capable of handling a job and is in a situation where he/she potentially can become poor. Deservingness criteria and public support for welfare policy The main effort of the deservingness literature has been to pinpoint which criteria the public use to judge whether a person or a group deserves help. In that regard, the studies conducted by Cook (1979), De Swaan (1988), Will (1993) and Oorschot (2000) are extremely helpful to our purpose. On the other hand, the studies have – in our opinion – not been very successful in finding the independent variables that influence the public perception, which is the raw material for the judgement of deservingness. The one expectation is the importance of the welfare clientele in question. The literature on deservingness seems capable of explaining the pattern of public support for welfare policy that Coughlin (1980) found in his pioneering crossnational study. He found what he calls ‘a universal dimension of support’ because the ranking of the deserving groups followed the same line in all the countries included in his study. As mentioned, the public was most in favour of support for old people, followed by support for sick and disabled, needy families with children and unemployed. The group given least support were people on social assistance. Petterson (1995), Oorschot (2000), Oorschot & Arts (2005) and others have confirmed this ranking. But besides this distinction between groups of recipients, the deservingness literature has been rather weak in terms of causality. At least the literature seems to overlook societal factors (and many of the proposed explanations at the individual level seem to operate in different directions, Lepianka 2004). Therefore we find it theoretically fruitful to combine these empirically deduced deservingness criteria with the regime theory as it delivers a theoretical description of the context in which the public makes these deservingness judgments. However, before doing so we need to take a closer look at the deservingness criteria found by these previous studies. If we follow the review in Oorschot (2000), we arrive at the following five criteria:

1 The extreme case of Northern Ireland illustrates this point. Ditch found that compared to respondents in the other EU countries, Northern Irish respondents indicated a very low level of people living in poverty, even though several international comparisons have found the province to be one of the poorest regions in the EU. One of the reasons given was that ‘socio-economic issues (such as poverty, unemployment, housing, low pay etc) are not part of the normal vocabulary of political literacy’ (Ditch 1984, p. 171) due to the absence of a social democratic movement and the overwhelming political importance assigned to security and terrorism. In other words, reference to a group of poor was simply unfamiliar to the Northern Irish respondents.

48

The Institutional Logic of Welfare Attitudes

(1) Control (the less control over neediness, the higher degree of deservingness). (2) Need (the greater level of need, the higher degree of deservingness). (3) Identity (the higher degree of group belonging, the higher degree of deservingness). (4) Attitude (the more grateful, docile and compliant, the higher degree of deservingness). (5) Reciprocity (the higher previous or future payback, the higher degree of deservingness). Both the previous studies and Oorschot’s empirical findings on the Dutch case show that the issue of control is especially important. Thus, the key to explaining modest support for unemployed is the perception that they are much more in control of their situation than, for example, the disabled, sick and pensioners. In De Swaan’s (1988) historical study of the modern welfare state, he labelled the criterion ‘disability’. In Cook’s (1979) study of Americans’ views on supporting the poor, she labels the criterion ‘locus of responsibility’. Finally, Will (1993) also found that the most important deservingness criterion was the degree to which the problems facing poor families were beyond the immediate control of the individual family. Naturally, the level of need also plays a part, but within the Anglo-Saxon poverty tradition need is more or less taken for granted; as welfare policy is all about support for those in need. The application of this criterion is more difficult in social democratic regimes where everybody is entitled, and in conservative regimes where those who pay contributions are entitled. The identity criterion refers to the importance of feeling a shared identity with the groups who are to be supported. Using the label of proximity, De Swaan argues that the boundary of the area can be defined by kinship relations, by place of residence, or more generally, by the boundaries of a certain identity group, like ‘our family’, ‘our town’, ‘our church’ or ‘our people’. The attitude criteria refer to the ways recipients respond to public support. De Swaan uses the term ‘docility’ to highlight that the poor who hide their misery and ask for nothing are seen as more deserving than those who make impudent demands. Cook uses the terms ‘gratefulness’ and ‘pleasantness’. Finally, the attitude criteria can be linked to a more general criterion of reciprocity, e.g. such behaviour as ‘the smile of thanks’, but also, in a modern context, actively looking for a job, and willingness to participate in re-insertion programmes. Oorschot furthermore argues that needy who at the moment are unable to reciprocate might fulfil this criterion if they have contributed to ‘us’ in the past, or are likely to do so in the future. The same emphasis on reciprocity is found in the writings about a ‘moral economy’ of the welfare state (Mau 2003). Thus, the large support for public assistance to the old found by Couglin and others might be explained by the perception that (1) they are not in control of their neediness, (2) they belong to ‘us’, (3) they are often grateful, docile and compliant, and (4) they have contributed to ‘us’ during their working age. At the other extreme, the low support for the group on social assistance might be caused by the perception that (1) they could get a job if they wanted, i.e. they are in control of their neediness,

Welfare Regimes and Deservingness

49

(2) they do not fully belong to ‘us’, (3) they are often ungrateful, and (4) they have often not contributed much to ‘us’ in the past. It is more difficult to see how the old and people on social assistance differ in terms of need. At the macro-level, one could also argue that it is probably no coincidence that many social reforms historically have been made after a war. Soldiers or families who have suffered during a war (1) are clearly in need, (2) are not in control of their neediness, as war is not decided by laymen, (3) they are identified as belonging to ‘us’, as they have fought for the interest of the nation state (and wars are probably one of the best ways to merge different groups into ‘one people’), and (4) they have earned support as they have been willing to pay the highest price, their lives, for the interest of the nation state, i.e. they fulfil the criterion of reciprocity. Taking the fulfilment of the four deservingness criteria into account, the public might tolerate the fact that veterans or families of killed soldiers often do not receive support with a grateful, docile and compliant attitude. Following this logic, it is not without reason that scholars have argued that immigration might threaten the public support for welfare policy. Immigrants may be expected to score particularly badly on the criterion of identity and reciprocity, i.e. they are not a part of the common ‘us’ and they have not contributed to the common ‘us’ in the past. Furthermore, they may be accused for having put themselves in a situation of welfare dependency and not receiving help with the right attitude. Measured as the concern respondents have about different groups’ living conditions, the relative low deservingness of immigrants compared to unemployed, sick/ disabled, elderly is clearly revealed by results from the European Values Survey (Oorschot 2005). In all 18 countries Oorschot analyses he finds the pattern of highest concern about the elderly followed by concern for sick/disabled, unemployed and finally immigrants. This finding supports the American studies that emphasise the importance of racial divides and support for welfare policy. However, Oorschot also finds that in Denmark, Sweden and the Netherlands the difference between the public concern about the living conditions of immigrants and unemployed is modest (Oorschot 2005, p. 15). Below we will argue that this probably has to do with variations in welfare institutions. The point is that these social democratic regimes have welfare institutions that through various mechanisms avoid identifying racial minorities, immigrants, the poor and unemployed as a distinct group in the welfare discussions. And the other way around; that the liberal regimes have mechanisms that generate such divides (unfortunately the countries closest to the ideal type liberal regime, USA and Canada, were not included in Oorschot’s 2005 study). Besides Oorschot’s (2005) study, the literature on deservingness has seldom been linked to institutional arguments. The previous studies that – besides the distinction between different groups of recipients – have looked for independent variables that influence the judgement of deservingness have primarily looked at characteristics at the individual level, e.g. the respondents’ age, class and education. This is no wonder, as it is, so to speak, ‘the natural way’ to analyse the variables provided by a national survey. But some of the results actually indicate that the judgement of deservingness might be the missing link between welfare regimes and public support

50

The Institutional Logic of Welfare Attitudes

for welfare policy. The empirical findings in the Dutch survey ‘that “social division” or “class” variables like income level and whether one is working or on benefit do not play a role in the conditionality of solidarity’ (Oorschot 2000, 40) fit nicely with our argument. The same lack of impact from individual background variables were found in Oorschot’s (2005, p. 16) study of 18 European countries. As class and transfer class did not manage to establish the link between regime and public support for welfare policy (cf. Chapter 3), we are searching for intervening variables that are not (strongly) correlated to class and transfer class. The Dutch results actually show that the so-called degree of conditionality is stronger among people with lower educational level and lower socio-economic status, which means that the working class actually seems to be against unconditional public support. Golding & Middleton (1981) also found that those who were socio-economically closest to beneficiaries and the poor had the strongest negative belief and feelings about these groups. However, as a point of departure we are not so much interested in explanations at the individual level. What we need is a theory that explains how the regime context influences the ability of poor and unemployed to fulfil these seemingly universal deservingness criteria. Figure 4.1 present the main causal reasoning, which will be elaborated in the following sections.

Figure 4.1. The theoretical link between welfare regime and cross-national differences in welfare attitudes

Welfare Regimes and Deservingness

51

The link between programmatic structure of welfare state policy and fulfilment of deservingness criteria The links between welfare regime and perception of the poor and unemployed are little explored, and the one cross-national study (to our knowledge) does not explain, why it expects a connection between regime and a perception of low-income group (Oorschot & Halman 2000). However, in order to strengthen the causal reasoning, but also to highlight factors that in empirical studies can ‘disturb’ the relation between regime and perception of the poor and unemployed, we find it important to make the theoretical argument more explicit. As recalled from Chapter 2, the three welfare regimes are distinguished by differences both in welfare state, labour market, and family structures. But it is especially on the state dimension we can find theoretical inspiration from previous studies. Within welfare state literature it is a classic thesis that systems dominated by universal programmatic structures of benefits and services, i.e. the ideal type policy of the social democratic welfare regime, and systems dominated by a selective programmatic structure of benefits and services, i.e. the ideal type policy of the liberal welfare regime, generate quite different public discussions and perceptions of recipients. The typical coverage of the incidence of long-term unemployment in liberal and social democratic welfare regimes respectively are illustrated in Figure 4.2.

Figure 4.2. The main structure of welfare services and benefits in case of longterm unemployment in an ideal-type liberal and social democratic welfare regime

52

The Institutional Logic of Welfare Attitudes

Following Rothstein (1998), the first step of the argument is simply to point out the fact that a selective policy that aims to provide ‘the needy’ with economic resources or what Rothstein calls basic capabilities must determine (1) who is needy, and (2) how much they need. Therefore, ‘the public discussion of social policy in a selective system often becomes a question of what the well-adjusted majority should do about the less well-adjusted, in varying degrees, socially marginalized minority’ (Rothstein 1998, p. 158). The general fairness of the policy is also open to challenge as the majority might start asking ‘a) where the line between the needy and the non-needy should be drawn, and b) whether the needy themselves are not to blame for their predicament’ (Rothstein 1998, p. 159). Relating this argument to the deservingness criteria presented in the previous section, one could say that a system dominated by selective welfare policies opens discussions of ‘need’ and ‘control’ (see Table 4.1). Table 4.1.

The effects from, respectively, selective and universal public policy on different dimensions of deservingness

Dimensions of deservingness Need

A welfare state dominated by selective benefits and services Open the discussion of whether recipients are in need

Control

Open the discussion of whether recipients are to blamed

Identity

Define the recipients as a special group distinguished from the well-adjusted majority

A welfare state dominated by universal benefits and services Close the discussion of whether recipients are in need Close the discussion of whether recipients are to be blamed Define the recipients as equal citizens who belong to a national ‘us’

Reciprocity

Highlight the boundary between those who give and those who receive Open the discussion of whether recipients receive benefits and services with a grateful, docile and compliant attitude

Blur the boundary between those who give and those who receive Close the discussion of whether recipients receive benefits and services with a grateful, docile and compliant attitude

Attitude

The identity dimension of deservingness is also influenced by this logic connected to selective policy as ‘the very act of separating out the needy almost always stamps them as socially inferior, as “others” with other types of social characteristics and needs’ (Rothstein 1998, p. 158). Even people or groups in favour of welfare policy are believed to contribute to this institutional logic, as the typical argument for better policy is that the group is even more different than initially presumed and therefore has even more special needs. In his influential work, Jordan (1996) also criticises poverty research for carrying this liberal logic, as it has been preoccupied with measuring and categorising the low-income groups along poverty lines, social

Welfare Regimes and Deservingness

53

minimum lines, or demographic and socio-economic categories. It could be perceived as a scientific effort to describe the size of the ‘needy’ group and how it differs from the ‘ordinary’ people. Furthermore, one could argue that boundaries between ‘them’ and ‘us’ generated by selective welfare policy also make it more clear who benefits from the welfare state (i.e. those who pay little or no tax and receive targeted benefits) and who loses on the welfare state (i.e. those who pay tax, but do not receive any benefits). Thus, the reciprocity of the system will be perceived as being very low, which probably increases the importance of grateful, docile and compliant attitudes among those who receive targeted benefits or services. This is basically a different way to present the visibility argument, which was discussed in Chapter 2 as an alternative to the class interest thesis of the power resource theory. Finally, one could argue that this logic of selective welfare policy might generate vicious circles or even self-fulfilling prophecies, because the ‘needy’ – exactly because they are labelled as not been ‘ordinary’ people – alter their behaviour. It creates a further division between ‘them’ and ‘us’ and probably makes it more difficult to find grateful, docile and compliant attitudes among recipients. This social mechanism is known from, for example, ethnic second-generation immigrants, who because they are labelled as outsiders sometimes start practising the religion and culture of their country of origin much more intensively than first-generation immigrants; even though they are born and educated in the new country and may never have visited their country of origin. The logic of universal welfare policy is, in all aspects, contrary to the logic of selective welfare policy. As discussed in Chapter 2, a number of countries had already, before the Second World War, adopted schemes that followed the universal principle (flat-rate benefits based on citizenship and given independent of need), but it was in the post-war period that Britain and especially the Nordic countries expanded these schemes. William Beveridge was one of the main architects behind the British welfare reforms in 1946–47 that made unemployment benefits, disability pension and old-age benefits more or less universal (Baldwin 1990, pp. 116–34). The Nordic countries that introduced flat-rate pension-systems in the late 1940s and 1950s followed the same path, but as the years went by, the generosity of the flat-rate benefits became much higher than in Britain. Furthermore, in the last two decades, this universal principle has been adopted in the services schemes, which the Nordic countries have so heavily expanded. Thus, in the ideal-type social democratic welfare regime, the poor and unemployed are more or less entitled to the same benefits and services and at the same level (flat-rate) as the rest of the population. It is this Scandinavian ‘Rolls Royce universalism’ we have in mind when we speak of welfare states dominated by universal benefits and services. In these systems, no line needs to be drawn between the needy and the nonneedy. Thus, the discussion of need and to what extent the poor and unemployed are in control of their neediness becomes more or less irrelevant. As Rothstein argues, ‘welfare policy does not, therefore, turn into a question of what should be done about “the poor” and “the maladjusted,” but rather a question of what constitutes general fairness in respect to the relation between citizens and the state. The question

54

The Institutional Logic of Welfare Attitudes

becomes not “how shall we solve their problem?” but rather “how shall we solve our common problem (healthcare, education, pensions, etc.)?’ (Rothstein 1998, p. 160). Instead of defining a line between ‘them’ and ‘us’, the universal benefits and services actually help define everybody within the nation state as belonging to one group. The vicious cycle of selective welfare policy seems to have been replaced by a positive circle where the whole population is given the basic capabilities that place them ‘on a more or less equal footing in respect to their ability to act as autonomous citizens’ (Rothstein 1998, p. 157). It is probably no coincidence that the link between welfare policies and a well-functioning democracy, often analysed within a socalled citizenship perspective, has been intensely explored by Scandinavian social scientists. Welfare policy is not just seen as a measure to avoid poverty, generate security or even economic equality, but also as a way to generate free, responsible and democratic citizens, who can produce and reproduce a democratic nation state. Following this line of reasoning, we should also expect universal welfare policies to have a high symbolic value for the nation state. One could say that the clear boundary between ‘them’ and ‘us’ within the ‘liberal’ regime is replaced by a clear boundary between citizens and non-citizens. What critics of systems dominated by universal welfare policy often describe as a system where ‘everybody gives to everybody’ is a characteristic that furthermore helps to break down a possible boundary between low-income groups and ‘ordinary people’. For the majority of people, it is not an easy task to calculate whether one is a net-winner or net-loser, even though welfare states dominated by universal policy have been shown to be the most redistributive nation states in the OECD area. If the cost-benefit analysis is done at the individual level in a given year, the calculation could be manageable. The market value of the universal benefits and services received in that year should be subtracted from the amount paid in VAT, income tax and different duties. But the calculation is complicated, and it becomes even more complicated if the cost-benefit analysis is done at the household level and within a lifetime perspective. In that case, the amount of VAT, income tax, duties etc paid by the family over a lifetime should be subtracted from the value of free education for the children, the old-age pension of one’s partner, the likely use of free hospitals, the likely use of unemployment benefits etc. The point is that the programmatic structure, but also the very size of the ideal-type social democratic regime, blurs the boundary between net-winners and net-losers, which makes it difficult to judge whether recipients of benefits and services have reciprocated. And combined with the influence on the other dimensions of deservingness, the most likely end result is that an ordinary citizen does not start to calculate at all.2 Finally, as the institutional logic suppresses the discussion of need, control, identity and reciprocity, the attitudes among recipients of benefits and services also become more or less irrelevant. Nobody expects citizens – including the poor and the unemployed

2 This argument is in line with Goul Andersen’s findings on the Danish case reported in Chapter 3, see Footnote 10.

Welfare Regimes and Deservingness

55

– to be grateful because they receive a basic old-age pension, have access to free hospital treatment, heavily subsidised childcare etc. Following this line of reasoning, which highlights the institutional logic, we have theoretical reasons to believe that the poor and unemployed (included ethnic minorities over-represented in this group of potential poor) in the liberal regimes will be asked to fulfil much harder deservingness criteria than the poor and unemployed in the social democratic regimes. Not because the citizens in the liberal regimes have very different egalitarian values, but because their welfare state is dominated by selective welfare policy. Now the question is how this reasoning applies to welfare policy conducted by the ideal type conservative regime. As described in Chapter 2, the primary programmatic structure of the conservative regime is compulsory insurance systems financed by social contributions from employers and employees. Entitlement of benefits and services through this insurance mechanism might suppress the deservingness discussion even more than a universal programmatic structure. As these people have paid the insurance, the discussion of need, control, and attitude becomes more or less irrelevant. And the boundaries of ‘us’ can simply be drawn between those who paid their contribution (and their families) and the others. The same goes for the question of reciprocity. An ‘earmarked’ compulsory contribution (in contrast to universal benefits and services typically financed through general taxes) makes it clear that recipients of welfare benefits primarily have paid themselves. Thus, in cases where the male breadwinner becomes long-term unemployed, he very easily fulfils the deservingness criteria. However, the combination of benefits in relation to contributions, strong job protection, and generous early retirement schemes for the insider male breadwinners has the result that the long-term unemployed and poor should mainly be found among the outsiders, i.e. the women and the young. Furthermore, in the goldenage ideal-type conservative regimes, these groups were covered by the male breadwinner’s income and his social security entitlements. This system is still in practise in the South European countries, where the European Labour Force survey shows that, in 1995, only 6.8 per cent of the Italian unemployed received public benefits (unemployed defined as people answering that they would like a job, they are looking for a job, and they can begin within two weeks). The share is 8.6 per cent in Greece, 23.8 per cent in Spain, and 27.3 per cent in Portugal (Gallie & Paugam 2000b, p. 8). Thus, most unemployed are not exposed to any welfare state policy at all.3 Those ‘bad risks’ that are not absorbed by the family are sometimes supported by rudimentary social assistance schemes, which are extremely selective. Thus, for

3 Within the family, those in need are probably confronted with some of the same deservingness criteria: is the family member really in need? Is it the family member’s own fault? Does the family member have a grateful, docile and compliant attitude? Has the family member reciprocated to the family previously, or is the member likely to do so in the future? The question of identity is probably easier to solve as the blood ties constitute a natural ‘us’. However, the point is that these intra-family discussions are disconnected from the discussion of what the state ought to do.

56

The Institutional Logic of Welfare Attitudes

this small minority, the effects of selective policy discussed above should be present, but still they do not live in systems where selective welfare policies dominate. What dominates is the family. However, in recent decades, the Northern Continental European countries have experienced a break-up in stable family patterns, which was a crucial premise for the effectiveness of the golden-age conservative welfare state regime. As a result, these countries have organised social assistance schemes, and in many countries the formal obligations towards children have been reduced. Gallie & Paugam (2000b, p. 17) speak of ‘advanced intergenerational autonomy’ in Germany and ‘relative intergenerational autonomy’ in France and Belgium. As a result, thoroughly organised social assistance schemes are set up to cover the risks of the outsiders and the insiders who are getting fired, do not find a job after the relatively long period on unemployment benefits, and are too young to enter the early retirement schemes. In this case, we have a more selective welfare policy than in the Mediterranean countries, but the poor and unemployed do not live in a system where selective welfare policy dominates. Thus, on a one-dimensional scale we have a highly selective welfare policy in the liberal regime, a somewhat selective welfare policy in the conservative regimes (with big internal variations), and only limited selective welfare policy in the social democratic regimes. It fits the pattern of low support in liberal regimes, moderate support in conservative regimes, and high support in social democratic regimes. However, welfare regimes also differ in terms of generosity – or call it the living conditions among the potential poor – and job opportunities. These characteristics are also likely to influence the judgement of deservingness. The link between generosity and fulfilment of the identity criterion The degree of generosity and the degree of selectivism of the welfare regimes are often not distinguished clearly from each other – probably because they go together. In terms of poverty risk, empirical studies have demonstrated that the risk tends to be highest in the liberal regimes, in-between in conservative regime, and lowest in social democratic regimes – at least if we take the countries closest to the ideal types. Naturally, poverty rates and, especially, overall inequality in disposable income in a given year are not only a consequence of welfare institutions. Förster & Pearson (2002, p. 8) describe the final distribution of income as ‘the result of a complex set of relationships, including family formation and dissolution, longevity and fertility, as well as the more obvious trends in earnings, taxes and the returns to capital’. Nevertheless, if we take the countries closest to the ideal type social democratic, conservative, and liberal regime we find the expected pattern. Measured by Ginicoefficients based on OECD data from the mid-1990s the disposal income distribution was most equal in Sweden (23.0), less equal in Germany (28.2), and most unequal in USA (34.4) (Förster & Parson 2002, p. 38). If we take the percentage of the total populations that have an income below 50 per cent of the median (equivalence) income we see the same pattern. In Sweden 6.6 per cent fell below this threshold in

Welfare Regimes and Deservingness

57

the mid-1990s, in Germany the share was 8.2 per cent, and in USA the share was 17.8 per cent.4 These conventional figures for the whole population also include poverty among children and the elderly. Therefore, we have also tried to find figures that more narrowly measure the poverty risk among the group of able-bodied persons of working age, which is our main concern. Relative poverty rates among the unemployed can be calculated from the European Household Panel Study and the results are very convincing even though the extreme cases of Sweden and USA are not included. In Denmark, another social democratic regime, only 8.1 per cent of the unemployed fell below the relative poverty threshold of 50 per cent of the mean income. In Germany, the share was 26.8 per cent and in Great Britain the share was 48.5 per cent (Gallie et al. 2000, 51). Thus, it is fair to conclude that speaking about the countries with the largest differences in attitudes towards welfare policy, i.e. countries within the cluster of liberal regimes and the cluster of social democratic regimes, we also find considerable differences in income inequality and poverty. Now it is a classic thesis that pursuing a welfare policy that allows recipients to continue a normal lifestyle reduces the risk of stigmatising (otherwise) poor and unemployed citizens. The basic argument is that reduced differences in economic resources between ‘the majority’ and ‘the bottom’ of society generates more similar living styles, which as a consequences makes it easier for ‘the bottom’ to fulfil the identity criterion. Thereby, we have a self-reinforcing feedback mechanism where policies that generate good living conditions among the potential poor produce public support for ‘more of the same’. And the other way around; if those at ‘the bottom’ of society are not provided with decent economic resources they are forced to have a way of living that is quite different from that of the majority. Thereby it becomes harder to fulfil the identity criterion and we have a ‘negative’ feedback mechanism on public opinion. It is quite symptomatic that the largest discussions about dependency culture took place in liberal regimes, which provide the least generous benefits and services (Murray 1984 for the US; Dean & Taylor-Gooby 1992 for the UK). To put it boldly, one can argue that the lack of identification with blacks in the USA it not only a matter of black being black. It is also a matter of black being poor. Thereby, we reach the same result as the power resource theory; namely that it basically becomes easier and easier to create good living conditions for ‘the bottom’ if a welfare state becomes more and more ‘institutionalised’ (Korpi & Palme 1998), but our line of reasoning is different. On a one-dimensional scale we suggest that the difference in economic resources between ‘the bottom’ or ‘the potential poor’ and ‘the majority’ is high in the ideal-type liberal regime, medium in the ideal-type conservative regime, and low in the ideal-type social democratic regime, see Table 4.2 later. Thus, we can expect a regime dependent effect, which should make it most difficult to fulfil the identity criterion in the liberal regimes and easiest in the social democratic regimes.

4 The estimates of the proportion of citizens living below the poverty line are heavily influenced by the method of calculation. Here we have again simply taken the overall figures provided on the webpage of the Luxembourg Income study, around 2000.

58

The Institutional Logic of Welfare Attitudes

However, one also could argue that if the pursued welfare policy – following whatever principle – manages to provide (potentially) poor and unemployed groups with good life conditions, one should expect the public to make tougher judgements about the fulfilment of the need criterion. As anecdotal evidence one could point to the fact that, in Denmark, not even a person concerned with welfare policy issues like myself gives money to ‘needy-looking’ beggars. The rationale is that these beggars are not really needy, as the Danish welfare policy provides ‘the bottom’ with some of the best living conditions in the world (e.g. Nolan & Hauser 2000). To put it boldly, it fosters the perception that if you give money to ‘hungry’ beggars at Danish railway stations, they will spend it on beer. The argument is supported by the fact that one can find a very strong relationship between social expenditures and the share of the public that finds the poor (R2=0.63) and unemployed (R2=0.74) to be sufficient protected by the welfare state (Taylor-Gooby 1995). This finding could be without relevance if the ISSP items used in the previous studies only measure welfare attitudes in absolute terms, i.e. if we assume that the public is not influenced by the actual pursued welfare policy. The questions are asked in absolute terms (e.g. should it be a governmental responsibility to….) but it is questionable whether the public can neglect the present level of welfare policy (see the previous chapter and Chapter 8). Therefore, Figure 4.1 includes a dotted arrow to the need criterion. The point is that more advanced welfare states (or at least welfare states believed by the public to be advanced) might experience a feedback process that influences the judgement of ‘need’. Thus, we probably have a ‘second order’ feedback process on deservingness that restrains what above seemed to be self-reinforcing ‘first-order’ feedback processes. Nevertheless, in order to explain the regime pattern in public support for welfare policy it is fair to assume that the first order ‘identity effect’ is more relevant than the second order ‘need effect’. The link between regime-dependent labour market trajectories and fulfilment of the control criterion Welfare regimes have not only a state dimension, but also a labour market and family dimension, c.f. Chapter 2. We will therefore expand the classic thesis of connection between the degree of selectivism and generosity with a discussion of the possible connection between the ‘regime-embedded’ labour market and the perception of the poor and unemployed. Economic textbooks as well as more scientific economic studies typically describe the labour market as an autonomous arena where a clearing takes place between the demands of enterprises and the labour supply of the households. However, as pointed out by Esping-Andersen and a number of ‘institutional economists’, one can also argue that the programmatic structure of the welfare state has a huge impact on the labour market. In his 1990 book, Esping-Andersen described how welfare policy creates important structures that influence how workers, through early retirement, can exit from the workforce, how workers can claim paid absence from work, and how women, in particular, can

Welfare Regimes and Deservingness

59

enter the workforce (Esping-Andersen 1990, chapter 6). However, his real interest was how these institutional regime differences have influenced the transformation from industrial to post-industrial economy (Esping-Andersen 1990, 1996a,b, 1999). The pressure on the industrial production structure comes from external factors such as increased economic integration and new technologies and from internal factors such as the women’s desire to participate in the workforce. The most discussed indication of these pressures is the high rates of unemployment that haunted most Western welfare states during the 1980s and 1990s. Very generally speaking, the situation is that the social democratic regimes have followed a trajectory where new jobs were generated in the public sector. It boosted employment considerably, created opportunities for women, and prevented declining wages in service jobs. At the same time, an active labour market policy was designed to train manual workers for new post-industrial jobs. The liberal regimes followed a neoliberal trajectory where new service jobs are provided in the service sector. It also boosted employment and created opportunities for women, but at the same time it established a large number of working poor. Finally, conservative regimes followed a labour reduction route. It did not boost employment in the service sector, but instead protected the male insider against the risk of unemployment through strict job protection and early retirement schemes (Esping-Andersen 1996a,b). Now the point is that these different employment trajectories associated with the different welfare regimes might also influence the public perception of poor and unemployed. In particular, the degree to which the poor and unemployed groups are believed to be in control of their neediness might be influenced by job opportunities. Therefore, the poor and unemployed in the conservative regimes should be seen as less in control of their neediness than the poor and unemployed in the social democratic and liberal regimes where job growth in the public and private sector respectively has generated job opportunities. Recent empirical studies actually support such an impact from labour market structures. Using Eurobarometer surveys Gallie & Paugam (2002, p. 21) found a clear connection between level of unemployment in European countries and a perception of poverty as being due to laziness among the unemployed. Using the World Value Study, the same finding is reached in Chapter 5. Blekesaune & Quadagno (2003), using the ISSP role of government module, also found a strong connection between level of unemployment and support for welfare policy. Finally the World values study from 1995–97 actually shows that the share answering that most poor people have very little chance of escaping poverty is very high in Germany (83 per cent) and Spain (73 per cent). We will argue that not only does the degree of service sector expansion (and thereby unemployment rates) matter but also so does the differences in the wage setting mechanisms in each regime. We expect the poor and unemployed to be perceived to be more in control of unemployment in countries where individuals are able to negotiate the wages themselves. This is possible in liberal regimes, where the importance of unions has always been modest, and where the position of the unions has been further undermined by the neo-liberal employment strategy during the 1980s and the 1990s. To put it boldly, this institutional setting is believed to facilitate

60

The Institutional Logic of Welfare Attitudes

the perception that everybody can get a job in the private service sector if they only are willing to reduce their wage demands. This is not possible in social democratic and conservative regimes, where the unions still have considerable influence on wage setting – especially in blue-collar sectors. So the poor and unemployed cannot escape unemployment by lowering their wage demands. Finally, the degree of job protection for the ‘insiders’ (Lindbeck & Snower 1988), which is very high in the conservative regimes, might also contribute to the perception of poor and unemployed being out of control. Alesina & Glaeser (2004) have rightly argued that the American perception of the poor having good chances to escape poverty does not coincide with the facts. Referring to Gottschalk & Spolaore (2002), who compare the USA and German, and Checchi et al. (1999), who compare the USA and Italy, reality seems to be that the poor are more ‘trapped’ in the USA than in Germany and Italy. In the USA, 60 per cent of the bottom quintile in 1984 were still in this quintile in 1993, compared with 43 per cent in Germany. In the USA, 25 per cent of the fathers in the bottom quartile have children who also are in this quartile. In Italy, the share is 21 per cent. To explain this paradox, Alesina & Glaeser (2004, p. 76) refer to different ideologies and later to a general negative perception of the poor caused by racial divides (Alesina & Glaeser (2004, pp. 133–183). In contrast, we point at regime-dependent labour market structures as a more straightforward explanation. Job growth in the private service sector and the ability of individuals to negotiate the salary might not, on average, increase the chance of moving out of the lowest quintile, but the possibilities of getting a job gives the public the impression that each individual has a decent chance. Thus, the perception of unemployed as ‘victims’ of the general labour market structure might be especially pronounced in conservative regimes, where strong unions not only prevent low wage jobs, but also obstruct employment of outsiders through high job protection of the insider male breadwinner. So, based on the argument of structural differences in service sector expansion, wage setting mechanisms and job protection, we would expect the labour markets to facilitate perceptions of little control of neediness among the poor and unemployed in conservative regimes, medium control of neediness in social democratic regimes, and high control of neediness in liberal regimes, see Table 4.2. The three regime dimensions, perception of the poor and unemployed and fulfilment of deservingness criteria In Table 4.2 we have summarised the variation in the independent variables or, in less technical terms, how each ideal-type welfare state regime is placed on the three dimensions, which we believe are important for the public judgement of deservingness. The logic of the table is that a low score on a given dimension should make it easier for the poor and unemployed to fulfil the deservingness criteria, and a high score should make it more difficult.

Welfare Regimes and Deservingness

61

The ideal type liberal welfare state regime has high scores on all three dimensions. The degree to which the welfare system is dominated by selective welfare policies is high, the difference in economic resources between ‘the bottom’ and ‘the majority’ is high, and the degree of perceived job opportunities for the poor and unemployed is high. Following the arguments above, this suggests that the poor and unemployed in the ideal-type liberal regime will have difficulties meeting the deservingness criteria. The only thing that, in the eyes of the public might speak in favour of the poor and unemployed in liberal regimes, is that they really are in need. Table 4.2.

Three regime dimensions that influence perceptions of the poor and unemployed and the theoretical position of the ideal-type countries of Sweden, West Germany and the USA

Degree to which welfare systems are dominated by selective welfare policies:

Ideal type social democratic regime

Ideal type conservative regime

Ideal type liberal regime

Low

Medium

High

Low

Medium

High

Medium

Low

High

(Primarily effect on ‘identity’, ‘attitude’ and ‘reciprocity’ discussion): Difference in economic resources of ‘the bottom’ / ’ the potential poor’ and ‘the majority’: (Primarily effect on ‘identity’ but also on ‘need’ in opposite directions) Degree of perceived job opportunities for poor and unemployed : (Primarily effect on ‘control’ and ‘need’)

62

The Institutional Logic of Welfare Attitudes

In the ideal-type social democratic welfare state regime, the degree of selective welfare policy is low, and so is the difference in economic resources between ‘the bottom’ and ‘the majority’. These positions should make it easier for the poor and unemployed to fulfil the deservingness criteria. The perceived job opportunities are believed to be lower than in the liberal welfare state regime, which should decrease the perception of the poor and unemployed being in control of their neediness, but still not as low as in the conservative regimes. The only thing that, in the eyes of the public, really speaks against the poor and unemployed in social democratic regimes is the fact that they are not very much in need (at least by comparative standards). Finally, we argued that the ideal-type conservative welfare state regime has a middle position in terms of degree of selective welfare policy and degree of generosity towards poor and unemployed. In terms of fulfilling the deservingness criteria in the ideal-type conservative regime, the position on these two dimensions suggests that it should be more difficult than in the social democratic regime and easier than in the liberal regime. However, on the dimension of perceived job opportunities, we argued that the ideal-type conservative regime has the lowest position, which should make it easier to meet the (not in) control criterion. Still, we believe that in the overall judgement of deservingness, the poor and unemployed are worse off in conservative regimes than in social democratic regimes. Hence, we expect public support for welfare policy to be higher in the ideal-type social democratic regime than in the ideal-type conservative regime. One could argue that these cross-regime expectations are part of a circular argument, because we already know the result; namely that welfare support should end up being high in the social democratic regime, medium in the conservative regime, and low in the liberal regime. However, just being able to pinpoint the relevant intervening variables between welfare regimes and the pattern of support for welfare policy found in the previous studies is a step forward. And in the following four chapters we will, by means of both existing and new cross-national and national survey data, prove the soundness of the causal reasoning presented in Figure 4.1. In particular, it needs to be clarified whether the degree of identification with the poor and unemployed is a product of the regime-dependent institutions, as we suggest, or simply reflects the fact that welfare regimes go together with different degrees of ethnic homogeneity. As mentioned in Chapter 1, we will use a combination of different analytical strategies. In Chapter 5 we use WVS and Eurobarometer surveys to prove the connection between welfare regimes and public perceptions of the poor and unemployed. Even though we have observations across most of the relevant countries and over a long period of time, we cannot avoid the ‘small-n problem’ that haunts all comparative research. Therefore, Chapters 6 and 7 supplement the ‘traditional’ comparative analysis in Chapter 5 with two so-called focused comparisons. The idea is to ‘isolate’ the institutional effects derived respectively from the degree of selectivism and the degree of generosity. By comparing the Nordic countries we hold the size of the countries, ethnic fractionalization, social trust, political systems, cultural legacy etc more or less constant and look at the effects of differences in welfare policy (which actually can be found across the Nordic countries). The data

Welfare Regimes and Deservingness

63

for these ‘most similar designs’ come from a Nordic module to the International Social Survey Program (ISSP) from 1999 and a large Danish and Finish dataset. The latter dataset includes both a general population sample and a large sub-sample of long-term unemployed. These focused comparisons enable us to conclude with very high certainty that the institutional effects suggested by our theoretical framework are present. However, as mentioned in the introduction, one of the main problems with ‘most similar designs’ is that one cannot assess the relative importance of the independent variable in play. In our case, one would, for example, like to know how strong the institutional effects are compared with the effects-derived ethnic divides. Therefore, Chapter 8 applies our theoretical framework on a national sample that has been established exclusively for this purpose. The sample enables us to analyse, in one model, how public perceptions of (1) immigrants, (2) the poor and unemployed in general, (3) the generosity of the welfare state, and (4) job opportunities, affect public support for welfare policy. The analysis shows that the perceptions, which by then we know are influenced by welfare institutions, are highly relevant in order to explain variations in individuals’ support for welfare policy. From that we infer that they may also be highly relevant in order to explain cross-national variations in public support for welfare policy.

This page intentionally left blank

Chapter 5

Welfare Regimes and Perceived Causes of Poverty In this chapter, we will analyse the connection between welfare regimes and the perception that poverty is caused by ‘laziness and lack of will power’. As we expected regime-dependent perceptions of poor and unemployed to be the intervening variable between welfare regimes and support for welfare policy, we should expect a regime pattern on this question. We should also expect a connection between cross-national differences in perception of poverty caused by ‘laziness and lack of will power’ and the cross-national differences in support of welfare policy found in previous studies. The chapter will show that both expectations can – with a number of exceptions – be confirmed by the available data. These overall findings are presented in sections two and five. The chapter also has a more explorative dimension as we will try to narrow down (1) which of the three regime dimensions discussed in the previous chapter is most important and what happens when we control for differences in racial fractionalisation (section three); and (2) how the welfare regime effect interacts with individual level variables (section four). The former task is complicated by the fact that we only have a limited number of cases, but the analyses suggest that the dimension of (perceived) job opportunities and the dimension of selectivism are the two most important. The degree of selectivism is measured by a very rough proxy, but the finding will be substantiated by analyses in Chapter 6. The degree of ethnic fractionalisation has impact in the direction we expect, but the effect seems less important. The argument will be substantiated in Chapter 8. Finally, the effect from level of generosity is uncertain, but will be further explored in Chapter 7. The latter, more explorative, task can be conducted with greater accuracy, as the number of cases is very high at the individual level. Our theoretical framework did not discuss the possibility that the regime effect could influence groups differently. We assumed that citizens in a given country would be equally affected by the regime effect, even the poor and unemployed themselves. As the regime effect is not reduced much by taking variations in individual level variables into account, this assumption of a general regime effect is confirmed. However, some interesting interaction effects are also present. For example, we find that the difference in perception of the poor between skilled blue-collar and white-collar workers is larger in the liberal regimes than in the social democratic and conservative regimes; white-collar workers are less inclined to perceive poverty as a consequence of laziness. This finding suggests that the lack of support in the liberal regimes for welfare policy (at least in the narrow

66

The Institutional Logic of Welfare Attitudes

sense) might not so much be a matter of the middle class being against, as expected by the power resource theory, but more a matter of the working class not being in favour. We also find that the difference between skilled and unskilled blue-collar workers in perceiving poverty as being caused by ‘laziness and lack of will power’ is greatest in the conservative regimes. The most obvious explanation is the insider–outsider divided conservative labour market. Finally, we find the difference between males and females to be largest in the social democratic regimes. The finding coincides with previous studies that showed that women in social democratic regimes especially favour welfare policy. But rather than being a matter of female self-interest, the higher support could be explained by larger gender differences in perception of control of neediness. These analyses at the individual level are presented in the fourth section. First of all, we need to discuss our operationalisation of the data. The question of control of neediness In the Eurobarometer surveys conducted in 1976, 1989, 1993 and 2001 among members of the European Union and in the World Value Study 1990–93, the public was asked:1 ‘Why in your opinion, are there people who live in need?’ (Eurobarometer), and ‘Why are there people in this country who live in need?’ (WVS)

These questions are almost identical, although the Eurobarometer question does not specify that it refers to people in the specific nation state. However, as the rest of the questionnaire refers to a national context, we believe that the respondents will also apply this national context when they answer the Eurobarometer question. Fortunately, the possible answers given to the respondents were also almost identical. The Eurobarometer survey offered a choice between ‘four opinions’ that come ‘closest to their opinion’. In the World Value Study 1990–93, the respondents were given ‘four possible reasons’ and were asked to select the reason they considered ‘to be the most important’ and ‘to be the second most important’. The four ‘opinions’ and ‘possible reasons’ were:

1 The second wave of the World Value Study (1995–97) also asked the question, but the number of possible answers was reduced. Therefore we do not find as clear a regime effect. The data are presented and discussed in Appendix 5.1. In the third wave of the World Values Study 1999–2001, the question was not asked. In the fourth European Value Study from 1999–2001, the question was asked, but the data were released after the analyses were conducted. In this chapter, it makes no big difference as the primary argument is built on the World Value Studies from 1990, which includes the USA and Canada.

Welfare Regimes and Perceived Causes of Poverty

1. 2. 3. 4.

67

Because they have been unlucky Because of laziness and lack of will power Because there is much injustice in our society It is an inevitable part of modern progress

Even though we only have a single indicator, the responses to the question leave us in a situation with truly comparative survey data that cover the period from 1976 to 2001 and include almost all countries of relevance for our purpose. The data have been explored before by Oorschot & Halman (2000) and by Gallie & Paugam (2002), but with a somewhat different focus.2 Following the theoretical discussion in the previous chapter, we operationalise the control and identity dimension as ‘laziness and will power’ on the one side versus ‘unlucky’, ‘injustice in our society’ and ‘an inevitable part of modern progress’ on the other. The ‘laziness and will power’ answer clearly refers to a situation where ‘the poor’ are in control of their neediness. If the poor only made an effort, they would be able to escape their neediness. The answer also ‘taps’ some of the identity dimension, as explaining poverty with ‘laziness and lack of will power’ also denotes that the poor are persons with deviant characteristics compared to ‘ordinary’ citizens. The three other possible answers refer to situations where the poor are not in control of their neediness. Whether it is due to individual fate (‘unlucky’), social fate (‘an inevitable part of modern progress’) or social blame (‘injustice in our society’) is not seen as important to our analysis. This it not to say that the cross-national differences on these three categories are not interesting (Oorschot & Halman 2000), but when it comes to explaining the impact on public support for welfare policy, we find our operationalisation more relevant. In other words, we do not believe that public support depends much on whether this ‘uncontrolled neediness’ is perceived to be caused by personal fate, social fate or social injustice.

2 Oorschot & Halman only include the World Value Study 1990–93 and the Eurobarometer study from 1976, whereas Gallie & Paugam only include the Eurobarometer studies. So far, the ‘complete’ data material has thus not been explored, and furthermore the focus of the two previous studies has been somewhat different from ours. Gallie & Paugam were not interested in regime differences (as their report to the European Commission had a more narrow European perspective), which to some extent explains why it was not important for them to include ‘the real’ liberal countries of the USA, Canada, New Zealand and Australia. Only Great Britain and Ireland are covered by the Eurobarometer surveys. Oorschot & Halman were interested in regime differences, but focused on the dimensions of blame versus fate and individual versus social, i.e. they did not operationalise the answers on a control versus not-control dimension.

68

Table 5.1.

Country

The Institutional Logic of Welfare Attitudes

Share explaining poverty with ‘laziness and lack of will power’ in selected Western countries in the period from 1976 to 2001 World Value Study 1990-1993 39 32 21 27 30

Eurobarometer 1976

Eurobarometer 1989

30 43 37

15 18 17

10 13 12

18 22 20

39 32 19 25 24

Germany (W) France Belgium Austria Spain Portugal Italy Luxembourg Greece Conservative mean:

23 15 22 37 25 24 27 25

23 16 22 20 31 22

19 14 14 15 14 23 25 25 19

15 6 8 9 15 16 20 19 14

17 16 18 22 12 29 15 24 20 19

19 13 17 30 15 21 20 25 21 19

Sweden Norway Denmark Finland The Netherlands Social democratic mean:

16 11 14 26 15

11 12

18 10

9 9 15 14

9 19 15 12

14 10 14 19 13

16

12

14

12

14

14

USA Canada Ireland GB Liberal mean:

EuroEuroCountry barometer barometer mean 1993 2001

Weight: Eurobarometer 1976 not weighted. Eurobarometer 1989 weighted (nation weight II, ‘v8’), Eurobarometer 1993 weighted (weight from target, ‘v8’, Finland ‘v17’, Norway ‘v16’), and Eurobarometer 2001 weighted (weight from target, ‘w.1’). Source: Own calculations on World Values Study first and second wave plus Eurobarometer 5, 31A, 40 and 56.1

Welfare Regimes and Perceived Causes of Poverty

69

The connection between welfare regimes and perception of poverty being caused by laziness and lack of will power The empirical findings from Eurobarometer and the World Values Surveys are summarised in Table 5.1. The figures represent the proportion that, within the given country at the given time, indicated that poverty was caused by ‘laziness and lack of will power’. Now the question is how we shall apply our theoretical framework to these findings. One strategy is to accept that the real world is a messy place and limit the analysis to the three countries that come closest to the ideal types described by Esping-Andersen, i.e. Sweden, the USA and West Germany. From West Germany, we have fully comparable survey observations from 1976, 1989, 1990, 1993 and 2001, from Sweden we have observations from 1990 and 2001, and finally from the USA we only have one observation, namely 1990. If we compare the mean of these observations (the last column in Table 5.1), we find the expected order (39 per cent in the USA, 19 per cent in West Germany, and 14 per cent in Sweden). If we believe time in itself makes a difference and only compare the three countries in 1990 (the first column in Table 5.1), we also find the expected order; 39 per cent in the USA, 23 per cent in West Germany, and 16 per cent in Sweden.3 A second strategy is to analyse the clusters of countries that, according to EspingAndersen, belong to the three regimes. By doing so we become more ambitious as the theory now should be able to explain not only survey observations from countries closest to an abstract ideal type description, but also survey observations from all the other Western countries. We also apply another logic. In the former strategy we took the three countries that should be easiest for the theory to explain, and deviant cases are therefore difficult to accept. If the theory could not predict the ranking of the USA, Sweden and West Germany, we cannot believe that the theory could work anywhere. In the latter strategy, we have a larger number of more ‘difficult’ countries, but at the same time accept deviant cases. This is the well-known ‘large-n logic’, where we look for the general trend without paying attention to the peculiarities of the individual observations. In Table 5.1 we have applied this logic by calculating an overall mean from the 12 ‘liberal observations’, the 37 ‘conservative observations’ and the 17 ‘social democratic observations’.4 The means are 24 per cent in the 3 Even though unemployment at that time was extremely low (1.7 per cent) in Sweden and the Swedish welfare state was extremely developed. In 1990 West Germany also experienced less unemployment (4.8 per cent) than the USA (5.6), i.e. we seem to have a strong regime difference that exists even though the situation on the labour market and the development of the welfare state according to our theoretical framework should moderate the differences. We will formally test the impact of unemployment level in the analyses below. 4 The countries are clustered according to Esping-Andersen (1999). Here Australia, Canada, New Zealand, the United States ‘and to a degree’ Great Britain are classified as having a residual programmatic structure. Denmark, Finland, Norway, the Netherlands ‘and to a degree’ Great Britain are classified as countries with a universal welfare state, i.e. low-income groups are to a large extent covered by the same schemes and services that are provided to the whole population. Even though the Netherlands come close to the ideal type social democratic

70

The Institutional Logic of Welfare Attitudes

liberal cluster, 19 per cent in the conservative cluster, and 14 per cent in the social democratic cluster. The strength of this ‘average’ finding is that it enables us (to some extent) to rule out rival explanations of the ranking between Sweden, West Germany and the USA. Especially in the case of the USA, many scholars would argue that we are dealing with an exceptional case, which differs from other Western countries in so many ways that it is difficult to find out whether the survey results reflect welfare regimes or one of the other peculiarities. However, even if we exclude the USA, the other ‘liberal observations’ (i.e. Ireland, Great Britain and Canada) still have a larger share answering ‘laziness’ (22.6 per cent) than in the ‘conservative’ and ‘social democratic’ observations. Now if we become even more ambitious, and copy the search for laws of nature within natural science, deviant cases become problematic. In other words, it would have been a problem for Newton if one of his apples did not fall to the ground. In social sciences, we typically have more modest expectations, but if a substantial number of apples do not fall to the ground – other things being more or less equal – we might begin to doubt our causal reasoning. In Table 5.1, we do find a substantial number of deviant observations, but, in particular, the differences in job opportunities help explain why some ‘apples’ do not fall. As seen in Figure 4.1, we expect the perception of control to be influenced by job opportunities, which in the following we measure as differences in standardised unemployment rates (see below). We start by looking at the simple country mean presented in the last column of Table 5.1. In the liberal cluster Ireland stands out. With an average of 19 per cent explaining poverty with ‘laziness’, Ireland matches the general average found in the conservative observations. However, considering the extreme level of unemployment in Ireland in the 1980s and 1990s, this low level of perceived control might not be so surprising. Looking at country means in the conservative cluster, we find two countries, Austria with 30 per cent and Luxembourg with 25 per cent, with values above the general mean of 24 per cent in the liberal observations. However, here one should take into account that these two countries have experienced the lowest level of unemployment in Continental Europe and at the same time have had some of the highest social expenditures. In the conservative cluster we also find country means just below (France with 13 per cent) or just above (Spain with 15 per cent) the general level of 14 per cent found in the social democratic observations. This perception of low control in Spain is no surprise as the country has been haunted by the highest level of unemployment

welfare regime, the country is in general described as a ‘Janus-headed welfare regime, combining both social democratic and conservative attributes’ (Esping-Andersen 2000, p. 88). Finally, Austria, Belgium, France, Germany, Italy, Japan and Spain are categorised as countries with a programmatic structure based on social insurances (Esping-Andersen 2000, p. 85). Using this Esping-Andersen (1999) classification we have covered most of the Western capitalist countries included in the Eurobarometer and World Value surveys but we would like to extend the group of social insurance welfare states to include Portugal and Luxembourg.

Welfare Regimes and Perceived Causes of Poverty

71

in the Western Countries.5 Even though unemployment has also been relatively high in France, the perception of low control is more surprising; especially considering that France also has some of the highest social expenditures in Continental Europe, and that the racial distinctiveness of the poor and unemployed is probably higher in France than in many other Continental European countries. If we take Alesina & Glaeser’s measure (see Table 4.2), there is a 10 per cent probability that two randomly selected French are from different races, which indicates a much higher racial fractionalisation than in the rest of Europe.6 Finally, we also find a deviant case in the social democratic cluster, where Finland with an average of 19 per cent matches the general average of the conservative observations. This relatively high perception of control becomes even more remarkable when we take into account that Finland experienced an extremely high level of unemployment in 1993 and at the same time, within a Nordic context, has had the least developed welfare state. The case of Finland will be further analysed in the next chapter. So far, we have argued that our theoretical expectation seems valid both if we analyse the three ideal type countries and if we analyse the average of the countries families. Furthermore, we argue that if we look at the overall country means and take the level of unemployment as well as the welfare state development into account, the number of deviant cases might be reduced to France and Finland. If we look at the individual observations, more deviant cases can be found, but some of these can also be explained by level of unemployment and level of welfare state development. In the next section we will analyse the three sub-dimensions of welfare state regimes (see Figure 4.1), which enables us (1) to discuss the relevance of each dimension and (2) to represent each country by three regime-dependent variables instead of the simple three-brackets regime classification.7 The connection between selectivism, generosity, job opportunities, fractionalisation, and perception of poverty Comparative research has always been, and always will be, challenged by the limited number of cases. It makes it difficult to get results that live up to the normal scientific standard for not making a mistake. Furthermore, it limits our possibilities to build advanced models that try to keep everything equal by including a large number of control variables. Nonetheless, in this section we will try to break the overall regime variable used in most previous studies down to four lower level variables, which represent the three dimensions discussed in the previous chapter and ethnic fractionalisation as a control variable. We have tried to operationalise the dimensions in such a way that we get comparable data from the whole period 5 OECD standardised unemployment levels at time of interview in Spain were 17 per cent (1989), 16 per cent (1990), 23 per cent (1993), and 13 per cent (2001). 6 For political reasons, country of origin of immigrants from the many France colonies is not registered, which makes it impossible to analyse their socio-economic status. 7 See Jæger for another attempt to break down the regime variable (forthcoming).

72

The Institutional Logic of Welfare Attitudes

covered by the World Value Study and Eurobarometer survey. This proved very difficult as comparable statistics – and especially comparable reliable statistics – are scattered and scanty. The four following variables are used: • The least problematic dimension is the job opportunity dimension, as we simply used the standardised OECD unemployment rates that are available for almost the whole period.8 This aggregated measure is naturally not a perfect measure for job opportunities within a given country because much depends on the in- and out-flows from unemployment over a year. This might speak in favour of using, for example, long-term unemployment as a measure for job opportunities. However, we stick to the aggregate unemployment level as it has a key position in the public debate. For laymen, the level of unemployment is probably much more important for the perceived economic situation of a given country than the level of long-term unemployment. • To measure the degree to which welfare policy produces good life conditions for poor and unemployed, we will use the rough measure of social expenditures, as done by Taylor-Gooby (1995). Poverty rates based on the Luxembourg Income Study is another option, but this would only allow us to analyse a limited number of years and a limited number of countries, i.e. the small-n problem would increase. Instead we use the OECD database of Social Expenditures (SOCX), which offers a good account of public and socalled mandatory private (i.e. primarily social contributions) expenditures back to 1980. The aggregate level of social expenditures is presented as a relation of GDP in order to take the general level of economic development within a given country and the number of people living in this country into account. Furthermore, we follow Castles (2002) and exclude expenditures on unemployment compensation because cross-national differences in these expenditures do not reflect differences in welfare state development (for a similar argument see Esping-Andersen 1990, p. 19). As we cannot get accurate figures from 1976 and 2001, we will use the 1980 and 1998 social expenditures as a rough proxy.9

8 Standardised unemployment rates 1983–2001 taken from OECD Economic Outlook, volume 2002/1, No. 71, June. Standardised unemployment rates from 1976–82 taken from OECD Economic Outlook, volume 1996, No. 59, June. In four cases, standardised unemployment rates are not available; Denmark 1976, Ireland 1976, Luxembourg 1976 and Austria 1990. In these cases, we have used the commonly used definitions from OECD historical statistics 1970–2000; Denmark 1975, Ireland 1975. 9 Taken online from OCED source. 1 October 2003. The observations in 1980 do have a number of shortages (Castles 2002) but as a rough indicator of welfare state development we find the data rather valid. As our aim is to capture feedback from generous welfare policy on the neediness of low-income groups it might seem odd to exclude expenditures on unemployment but it is clear that, for example, a decrease in unemployment expenditures in Denmark from 5.3 per cent of GDP in 1993 to 3.4 per cent in 2001 does not reflect retrenchment of the Danish welfare state. It can be discussed whether expenditures on active labour market policy should also be excluded but, as done by Castles, we choose to include them. Furthermore, by excluding expenditures on unemployment we prevent one of the major sources of error, as these expenditures in 1980 not are available from Ireland and France. One could also argue that expenditures per unemployed could have been used to capture the ‘produced’ living

Welfare Regimes and Perceived Causes of Poverty

73

• Ethnic fractionalisation is also difficult to measure. The optimal indicator could, for example, have been the share of ethic minorities among the group below the relative poverty line or among the unemployed, but even finding the national stock of such minority groups is difficult. Recently, the OECD has made a large scale effort to provide a simple statistics of the size of the foreign or foreign-born populations, but even with these recent data it is more or less impossible to compare Anglo-Saxon countries with Continental European and Nordic countries (OECD 2003).10 Therefore we will use the measure of ethnic fractionalisation provided Alesina et al. (2003) as a (very) rough proxy. Their figures are based on information from the Encyclopaedia Britannica and they calculate the probability of two randomly selected individuals belonging to different ethnic groups. The boundaries of an ethnic group are very blurred as they not only include language differences ‘but also other cleavages such as racial characteristics’ (Alesina et al. 2003, p. 157). Furthermore, in order to make a best case for the argument of ethnic differences we also use Alesina & Glaeser’s (2004) measure of racial fractionalisation, even though we do not have data on all the relevant countries.11 We then found this racial fractionalisation measure to be the strongest predictor of social spending. • Finally, it is also very difficult to measure the degree to which the poor and unemployed are covered by targeted benefits/services. The Social Citizenship Indicator Program (SCIP) established at SOFI in Sweden might solve this problem in the future, but at the moment the database is not accessible. Therefore, we will simply reuse EspingAndersen’s rough distinction between the three regimes, as the distinction from the beginning relied most on the state dimension. The logic is that what is left of the regime effect after taking level of unemployment, social expenditures, and ethnic fractionalisation into account must be assigned to the degree of selectivism. The effect is captured by two dummy variables: belonging to the social democratic regime (0 or 1) and belonging to the conservative regime (0 or 1), i.e. the liberal regime functions as a reference category.

conditions among low-income groups. However, such a measure do not capture the living conditions ‘produced’ by all the other types of benefits and the service sector. 10 Since 2000, the OCED has tried to deliver comparable data on immigrations but even on this more narrow issue we have considerable data problems. Comparison between Anglo-Saxon countries and European countries is more or less impossible as the former group defines immigrants as people born in another country, whereas the latter group defines immigrants as people without citizenship. Therefore, we have also tried to look for possible operationalisation within the survey data. As we shall see, the further analyses relied to a large extent on the first wave of the World Value Study. But fortunately the question of ethnic background has only been asked in the USA, Canada and Spain, which makes it more or less impossible just to establish a rough proxy. 11 We have the figures on ethnic fractionalisation from all the relevant countries but we do not have separate figures for the former East and West Germany. Therefore, West Germany will be given the overall German value (0.17). We do not have the racial fractionalisation figures from Ireland, Finland, Canada, and Luxembourg. Instead of deleting these cases we have used the figures on ethnic fractionalisation. We have tried to obtain these figures from professor Alesina but without success.

74

The Institutional Logic of Welfare Attitudes

As a point of departure we use the World Values Study from 1990 because it contains most countries, and because it includes the ‘true’ liberal regimes of the USA and Canada. A simple linear regression model is used, and in model I (see Table 5.2) we have only included the two dummies, which gives us clear indications of a regime effect. Belonging to a social democratic regime is estimated (OLS) to reduce the aggregated level that believes the poor are in control by 13.4 percentage points. As expected, belonging to the conservative cluster also reduces the share that believes the poor are in control and, with an estimated 5.8 percentage point decrease, the effect is smaller than in social democratic countries. So far, however, the two dummies catch the whole regime effect and not only the degree of selectivism. In the following models we introduce the other dimensions. Table 5.2.

Prediction of proportion answering in poverty caused by ‘laziness and lack of will power’ based on degree of selectivism, level of unemployment, social expenditures, and ethnic fractionalisation (OLS). World Value Study (1990)

Model: Belonging to social democratic regime (dummy) Beta coefficient unstandardised Beta coefficient standardised Sig. Belonging to conservative regime (dummy) Beta coefficient Beta coefficient standardised Sig. Level of unemployment Beta coefficient Beta coefficient standardised Sig. Social expenditures (without unemployment) / GDP Beta coefficient Beta coefficient standardised Sig. Ethnic fractionalisation Beta coefficient Beta coefficient standardised Sig. R2 N Coefficients significant at 0.10 are highlighted.

I

II

III

IV

-13.4

-16.6

-12.3

0.01

0.03

0.07

-10.9 -0.65 0.12

-5.0

-5.8

-4.1

0.25

0.16

0.36

-0.8

-0.9

0.10

0.08

-0.5 0.34

-3.2 -0.20 0.48 -0.10 -0.43 0.08

-0.4 -0.24 0.45 8.9 0.23 0.35

0.43 16

0.54 16

0.58 16

0.62 16

Welfare Regimes and Perceived Causes of Poverty

75

Model II in Table 5.2 introduces the unemployment level, and as expected the effect from the regime dummies increases – from -13.4 to -16.6 on the social democratic dummy and from -5.0 to -5.8 on the conservative dummy. Furthermore, the risk of there being no effect from belonging to the conservative cluster (compared to belonging to the liberal cluster) decreases from 0.25 to 0.16 which, in this kind of analysis, we find quite acceptable. The level of unemployment has a beta-coefficient at -0.8, i.e. an increase in unemployment at one percentage point is estimated to reduce the perception of the poor being in control by 0.8 percentage points. This coincides with previous studies that found a strong connection between level of unemployment and support for welfare policy (Blekesaune & Quadagno 2003). But rather than being a matter of self-interest (among unemployed and among those who think they are in risk of unemployment), our theoretical framework suggests that the effect is caused by the influence of unemployment levels on the perception of the poor and unemployed as being in control of their neediness. Model III introduces the level of social expenditures, which was expected to have an influence on the perception of the poor and unemployed as being in need. Figure 4.1 showed a strong connection between level of social expenditures and proportion in a given country that perceived ‘the poor’ and ‘the unemployed’ as being sufficiently well covered. So, on the one hand, we believe that the level of social expenditures could increase the proportion that answers ‘laziness and lack of will power’. But on the other hand, the theoretical framework also suggests a decrease because a generous welfare system allows the poor and unemployed to uphold a lifestyle close to that of the majority, which makes ‘the bottom’ less culturally distinct. All together, model III estimates a negative effect (-0.5) from the level of social expenditures on the proportion that answers ‘laziness and lack of will power’, i.e. the identity effect connected to generous welfare policy seems to overrule the need effect. As a consequence, the effect from the two dummies that capture degree of selectivism is reduced from -16.6 to -12.3 and from -5.8 to -4.1. So far, we should not perceive this as a bullet-proof argument, but the influence from degree of selectivism and generosity will be substantiated by the analyses in the next chapter. Finally, model IV introduces our measure for overall ethnic fractionalisation within a given country. As expected, a higher probability of drawing two individuals with different ethnic backgrounds is estimated to increase the proportion that answers laziness and lack of will power. The unstandardised beta-coefficient is 8.9, which is the increase one could expect if we had a change from everybody belonging to the same ethnic group (i.e. the probability of picking persons with different ethnic backgrounds would be zero) to everybody belonging to a different ethnic group (i.e. the probability of picking persons with different ethnic backgrounds would be one). Both situations are unlikely, and if we look at the standardised coefficients (0.23), we see that the effect is a little below the effect from social expenditures (-0.24), and in both cases the level of significance is problematic.12 Still the coefficients for ethnic 12 So the significance level for the conservative regime dummy in model IV, is 0.48. However, as the two dummies are meant to measure a one-dimensional scale of selectivism,

76

The Institutional Logic of Welfare Attitudes

fractionalisation point in the expected direction, and as expected the effect from selectivism is further reduced from -12.3 to -10.9 and from -4.1 to -3.2. This estimate suggests that the degree of ethnic fractionalization is part of the overall regime effect, but in contrast to Alesina & Glaeser (2004) it does not overrule the other dimensions we have deduced from the welfare regime theory.13 Looking at the standardised betacoefficients, the two most important variables are being minimum selective or not (social democracy dummy; -0.65) and level of unemployment (-0.43). The same kind of simple OLS regression can be applied to the Eurobaromenter surveys, but the results will be influenced by the fact that (1) they have fewer cases, i.e. the chances of significant results are lower, and (2) they do not include the countries that come closest to the ideal type liberal regime type. However, the results from 1976 actually support our expectations. We find a negative strong effect on the proportion that answers laziness (-29.4) in the countries with a minimum of selective policy (the social democratic regimes) and a smaller negative effect (-19.1) in the countries with medium selective policy (the conservative regimes). The impact of unemployment level is negative (-0.85), the impact of social expenditures is positive and, finally, as expected, the effect from ethnic fractionalisation is positive (+15.9), but also uncertain (significance 0.65). The regime effects are more blurred in the other Eurobarometer surveys, which could be interpreted as the regime theory being most valid for the golden age. However, taking the results from the World Values Survey from 1990 into account, one could also argue that the findings are highly influenced by the fact that Ireland and Great Britain are not good representatives for the ideal type description of the liberal regime.14 If we compare the countries with minimum (social democratic) and medium (conservative) selectivism, we find that belonging to the former group decreases the proportion that answers laziness. We also find that the importance of unemployment level seems to increase. The impact of social expenditures shifts direction from 1976/1989 to 1993/2001, but most of all it reflects the very modest impact, i.e. the coefficient is around zero. More surprisingly, the role of ethnic fractionalisation also we could have made a scale with three categories (1 = very selective = liberal; 2 = medium selective = conservative; 3 = minimum selective = social democratic). If we do that, we get an unstandardised beta at -5.3 with an acceptable significance level at 0.11. 13 We have also tried to include Alesina & Glaeser’s measure for racial fractionalisation, which turned out to have the strongest effect in their analyses. In model IV the unstandardised beta-coefficient for racial fractionalisation became 10.4 (compared to 8.9 for ethnic fractionalisation), but the level of significance increased to 0.44 (compared to 0.35). Alesina & Glaeser’s analyses do not include the regime variables at all. Furthermore, they include many non-Western countries (even though they primarily want to explain the difference between USA and Europe), and finally they primarily have social expenditures as a dependent variable. However, they use the perception of the poor and unemployed as a crucial intervening variable in their causal reasoning. 14 Furthermore, it has been argued that, especially in the British case, the Thatcher period generated a very positive attitude towards the welfare state, i.e. a counteraction to the pursued policy (e.g. Pierson 1994, pp. 146–49; Mau 2003, p. 21). See Chapter 9 for a further discussion of this argument.

Welfare Regimes and Perceived Causes of Poverty

Table 5.3.

Prediction of share answering ‘in control’ based on regime belonging, level of unemployment and social expenditures (OLS) EURO EURO EURO EURO 1976 1989 1993 2001

Model: Belonging to social democratic regime (dummy) Beta coefficient Sig. Belonging to conservative regime (dummy) Beta coefficient Sig. Level of unemployment Beta coefficient Sig. Social expenditures (without unemployment) / GDP Beta coefficient Sig. Ethnic fractionalisation Beta coefficient Sig. R2 N

77

Pooled WVS 1990, EURO 1976, 1993, 2001 II III IV

IV

IV

IV

IV

-27.2

-5.0

+0.76

-4.5

-12.0

-8.7

-8.3

0.14

0.52

0.88

0.31

0.00

0.00

0.01

-18.1

+0.80

+1.23

+3.2

-5.5

-4.1

-4.4

0.10

0.87

0.77

0.43

0.01

0.06

0.04

-0.85

-0.39

-0.36

-0.90

-0.8

-0.8

-0.8

0.73

0.41

0.25

0.05

0.00

0.00

0.00

+0.19

+0.11

-0.26

-0.19

-0.4

0.4

0.91

0.87

0.56

0.59

0.07

0.08

+15.9

-0.50

-0.62

-5.7

+5.0

0.65

0.96

0.95

0.48

0.28

0.88 9

0.26 12

0.22 14

0.55 15

Coefficients significant as 0.10 has been highligted

0.38 66

0.41 66

0.42 66

78

The Institutional Logic of Welfare Attitudes

shifts direction from 1976 (+15.9) to 2001 (-5.7), which actually suggests that higher ethnic homogeneity in 2001 helps to decrease the proportion that answers laziness. This is a small mystery, but we should keep in mind that ethnic fractionalisation is only a very rough proxy for what we really want to measure; namely the cultural distinctiveness of the poor and unemployed. The problems of significance are primarily caused by the limited number of cases, which constitutes the biggest problem for applying the quantitative ‘logic’ to cross-country research. One method to ‘boost’ the number of cases is to imagine that the observations made in the same country, but at different times represent independent cases. Without being explicit about it, that was what we did when we calculated an overall regime mean in Table 5.1. There are many problems associated with analyses conducted on such pooled data, as the assumption of independent observations is easily violated (Sayers 1989). It is obvious, for example, that the level of unemployment or social expenditures in 1989 strongly affects the level of unemployment or social expenditures in 1990, i.e. the chance of autocorrelation is high.15 However, if for a moment we ‘boost’ the number of cases in artificial ways (and afterwards have a detailed look at eventually broken assumptions, see Footnote 17) we end up with 66 cases. These 66 cases were entered in a simple constant coefficients model, i.e. we assume that all coefficients are the same for each cross-section in the pool. This is a rather strict assumption, but as our previous theoretical arguments do not imply changed effects over time, and as our number of observations still is rather limited, we find the constant coefficient model sufficient for our purpose.16 The coefficients of such a model confirm the results found in the first World Value Study. If we take model IV based on the pooled dataset, pursuing a minimum of selective policy 15 The same problem exists on the regime dummies and the ethnic fractionalisation measures, as we do not have any variation at all. If a country is grouped as having minimum selective policy in 1976 (social democratic), it falls in the same group in 1990, 1993 and 2001. But as already discussed, there is no easy way to solve this problem as indicators simply do not exist. 16 However, it should be mentioned that Gallie & Paugam on the Eurobarometer data find a time effect; even taking unemployment level into account, time seems to decrease the proportion that answers laziness from 1976 to 1993, but then again to increase the proportion from 1993 to 2001 (Gallie & Paugam 2000, p. 22). Gallie & Paugam give no explanation for this effect. The time effect from 1976 to 1993 suggests that not only the actual level of unemployment, but also the ‘cumulative experience’ of unemployment matters, i.e. as the year went by after the oil crisis of the 1970s it might have become more obvious that unemployment was structural (especially in the mid-1990s) and therefore the poor were not in control. The better employment situation from 1993 to 2001 could have softened this situation, but it should mainly be captured by the unemployment variable. Furthermore it should be mentioned that our attempt to model this time effect has not been very successful. If we include the percentage change in unemployment from 1993 to 2001 as well as the percentage change in unemployment prior to the other surveys (1968–76, 1981–89, 1985–93) (commonly used definitions) in the pooled model, some of the time effect is explained, but we still have an increase from 1993 to 2001 that cannot be explained.

Welfare Regimes and Perceived Causes of Poverty

79

(social democratic regimes) has a strong negative effect on the proportion answering laziness (-8.3), pursing medium selective policy (conservative) has a medium effect (-4.4), and a higher level of unemployment (-0.8) reduces the perception of control. These three variables are now significant. As in the World Value Study 1990, the effect of social expenditures is negative (-0.4) and less significant (0.08). And finally higher ethnic fractionalisation is estimated to increase the proportion answering laziness (+5.0), but the effect is not significant (0.28). These models, even though based on pooled data, do not seem to violate the basic assumption of OLS-regression, but it is an artificial way to increase the number of cases.17 Together, these analyses, which tried to replace the overall regime category with the four deduced sub-dimensions, showed that, in particular, the level of unemployment and the degree of selectivism turn out to be important for the proportion answering laziness. Naturally, we should take into account that the measure of selectivism is very rough, but as mentioned the finding will be substantiated by the analyses in the next chapter. So far, we can at least conclude that the effect from the regime dummies did not disappear after including ethnic fractionalisation, level of unemployment, and level of social expenditures. However, the effects from the regime dummies did decrease, which suggests that the overall regime effect does include effects from all four sub-dimensions. Thus, we should not make the extreme stat-centred institutional argument, which claims that only institutions matter. In the next section we turn to the more solid ground of the individual level. The individual level variation, the relative size of regime effect and the existence of possible interaction and composition effects So far we have described and analysed the differences between countries and not looked at the differences between individuals within each countries. Naturally, differences between countries are our main concern, as we are looking for the possible intervening variables that will enable us to explain the link between welfare regimes and cross-national differences in public support for welfare policies. Furthermore, our theoretical framework did not pay much attention to the individual level; for example it was not discussed whether the four regime dimensions would influence certain groups and not others. Thus, implicitly, the thesis is that the regime effects 17 For model IV, the Durbin–Watson test gives the value of 2.3, i.e. the correlation of the residuals is positive, but fortunately the value is not below 1, which indicates a serious problem. The tolerances of the variables are 0.96, 0.86, 0.59, 0.48, and 0.33. The latter is the tolerance of the social democratic dummy. The relative low value is caused by the fairly high correlation with level of social expenditures, but in statistical terms 0.33 is not an indication of serious multicollinarity problems. The histograms of standardised residuals show that the residuals seem to be rather normally distributed. The q-q plots do not perfectly follow the diagonal, but it is difficult to judge whether there is serious deviation from normality. Finally the ZPRED-ZSRID plots give us ‘shotgun pictures’, which indicate no serious violation with the assumption of linearity and heteroscedasticity.

80

The Institutional Logic of Welfare Attitudes

influence all inhabitants in a given country, even the poor and the unemployed themselves. Whether this is the case or not will be explored below. First we explore the so-called compositional effects, and then we explore the so-called interaction effects. We will limit ourselves to the World Value Study 1990–93 as it has the largest number of countries and representatives for the ‘real’ liberal countries. Furthermore, we limit ourselves to a relatively simple binary logistic model, which we have run for the whole sample and afterwards run for the clusters of liberal, conservative and social democratic countries (see Table 5.3).18 Compositional effects In contrast to our thesis of a general regime effect on support for welfare policy caused by degree of selectivism, generosity, job opportunities, and cultural distinctiveness of poor and unemployed, one could suggest that the regime effect seen above is (or partly is) an outcome of difference in the sizes of the groups, which in all regimes have more negative perceptions of the poor and unemployed. As a point of departure we find the argument very speculative. Of course there are differences in terms of number of old people, highly educated, unemployed etc, but it is difficult to see how they should be so strongly related to the welfare regime that they can account for the regime effect found in previous sections. These composition effects might be more plausible if we consider, for example, the number of left-wingers in a given regime, but then one can argue that we actually see the effect of regime belonging. In the latter case, it is at least possible to argue both ways, which for example made Svallfors (1999, p. 97) argue that Bean & Papadakis (1998) had erroneously included 18 In recent years, multilevel models that can analyse both the individual level and the national level at the same time have been developed, but for the time being we stick to a simple binary logistic model on the individual level with the regime variables introduced as dummies. First of all, the multilevel model secures that the level of significance of the aggregated variables is not inflated, as was the problem in Gallie & Paugam (2002). However, we have already conducted analysis on the aggregated level and seen that the regime effect is indeed significant. Aside from ignoring some interesting information, the main problem with conducting the analysis on the aggregated data is the risk of ecological fallacy, i.e. inferring from connection at the national level to connection at the individual level. However, in our case we have not made such inferences as degree of selectivism, unemployment level, level of social expenditures, and ethnic fractionalisation are ‘through’ second level variables. Finally, the multilevel model can normally give us a number of additional information as the variation in the data can be decomposed into individual level variation and national level variation. However, in our case we have binary dependent variables, which force us to use logistic regression. This can also be done within the framework of multilevel modelling, e.g. with Mlwin software, but most of the normal additional information provided by a multilevel model is lost. Therefore we will stick to the simple binary logistic regression on the individual and keep in mind that the level of significance of the regime effects in this model is inflated. Anyhow, we need more than 16 level-two cases to use the advantage of multilevel analyses, and when it comes to the estimated coefficients those obtained above should vary substantially from those obtained from a multilevel model (Rasbach et al. 2000, p. 95).

Welfare Regimes and Perceived Causes of Poverty

81

the political orientation variables in the regression models and therefore did not find a regime effect. Below we will take the most ‘conservative’ approach and show that even with a political orientation variable included we find a strong connection between welfare regimes and perception of control of poverty. In the first model (see Table 5.4), we have simply introduced the macro-level variables of regime belonging and unemployment rate, which have already been studied. As the level of social expenditures and degree of ethnic fractionalisation turned out only to have modest effects on the proportion answering laziness, these variables have not been included. The logistic technique, which estimates the odds of a person answering laziness, also indicates a very strong connection (at least in 1990 and for the 16 Western countries available in the World Value Study). The chance of a person living in a liberal regime answering ‘laziness’ is estimated to be 2.6 times that of a person living in a social democratic regime. The chance of a person living in a conservative regime answering ‘laziness’ is around twice as high (1.98) as that of a person living in a social democratic regime – taking differences in unemployment levels into account. In the following four models, we enter individual level variables; first ‘the usual suspects’ i.e. sex, age, education, and employment status (model II), then occupation (model III), left/right-wing orientation (model IV) and finally attitudes towards minority groups (model V). The overall conclusion is that taking differences in all these variables into account does not reduce the estimated regime effect at all, i.e. differences in sizes of these groups cannot explain the regime effect found. The chance of respondents answering ‘laziness’ remains around 2.6 times higher in the liberal regimes and 1.9 times higher in conservative regimes in all the models. And the impact from level of unemployment also remains stable.

82 Table 5.4.

The Institutional Logic of Welfare Attitudes The connection between welfare regime, unemployment level, individual background variables, political orientation and perception of poor being in control of poverty in the World Value Study 1990–93, 16 Western countries. Binary logistic regression, odds-ratio presented. (1= ‘laziness’ and 0=one of the three other answers). (* significant at 0.05, ** significant at 0.01)

Model: Liberal regime Conservative regime Social democratic regime Unemployment rate Female Male 18-24 25-34 35-44 45-54 55-64 65Completed education below 15 year old 15-16 years old 17-18 years old 19-20 years old 21 years old or above Employed (full-time, part-time, self-employed) Unemployed Retired Other out of labour force Employer/manager of establishment with 10 or more employees. Employer/manager of establishment with less than 10 employees. Farmer (employer, manager on own account) Professional worker, lawyer, accountant, teacher etc. Mid-level non-manual office worker etc. Junior level non-manual office worker etc. Foreman and supervisor

I 2.60 1.98 Ref. 0.97

II 2.44 1.77 Ref. 0.97 Ref. 1.28** 1.10ns 1.06ns Ref. 1.19** 1.48** 1.61** 1.09ns

III 2.63 1.91 Ref. 0.97 Ref. 1.19** 1.12ns 1.07ns Ref. 1.19** 1.46** 1.55** 1.06ns

IV 2.58 1.96 Ref. 0.97 Ref. 1.19** 1.13ns 1.09ns Ref 1.14* 1.37** 1.43** 1.08ns

V 2.62 1.92 Ref. 0.98 Ref. 1.18** 1.13ns 1.09ns Ref 1.14* 1.33** 1.39** 1.08ns

0.97ns Ref. 0.93ns 0.74** Ref.

0.93ns Ref. 0.95ns 0.77** Ref.

0.95ns Ref. 0.95ns 0.78** Ref.

0.95ns Ref. 0.95ns 0.79** Ref.

0.62** 0.91ns 0.94ns

O.65** 0.92ns 0.94ns 1.15ns

0.67** 0.94ns 0.93ns 1.04ns

0.67** 0.94ns 0.92ns 1.04ns

1.50**

1.40**

1.40**

1.78**

1.63**

1.61**

0.84*

0.82**

0.84**

0.86*

0.84*

0.85*

0.81**

0.79**

0.80**

1.23ns

1.23ns

1.24*

Welfare Regimes and Perceived Causes of Poverty

83

Table 5.4 continued Skilled manual worker Semi-skilled manual worker Unskilled manual worker Agricultural worker Member of armed forces Never had a job Left/right-wing political orientation on 1 – 10 scale (subjective). Dislike in minority groups as neighbours1 Do not mention N (not weighted)

Ref. 1.11ns 0.91ns 1.44* 1.43ns 1.03ns

Ref. 1.10ns 0.91ns 1.34ns 1.28ns 0.98ns 1.12**

Ref. 1.09ns 0.91ns 1.32ns 1.25ns 0.99ns 1.13**

1.39** Ref. N= 17.144

Note: We have not used the weight provided in the World Value Study because N in small countries has been artificially reduced, so that e.g. Norway does not have the same impact as USA in the overall analysis. However, in our case we are interested in treating the countries with equal weights (analyses have also been conducted with weighted data and except lower significance levels in the sub-sample of social democratic countries due to smaller N, the results are more or less identical. 1 Either mentioned ‘people with different race’, ‘Muslims’, or ‘immigrants/foreign workers’. These findings support our line of reasoning in Chapter 4, namely that cross-national differences in support of welfare policy should be explained with macro-level variables and not with micro-level variables. However, in order to understand differences in individuals’ support for welfare policy, which have also been a topic in the previous comparative research, it might be interesting to take a brief look at the impact of different individual level variables on perceptions of poor being in control of neediness. In model II we find a significant effect of men being more inclined than women (odds-ratio 1.28) to perceive the poor as being in control of poverty. We also find a substantial age effect. Taking the middle-aged (35–44) as a reference, we see that the 45–54 year-olds (odds-ratio 1.19), but especially the 55–64 year-olds (oddsratio 1.48) and those above (odds-ratio 1.61) are much more inclined to perceive the poor as being in control of poverty. There are no significant differences between the middle-aged and the younger groups. It is extremely difficult to build a comparable measure for education across countries, but if we include the rough measure based on the year the respondent finished his/hers education provided in the World Value Study, we see a ‘high education effect’; those who finished their education at age 21 or later are significantly less inclined to perceive the poor as being in control of

84

The Institutional Logic of Welfare Attitudes

poverty (odds-ratio 0.74). Finally, we see that, compared to the employed, those currently unemployed are much less inclined to perceive the poor as being in control of poverty (odds-ratio 0.62). However, the effect of living in a liberal regime compared to living in a social democratic regime is still estimated to be much more important than whether a person is unemployed or not. Model III enters occupational status, which in the World Value Study has been ordered in 13 categories. We have chosen not to recode this variable, as the original categories actually give us more information than the typical distinction between working, middle and upper class. The results in Table 5.3 clearly reveal that the perception of the poor being in control of poverty cannot be reduced to a classic ‘capitalist’ conflict between working class and bourgeoisie. Using a skilled manual worker as a point of reference, we actually see that employers with 10 or more employees are not significantly more inclined to answer ‘laziness’ (odds-ratio 1.15, ns). But we do find a very strong tendency of especially farmers (odds-ratio 1.78) and employers with less than ten employees (odd-ratio 1.50) to be more inclined to answer laziness. The effect from this group of more or less independent ‘hard workers’ even seems to be followed by the agricultural workers, which in this model is significantly more inclined to answer ‘laziness’ (odds-ratio 1.44). In contrast, all three categories of white-collar workers are less inclined than skilled workers to perceive the poor as being in control of poverty (professional workers etc, odds-ratio 0.84; mid-level nonmanual office worker etc, 0.86; and junior level non-manual office worker etc, oddsratio 0.81). So if our ‘political men’ are more inclined to base their policy attitudes (at least towards welfare policy in the narrow sense) on these kinds of perceptions of poor and unemployed than on class interest, it is understandable that previous studies have had difficulties finding clear-cut class effects. In model IV we have entered the self-reported left/right-wing orientation of the individuals. On the 10-point scale, 1 meant far left, and 10 meant far right. Basically, the variable behaves as expected, one step towards the right is estimated approximately to increase the chance of answering ‘laziness’ by 12 per cent, but the effects found in the previous models are not altered substantially (except that the effect from being an agricultural worker becomes insignificant).19 This means that we have an independent effect from the background variables that is not mediated from the left/right-wing orientation. For example, men being more inclined to answer ‘laziness’ is not simply an effect of men being more right-wing. Finally, in model V, 19 Not even taking political orientation into account changes the results. However, it should be remembered that left–right orientation typically is answered within a national context (e.g. Americans having more less the same average left–right orientation (5.74) as Swedes (5.69) does not necessarily mean that the former group of individuals is not more rightwing). We have looked for other variables that could capture general egalitarian values, but unfortunately only relative measures are found in the WVS. Taking variations in the two 10point scales, ’Private ownership of business and industries should be increased – Government ownership of business and industry should be increased’ and ‘Individuals should take more responsibility for providing for themselves – The state should take more responsibility to ensure that everyone is provided for’ does not reduce the estimated regime effect either.

Welfare Regimes and Perceived Causes of Poverty

85

we entered whether respondents have a problem with having ‘people of a different race’, ‘Muslims’ or ‘immigrants/foreigners’ as neighbours. It is estimated that those who mentioned one or more of these three groups are 1.39 times more likely to answer ‘laziness’ than those with no problems having these groups as neighbours. On the one hand, this fairly strong effect supports the argument that the negative perception of these minority groups goes together with a negative perception of the poor. On the other hand, the regime effect (modelled this way) is not reduced at all, so one should not point to ethnic conflicts as the one and only variable (which, for example, Alesina & Glaeser, 2004, tend to),20 see also below. Interaction effects As mentioned, our theoretical framework in Chapter 4 did not discuss whether regime characteristics, e.g. the high degree of selectivism, the low generosity, the good (perceived) job opportunities, and high cultural distinctiveness found in the ideal type liberal regime, should have a special impact on some groups and not on others. However, an explorative analysis of these so-called interaction effects can serve (1) as inspiration for further theoretical development and (2) as a possible explanation of the interaction effects found in the previous comparative studies of support for welfare policy. The attitudes of white-collar workers are theoretically most important, and running the above statistical model for each regime cluster actually reveals some differences. The described difference between skilled workers and white-collar workers appears to be largest in liberal regimes (odds-ratio 0.57, 0.76, and 0.63), whereas the differences more or less disappear in social democratic regimes (odds-ratio 1.16, 0.98 and 0.98) and conservative regimes (odds-ratio 1.10, 0.88 and 0.89). Thus, the tendency that white-collar workers are less inclined than blue-collar workers to answer ‘laziness’ is especially found in liberal regimes.21 This finding suggests that the liberal regime characteristics specifically generate negative perceptions of poor among blue-collar workers.22 If this is the right interpretation, the implication seems to be that the lack of support for welfare policy (at least in the narrow sense) 20 In the final chapter they conclude that ‘this importance of ethnic fractionalization cannot be overemphasized. … Racial differences between rich and poor facilitated the propagation of views such as “all poor are lazy” precisely because racist views associate laziness with different skin colors. In Sweden where, say, 95 per cent of the population has the same race, ethnicity, and religion, it is much more difficult to identify the poor with some racial characteristics’ (Alesina & Glaeser 2004, p. 218). 21 The difference has formally been examined by testing the interaction effect between belonging to one of these three white-collar groups and living in a liberal regime or not. The difference in behaviour of this white-collar group in liberal regimes is significant at the 0.01 level. 22 Such an effect could, in the power resource tradition, be explained by a weaker labour movement unable to convince skilled workers that poverty is a dysfunction of capitalism. Furthermore, it should be mentioned that the effect is not simply caused by the USA, but is also prevalent in Great Britain, Ireland and actually strongest in Canada.

86

Table 5.5.

The Institutional Logic of Welfare Attitudes

The connection between welfare regime, unemployment level, individual background variables, political orientation and perception of the poor being in control of poverty in the World Value Study 1990–93. Binary logistic regression, odds-ratio presented (1=’laziness’ and 0=one of the three other answers)(* significant at 0.05, ** 0.01) The liberal countries1

Unemployment rate Female Male 18-24 25-34 35-44 45-54 55-64 65Completed education below 15 year old 15-16 years old 17-18 years old 19-20 years old 21 years old or above Employed (fulltime, part-time, self-employed) Unemployed Retired Other out of labour force Employer/manager of establishment with 10 or more employees. Employer/manager of establishment with less than 10 employees. Farmer (employer, manager on own account) Professional worker, lawyer, accountant, teacher etc. Mid-level non-manual office worker etc. Junior level non-manual office worker etc. Foreman and supervisor Skilled manual worker Semi-skilled manual worker Unskilled manual worker Agricultural worker Member of armed forces Never had a job

The conservative countries1 1.00 Ref. 1.11* 1.00ns 0.98ns Ref. 1.04ns 1.24* 1.25* 1.16* 1.11ns

The social democratic countries1 0.95 Ref. 1.54** 1.66** 1.35* Ref. 1.47** 1.69** 1.75** 1.16ns 9.11ns

1.00ns 0.74** Ref. 0.63** 1.01ns 0.94ns 1.23ns

0.77ns 0.74* Ref. 0.72ns 0.77ns 1.05ns 1.44ns

1.29ns

1.36**

1.56ns

1.53* 0.57**

1.44** 1.10ns

2.20** 1.16ns

0.76* 0.63** 1.17ns Ref. 1.00ns 0.94ns 1.33ns 2.39** 1.17ns

0.88ns 0.89ns 1.23ns Ref. 1.10ns 0.76ns 1.35ns 0.64ns 0.93ns

0.98ns 0.98ns 1.17ns Ref. 1.53ns 1.07ns 0.75ns 2.5ns 1.13ns

0.89 Ref. 1.11ns 1.28* 1.19ns Ref. 1.22ns 1.35* 1.48** 0.99ns 0.79** Ref. 1.01ns 0.76** Ref. 0.70* 0.91ns 0.98ns 0.74ns

Welfare Regimes and Perceived Causes of Poverty

87

Table 5.5 continued Left/right-wing political orientation on 1 – 10 scale (subjective). Dislike in minority groups as neighbours2 Do not mention N (not weighted)

1.11**

1.13**

1.20**

1.45** Ref. 4446

1.34** Ref. 9072

1.54** Ref. 3626

Note: We have not used the weight provided in the World Value Study because N in small countries has been artificially reduced, so that e.g. Norway does not have the same impact as USA in the overall analysis. However, in our case we are interested in treating the countries with equal weights (analyses have also been conducted with weighted data and except lower significance levels in the sub-sample of social democratic countries due to smaller N, the results are more or less identical. 1 The same clusters as seen in table 5.1. 2 Either mentioned ‘people of different race’, ‘Muslims’, or ‘immigrants/foreign workers’.

in liberal regimes is more a matter of the working class being against, than the middle class being against. Looking at the other occupational groups, we find a number of similarities. In all three regimes, farmers are significantly more inclined to answer laziness, and employers with less than ten employees and agricultural workers have an odds-ratio in the same direction (but the only significant effect is among (small) employers in the conservative regime). In contrast, those who differ most from skilled workers turn out to be members of the armed forces in the liberal (odds-ratio 2.4) and social democratic (odds-ratio 2.5) regimes. In the model covering all countries, this effect was hidden by the opposite tendency in the conservative countries (odds-ratio 0.64). More interesting is the effect of unskilled workers differing significantly from the skilled-workers in the conservative regimes, whereas there is no difference in the liberal and social democratic regimes. The difference – which just fails to be significant23 – could be a consequence of the stronger insider/outsider divide found on the labour market in conservative regimes, i.e. unskilled manual workers are much more exposed to the risk of unemployment than skilled manual workers, which might contribute to the difference in perception of control of poverty. This fits nicely with the finding that the 18–24 year-olds in conservative regimes (odds-ratio 1.00) are less inclined to answer laziness than the 18–24 year-olds in liberal (odds-ratio 1.28), and especially in social democratic, (odds-ratio 1.66) regimes.24 23 The difference was tested by introducing an interaction effect between belonging to a conservative regime or not and being an unskilled manual worker or not. The difference had a significance level of 0.056. Looking at each conservative regime, we fine modest differences except in Italy (odds-ratio 0.38 significant at 0.01 level). 24 Again the difference was tested by introducing an interaction effect between belonging to a conservative regime or not and being between 18 and 24 or not. The difference had a significance level of 0.02. Looking at the liberal countries, we see that the effect is significant

88

The Institutional Logic of Welfare Attitudes

Looking at the other variables we see that education, labour market status and left-right orientation have more or less the same effect in the three regimes. It is also a general trend that the 55–64 year-olds and those above are much more inclined than the 35–44 years old to perceive the poor as being in control of poverty. We also see that the influence from negative perception of minority groups (dislike having ‘people with different race’, ‘Muslims’, and ‘immigrants’ as neighbours) on the share answering ‘laziness’ is present in all three welfare regimes. This suggests that negative perceptions of minority groups are important for perceptions of the poor, but it is not a special liberal or American phenomenon. Actually the impact seems to be strongest in social democratic regimes (odd-ratio 1.54).25 The topic will be further discussed in Chapter 8. Finally, we see that the gender difference is especially large in the social democratic regimes (odd-ratio 1.54).26 Why this is the case is difficult to explain, but perhaps it is a ‘soft-woman effect’ due to socialisation through public sector employment (the sector is unfortunately not available in the WVS) or more independence from their husbands’ attitudes. However, it could also be a ‘tough guy effect’ caused by a stronger ‘negative’ feedback from the level of generosity on the judgement of need among men than among women. The data used in Figure 4.1 actually indicate such a pattern.27 In any case, the finding suggests that the stronger gender divide in support for welfare policy in social democratic regimes found in most previous studies might be caused by larger gender differences in perceptions of the poor and unemployed instead of larger differences in self-interest.28 in the USA and Canada, but not in Ireland and Great Britain. In the social democratic countries the effect was significant in Denmark and Sweden, but not in Finland and the Netherlands. In Norway, the direction was actually the opposite – the young group was less inclined to answer ‘laziness’, but the effect was not significant. 25 Further analyses show that the impact is not even stronger in the USA (odds-ratio 1.12), which is a mystery. If we only use dislike of neighbours of different races, the effect in the USA increases (odd-ratio 1.37), but still the level is not exceptional. 26 Again formally tested by introducing an interaction term that becomes significant at the 0.01 per cent level. Looking at the social democratic countries, we see that the effect is especially strong in Norway (odd-ratio 2.1) and more modest in Denmark (odds-ratio 1.60 ns), Finland (odds-ratio 1.56, ns), The Netherlands (odds-ratio 1.35, ns) and Sweden (oddsratio 1.30, ns). 27 As described in Chapter 4, the level of social expenditures in the 12 countries covered by the Eurobarometer study (37.1) were highly correlated with the share that disagrees that unemployed were well-protected. However, the correlation is higher among males (0.87) than among females (0.82). The same goes for the question about coverage of the ‘poor’. The correlation with social expenditures is again higher among males (0.80) than among females (0.70). Thus, males seem to become ‘fed up’ with the welfare state faster than females. 28 After all, the argument that women are more in favour of welfare policy because they are more dependent on the welfare state in terms of employment, as family members relieved of care work, and as recipients of benefits, is weakened by the fact the that maximising household income is probably much more likely than maximising individual income. Thus, as women and men tend to live together, the rational choice argument seems to have severe problems.

Welfare Regimes and Perceived Causes of Poverty

89

Control and identity and the public support for welfare policy In the previous sections we found evidence of a connection between welfare regimes and the perception that poverty is caused by laziness and will power. As this item only tapped the need and partly the identity criteria, it is not a perfect measure for the overall judgement of deservingness, but it is probably the best we find in international surveys. So, if perceived control of neediness is the missing link between the welfare regimes and public support for welfare policy found in previous studies, we also need to show that perceived control of neediness actually influences the level of public support for welfare policy. From a theoretical point of view, we find it quite logical that if the ‘poor’ are perceived as being in control of their neediness, the public support for policy aimed at helping this group (and the public objections against retrenchment) will be lower, other things being equal. This reasoning is believed both to follow a moral logic, if in control of neediness the poor do not deserve to be helped, and a functional logic, if in control of neediness (further) public policy is not necessary. Early Anglo-Saxon empirical studies confirm such a connection (Feagin 1972; Feather 1974; Furnham 1982), and we also have comparative evidence from Kluegel & Miyano (1995), who analyse the connection between items that tap control of neediness and to what extent (1) ‘the government should guarantee everyone a minimum standard of living’, and to what extent (2) ‘the government should place an upper limit on the amount of money any one person can make’.29 If control of neediness is the missing link between welfare regimes and public support for welfare policy found in previous studies using the ISSP inequality module (see Chapter 3), the optimal procedure would be to include this variable in the models of previous studies, but – as mentioned – the question was simply not asked in the ISSP modules. As a simple alternative, we will match the aggregated values for public support for welfare policy from the eight Western countries found both in the ISSP 1992 social inequality module and the aggregated level of perception of control of neediness measured in the Eurobarometer 1993 and World Value Survey 1990 (see Table 5.1).30

29 Kluegel & Miyano (1995, p. 97) find a clear connection on both dependent variables in Great Britain, West Germany, the Netherlands and the USA, and on one of the dependent variables in Japan. However, these finding have not been related to the regime discussion, partly because three (Great Britain, the Netherlands and Japan) of the five countries studied by Kluegel & Maiyano are seen as deviant cases within Esping-Andersen’s framework, and partly because the question of deservingness was not asked in the ISSP surveys. 30 The first choice of the latter variable is the Eurobarometer 1993 survey conducted a year after the ISSP survey. Here we have results from Germany (W), Great Britain, Italy and Norway. The second choice is the World Value Study 1990 that was conducted before the ISSP survey. Here we have results from Canada, USA (ISSP in 1993), Austria (ISSP 1993) and Sweden (ISSP 1991).

The Institutional Logic of Welfare Attitudes

90

Table 5.6.

Cross-national differences in perception of control of neediness and public support for welfare state policy in countries where the ISSP social inequality module II was conducted. Shares, index and regression (OLS) between ‘laziness’ and dependent ISSP variables Cross-national differences in perception of control of neediness

39 32 13

Share that Share that Share that Index 0-6 Source: ISSP agrees agrees agrees with (Svallfors ‘provide 1997) inequality with with basic module II ‘redistri- ‘provide income’ bution’ jobs to all’. WVS 1990 38 47 34 2.86 1993 WVS 1990 48 40 48 3.24 1992 Euro 1993 65 56 66 4.14 1992

15

Euro 1993

66

66

58

4.21

1992

37 16 16 9

WVS 1990 Euro 1993 WVS 1990 Euro 1993

70 80 53 60

72 86 72 78

51 69 43 78

4.29 4.98 3.93 4.72

1993 1992 1991 1992

-0.44

-0.55

-0.75

-0.69

0.27 0.20

0.16 0.30

0.03 0.56

0.06 0.47

Share that explains poverty with ‘laziness’ USA Canada Great Britain Germany (W) Austria Italy Sweden Norway Standardised Beta Significant R2

Cross-national differences in public support for welfare state policy (dependent variable).

Source:

Note 1: Wording of ISSP questions: ‘It is the responsibility of the government to reduce the differences in income between people with high incomes and those with low incomes.’ ‘The government should provide jobs for everyone who wants one.’ ‘The government should provide everyone with a guaranteed basic income’. Share that ‘strongly agrees’ or ‘agrees’. Share that does not know excluded. Note 2: A simple additive index. Those who ‘agree’ or ‘strongly agree’ were given the value 2, those who ‘neither agree nor disagree’ were given the value 1, and those who ‘disagree’ or ‘strongly disagree’ were given the value 0. For Cronbach’s alpha values see Svallfors 1997:289.

Welfare Regimes and Perceived Causes of Poverty

91

We will not describe the national levels in any detail, as the pattern is already known from the sections above and the review of previous studies in Chapter 3. We will concentrate on the connection between cross-national differences in the perception of control of neediness and cross-national differences in public support for welfare policy found in the ISSP data. If we take a simple additive index of the three items used by Evans (1995), Svallfors (1997), Heien & Hofäcker (1999) and Andress & Heien (2001), i.e. the studies that confirmed a connection between regime and public support, we find a rather strong connection with the level of perceived control of neediness. As seen in Figure 5.1, the relation is negative as expected, i.e. the larger share that explains poverty with ‘laziness’ and ‘lack of will power’, the less support for welfare policy. The eight countries more or less seem to follow a straight line except Austria – which is too much in favour of public policy considering the perception of control of neediness.31 In statistical terms, the cross-national differences in perceived control of neediness as measured by the question of explanations of poverty ‘explain’ 47 per cent of the cross-national differences in level of public support to welfare policy. The level of significance is acceptable (0.06, see Table 5.6), even though we only have eight cases in the regression. Furthermore, one might add that if Austria is excluded, the explained variation is as high as 77 per cent, and the correlation becomes significant at the 1 per cent level).32

31 In Austria, the support for welfare policy seems too high considering the perception of control of neediness matches the level found in Canada and the USA. One explanation could be that the level of unemployment in Austria increased from 1990, when the World Values Survey was conducted, to 1993, when the ISSP survey was conducted. In absolute terms, the increase in unemployment was modest (commonly used definitions from 4.1 per cent to 5.4 per cent), but in relative terms the increase was around 32 per cent, which probably has weakened the perception of the poor being in control of their neediness. At least we see a very strong decrease from 1990 to 2001 in the case of Austria, see Table 5.1. If Austria is excluded, the R-square increases from 0.47 to 0.77 (adjusted from 0.38 to 0.72). 32 Even though this simple matching of aggregated level of perceived control of neediness and policy attitude supports previous findings and seems very logical from our theoretical point of view, we will stress that we do not have a deterministic relationship. If we leave the surveys conducted in the early 1990s and turn to the Eurobarometer survey conducted in 2001, we actually find a much weaker connection between perception of control and policy attitudes (see Appendix 5.2.)

92

The Institutional Logic of Welfare Attitudes

Share that explains poverty with ‘laziness’ and ‘lack of will power’.

Figure 5.1. The connection between cross-national differences in perception of control of neediness found in Eurobarometer 1993/World Value Study 1990 and public support for welfare policy found in ISSP social inequality 1992 (N=8) If we take a closer look at Table 5.6, it also reveals that the cross-national differences in perception of control of neediness influence the three ISSP items differently. The standardised beta-coefficients for support for ‘redistribution’, ‘jobs to all’ and ‘basic income’ are respectively -0.44, -0.55 and -0.75, which we explain as follows: the question of ‘reduce differences between people with high incomes and those with low incomes’ is the most general and includes a judgement of ‘those with high incomes and those with low incomes’. In contrast, the question of ‘providing everyone with a guaranteed basic income’ does not include a judgement of ‘those with high incomes’, and a ‘guaranteed basic income’ is primarily relevant only for groups that have no income at all. Therefore we believe this latter indicator of public support for welfare policy to be most dependent on the perception of control of neediness. Welfare regimes, perception of poor and unemployed and a sceptical mind In this chapter, we analysed the suggested connection between welfare regimes and perception of poor and unemployed by means of the question of ‘poverty being caused by laziness and lack of will power’ asked in the World Value Study and in a number of Eurobarometer surveys. The overall result is that we find a rather strong connection between welfare regimes and the perception that the poor are in control of their poverty. Naturally, the limited number of cases makes it difficult to fulfil the standards of the large-n logic, but at least the most comprehensive crosscut, the World Value Survey

Welfare Regimes and Perceived Causes of Poverty

93

1990–93, and the Eurobarometer cross-cut from 1976 revealed the expected pattern. Thus, a high proportion in the liberal regimes believed that the poor are in control of poverty compared to a medium proportion in the conservative regimes, and a low proportion in the social democratic regimes. There are a number of deviant countries – especially Finland and France – and the differences between the European Liberal regimes, i.e. Ireland and United Kingdom, and the other European countries seem modest in the Eurobarometer surveys conducted in 1989, 1993 and 2001. Breaking this overall regime variable down to the four dimensions discussed in Chapter 4, we found, in particular, the degree of selectivism and the degree of job opportunities to be most important for the perception of poverty being caused by laziness and lack of will power. The degree of ethnic fractionalisation and especially the level of social expenditures seem less important, but it should be remembered that very rough proxies were used. Including individual level variables in the analysis did not change this regime effect much. Even with gender, age, education, occupation, employment status, left/ right-wing orientation, and dislike of minority groups taken into account, living in a liberal regime was estimated to more than double the chance of an individual answering ‘laziness’ (compared to living in a social democratic regime). The effect from living in conservative regimes also remained stable. Thus, the regime effect seems to be ‘imposed’ on everybody living in a given regime and not a matter of ‘positive’ or ‘negative’ groups being bigger or smaller in some regimes. However, based on the World Value Survey we did find some significant differences in the way ‘the same’ group across the three regimes ‘reacts’ in the three regimes. The difference between men and women was especially pronounced in the social democratic regimes, the difference between skilled workers and unskilled workers was pronounced in the conservative regimes, and the difference between blue-collar workers and white-collar workers was pronounced in the liberal regimes. In the latter case, the white-collar workers perceived the poor to be much less in control than the blue-collar workers. These findings might help explain why previous studies have had difficulties finding ‘the middle class effect’ in liberal regimes suggested by the power resource theory, why the gender difference in support for welfare policy is larger in the social democratic regimes, and why labour market insiders and outsiders hold different attitudes in the conservative regimes. Finally, we investigated the link between the perception of poverty being caused by ‘laziness and lack of will power’ and support for welfare policy. By combining the share answering ‘laziness’ and the share in favour of welfare policy measured by the ISSP surveys on social inequality, we found the expected connection, which supports the finding in previous studies. Overall, the main point is that the theoretical framework presented in Chapter 4 is operational and at least can ‘make sense’ of the findings in the available comparative survey material. Whether we have proved the causal reasoning to be right is more debatable, as the small-n problem and the use of proxies rightfully leave behind a sceptical mind. In the next chapter, we will try to isolate the effect from selectivism and level of generosity by applying another analytical strategy.

The Institutional Logic of Welfare Attitudes

94

Appendix 5.1. Perceptions of causes of poverty found in the second wave of the World Value Study The question of causes of poverty used in this chapter was also asked in the second wave of the World Value Study. It only includes eight of the countries we are interested in, which makes the survey less interesting than the first wave. However, first of all the second wave was not used in this chapter because the possible answers to the question of ‘why, in your opinion, are there people in this country who live in need’ was reduced from the four options to the following two options: 1. They are poor because of laziness and lack of will power. 2. They are poor because society treats them unfairly. We believe these answers still tap the control dimension and part of the identity dimension. But it is also clear that respondents will be much more inclined to answer ‘laziness and lack of will power’ because only one other alternative is given. As we shall see below, this does have a clear impact on the answers. Table 5.7 shows the share that answers laziness and lack of will power. Table 5.7.

Share explaining poverty with ‘laziness and lack of will power’ in Western Countries available in the second wave of the World Value Study

Country

World Value Study 1995-1997

USA Australia New Zealand Liberal regime mean:

61 50 37 49

Germany (W) Spain Conservative regime mean:

12 16 14

Sweden Norway Finland Social democratic regime mean:

17 31 20 23

Welfare Regimes and Perceived Causes of Poverty

95

Compared to the analyses conducted in above, we see that even the ranking of the countries closest to the ideal type becomes blurred in the second wave. With 61 answering ‘laziness and lack of will power’, the Americans still distinguish themselves as a population that believes that the poor are in control of their neediness; but with 12 per cent and 17 per cent using this explanation respectively in Germany and Sweden, the ranking between the ideal-type conservative and social democratic regime is unexpected. However, the validity of these findings in the second World Value Study is questioned by the fact that a large number of respondents, 35 per cent in West Germany and 22 per cent in Sweden, answer ‘don’t know’ to the question. Thus, the reduction of the possible answers to two makes the categories less exhaustive, which generates questionable results. One possibility is to exclude the ‘don’t knows’ from the analysis, which increases the share answering ‘laziness’ in Germany from 12 to 18 per cent, and in Sweden from 17 to 22 per cent This reduces the difference between Germany and Sweden, but we still do not find the expected ranking. Furthermore, the finding cannot be explained by differences in unemployment levels as both countries at the time of interview experienced high unemployment: 9.9 per cent in Germany and 9.6 per cent in Sweden (OECD standardised). A better explanation is that the feedback from a generous welfare state becomes more pronounced when respondents can only choose between the explanation of ‘laziness’ and ‘because society treats them unfairly’. Living in a highly developed welfares state that especially benefits the poor and unemployed makes it difficult for Swedes to explain poverty with unfair treatment by society. Thus, we argue that the unexpected ranking between West Germany and Sweden in the second wave of the World Values Study can partly be explained by methodological problems and the fact that the item taps much more of the expected feedback from the level of welfare state development (see Chapter 4). In relation to the overall argument presented above, the most interesting finding is that Australia (50 per cent) and New Zealand (37 per cent) also clearly distinguish themselves from the conservative and social democratic regimes. This supports the argument that we have a general effect connected to the liberal regimes and not a specific American effect caused by some kind of exceptionalism. Appendix 5.2 The relationship between perception of poverty and support for welfare policy found in Eurobarometer 2001 In this chapter we showed a strong connection, on the aggregated level, between the proportion answering laziness and the proportion that supports welfare policy, which confirmed our theoretical expectations and previous empirical findings in the field. However, if we turn to the Eurobarometer survey conducted in 2001, we actually find a much weaker connection between perception of control and policy attitudes. Table 5.8 summarises the connection between the ‘control’ variable and the items

96

The Institutional Logic of Welfare Attitudes

measuring public support for welfare policy both at the individual level (gammacoefficients) and at the aggregated level (Pierson’s r). The Eurobarometer study has the ISSP items as well as a number of additional items that more narrowly measure anti-poverty policies. Starting with aggregated data, we do not find the relatively clear-cut pattern found at the end of this chapter. To some extent, this is understandable as the cross-national variation in aggregated level of ‘control’ is much smaller than in the first wave of the World Value Study. This is partly because the countries closest to the ideal-type liberal regime, i.e. USA and Canada, are not covered by Eurobarometer. Still, it is interesting that measured on the three ISSP items, the correlation coefficients at the aggregated level are as low as +0.01, +0.25 and +0.12 and none of them are significant. They even have the wrong signs, as we expect a connection between agreeing that laziness is the cause of poverty (coded 0/1) and support for welfare policy (coded 5,4,3,2,1 beginning with ‘strongly agreeing’). Thus, among countries covered by the Eurobarometer in 2001, cross-national differences in support for public policy cannot be explained by cross-national differences in perception of ‘control’.

Welfare Regimes and Perceived Causes of Poverty

97

Table 5.8.

Correlations between perception of control of neediness and public support for welfare policy measured in the Eurobarometer 2001. Correlations at individual level within each country and overall correlation at individual level (gamma-coefficients and level of significance).a Correlation at aggregated level (Pierson’s r and level of significance)b

Item:

1 2 3 ‘provide ‘provide ‘Income distribution’ job to a basic all’ income’

Belgium Denmark Germany (W) Greece Italy Spain France Ireland Luxembourg Netherlands Portugal Great Britain Finland Sweden Austria

4 5 6 ‘provide ‘the ‘the decent government unemployed standard should should be of living spend less forced to for unon benefits take job’ (-) employed’ for the poor’ (-) -0.17** -0.08ns -0.21** -0.36** -0.14* -0.23** -0.28** -0.27** -0.16* -0.20** -0.15** -0.30**

-0.05ns -0.12ns -0.22ns

+0.09ns -0.04ns -0.05ns

-0.16ns -0.20ns -0.10ns -0.00ns -0.03ns -0.09ns -0.16ns -0.04ns -0.14ns -0.21ns -0.04ns -0.11ns

-0.12ns -0.05ns -0.10ns -0.21** -0.04ns -0.13ns +0.01ns -0.10ns +0.05ns +0.08ns -0.01ns -0.05ns +0.01ns -0.04ns +0.04ns -0.14** +0.05ns -0.12* -0.02ns -0.21** -0.13ns -0.14ns -0.14* -0.20**

-0.13ns -0.31** -0.11ns -0.26** -0.08ns -0.12ns -0.27** -0.00ns -0.22** -0.24** -0.28** -0.24**

-0.14* -0.17** -0.20** -0.15* -0.03ns -0.17* -0.38** -0.08ns -0.25** -0.14* -0.36** -0.17**

-0.08ns -0.09ns -0.19** -0.47** -0.17** -0.24** -0.30** -0.14** -0.15** -0.39** -0.32** -0.18**

Individual level all countries

-0.11**

-0.01ns

-0.11**

-0.19**

-0.20**

-0.23**

Correlation at aggregated level

+0.01ns

+0.25ns +0.12ns

-0.24ns

-0.29ns

-0.13ns

98

The Institutional Logic of Welfare Attitudes

Notes to Table 5.8 NS: Not significant, * significant at 5 % level, ** significant at 1% level. The question of explanation of poverty was coded as a dummy (lazy / other answers) and ‘don’t know’ was coded as missing value on dependent variables. 2 The data was aggregated to the share answering ‘lazy’ and the national mean of the other items. 3 Wording item 1-3, see chapter 3. ‘The government should provide decent housing for all who can’t afford it’ (4), ‘The government should provide a decent standard of living for the unemployed’ (5), ‘The government should spend less on benefits for the poor’ (6), and finally ‘The unemployed should be forced to take a job quickly, even if it is not as good as their previous job’ (7). 1

If we look within countries, we find the expected direction of the relationship in most cases, even though the first ISSP item ‘redistribution’ is not significantly correlated with perception of ‘control’ in any of the 15 countries. The second ISSP item ‘provide jobs to all’ is only significant in Austria, whereas the third ISSP ‘provide a basic income’ is significantly influenced by ‘control’ in eight countries. Thus, on the individual level, we still find support for a connection between perception of ‘control’ and support for welfare policy – especially when it comes to the third ISSP item. Furthermore, the result comes more in line with previous findings if we measure public support for welfare policy on items that more directly refer to the poor and unemployed. In the Eurobarometer 2001, we have chosen ‘provide decent standard of living for unemployed’ (4), ‘the government should spend less on benefits for the poor’ (5) and ‘the unemployed should be forced to take a job quickly, even if is not as good as their previous jobs’ (6). Measured at the aggregated level, we now find the expected direction of the relationship and somewhat stronger correlations (respectively -0.24, -0.29 and -0.13) even though the connections do not turn significant. Furthermore, if we look within the individual countries, we now find relatively strong correlations between the perception of control and new measures of policy attitudes. On the latter three items, we do not find unexpected directions of the relationship in any country, and in most countries the correlations are highly significant. So, based on the empirical results from the Eurobarometer 2001 study, we will argue that the expected relationship can still be found, but it does not seem as strong as previous research suggests. One plausible explanation is that in a European context the awareness of sustainability of the overall welfare state, so heavily discussed during the late 1990s, has increased the complexity of policy attitudes (see further discussion in Chapter 9).

Chapter 6

Selectivism and Stigmatisation In this chapter, we will elaborate on the thesis that selective welfare policy generates an ‘us–them divide’, which makes it more difficult for the poor and unemployed to fulfil the identity criterion. In the previous chapter we found evidence of such an effect, but a very rough proxy measured the degree of selectivism, and the number of countries was modest. In this chapter we will measure the ‘us–them divide’ by the extent to which recipients of welfare benefits experience stigmatisation. It is actually a rather valid measure as stigmatisation precisely denotes a situation where persons or groups are given negative characteristics that distinguish them from the majority. By means of survey data from the social democratic welfare regimes this chapter reaches two overall conclusions. Firstly, it is shown that if we look at different welfare benefits and services within countries, the most selective welfare policies are clearly the most stigmatising. Secondly, it is shown that when we look at welfare services and benefits that cover the same social risks across the Nordic countries, but vary in terms of selectivism, there is also a link between selectivism and stigmatisation. In particular, the latter finding is interesting as it is based on a similar design where (in principle) the only thing that varies across countries is the variable we are interested in. It is probably the closest comparative research comes to a real experiment, and the great similarities between the Nordic countries, e.g. in terms of size, religious/linguistic/ethnic fractionalisation (see Chapter 4), and egalitarian values (see Chapter 3), make these countries especially suitable for our purpose. The chapter is organised in four main sections. The first section reviews the previous studies and further discusses our analytical strategy. In the second section, a number of theses are deduced from our main theoretical framework in Chapter 4. As we analytically go from overall welfare regime characteristics to the specific characteristics of individual welfare policies, we need to clarify how we, at the programme level, can distinguish between selectivism and universalism. In the third section we test the deduced theses, and the fourth section summarises the findings. Previous studies and the analytical strategy A number of national studies support they thesis that selective welfare policies produce a higher degree of stigma than universal policies. Most of these studies have non-take-up of welfare benefits to which citizens are entitled as their point of departure. In the US it has especially been analysed how the Food Stamp program,

100

The Institutional Logic of Welfare Attitudes

Medicaid and (the former) AFDC have very low take-up rates. It has, for example, been shown that among those eligible for benefits, around 25 per cent do not receive AFDC, and 50 per cent do not receive Food Stamps (Kim & Mergoupis 1997, p. 708). We also have a number of studies of non-take-up rates among British pensioners entitled to means-tested schemes that supplement the low basic pension. According to official estimates in 2001, 28 per cent of pensioners entitled to (formerly) Income Support did not receive it, 35 per cent of pensioners entitled to Housing Benefits did not receive it, and 10 per cent of pensioners entitled to Council Tax Benefit did not receive it (Department of Work and Pensions 2003). The subject is, in its nature, more relevant in the liberal regimes, but we also have a few studies of take-up rates in European countries (Riphahn 2001; Oorschot 1991). In the following, we will supplement these analyses with survey results, which more directly try to measure the extent of the relation between social security benefits and stigma. The questions were posed in addition to the 1999 ISSP inequality module in Norway, Denmark and Sweden. Furthermore, the identical question was posed in the Danish and Finnish Unemployment, Early Retirement and Citizenship (UEC) surveys, which include, besides a general population sample (18–66), a large subsample of long-term unemployed, which we will make use of in the next chapter (in Denmark we used the population sample from the ISSP 1999 data). In the take-up studies – usually conducted by building an econometric model that explains why the entitled do not take up benefits from a given scheme – we get a rather indirect measure of stigma. As pointed out in these studies, the ‘hidden costs’ that outweigh the potential benefits partly include ‘the cost of stigma’ associated with a given programme, but also the unpleasantness entailed in the claim process, the effort it takes to gather information, lack of knowledge etc (e.g. Hancock et al. 2003). Furthermore, we will take a more comparative approach as we compare stigma connected to programmes with different programmatic structures within countries and between countries. In contrast, most econometric studies only deal with a single programme, which generates methodological programmes. Bay & Pedersen (2003, p. 8) argued that ‘it is often pointed out that means-tested programs are less popular than universally provided benefits, but it is difficult to say whether this is a true design effect or simply an effect of attitudes towards the social risk and the composition of the beneficiaries covered by the different schemes’. By comparing schemes that address the same social risk, but vary in terms of design, we should at least be able to rule out the former objection. As to the latter objection, i.e. the effect from the composition of the beneficiaries, we find it less relevant. That recipients of welfare benefits become less culturally distinct as the number of recipients increases is actually a part of our theoretical reasoning. Deducing the theses at the programme level In Chapter 4 we argued that selective welfare policy would (1) open the discussion of need, (2) open the discussion of control, (3) define the recipients as a special

Selectivism and Stigmatisation

101

group, (4) highlight the boundary between ‘net-winners’ and ‘net-losers’ and (5) open the discussion of recipients’ attitudes towards those who give. On the contrary, a universal welfare policy would (1) close the need discussion, (2) close the control discussion, (3) define the recipients as belonging to ‘us’, (4) blur the boundary between ‘net-winners’ and ‘net-losers’ and finally (5) close the discussion of attitudes (see Table 2.1). Following Esping-Andersen, we argued that on this dimension the social democratic countries were at the universal end, the liberal regimes at the selective end, and the conservative regimes in between. However, it is well known that if we take a closer look within each regime type or within each country, we do not find one principle, but a variety of principles behind benefits and services. In this section, we will take such a look within the social democratic regimes of Sweden, Norway, Denmark and Finland in order to analyse to what extent the programmes guided by universal principles and the programmes guided by selective principles produce different degrees of stigmatisation. In order to make such a classification of the different benefits/services, we need to clarify what characterises a universal programme versus a selective programme.

Figure 6.1. The distinction between the insurance track, the universal track, and the residual track at programme level

102

The Institutional Logic of Welfare Attitudes

Following Goul Andersen, we argue that the ‘purest’ universal benefit is (1) constituted by law, (2) citizenship oriented, (3) given to all citizens (experiencing the same ‘social event’ e.g. old age), (4) adequate, and finally (5) flat-rate. In the Nordic countries, the basic pension scheme and child allowances are the programmes that come closest to this pure form of universalism. With these kinds of benefits, it is not discussed whether recipients really need them, whether they are in control of their possible neediness, and whether they receive the benefits with a grateful attitude. Thus, we should expect very low stigma associated with receiving these benefits in the social democratic regimes. One of the rough quantitative measures of universality is the share of the population that receives these benefits. All citizens of 67 years and above in Denmark and Norway receive the basic old-age pension, which adds up to approximately one of five adult citizens. Approximately one out of five adult citizens also receive a basic pension in Finland (19.9 per cent) and Sweden (20.5 per cent) even though the pension systems have recently been changed.1 Likewise, child allowance is given as a flat rate to all families with children (up to the age of 16 in Norway, 17 in Finland, 18 in Denmark and between 16 and 20 in Sweden (1999 rules)), which adds up to 20–30 per cent of all families (see Table 6.1). As it is quite easy for the administration to identify who is eligible for these benefits and therefore be proactive; the take-up rate is roughly 100 per cent. At the other end of the ‘selective-universal scale’ we find the social assistance schemes that are (1) constituted by law, (2) citizenship-oriented, but at the same time (3) targeted at ‘the weak’, and (4) means-tested. Thus, in the case of social assistance, it is necessary to develop standards for distinguishing the ‘weak’ or ‘needy’ group and for evaluating how much is needed to overcome this ‘neediness’. In contrast to the old poor laws or simple charity, the municipalities, which administrate the social assistance schemes, are not free to choose their own standards, but are guided by legislation. Nevertheless, the front-line workers have to make a large number of estimates. Thus, the Nordic social assistance schemes come close a pure ‘selective policy’, which we expect to be associated with the highest degree of stigma. Measured on the share of the adult population that (at some point in time) in 1999 has received social assistance, the figures are 5.8 per cent in Denmark, 9.0 per cent in Finland, 4.5 per cent in Norway and 5.5 per cent in Sweden, i.e. well below the share receiving basic old-age pension and child allowances. The larger share in Finland is partly due to a higher level of unemployment, and partly due to a shorter unemployment benefit period. It is therefore more common to receive social assistance in Finland, which we expect to reduce the level of stigma compared to the other Nordic countries. 1 In the Swedish case, the new pension system abolishes the old basic pension and only delivers a so-called guarantied minimum pension to those who have not qualified for a sufficient income-related pension. However, the changes do not disturb our analyses as the system was not in place in 2000. In the Finish case the pension reform from 1996 also abolished the basic old-age pension and introduced a guaranteed minimum pension but, in 2000, most of the pensioners were still covered by the old rules. In 2000 773,000 citizens still received the basic pension. 197,000 received a public pension without receiving the new minimum pension (Nososko 2004, Table 7.6).

Selectivism and Stigmatisation

103

Housing allowances is another benefit that is means-tested in all four countries, i.e. we expect more stigmas to be associated with this scheme than with the basic old-age pension and child allowances. Furthermore, we expect the stigmas to be larger in Norway than in the other countries, as housing benefits in Norway are much more targeted at the ‘weakest’. Thus, in Norway, only 1.1 per cent of families receive housing benefits compared to 5.7 per cent, 7.5 per cent, and 8.0 per cent in the other three countries. Table 6.1.

Share of population (18–) or share of families receiving various social benefits in four social democratic regimes in 1999/2000. Unemployment rate also included Denmark

Finland

Norway

Sweden

Cross-country thesis

Basic old-age pension1

16.9

19.9

18.3

23.2

No cross-country difference expected as (still) almost 100 pct. coverage

Child allowances (families in relation to families)2

23.0

22.1

27.6

31.7

No cross-country difference expected as 100 pct. coverage

Disability pension4

4.0

7.0

7.9

6.1

Stigma predicted to be highest in Denmark

Housing allowances (families in relation to families)5

5.7

7.5

1.1

8.0

Stigma predicted to be highest in Norway

Early retirement6

3.7

0.6

-

-

Stigma predicted to be highest in Finland

Social assistance7

5.8

9.0

4.5

5.5

Unemployment benefits3

12.5

9.4

4.8

9.9

Stigma predicted to be lowest in Finland Stigma predicted to be highest in Norway and lowest in Denmark

Level of unemployment (1999, OECD standardised)

5.2

10.2

3.2

7.2

104

The Institutional Logic of Welfare Attitudes

Notes to Table 6.1

The three remaining schemes presented in Table 6.1, i.e. unemployment benefits, 1

Number of citizens receiving basic pension (‘grundpension/ garanteret mindste pension’) in 2000 in relation to grown up population (18-). 2 The share of families with children below 18 in relation to all families in 2000. In Norway and Finland the child allowance is only paid up to the age of 16 and 17 respectively, i.e. the figures for these two countries are slightly overestimated. In Sweden the child allowance is paid only up to age 16 for all, but up to age 20 if the child is under education. 3 Number of persons in 1999 who received unemployment benefits one or more times in relation to population (18- in 2000). 4 Number of persons in 1999 who received disability pension one or more times in relation to population (18- in 2000). 5 Numbers from 1999. 6 Denmark: The number of persons on ‘Efterløn’ in 1999 in relation to population (18- in 2000). In Finland the number of persons on ’Unemployment pension’ in 2000 in relation to population. These figures taken in order to match survey data. 7 Number of persons who one or more times received social assistance in 1999. In Denmark 65,592 persons who had the main social security from ‘activation allowance’ from municipalities has been added (for political purposes the groups in ‘activation’ are in the official statistics not counted as being unemployed or receiving social assistance). Source: Calculation based on Nososko 2002: 29, 36, 70, 118, 135, 167; 2004; table 7.6.

disability pension and early retirement benefits, are a bit more difficult to classify. In the Nordic countries, except Norway, the unemployment benefits are in principle a voluntary insurance scheme, but in all countries it is heavily regulated by law, heavily financed by the state and, except in Finland, there is a rather loose connection between contribution and compensation. Thus, in relation to Figure 6.1, the unemployment schemes are somewhere between the insurance track and the universal track, which makes it difficult to make predictions about the ‘universal-selective scale’. However, based on the share of the adult population that (at some point) in 1999 received unemployment benefit – 12.5 per cent in Denmark, 9.4 per cent in Finland, 4.8 per cent in Norway, and 9.9 per cent in Sweden – we should expect the level of stigma to be somewhere between the selective programmes of social assistance and housing allowances and the universal programmes of basic old-age pension and child allowances. Furthermore, one could expect receiving unemployment benefits in Norway to be more stigmatising than in Denmark. In the former country, only 4.8 per cent received unemployment benefits during 1999 compared to 12.5 per cent in the latter country. However, looking at the level of unemployment, we end up with other expectations. The level of unemployment in Finland in 1999 was three times higher than in Norway and almost twice as high as in Denmark. Thus, if the effect from the level of unemployment is strong – which the results in the previous chapter suggest – we should expect recipients of unemployment benefits to be more stigmatised in Norway and Denmark than in Finland.2 2 A number of factors explain why the number of people who at some point in 1999 received unemployment benefits does not go together with the level of unemployment. The

Selectivism and Stigmatisation

105

The disability pension in the Nordic countries comes closer to a universal scheme as it is (1) constituted by law, (2) citizenship-oriented, (3) given to all (with disability) at (4) an adequate level. In Denmark it come close to a ‘pure’ flatrate benefit, whereas the size of the disability pension in the other three countries depends on previous earnings, i.e. disabled high earners will (to a modest extent) get a higher disability pension than low earners. We do not see variation across the four countries – maybe except Denmark, where the disability pension (with 4.0 per cent of the adults receiving it in 1999) is a little less common. Finally, we have included the shares that receive early retirement in Denmark (‘efterløn’) and Finland (unemployment pension) in 1999. The Danish early retirement scheme ‘Efterløn’ is open to all employees (being 60 years old and long-time members of the unemployment insurance), whereas the Finnish schemes covered by the survey data are only open to the unemployed (being 60 years old). Measured by the share that received these benefits, 3.7 per cent in Denmark and 0.6 per cent in Finland, we should expect the Finnish scheme to be more associated with stigma than the Danish scheme. The last column in Table 6.1 summarises the cross-country theses. Level of stigmatisation across programmes and across countries among the social democratic regimes Measuring the level of stigmatisation is not straightforward. To increase the validity of the answers, the surveys did not ask directly whether the respondents look down on citizens receiving social benefits. A direct question could, especially within a Nordic context, lead respondents to give ‘the right answer’ instead of ‘the true answer’. Therefore the respondents were asked: ‘It has been discussed if people are inclined to look down on those who receive social security benefits. Would you say that it happens very often, often, quite rarely or very rarely?’. Table 6.2 summarises the results.

two extremes are Denmark and Finland. In the former case, many received unemployment benefits even though the unemployment rate was low. This is caused by the fact that the job turnover is very high, partly as a consequence of low job protection, and access to benefits is rather open. In the latter case, few have received unemployment benefits even though the unemployment rate is high. This is primarily due to the fact that access to benefits is more difficult. As seen in Table 6.1, a large share is covered by social assistance.

The Institutional Logic of Welfare Attitudes

106

Table 6.2.

Judgement of how often people tend to look down on recipients of different social benefits. Percentage answering ‘very often’ or ‘often’ and percentage difference (‘very often’ and ‘often’ minus ‘quite rarely’ and ‘very rarely’)

Sweden1

Norway1

Denmark1

Finland2

Pct. often

Pct. dif.

Pct. often

Pct. dif.

Pct. often

Pct. Dif.

Pct. often

Pct. dif.

4

-93

9

-83

2

-96

4

-93

3

-94

ISSP 1999 (additional) / Finland UEC survey 1999 Old-age basic pension Child allowances Housing allowances

12

-76

23

-55

10

-80

17

-66

Disability pension

19

-62

6

-89

21

-58

8

-85

Unemployment ben.

19

-61

33

-35

29

-41

30

-41

Social assistance

68

+35

70

+41

73

+46

49

-2

5

-90

11

-78

1823

1823

1386

1386

Early retirement N (min) 1

1143

1143

1299

1299

Additional questions posed in the third ISSP social inequality module from 1999. In the Danish case except the question on early retirement, which was only posed in the UEC survey. 2 Based on the Unemployment, Early Retirement, and Citizenship survey (UEC) from 1999.

Selectivism and Stigmatisation

107

Table 6.2 allows us to make a number of comparisons, but as a point of departure we will take a look at the absolute numbers. As expected, the overall picture is that receiving welfare benefits is rarely followed by stigma in these four social democratic regimes (the same goes for welfare services, but they are not covered by these data). All the schemes in all four countries, except social assistance, have a negative percentage difference, i.e. the group answering that beneficiaries are ‘quite rarely’ or ‘rarely’ looked down on is larger than the group answering that beneficiaries are ‘often’ or ‘very often’ looked down on. This supports the theoretical argument about systems dominated by universal welfare policies, which was discussed in Chapter 4. If we compare the schemes within the countries, it is also quite clear that receiving the most universal schemes, basic-old age pension and child allowances, is much less associated with stigma than receiving the most clear-cut selective schemes, i.e. social assistance and housing allowances. The percentage responding that receiving old-age pension is often associated with stigma is 4 per cent in Sweden, 9 per cent in Norway, 2 per cent in Denmark, and 4 per cent in Finland. The question of child allowances was only asked in Finland. The result is that only 3 per cent think receiving these allowances is associated with stigma. At the other end we find social assistance. In Sweden, 68 per cent indicate that receiving social assistance is associated with stigma. The share is 70 per cent in Norway, 73 per cent in Denmark, and 49 per cent in Finland. The stigmas associated with housing allowances are also above those for old-age basic pension and child allowances, but considerably lower than for social assistance; the share answering often associated with stigma is 12 per cent in Sweden, 23 per cent in Norway, 10 per cent in Denmark, and 17 per cent in Finland. Thus, even though we can safely conclude that the clear-cut universal schemes found in the social democratic welfare regimes are less associated with stigma than the selective schemes, the difference between social assistance and housing allowances warn us that differences in stigma are not solely related to the degree of selectivism, but also to the ‘social event’ that makes specific social security benefits relevant. One could argue that the difference in stigma associated with basic old-age pension and, for example, social assistance is caused by the fact that it is much more acceptable to become old than to become unemployed (without unemployment insurance). In contrast, our basic thesis is that receiving an old-age basic pension would be much more associated with stigma if it was a selective policy, or being without job would be much less associated with stigma if the whole population were covered by a citizens’ wage, as has been discussed in a Nordic context. One solution to the social risk objection is to compare the stigma associated with receiving unemployment benefits and social assistance, as they both cover more or less the same ‘social event’; namely citizens of working age who cannot find a job. If we accept such a comparison, the results in Table 6.2 clearly demonstrate that the selective social assistance scheme in all four countries is much more associated with stigma than the ‘half universal’ unemployment benefit scheme. Only 19 per cent in Sweden, 33 per cent in Norway, 29 per cent in Denmark and 30 per cent in Finland answer that ‘people’ tend to look down on recipients of unemployment benefits. It is much lower than the level reported for social assistance. However, one can still

108

The Institutional Logic of Welfare Attitudes

argue that the comparison is tentative as the social assistance schemes also pick up a number of other ‘social events’ besides unemployment. This leads us to the next solution, which is to compare differences in stigma between countries with different programmatic structure (e.g. different pension systems) in the case of the same social event (e.g. old age). This is the ‘most similar design logic’ that we discussed in the introduction.3 One of the most marked differences was that housing allowances were much more selective in Norway – only 1.1 per cent of families received allowances – and the survey results actually confirm that the stigma associated with housing allowances is higher in Norway. Twenty-three per cent answer ‘often’ compared with 12 per cent in Sweden, 10 per cent in Denmark, and 17 per cent in Finland. Another marked difference was that more people in Finland received social assistance, which we expected to lead to fewer stigmas. This thesis is also confirmed as ‘only’ 49 per cent in Finland answer that people often look down on recipients of social assistance compared to 68 per cent in Sweden, 70 per cent in Norway and 73 per cent in Denmark. Finally, we are partly confirmed in the thesis that disability should be more stigmatising in Denmark, as there are relatively few recipients (although the difference between Denmark and Sweden is insignificant) and fully confirmed in the thesis that the Danish early retirement scheme ‘Efterløn’ is associated with less stigma than the Finnish unemployment pension. However, the large proportion of Danes that received unemployment benefits (once) in 1999 (12.5 per cent compared with 4.8 per cent in Norway, 9.4 per cent in Finland, and 9.9 per cent in Sweden) did not seem to reduce the associated stigma in Denmark. Twenty-nine per cent answered ‘often’ compared with 33 per cent in Norway, 30 per cent in Finland and 19 per cent in Sweden. As discussed, the level of unemployment probably ‘disturbs’ these results. Selectivism and stigmatisation. Confirming the old thesis in a new way The overall conclusion is that Nordic survey data on stigmatisation support the classic thesis of selective welfare policy being associated with stigma and thereby generating an ‘us–them divide’, which we believe is important when it comes to the overall judgement of deservingness. Asked about the tendency to look down on various groups of recipients, the answers clearly showed a difference between the ‘purest’ universal schemes, basic old-age pension and child allowances, and the purest selective schemes, social assistance and housing allowances. Furthermore, we saw a clear difference in stigma associated with receiving the ‘semi-universal’ unemployment benefits and the ‘pure’ selective social assistance. These findings are not – and have never been – without methodological problems as the social 3 As a small methodological remark, we should keep in mind that comparison between the ISSP population samples in Sweden, Denmark and Norway and the Finnish UEC-sample should be treated with some caution as the latter sample is slightly different (18 to 66 and not 18 and above), but most of all because the respondents had been through another questionnaire before the question was posed.

Selectivism and Stigmatisation

109

risk covered by the individual programmes may matter more than the degree of selectivism. We tried to solve this problem by comparing welfare benefits that in the Nordic countries cover the same risk, but at the same time differ in degree of selectivism. As the Nordic countries are very similar in almost all other aspects, this most similar design is probably the closest we get to ‘isolating’ the effect from selectivism, and basically the data confirmed our expectations. The larger share receiving social assistance in Finland did coincide with a lower level of stigmatisation, and the higher degree of targeting in housing allowance in Norway did coincide with a higher level of stigmatisation. We also found the expected difference between the Danish and Finnish early retirement. The expectation of cross-national differences in stigmatisation of recipients of unemployment benefits were blurred because the number of recipients within a year and level of unemployment did not go together. Compared to analyses conducted in the previous chapters, this evidence based on a most similar design gives us a much more solid basis for arguing that the institutions connected to the welfare regimes matter. Thereby, we also have an institutional argument that cannot be disproved by the fact that entering a number of structural variables in the aggregated analyses, e.g. size (Cameron 1978) or economic openness (Rodrik 1998), probably could reduce the impact of the suggested regime-dependent variables included in our theoretical framework. We also have an institutional argument that cannot be disproved by reference to differences in culture, dominant welfare state ideology, and degree of religious, linguistic or ethnic fractionalisation. In the next chapter, we try to ‘isolate’ the effect from generosity by using the same line of reasoning.

This page intentionally left blank

Chapter 7

Generosity and Stigmatisation In the theoretical framework in Chapter 4, we expected the level of generosity – the extent to which the welfare state provides economic resources to the potential poor – to influence the judgement of deservingness in two ways. A high level of generosity could question whether the poor and unemployed really were in need, which makes it more likely to categorise them as undeserving. But it could also allow the poor and unemployed to obtain a life standard close to that of the majority, which makes it more likely to categorise the poor and unemployed as deserving because they are part of the ‘common us’ and not a culturally distinct subgroup. And the other way around – very tight-fisted benefits are believed to generate culturally distinct subgroups, which makes it difficult to fulfil the identity criterion. So we have two opposing mechanisms, which might explain the finding in Chapter 5 that the level of social expenditures had little impact on the proportion answering that poverty was caused by laziness and lack of will power. The expected connection between generosity and cultural distinctiveness of the poor and unemployed might sound a bit speculative, but this line of reasoning is actually at the heart of modern welfare policy, and recent research underlines that the impact of economic resources has been underestimated (e.g. Goul Andersen 2002a; Halvorsen 1999). This chapter will try to ‘isolate’ the identity effect by means of a most similar design based on data from two social democratic regimes, Denmark and Finland. Everything else being more or less being equal, it will be shown that these two countries vary substantially in terms of generosity towards long-term unemployed. Furthermore, it will be shown that this variation accompanies the fact that Finnish unemployed are more culturally distinct from the majority than are Danish unemployed. Again, the ‘us–them divide’ will be measured by level of stigmatisation, but in contrast to the previous chapter we use the actual stigmatisation felt by the long-term unemployed themselves as the dependent variable. The chapter is organised into four main sections. In the first section, we describe the different levels of generosity in Denmark and Finland and how it is measured in our two samples of long-term unemployed. In the second section, we analyse to what extent this difference in generosity influences the degree of stigmatisation felt by the long-term unemployed. In the third section, we will try to rule out other plausible explanations for the country difference. Here we look at differences in the composition of longterm unemployed (sex, age, health, share on social assistance, and family status) and in self-esteem among the long-term unemployed in the two countries. We also look at the impact from the level of unemployment, which at the time of the interview was twice as high in Finland as in Denmark. In line with the findings from Chapter

112

The Institutional Logic of Welfare Attitudes

6, we find – after taking differences in generosity, self-esteem, and composition into account – that the lower level of unemployment in Denmark, other things being equal, generates a larger degree of stigmatisation. The variation in level of generosity Denmark and Finland are extreme cases among the social democratic regimes in terms of economic resources given to low-income groups in unemployment (Nososco 2002, p. 77). The difference is partly caused by the fact that a larger proportion of Danish unemployed receives unemployment benefits (cf. Chapter 6) and partly caused by the different rules of compensation. In the Danish case, the unemployed are compensated by 90 per cent of previous earnings up to a flat-rate ceiling, which gives a relatively high compensation for the lowest income groups. The high ‘low-income compensation rate’ is also reflected in the results from the European household panel study. Based on these data, it has been shown that an extremely small share of Danish households experiencing unemployment falls below the relative poverty line (Hauser et al. 2000). In the Finnish case, the unemployed are (in general) only compensated by 58 per cent of previous earnings without a ceiling (Nososco 2002, p. 75). Compared to Denmark, this gives low-income groups a much lower compensation. Within our two samples of long-term unemployed, respectively from Denmark and Finland, we have tried to capture the experienced financial hardship on three items: (1) the ability to pay current expenses, (2) the ability to pay an unexpected bill, and (3) the respondents’ judgement of their economic security in the future (cf. Table 7.1).

Generosity and Stigmatisation

Table 7.1.

113

Experience of financial hardship among Danish and Finnish longterm unemployed Denmark

Finland

‘Often’ or ‘sometimes’ difficulties in paying current expenses

24 %

50 %

‘Rarely’ or ‘never’ difficulties in paying current expenses

76 %

50 %

100 %

100%

Would be able to pay unexpected bill (3000 Danish kroner / 7000 Finnish mark)1

61 %

26 %

Would not be able to pay unexpected bill

39 %

74 %

100 %

100 %

‘Very’ or ‘rather’ confident in economic situation the next 3-5 years

56 %

29 %

‘Very’ or ‘somewhat’ insecure in the economic situation the next 3-5 years

44 %

71 %

100 %

100 %

1.01

1.91

N (min.) = 467

N (min.) = 753

Hardship index (0 –3 scale), mean2

1

Unfortunately the amount of money does not equal each other. 3000 Danish kroner roughly equals 400 euros whereas 7000 Finnish mark roughly equals 1300 euros. 2 Ordinary sum index. On each of the three indicators in table 3 on financial hardship, the value 1 is given to difficulties, and the value 0 to no financial difficulties. Thus we get an index from 0 to 3 (Cronbach alpha = 0.70). Measured in this way we see, as expected, a quite remarkable difference in financial hardship among Danish and Finnish long-term unemployed. Fifty per cent of the Finnish unemployed indicate that they ‘often’ or ‘sometimes’ have had difficulties paying current expenses compared to 24 per cent in Denmark. Likewise, 61 per cent of the Danish long-term unemployed answer that they would be able to pay an unexpected bill compared with 26 per cent among Finnish long-term unemployed; however, it should be taken into account that the size of the unexpected bill is larger in the Finish survey, see note to Table 7.1. Fifty-six per cent of the Danish long-term unemployed are ‘very’ or ‘rather’ confident in their future economic

The Institutional Logic of Welfare Attitudes

114

situation, compared with 29 per cent in the Finnish sample. Added into a simple sum index from 0 (no problem paying current expenses, able to pay unexpected bill, and confident in economic situation) to 3 (problem paying current expenses, not able to pay unexpected bill and insecure about future economic situation) the average is 1.01 in Denmark and 1.91 in Finland. Now, the thesis is that this difference in financial hardship causes higher stigmatisation in Finland than in Denmark or, to use another term, a greater ‘us–them divide’. The much higher Danish unemployment level at the time of interview (5.2 per cent in Denmark compared with 10.2 per cent in Finland) should work in the opposite direction. So if we can find a difference in level of stigmatisation, we actually have a rather strong argument. The connection between level of generosity and stigmatisation The level of stigmatisation of long-term unemployed in Denmark and Finland is measured by two questions. In the first question the long-term unemployed were asked ‘have you sometimes felt that people looked a little down on you because you receive unemployment benefits/social assistance’. The answers to this question are shown in Table 7.2. Table 7.2.

The experience of being ‘looked down on’ among long-term unemployed due to receiving unemployment benefits or social assistance. Denmark and Finland. Percentages

Very often

Quite often

Quite rarely

Hardly ever1

Percentage difference

N

Denmark

6

13

17

64

-62

390

Finland

15

19

28

38

-32

818

1

The Danish questionnaire also had a category label ‘not at all’, which was merged with the ‘hardly ever’ category in order to make the data comparable.

If we first of all look at the overall distribution, we see that a majority of the longterm unemployed in the two social democratic regimes answer that they ‘quite rarely’ or ‘hardly ever’ have felt that people looked a little down on them because they receive unemployment benefits or social assistance. As we do not have comparable data from other regimes, we cannot say anything with certainty, but it seems to support the argument made in Chapter 4 that in systems dominated by universal benefits and services the poor and unemployed are not perceived as a distinct subgroup. However, more interesting is the difference between Denmark and Finland, which coincides with our expectation. Among Danish long-term unemployed, only 6 per

Generosity and Stigmatisation

115

cent feel stigmatised ‘very often’ compared with 15 per cent in Finland. Thirteen per cent of Danish long-term unemployed feel stigmatised ‘quite often’ compared with 19 per cent in Finland. If we look at the percentage difference (subtracting those who answer ‘very often’ or ‘quite often’ from those answering ‘quite rarely’ or ‘hardly ever’), the figure is -62 in Denmark and -32 in Finland. Thus, even in a situation where the Finnish unemployment level is about twice as high as in Denmark (which should make more ‘normal’ people unemployed in Finland and decrease the perception of unemployed being in control of their neediness) do the Finnish longterm unemployed feel more stigma than the Danish unemployed. Table 7.3.

Personal feeling of being looked down on as a problem of being unemployed. Finland and Denmark. 10 point scale Not a problem at all

1

2

3

4

5

6

7

8

Long-term unemployed in Denmark

41

3

5

3

2

23

5

5

Long-term unemployed Finland

29

7

6

5

5

11

6

6

Total average

9

A very big problem

Average

N

5

2

7

3.3

343

9

5

13

4.2

840

3.9

1183

The same results were found on our second measure for stigmatisation. Here, the long-term unemployed were asked to what extent the feeling of being looked down on has been a problem of being unemployed. The respondents could answer on a scale from 0 to 10 where 0 indicated ‘do not experience it as a problem at all’, and 10 indicated ‘experience it as a very big problem’. The answers from this question are shown in Table 7.3. Again we see that the Finnish unemployed experience stigmatisation as a larger problem than the Danish unemployed. On the ten-point scale, the Danish average is 3.3 compared with the Finnish average at 4.2. The most common answer is ‘not at problem at all’ (41 per cent in Denmark and 29 per cent in Finland), which again supports the argument that in systems dominated by universal policies, the poor and unemployed are not seen as a special subgroup. However, the solid finding is that the cross-national difference in terms of generosity goes together with larger stigmatisation. Nevertheless, one could argue that ‘going together’ could be a matter of chance. After all, the world, even across the very similar Nordic countries, is a messy place. Therefore we have analysed on the individual level to what extent economic hardship influences the feeling of stigmatisation as a problem connected to unemployment. Table 7.4 shows the simple (uncontrolled) relationship between experienced financial

116

The Institutional Logic of Welfare Attitudes

hardship and the 10-point scale of personally felt stigmatisation among both Danish and Finnish long-term unemployed. Table 7.4.

Relationship between financial hardship and feeling of stigmatisation as a problem of being unemployed. Share between 6 and 10 on 10-point scale and average Share between 6 and 10

Average

N

‘Often’ or ‘sometimes’ difficulties paying current expenses

46

5.0

462

‘Rarely’ or ‘never’ difficulties paying current expenses

26

3.2

663

Would be able to pay unexpected bill (3000 Danish kroner / 7000 Finnish mark)1

23

2.9

423

Would not be able to pay unexpected bill

40

4.6

701

‘Very’ or ‘rather’ confident in economic situation the next 3-5 years

19

2.6

401

‘Very’ or ‘somewhat’ insecure in the economic situation the next 3-5 years

42

4.8

662

1

Unfortunately the amount of money does not equal each other. 3000 Danish kroner roughly equals 400 euros whereas 7000 Finnish mark roughly equals 1300 euros.

Those Danish and Finnish long-term unemployed who ‘often’ or ‘sometimes’ have difficulties paying current expenses experience stigmatisation as a much larger problem than the group without such difficulties. The average on the 10-point scale is 5.0 for the former group and 3.2 for the latter group. Measured as those who answer between 6 and 10, i.e. stigmatisation is to some extent perceived as a problem, the shares are respectively 46 per cent and 26 per cent. More or less the same relationship is present between the two other measures of financial hardship and stigmatisation. Taken together, the correlation between the index of financial hardship in Table 8.1 (0–3) and stigmatisation (0–10) is 0.32, which is relatively strong and highly significant. Taking the national context into account does not change this relationship much. The correlation within each country is 0.29 in Denmark and 0.31 in Finland.

Generosity and Stigmatisation

117

Thus, the relationship ‘isolated’ in the most similar design can also be found at the individual level in both countries. Exploring other possible explanations Even though the effect ‘isolated’ by the most similar design is also found at the individual level, one could still argue that everything else might not be equal, i.e. maybe other important variables differ across the two countries. Thus, the task is to pinpoint other variables that could be important for the experience of stigmatisation as a problem and see whether they make any difference. From a psychological perspective, one of the main theses would be that the feeling of stigmatisation during unemployment is dependent on alternative legitimate identities that are not related to work (Wadel 1973). Already the term ‘unemployed’ implies an empty identity defined by the absence of work. It is a common assumption that women cope more easily with unemployment because the housewife identity is open to women – especially women with small children (Christensen 1987; Halvorsen 1999, p. 46). One can question whether this identity is still legitimate after the Nordic women’s full-scale entrance into the labour market, but many would expect a gender effect. It might therefore be relevant to take differences in gender composition in the two samples into account. Another possibility is a pensioner identity where age, previous years on the labour market and the general acknowledgement of reduced job opportunities function as legitimising factors, cf. the deserving literature discussed in Chapter 4. Sickness might also function as a legitimising factor (Meyer 1999). The point is that the unemployed will try to cope with unemployment by offering themselves more legitimate identities. This understanding is also found in Goffman’s (1963) classic Stigma – Notes on the Management of Spoiled Identity, which describes how individuals by means of ritual role playing try to keep up a facade towards their social surroundings. Finally, the share that receives social assistance is important. From the previous chapter we know that a large share receives social assistance in Finland and we also know that receiving social assistance is connected to more stigma than receiving unemployment benefits. Now the trick is to see whether variations in these variables can reduce the country difference, which in model 1 is estimated to be 0.09 (standardised beta-coefficient). We use the 10 point scale from Table 7.3.1 Model II takes possible differences in sex, 1 This measure of stigmatisation is used as it comes closer to an interval scale than the measure used in Table 7.2, which only has four categories. However, there is a difference between ‘being looked down on has been a problem of unemployment’ (the 10-point scale) and ‘sometimes felt that people looked a little down on you because you receive unemployment benefits/social assistance’ (the 4-point scale). Therefore, we have also conducted the analyses on the 4-point scale. In general, the effect is stronger, e.g. simply including the country variable is able to explain 5 per cent of the variation, and the standardised coefficient is 0.23, but they follow the same pattern. However, using this variable we cannot change the direction of the country effect, which we below interpret as the effect from level of unemployment. Using

The Institutional Logic of Welfare Attitudes

118

Table 7.5.

Relationship between sex, age, number of children, health, type of social benefits, country and feeling of stigmatisation as a problem of unemployment (Model 1). Added index of psychological selfesteem (model 2) and financial hardship (model 3). Ordinary OLS-regression. Standardised beta-coefficients Model 2 0.07* Ref.

Model 3 -0.03ns Ref.

Model 4 -0.08* Ref.

Male Female

0.05ns Ref.

0.06* Ref.

0.06n* Ref.

Under 50 years 50 years and over

0.11** Ref.

0.09** Ref.

0.06ns Ref.

Number of children below 181

0.03ns

0.05ns

0.05ns

Suffer from vital disease Other

Ref. -0.04ns

Ref. -0.03ns

Ref. -0.02ns

Recipients of ordinary unemployment benefits3 Other

-0.08*

-0.05ns

0.02ns

Ref.

Ref.

Ref.

-0.42**

-0.38**

Finland Denmark

Model 1 0.09** Ref.

Index of self-esteem (0-6)2 Index of financial hardship (0-3) Explained variation (R2) N 1

1. pct.

4 pct.

20 pct.

0.23** 24 pct.

797

797

797

797

Number of children in household below 18. 0, 1, 2, 3, 4 or more. Each indicator of self-esteem (see text) is given the values 1, 2, 3 and 4 (second question turned around) and afterwards added together and subtracted by two. We then get a sum index from 0 to 6, where 0 is lowest and 6 is highest self-esteem. Due to this rather low alpha we constructed an index based on 7 additional items with an alpha value of 0.85. But almost half of the unemployed were excluded due to missing answers on one or more of the nine items. Therefore we kept the original index of psychological resources. 3 In addition to the groups receiving regular unemployment benefits and not being in activation, we include 26 respondents on so-called unemployment in the Danish case, 56 respondents receiving no benefits at all, and 21 respondents on unemployment pension in the Finnish case. 2

Generosity and Stigmatisation

119

age, number of children below 18 in the household (0, 1, 2, 3 or more), sickness, types of benefit (regular unemployment benefits/other) into account. This reduces the country effect slightly from 0.09 to 0.07, but it remains significant. Taking a brief look at the control variables, we see a weak tendency towards more stigmatisation among men, but it is modest (0.05) and not significant in model 2. Nor do we see any relationship between number of children below 18 and feeling of stigmatisation as a problem. It might have been more relevant to look at children of nursing age, e.g. under age 5, but it is not possible due to the limitations of the data material. Nevertheless, the results indicate that effects from the housewife identity are limited. The same is the case for sickness (based on the respondents’ answers). Actually, those who indicated that they suffer from a vital disease felt stigmatisation as a larger problem than the healthy group. But again the relationship is weak (-0.04) and not significant. We do find a significant relationship between age and the feeling of stigmatisation as a problem. The group of long-term unemployed above 50 does feel stigmatisation to be less of a problem than the group below 50, which supports the thesis of a pensioner identity. Type of benefits is also included in the model, and we see the expected relationship. Those on regular unemployment benefits felt less stigmatised (-0.08) than the groups of ‘others’, which include the groups on social assistance, ‘wage’ from activation schemes etc.2 Further analyses – not presented in the table – show that the reduced country effect is caused by this variable; if it is left out, the country effect is not changed at all. Thus, in line with the findings from the previous chapter, it is not only the lower generosity, but also the higher degree of selectivism that explains the higher stigmatisation of long-term unemployed in Finland. Thus, except for this difference in degree of selectivism, which we have now controlled for, we have no reason to believe that everything else should not be equal in our most similar design. However, in model III we have tried to make the ‘best case’ for a psychological perspective by including a measure of self-esteem, i.e. we try to take into account that the Danish long-term unemployed have a higher self-esteem than the Finnish longterm unemployed. Now this difference could in itself be an outcome of differences in stigmatisation,3 or at least the causal relationships goes in both directions, but if we perceive self-esteem as an exogenous variable, it is a way to take a possible ‘Finnish melancholy’ into account. In the survey, the long-term unemployed were asked whether they agree or disagree with the following two statements, which go in opposite directions: (1) ‘Sometimes I feel I am no good at all’, and (2) ‘I am almost always quite satisfied with myself’. In order to get a better and more compact the 4-point scale, we can only make the country effect disappear (from a standardised betacoefficient of 0.23 in model I to 0.04 in model IV) from level of unemployment, see below. 2 The data material does not allow a clear distinction between insured and non-insured. Many long-term unemployed are in activation schemes, and therefore we do not know to what category they belong. This forces us to make the rough distinction used above. 3 Studies such as, for example, Aalport (1954, 1979), Rosenberg (1979), Crocker & Major (1989) assume that destructive societal stereotypes are more or less ‘automatically’ internalised into the deviants leading to psychological breakdowns.

120

The Institutional Logic of Welfare Attitudes

measure, these two self-esteem items are merged into a simple sum index, which we included in model III (see note to Table 8.5). On this 7-point scale, the Danish average is 4.6 and the Finnish average is 3.8, so the Danish long-term unemployed have higher self-esteem. Furthermore, model III shows a strong relationship (0.42) between this measure of what could be labelled psychological resources and a feeling of stigmatisation as a problem. Introducing this index actually increases the explained variance from 4 per cent to 20 per cent and eliminates the differences between Denmark and Finland (-0.03). Thus, one could suggest that the effect from generosity found in the most similar design could also be an outcome of crossnational differences in self-esteem among long-term unemployed. In order to rule out this explanation, we finally introduced the individual level index of financial hardship in model IV (in the previous models the generosity effect was captured by the country dummy), which even with background variables and psychological variables included turn out to have a strong (0.23) and highly significant impact on feeling of stigmatisation. The impact from financial hardship does not fully match the impact from self-esteem, but we can be more certain that financial hardship influences stigmatisation and not the other way around, i.e. the causal reasoning is much clearer than in the case of self-esteem. It is easier to explain cross-national differences in financial hardship than to explain why the Finnish unemployed have less self-esteem. Anyway, even with self-esteem included, the degree of financial hardship turned out to be important, which supports the causal reasoning in our most similar design. Furthermore, model IV shows that taking both differences in self-esteem and financial hardship (now tapped at the individual level) into account, being Danish is now estimated to have a significant negative impact on stigmatisation (from +0.09 in model I to –0.08 in model IV). Thereby, we actually end up with the country effect, which we expected the differences in unemployment levels to produce. Everything else being equal, the lower level of unemployment in Denmark seems to cause increased stigmatisation of the long-term unemployed. This fits nicely with the findings in Chapter 5.4 Generosity and cultural distinctiveness of the poor and unemployed. Proving the old assumption in a new way In this chapter, we tried to isolate the effect from the degree of generosity on the degree of ‘us–them-divide’ or the cultural distinctiveness of the long-term unemployed. This de-stigmatising effect from generous benefits or the stigmatising effect from tightfisted benefits is a basic assumption in modern welfare policy, but (to our knowledge) there is not much literature in comparative research on the topic. Following the logic from the previous chapter, we tried to ‘isolate’ the effect by means of a most similar 4 We also see that the effect from type of benefits disappears (0.02), but it does not suggest that it is not important. We would rather argue that it is caused by the fact that lower self-esteem is part of the causal reasoning, i.e. selectivism generates lower self-esteem so the latter variable captures the effect.

Generosity and Stigmatisation

121

design based on the comparison between two countries in the social democratic regime cluster. The most similar design enables us to hold such variables as country size, religious/linguistic/ethnic fractionalisation, and egalitarian values constant and thereby ‘isolate’ the suggested effect from degree of generosity. In terms of generosity, we saw a substantial difference between Denmark, being generous, and Finland, being less generous, which was clearly reflected in measures of economic hardship within our two samples of long-term unemployed. In the next step, we measured the experience of stigmatisation among the longterm unemployed, which showed the expected cross-national difference. Thus, the Finnish long-term unemployed felt more stigmatised than the Danish long-term unemployed, even in a situation were the unemployment level was about twice as high in Finland. Furthermore, we showed that the relationship was present at the individual level and could not be ruled out by other plausible explanations. Cross-national differences in the number of long-term unemployed on regular unemployment benefits reduced the country effect a little, but did not rule out the impact of generosity; nor could differences in self-esteem rule out the impact of generosity. Finally, we confirmed the finding from Chapter 5 that, everything else being equal, a lower level of unemployment leads to higher degree of stigmatisation among the poor and unemployed.

This page intentionally left blank

Chapter 8

Regime-dependent Perceptions and Social Assistance In the previous two chapters we have used ‘most-similar designs’ to isolate and thereby prove the existence of the institutional effects suggested by our theory. However, as already mentioned, ‘most-similar designs’ do not give us any measure of the relative importance of the independent variable. The cross-regime analyses in Chapter 5 indicated that the institutional effects were more important than ethnic fractionalisation, but as always such analyses are haunted but the ‘small-n problem’ and the fact that one has to use very rough operationalisations in order to get comparative measures. In order to substantiate our argument of substantial institutional effects we have conducted a national study. The previous chapters have not paid much attention to national surveys, as our purpose has not been to explain differences across individuals, but rather differences across countries. However, in this national study the analytical strategy is to measure perceptions – which according to the previous chapters are likely to be regime-dependent – and see how strongly they affect the individual’s support for welfare policy. The sample enables us to pinpoint perceptions of job opportunities, generosity, and perception of cultural distinctiveness of claimants, and basically it will be shown that variations in these (regime-dependent) perceptions can explain much of the variation in support for welfare policy. The sample also enables us to control for the impact from perceptions of immigrants. The chapter is divided into five sections. The first section presents the dependent variable, which we have operationalised in both relative (more, less or the same spending on social assistance) and absolute terms (how high a living standard should the public provide recipients of social assistance). The second section analyses the bivariate connections between perception of job opportunities of social assistance claimants, perception of work ethic among claimants, perception of generosity of social assistance, and relative and absolute support for social assistance. The third section analyses the impact of perceptions of immigrants on attitudes to social assistance. The American findings suggest a strong connection between ethnic stereotypes and lack of support for welfare policy; the question is whether this can also be found in the context of a social democratic welfare regime. Denmark actually seems to be the best case for a replication of the American experience as (1) the number of immigrants from third world countries has increased substantially over the last two decades, (2) the immigrants and their descendants are highly overrepresented among claimants of social assistance, and (3) the political

124

The Institutional Logic of Welfare Attitudes

mobilisation around the issue of immigration has been very strong in the last decade. In the fourth section we discuss problems concerning the directions of causality and conduct a simple control analysis that takes political affiliation into account. Finally we discuss how inferences from these analyses contribute to the overall argument of the book. Absolute and relative support for social assistance – the dependent variable In the previous chapters we have implicitly criticised the national studies of welfare attitudes, especially the American studies, for their ‘institutional blindness’. However, following the discussion at the beginning of Chapter 3, it should not be forgotten that there are also many advantages to national survey studies. First of all, the validity is increased because respondents are asked in the same language, and one can be much more specific about the welfare programme about which respondents are asked to express their opinion. In our national Danish sample, we asked the respondents to express their opinion about the social assistance programme, i.e. the programme that covers those of working age who cannot obtain or have lost their entitlement to regular unemployment benefits. It is a very mixed group, but it is the closest we get to the group respondents will label ‘poor’ or ‘being on the dole’. We have one relative measure of the dependent variable: (1) ‘Do you think that the public spends too much money, the right amount of money, or too little money on social assistance for the individual?’; and two absolute measures: (2) ‘Is it a public responsibility to provide a decent living standard for claimants of social assistance’, and (3) ‘How high a living standard do you think the public should provide claimants of social assistance’. The distributions of answers on these questions are shown in Table 8.1.

Regime-dependent Perceptions and Social Assistance

Table 8.1.

125

One relative and two absolute measures of public support for social assistance. Denmark, 2004, per cent

Spending on social assistance:

Too little money

14

The right amount of money

49

Too much money

27

A public responsibility to provide decent living standard for recipients of social assistance:

How high a living standard should the public provide recipients of social assistance:

Fully agree

39

7: The public should provide a living standard that equals that of an ordinary wage earner

5

Partly agree

44

6

9

Partly disagree

11

5

15

4

22

3

18

2

16

1: The public should provide a living standard that just enables one to get by

13

Don’t know

2

Fully disagree

Don’t know

4

Don’t know

11

2

Total

100

100

100

N

1148

1148

1148

Wording: 1) Do you think that the public spends too much money, the right amount of money, or too little money on social assistance for the individual? 2) How much do you agree or disagree with the following statement: It is a public responsibility to provide a decent living standard for claimants of social assistance? 3) How high a living standard do you think the public should provide claimants of social assistance? Please answer on a scale from 1 to 7, where 1 is a living standard where one can just get by, and 7 is a living standard that equals that of an ordinary wage earner?

126

The Institutional Logic of Welfare Attitudes

The most important point in Table 8.1 is that there is variation on the dependent variables, i.e. the respondents actually hold different attitudes; one could also say that there is something to explain. On the relative variable, most respondents (49 per cent) indicate that the public spends the right amount of money on social assistance for the individual. Twenty-seven per cent think too much money is spent, and only 14 per cent think too little money is spent. This latter figure might sound surprising as a social democratic regime should generate very positive attitudes towards the poor and unemployed, but it is of course affected by the relative nature of the question; when asked in absolute terms, 39 per cent ‘fully agree’ and 44 per cent ‘partly agree’ that it is a public responsibility to provide a decent standard of living for claimants of social assistance. Only 4 per cent ‘fully disagree’. These figures are more in line with our expectations about attitudes in a social democratic regime. However, from a methodological point of view, it can be problematic that all respondents cluster at one end of the scale. We will therefore primarily make use of the second absolute measure of public support for social assistance, i.e. the question that asks respondents to specify the living standard they think should be provided. At one end of the 7point scale we find 5 per cent who think that the public should provide claimants of social assistance with a living standard that equals that of an ordinary wage earner. At the other end we find 13 per cent who think that only a living standard that enables one ‘just to get by’ should be provided. As is often the case using such a scale, the middle bracket is the most common answer, but with only 22 per cent using it (i.e. answer 4), the ‘middle-effect’ is not very strong. Actually, we find a nice (from a methodological point of view) distribution across all the brackets. The average of the scale is 3.6. As expected, the three variables are quite strongly correlated, also the relative and absolute measures, which indicates (as argued in Chapter 3) that respondents asked in absolute terms, might not be able to overlook the policies that actually are in place.1 In the following sections we elaborate on this point as we use both the relative and the absolute measure in the analyses. Perception of generosity, job opportunities, work ethic, and support for social assistance As a point of departure this section will show the bivariate correlation between perception of job opportunities, generosity, work ethic, and support for social assistance. Each variable is connected to one of the regime dimensions presented in the causal model in Figure 4.1. It is not without methodological problems to substitute these ‘structural variables’ with ‘perceptions’, but this will be discussed in section four. Table 8.2 shows the distribution of the three measures.

1 The correlation (gamma-coefficients) between the two absolute measures is 0.42. The correlation (gamma) between the relative measure and the first absolute measure (decent standard) is 0.60, and the second absolute measure (which standard) is 0.52.

Regime-dependent Perceptions and Social Assistance

Table 8.2.

127

Distributions of perception of job opportunities, level of generosity, and work ethic. Denmark, 2004, per cent

Assessment of opportunities of claimants of social assistance to get into the Danish labour market

Assessment of living conditions of claimants of social assistance in Denmark compared to other countries

Assessment of work ethic of claimants of social assistance compared to work ethic of employed

Very good opportunities

Much better conditions in Denmark

Much lower work ethic

18

Lower work ethic

52

More or less the same

20

Higher work ethic

1

Don’t know

9

Good opportunities Neither good nor poor opportunities Poor opportunities

Very poor possiblities Don’t know

4 22

33 30

5 5

Better conditions in Denmark More or less the same conditions

25

35

23

Worse conditions in Denmark

2

Much worse conditions in Demark

0

Don’t know

15

Total

100

100

100

N

1150

1149

1149

Wording: 1) How do you believe the opportunities are for recipients of social assistance to get into the Danish labour market? 2) How do you think the conditions for recipients of social assistance are in Denmark as compared to the conditions in the countries with which we normally compare ourselves? 3) How do you believe the work ethic, i.e. the desire and willingness to work, is among claimants of social assistance as compared to the work ethic among employed?

128

The Institutional Logic of Welfare Attitudes

We expected the public perception of job opportunities in social democratic regimes to be higher than in conservative regimes and lower than in liberal regimes. We do not have figures from other countries, but in the first column in Table 8.2 we can see that the Danish respondents actually indicate a middle position; 33 per cent answer ‘neither good nor bad opportunities’, 22 per cent answer ‘good opportunities’, and 30 per cent answer ‘bad’ opportunities. Only 4 per cent use the bracket ‘very good opportunities’, and only 5 per cent use ‘very bad opportunities’. In terms of perception of generosity, we expected the respondents in social democratic regimes to be aware of the high redistribution that takes place. Again we do not have results from other countries, but the second column in Table 8.2 shows that most Danes actually think that claimants of social assistance are better off in Denmark than in the countries normally used for comparison; 25 per cent answer ‘much better conditions’, 35 per cent ‘better conditions’, and 23 per cent ‘more or less the same conditions’. Fifteen per cent answer ‘don’t know’ (which actually is a modest share, considering the character of the question), and only 2 per cent answer ‘worse conditions’. Finally, we expected the perception of cultural distinctiveness of the poor and unemployed to be low in social democratic regimes. The perception of cultural distinctiveness is operationalised as perceived difference in work ethic between social assistance claimants and employed. The answers in the last column of Table 8.2 indicate that people do perceive a difference to be present; 18 per cent indicate ‘much lower work ethic’ and 52 per cent ‘lower work ethic’ among claimants of social assistance. Only 20 per cent answer ‘more or less the same’, 1 per cent answer ‘higher work ethic’, and finally 9 per cent ‘don’t know’.

Figure 8.1. The connection (gamma-coefficients) between perceptions of job opportunities, level of generosity, work ethic, and public support for social assistance. Attitudes towards spending and the living standard that should be provided

Regime-dependent Perceptions and Social Assistance

129

Figure 8.1 shows how strongly these three variables are linked to public support for social assistance. On the left-hand side we use attitudes towards spending (the relative measure) as the dependent variable. We see that all three relationships are strong and highly significant. Using gamma-coefficients, the strongest relationship (0.58) is between perceptions of work ethic and spending on social assistance, followed by perception of generosity (0.55), and perceptions of job opportunities (0.40). If we treat the variables as interval variables and conduct a simple OLS-regression, the result is that these three variables are able to explain 28 per cent of the variation in the individual’s attitudes towards spending on social assistance. On the right-hand side we use attitudes toward the living standard that the public should provide for claimants of social assistance as the dependent variable. Again we see three highly significant relationships, but the strength is somewhat reduced.2 As expected, the impact of perceptions of generosity (compared with the impact of the other two variables) is reduced when the support for social assistance is measured in absolute terms, but it is still fairly strong (0.31) – actually still stronger than the perceptions of job opportunities (0.18). It again suggests that respondents cannot ignore the present level of generosity, even when they are asked an absolute question. Nevertheless, again the perceptions of work ethic among claimants of social assistance come out as the strongest predictor of support for welfare policy (0.41). If we again conduct a simple OLS-regression, the result is that these three variables are able to explain 17 per cent of the variations in the respondents’ attitudes towards the living standard that the public should provide claimants of social assistance. Thus, these findings do indeed confirm that the four regime dimensions covered in Chapter 4 are highly relevant in order to explain public support for welfare policy (in the narrow sense). Before discussing the causality behind these relationships, we will discuss to what extent the strong impact of work ethic, as in the American survey studies (Gilens 2000), is rooted in ethnic stereotypes. The relationship between perception of immigrants and support for social assistance In Chapter 4, the social democratic welfare regimes were presented as small and homogeneous nation states. Measured in terms of religious, linguistic, and ethnic fractionalisation (Table 4.2), the social democratic regimes were on the average more homogeneous than the conservative and especially the liberal regimes. Nevertheless, as mentioned in Chapter 4, the increased immigration, especially within the last decade, potentially erodes these preconditions for welfare state developments. Based on the American experience, Alesina & Glaeser (2004) suggest that it is only a matter of time before support for welfare policy will also erode in Europe.

2 And the reduced strength is not simply an effect of the dependent variable having more brackets. Even if the 7-point scale is reduced to a 3-point scale (as the spending variable), the gamma-coefficients remain lower.

130

The Institutional Logic of Welfare Attitudes

In many respects we should expect such an erosion of support for welfare policy caused by ethnic divides to be very strong in Denmark. As mentioned, we will point out three factors. First of all the proportion of immigrants from non-western countries and their descendants has increased rapidly over the last two decades; in 1990 the share was 2.0 per cent, and by 2003 it had increased to 5.7 per cent of the population (Danish Statistics 2000, 2003). Secondly, immigrants and their descendants are highly overrepresented in the social assistance system. In 2000, the Ministry of Social Affairs calculated that the share of immigrants and their descendants on social assistance was 21 per cent in 1994, which in 1999 had increased to 36 per cent (Ministry of Social Affairs 2000, chapter 3). Furthermore, we also know that immigrants and their descendants in general stay on social assistance for a longer period (Ministry of Refugee, Immigration and Integration Affairs 2004). Thirdly, the immigrant issue has climbed to the top of the political agenda. Thus, Denmark is the social democratic regime with the largest anti-immigrant party; the Danish People’s Party was established in 1995 and got 13.3 per cent of the votes in the last (2005) election. Thus, on the one hand, Denmark has everything that potentially could erode public support for social assistance. But on the other hand Denmark has a very universal and generous welfare policy, which, following the previous chapters, is believed to generate positive images of the poor and unemployed. Thus, we have opposing forces that could push support for social assistance in different directions. In the following we will analyse three pieces of empirical evidence. First of all we look at the connection between perceptions of the share of immigrants among claimants of social assistance and support for social assistance. Naturally, a good number of respondents, 18 per cent, answered ‘don’t know’ to the question about the share of immigrants among claimants of social assistance. But still we have one third that indicates a share below the actual level (0 to 29 per cent), one third that indicates a share above the actual level (40–100 per cent), and 14 per cent that indicate approximately the right share (30–39 per cent). If we exclude the ‘don’t knows’and look at the correlation with our dependent variables, we find no connection between perception of share of immigrants on social assistance and support for social assistance; roughly speaking, the gamma-coefficients are zero (-0.02 and -0.02). Thus, in Denmark, the immigration issue does not seem to be directly connected to public support for social assistance as it is in the US context.3 3 Such a ‘knowledge question’ (the proportion) could be influenced by educational level, but even controlled for such differences, an association does not emerge. We have also looked at the relationship among respondents affiliated with different parties. But again no clear connection was found, not even among individuals affiliated with the Danish People’s Party. Finally we controlled for the general attitudes towards immigrants using the question ‘to what extent should people who have condemnatory opinions about people with different race or ethnicity be allowed to hold public meetings?’. It is not a perfect measure (as opinions about free expression is also at stake), but among those who answer ‘should definitely be allowed’, we actually find the expected connection to be significant, i.e. within this limited group, those who perceive the share of immigrants on social assistance as large indicate less

Regime-dependent Perceptions and Social Assistance

131

Figure 8.2. Connection between perception of the share of immigrants on social assistance and relative and absolute support for social assistance in a social democratic regime. Denmark, 2004, gamma-coefficients

Figure 8.3. The connection between general perception of work ethic among claimants of social assistance, work ethic among immigrants on social assistance and absolute and relative support for social assistance. Denmark, 2004, gamma-coefficients

Secondly, we will analyse the connection between perception of work ethic among immigrants on social assistance and support for social assistance. In the sample we have measured the perception of work ethic among immigrants by asking respondents support for social assistance. Thus, in this small segment, which is strongly against immigrants, the perceived proportion of immigrants among social claimants does influence support social assistance.

132

The Institutional Logic of Welfare Attitudes

to take a stand on the statement; ‘many immigrants on social assistance actually do not want to work’. In Figure 8.3, we see that the bivariate correlation between this perception and support for welfare policy is present and substantial; both in terms of relative support (spend more money, gamma-coefficient 0.30) and absolute support (provide a high living standard, gamma-coefficient 0.17). However, the connection between perception of work ethic among ethnic minorities and support for welfare policy does not seem to be as decisive as in the American findings. Furthermore, we argue that the respondents’ assessments of the work ethic of claimants of social assistance in general need to be taken into account. In Figure 8.3, this general assessment is indeed correlated with the assessment of the work ethic of immigrations on social assistance (gamma-coefficients in both cases 0.45), and at the same time the general assessment influences the support for social assistance (respectively 0.58 and 0.41). If we control for this general assessment of work ethic among claimants of social assistance, the effect from assessment of work ethic among immigrants on social assistance is halved in both models (from 0.30 to 0.16 and from 0.17 to 0.08), i.e. again the impact of ethnic stereotypes seems to be limited in the social democratic regime.4 Finally, we will take at look at the development of public attitudes in Denmark within the last decade. The question is whether we at the aggregated level can see that the increased immigration, the increased overrepresentation of immigrants in the social assistance system, and the increased political mobilisation began to erode public support for immigrants’ entitlement to social assistance. The number of available indicators is very limited, but we have found two items that more or less capture what we are looking for. The first question asks respondents what they think about the statement, ‘immigrants who have not learned Danish should not be able to get support during unemployment’. It is indeed not a perfect item as it asks about entitlement to support during unemployment and not specifically about social assistance. And besides attitudes towards entitlement in general, it also taps attitudes towards the importance of learning the language. Nevertheless, the advantage is that the question was asked prior to the strong political mobilisation around the immigration issue, which appeared on the public agenda in the 1994 election (8 per cent mentioned it as one of the most important problems politicians should deal with) and received increasing attention in the following elections.5 4 The assumption is that the general assessment of work ethic among claimants of social assistance is generated by the regime-specific mechanisms described in Chapter 4, general left/right-wing political orientation etc. However, one could naturally argue that the general assessment of work ethic might be caused by the assessment of work ethic among immigrants on social assistance. If this is the case, we should expect the connection between the work ethic variables to be increasingly correlated the higher the share that the respondents think immigrants make up of the claimants of social assistance. This connection cannot be found, which supports the idea of the general assessment of work ethic logically being prior to the assessment of work ethic among immigration on social assistance. 5 In the 1998 election, the share was 14 per cent, and in the 2001 election the share was 20 per cent (Goul Andersen 2003, p. 138). It should be mentioned that the issue appeared on

Regime-dependent Perceptions and Social Assistance

Table 8.3.

Attitudes towards immigrants’ entitlement to social support during unemployment in a social democratic regime. Denmark in 1993 and 2004, per cent

Statement: Immigrants who have not learned Danish should not be able to get support during unemployment: Completely agree Partly agree Partly disagree Completely disagree Don’t know Total N

133

1993

2004

40 26 15 17 3

40 33 14 11 3

100 1666

100 1161

Source: 1993: Togeby & Gaasholt, 2000:130; 2004: Own national sample.

Basically, Table 8.3 indicates that the increasing political mobilisation around the immigrant issue from 1994 and onwards has not substantially lowered public support for immigrants’ entitlement to support during unemployment. The share that ‘completely agreed’ that immigrants who have not learned Danish should not be able to get support during unemployment was 40 per cent in both 1993 and 2004. The share that ‘partly agrees’ has increased, from 26 per cent in 1993 to 33 per cent in 2004, so it is fair to say that there has been a modest decline in public support. Nevertheless, if we take a look at our second indicator, which more specifically deals with social assistance, there is no sign of eroding public support for immigrants’ entitlement. In the 1998 election survey, the public was asked what they thought about the statement, ‘immigrants should have the same right to social assistance as Danes even though they are not Danish citizens’. Again it is not a perfect measure as it involves, besides attitudes towards entitlement, the question of citizenship. But for our purpose it is better than the first indicator (in addition to the problem that we can only go back to 1998). In 1998, 11 per cent ‘completely agreed’ that immigrants should have the same right to social assistance as Danes. The number was 17 per cent in 2000 and 15 per cent in 2004. In 1998, 16 per cent ‘agreed’, a figure that in 2000 and 2004 increased to, respectively, 21 per cent and 22 per cent, see Table 8.4. So actually we see a small increase in public support for immigrants’ entitlement to social assistance from 1998 to 2004.

the political agenda around 1984/1985 but it first became ‘high politics’ in the 1990s.

134

Table 8.4.

The Institutional Logic of Welfare Attitudes

Attitudes towards immigrants’ entitlement to social assistance in a social democratic regime. Denmark 1998, 2000, and 2004, per cent

Statement: Immigrants should have the same right to social assistance as Danes even though they are not Danish citizens. Completely agree Partly agree Neither agree nor disagree Partly disagree Completely disagree Don’t know Total N

1998

2000

2004

11 16 12 25 32 3

17 21 14 26 22 1

15 22 12 22 25 4

100 2001

100 1234

100 1167

Source: 1998: Danish election survey; 2000: ACNielsenn AIM for Mandag Morgen; 2004: Own national sample.

Naturally, we should remember that we only have two indicators and that comparing the same question in different surveys can sometimes generate methodological problems (the problem is that respondents might give different answers to the same question because they have been through another questionnaire), but combined with the analyses above it seems safe to conclude that so far the increased immigration has not eroded public support for social assistance. Why this is the case is naturally open for discussion, but based on the previous chapters we will suggest that the institutional structures of the social democratic welfare regimes dampen the potential ‘them–us conflict’ found in liberal welfare regimes. Controlling the analyses for directions of causality Above, we assumed that the individual’s perceptions of claimants of social assistance come prior to their support or lack of support for social assistance. Following the discussion at the end of Chapter 2, the idea is that variations in support for welfare policy are an outcome of ‘political man’s’ different perceptions of reality. We contrasted this position with the ‘economic perspective’ (which basically would explain variations in support for social assistance by variations in individuals’ selfinterest in social assistance) and the ‘sociological man perspective’ (which basically would explain variations in support for social assistance by variations in basic

Regime-dependent Perceptions and Social Assistance

135

egalitarian values). The national sample does not allow us to analyse the economic man explanation in any detail. However, it is clear that most respondents do not have much self-interest in a generous social assistance system. Furthermore, the 27 unemployed in our sample do not express remarkably different attitudes. On the 7point scale of living standard, the average of this sample is 4.0 compared with 3.6 in the overall sample. We will primarily pay attention to the sociological man perspective within which one could contest the idea that perceptions of claimants of social assistance come prior to support for social assistance. Instead, the line of reasoning would be that respondents with basic egalitarian attitudes will support social assistance, and in order to legitimatise and defend this position they will tend to claim that job opportunities are limited, work ethic of claimants is high, and the level of generosity is not higher than in other countries. And the other way around for individuals holding antiegalitarian values; they will be less supportive for social assistance and legitimise and defend their position by claiming that job opportunities are good, work ethic is low, and the level of generosity is already high. We have already delivered a criticism of the culture explanations, which more or less can be seen as the aggregated edition of this line of reasoning. However, we may try to substantiate our causal reasoning by taking the political affiliation of the individuals into account. Based on the votes in the last national election the individuals have been sorted in (1) a group that voted for socialist parties, (2) a group that voted for the centre parties, (3) a group that voted for the bourgeois parties, and finally (4) a group that voted for the anti-immigrant party (see note to Table 8.5 for operationalisation). If both support and the measured perceptions are an outcome of this basic political affiliation of the individuals, we should expect the connection between perceptions and support for social assistance to vanish or at least become much weaker when we look at the correlation within each of these four groups.

136

Table 8.5.

The Institutional Logic of Welfare Attitudes

Causal model between perceptions of job opportunities, generosity, work ethic and support for social assistance controlled for political affiliation. Gamma-coefficients and explained variation

Voters of left-wing parties: (N min. 295) Perceptions of job opportunities Perceptions of generosity Perceptions of work ethic Explained variation: Voters of centre parties: (N min. 80) Perceptions of job opportunities Perceptions of generosity Perceptions of work ethic Explained variation: Voters of bourgeois parties: (N min. 332) Perceptions of job opportunities Perceptions of generosity Perceptions of work ethic Explained variation: Voters of anti-immigrant parties: (N min. 74) Perceptions of job opportunities Perceptions of generosity Perceptions of work ethic Explained variation:

Spending on social assistance

Provided living standard

0.40** 0.54** 0.47** 25 pct.

0.19** 0.23** 0.33** 12 pct.

0.35* 0.49** 0.59** 25 pct.

0.18ns 0.24ns 0.50** 16 pct.

0.28** 0.42** 0.50** 15 pct.

0.06ns 0.37** 0.38** 15 pct.

0.30* 0.66** 0.72** 29 pct.

0.21ns 0.22ns 0.15ns 7 pct.

Note: Left-wing parties: Socialdemokratiet, Socialistisk Folkeparti, Enhedslisten. Centre parties: Det radikale venstre, Centrum demokraterne, Kristdemokraterne. Bourgeois parties: Venstre, Konservativt Folkeparti. Anti-immigrant parties: Dansk Folkeparti og Fremskridtspartiet. * significant at 0.05, ** significant at 0.01.

Regime-dependent Perceptions and Social Assistance

137

In most cases, Table 8.5 shows that the correlations for the overall sample presented in Figure 8.1 are replicated when we look within the political groups. If, for example, we take the voters of left-wing parties, Table 8.5 shows that the correlations between perceptions of job opportunities (0.40), generosity (0.54), work ethic (0.47) and support for spending on social assistance replicate the pattern presented in Figure 8.1 (0.40, 0.55, and 0.58). The same goes for the two groups that voted for centre parties and the anti-immigrant party, respectively. Thus, the correlations presented in Figure 8.1 do not disappear or decrease substantially when we take political orientation into account. It supports the argument that differences in the perceptions of reality actually influence support for a given welfare policy. However, there is an exception to the rule. Among the voters of the two bourgeois parties, the correlations between attitudes towards spending on social assistance and perceptions of job opportunities (from 0.40 in the whole sample to 0.28) and perceptions of generosity (from 0.55 to 0.42) are somewhat reduced. In terms of attitudes towards provided living standard, it is only the impact of perceptions of job opportunities that is reduced (from 0.18 to 0.06). Thus, among bourgeois voters, our causal model is able to predict less of the variation in support for social assistance (respectively 15 per cent and 15 per cent), but still the model is not at all without relevance. Inferences from the individual level The chapter has shown that the perceptions of the job opportunities, generosity, and cultural distinctiveness of the poor and unemployed, which we have already argued are regime-dependent, have a strong influence on public support for welfare policy. On a Danish sample it was shown that differences in these perceptions measured by three items were able to explain 28 per cent of the differences in support for spending on social assistance and 19 per cent of the differences of the attitudes towards the living standard that the public should provide claimants of social assistance. Within our ‘political man perspective’, these findings support the overall argument that cross-national differences in support for welfare policy are likely to be an outcome of the three dimensions we pinpointed in Chapter 4 by combining the welfare regime theory and the deservingness literature. However, from a ‘sociological man perceptive’, one could argue that both support for social assistance and the perceptions we measured could be an outcome of differences in basic egalitarian values. In the previous section, we tried to rule out that explanation by taking the individual’s affiliation with political parties into account; the conclusion was that looking within four political segments the relationship between the regime-dependent perceptions and support for social assistance was more or less unaffected. The exception was the group that voted for one of the two bourgeois parties; here relationships were somewhat reduced, especially the impact of perceptions of job opportunities. As a point of departure we measured cultural distinctiveness as the perceived difference in work ethic among the employed and claimants of social assistance,

138

The Institutional Logic of Welfare Attitudes

which in fact turned out to be the strongest predictor of support for social assistance. At first glance, this finding could suggest that we have a replication of the American studies, which have shown a strong connection between the perception of work ethic among blacks and support for welfare policy. Actually such a scenario would be easy to imagine in Denmark as (1) the share of immigrants from third world countries and their descendants has increased substantially within the last two decades, (2) they are highly overrepresented in the social assistance system, and (3) the political mobilisation around the immigrant issues has been very intense since the 1994 election. However, our analyses found no support for this scenario. Firstly, no connection could be found between the public perceptions of the share of immigrants among claimants of social assistance and support for social assistance. Secondly, a connection was found between perception of work ethic among immigrants on social assistance and support for social assistance, but when the general assessments of work ethic among claimants was taken into account, the effect was halved. Thirdly, we could not find a decrease in public support for the entitlement of immigrants to social assistance (from 1998 to 2004) and to social support during unemployment in general (from 1993 to 2004). Overall this leads to the conclusion that due to the institutional structure of the welfare regimes, the American experience with ethnic stereotypes and public support for welfare policy cannot directly be transferred to the conservative and social democratic regimes. This will be discussed further in Chapter 9.

Chapter 9

Deservingness and Welfare State Development One of the founding fathers of modern welfare state research, Richard Titmuss (1907–1973), posed the following questions at the end of his last book: ‘What effect does the system have on the social and psychological sense of community? Does it have divisive or unifying effects and in what sense and for what groups? Does it, in short, widen or diminish the concept and consciousness of “who is my neighbour”’ (Titmuss 1974). The same kind of interest in a connection between the system – or call it institutions – and attitudes of the citizens – or call it actors – was the point of departure for this book. Based on the insights from comparative welfare studies, the book has tried to open the ‘black box’ of the formation of support for welfare policy in different welfare regimes. Most people are probably not aware that they live in a distinct welfare regime, and the majority of the many people who have answered our questionnaires would probably reject the idea that their answers are heavily influenced by this institutional arrangement. However, they may grant us that it is not easy to pinpoint the origin of one’s either positive or negative feelings about the poor and unemployed. To use the words in Charles Lemert’s introduction to ‘the sociological life’, we are dealing with ‘the mysterious power of social structures’, which ‘appear to people in the course of daily life as through a mysterious fog’ (Lemert 2002, p. 118). Theoretical reasoning has been our tool to shed some light on this mysterious fog, which influences the welfare attitudes of the individuals. The point of departure has been Esping Andersen’s influential welfare regime theory presented in Chapter 2, which basically claims that the western countries can be divided into clusters of social democratic, conservative and liberal welfare regimes. Esping Andersen’s theoretical reasoning both draws on the power resource theory, often associated with Walter Korpi, and on insight from modern historical institutionalism. One of the pivotal elements of the theory is that the institutional settings of a given welfare regime create a feedback-mechanism, Esping-Andersen use the term stratification mechanism, which enables us to explain regime specific trajectories in the transition from industrial to post-industrial societies. Following this line of reasoning the concept of path-dependency has become widespread within comparative research. However, it is also widely recognised that claims of path-dependency are often under-specified (e.g. Pierson 2000), i.e. it is not described in detail how this feedback mechanism actually works. This criticism not only applies to newcomers’ use of a new fashionable concept, but also to the well-established power resource and welfare

140

The Institutional Logic of Welfare Attitudes

regime theory. At least it is recognised that in both theoretical traditions, the macromicro link – the feedback from structure to actor – has not been the object of much empirical research. In Chapter 3, we quoted Walter Korpi as saying that ‘the empirical testing of macro-micro links among institutions and the formation of interest and coalitions provides a major challenge for social scientists, but comparative microdata currently are lacking’ (Korpi & Palme 1998, p. 682). Indeed the lack of data on welfare attitudes is a major obstacle, but over the last decade more international survey data have become available. Especially based on the ISSP social inequality and role of government module, a number of studies have analysed the connection between welfare regimes and welfare attitudes, but the results are contradictory. Depending on the countries included and the selection of items, some studies report a regime effect, and others do not. The review of these studies in Chapter 3 showed that the studies that find the expected pattern measure attitudes towards public policies that primarily concern the living conditions of the poor and unemployed. This is part of the reason for the focus of this book. Another central finding from the review in Chapter 3 was that the previous studies were not able find the expected effects from class interest, self-interest, and egalitarian values. Thus, on the one hand, previous studies find the expected crossnational differences (when measuring support for welfare policy in the narrow or American sense), and on the other hand they demonstrate an inability to explain the mechanism that leads to these cross-national differences. In order to substantiate this argument and especially in order to highlight the deficiency of the argument of American exceptionalism, we conducted a number of simple analyses that pinpointed the egalitarian attitudes found in each regime. We argued that the best measures are the perceptions of just income differences captured by the ISSP social inequality module. By means of these data we showed that, measured as the just difference between skilled workers and four top occupations, the most anti-egalitarian values were actually found in the conservative welfare state regimes. So instead of a general American culture of inequality, we argued that what distinguishes Americans and others living in liberal welfare state regimes is anti-egalitarian attitudes towards the very ‘bottom’ of society. Why this is the case was given an institutional explanation in Chapter 4. Chapter 3 showed that we need a new theoretical explanation that links the macro level – welfare regimes – and the micro level – the welfare attitudes of the electorate. Chapter 4 made an attempt to establish such a theory. The overall idea was to combine the welfare regime theory with the literature on deservingness criteria, which for strange reasons have lived separated lives. In doing so, we described three dimensions of welfare regimes – the degree of selectivism in welfare policy, the degree of generosity in welfare policy, and the job opportunities on the labour market – which were likely to influence the degree to which the poor and unemployed were able to fulfil five central deservingness criteria. Within the deservingness literature, a number of studies have shown (1) how the public judgement of deservingness influences policy attitudes, and 2) how the criteria for deservingness are stable over time and across countries. Following Oorschot (2000), the five criteria concern the

Deservingness and Welfare State Development

141

question of need, control, identity, attitude, and reciprocity. This combined theory leads to the expectation that the perception of the poor and unemployed was most negative in liberal regimes (maybe moderated by the fact that the recipients really are in need), more positive in conservative regimes (especially caused by the modest job opportunities), and most positive in social democratic regimes (maybe moderated by the fact that many of the potential poor have rather good living conditions). In the following chapters we substantiated this theoretical argument. By means of different analytical strategies we found empirical evidence for the theses deduced from our theoretical reasoning in Chapter 4. Chapter 5 analysed the question of ‘poverty being caused by laziness and lack of will power’, which was asked in the first and second wave of WVS and Eurobarometer studies from 1976, 1989, 1993 and 2001. This question was believed both to ‘tap’ the question of control and the question of identity. In general, we found a rather strong connection between welfare regimes and the perception of poor being in control of their poverty. The limited number of cases made it, as always, difficult to apply a ‘large-n logic’, but if we stick to the dataset with most countries included – WVS from 1990 – the expected pattern was found. We found a clear connection between (1) the liberal welfare regimes and a high share answering that the poor were in control of poverty; (2) the social democratic regimes and a low share answering that the poor were in control of poverty, and (3) the conservative regimes and a medium share answering that the poor were in control of poverty. There are a number of deviant cases – especially Finland and France – and the differences between the European liberal regimes, i.e. the UK and Ireland, and the other European countries seem modest in the Eurobarometer surveys conducted in 1989, 1993 and 2001. In the Eurobarometer survey from 1976 the picture was more clear-cut. We also found a strong relationship between unemployment level and public perception of poverty being caused by ‘laziness and lack of will power’; the higher level of unemployment, the lower perception of poor being in control of poverty. The impact of generosity and ethnic fractionalisation turned out to be less important in the multivariate analyses at the aggregated level. Furthermore, the analyses in Chapter 5 showed that the introduction of variables at the individual level such as gender, age, education, occupation, employment status and left/right-wing orientation did not change the regime effect much. The chance of a respondent living in a liberal regime answering that poverty is caused by laziness and lack of will power was still estimated to be twice as high as for a respondent living in a social democratic regime. Hence, the regime effect is not a matter of differences in the size of groups that (in all regimes) tend to hold ‘negative’ or ‘positive’ attitudes. In general, we found rather similar patterns at the individual level (which suggest that the regime effect is ‘imposed’ on all members living within each regime) but we also found groups that ‘react’ differently across regimes. The gender effect seems to be especially strong in the social democratic regimes, which is in line with findings from previous studies but we suggested a somewhat different interpretation; namely that this gender effect could be explained by a ‘tough guy effect’ rather than a ‘soft women effect’. The divide between skilled and unskilled workers was most pronounced in the conservative regimes, which coincides nicely

142

The Institutional Logic of Welfare Attitudes

with the insider–outsider characteristic of their labour markets, and the divide between blue-collar and white-collar workers was most pronounced in the liberal regimes. In the latter case, the blue-collar workers were more inclined to answer that poverty was due to laziness (compared to white-collar workers) than blue-collar workers in the conservative and social democratic welfare regimes. The findings in Chapter 5 support the argument that the institutional set-up of the welfare regime to a large extent influences the public perception of poor and unemployed, but still the impact of ethnic composition prior to the building of modern welfare states remained a possible alternative explanation. Thus we have to be cautious with extreme institutional arguments such as ‘only institutions matters’. In the US case, the slave economy and the long history of immigration are historical conditions established prior to the welfare state building of the 20th century cannot be ignored. Following the same line of reasoning, the high public support for welfare policy in the social democratic regimes could be routed in the absence of ethnic minorities and the small size of the countries. To some extent, it is a ‘chicken and egg discussion’, but by means of mostsimilar design, we tried to isolate the effect from degree of selectivism and degree of generosity (respectively in Chapters 6 and 7), i.e. the effects caused by the programmatic structure of welfare policy. In Chapter 6, additional questions in the ISSP surveys from 1999 in the Nordic countries were used to make these ‘within regime comparisons’ where degree of religious, linguistic and ethnic fractionalisation, egalitarian values, political institutions, country size etc, was kept constant. On questions about the level of stigma associated with receiving different types of welfare benefits, the survey results clearly showed that the most universal schemes, e.g. basic old-age pension and child-allowances, were associated with little stigma, and the most selective schemes, e.g. social assistance, with a high level of stigma. The finding is almost trivial, and the problem is that we cannot really tell whether we see a true institutional effect caused by the programmatic structure of the welfare policy or simply an effect of the social risk covered by the schemes. However, the comparisons between the programmes that cover the same risks in the Nordic countries, but differ in terms of the number of people covered enable us to rule out this objection. As expected we found (1) social assistance to be least stigmatising in Finland (because the number of recipients is quite high); (2) housing allowances to be most stigmatising in Norway (few recipients); and (3) the early retirement schemes to be more stigmatising in Finland than in Denmark (as few receive it). The Nordic comparison of stigma associated with unemployment benefits did not give clear-cut results, but the analyses generally confirmed one of the theoretical axioms in welfare state research; namely a relationship between selectivism and stigma. It is more difficult to isolate the effect from degree of generosity on the perceived cultural distinctiveness of the poor and unemployed, which was the topic of Chapter 7. Basically, the idea is that generous benefits enable groups in risk of poverty to maintain a way of life that is fairly similar to that of the rest of the population. This should make it easier to fulfil the criterion of identity. Tight-fisted benefits were believed to work the other way around. Again the analytical strategy was to isolate

Deservingness and Welfare State Development

143

the expected effect by a most similar design, and again we used the Nordic countries, which in so many respects are very similar. However, in terms of generosity towards the unemployed, Denmark and Finland differ quite substantially. The fact that Denmark has the most generous unemployment benefits in the Nordic countries (for low-income groups), and Finland has the least generous schemes in the Nordic countries (for low-income groups), led to the book that the Finnish unemployed should be perceived as more culturally distinct than the Danish unemployed. Using two large samples of long-term unemployed in Denmark and Finland, we measured the cultural distinctiveness as the level of stigmatisation felt by the unemployed themselves. As expected, the Finnish long-term unemployed experienced much more stigmatisation than the Danish long-term unemployed – even in a situation where unemployment was twice as high in Finland. On the individual level, we also found a very strong connection between the experiences of economic hardship (measured as the ability to pay current expenses, an unexpected bill, and the prospect of an economic future) and the experiences of stigmatisation. The relationship was stable even if we controlled for sex, age, type of benefit, health, number of children living at home, and self-esteem. Finally, taking these control variables and the experiences of financial hardship into account, the multivariate analyses end up with the expected effect from the unemployment level; i.e. with everything else equal, the high unemployment level in Finland generates less stigmatisation of the unemployed. Using the terms from our theoretical framework, the overall conclusion in Chapters 6 and 7 was that the programmatic structure of the welfare policy found in the different regimes is likely to suppress or facilitate the public perception of the cultural distinctiveness among the poor and the unemployed. Thus, still without making the extreme institutional argument, we can conclude with much more confidence that cross-national differences in the perception of the poor and unemployed are not only a matter of conditions given prior to the building of the modern welfare state, e.g. the size of countries, ethnic composition, egalitarian values etc. They are also a matter of the pursued welfare policy. At the end of this chapter, we will return to the question of ethnicity and welfare attitudes, but it is clear that Chapters 6 and 7 point at mechanisms that might prevent increased immigration in Europe from eroding public support for welfare policies (in contrast to the prediction of Alesina & Glaeser 2004). According to the analyses conducted in Chapters 6 and 7, the high public support, especially in the social democratic regimes, is rooted in more than mechanical solidarity in small and homogeneous nation states. Chapters 5, 6 and 7 mostly focused on the left-hand side of the causal model presented in Chapter 4. Therefore Chapter 8 explored the right-hand side of the causal model (see Figure 4.1) by means of a national sample. The analytical strategy was to measure the perceptions, which we had already argued were regime-dependent, and see how strongly they influenced individual support for welfare policy. We used a Danish sample and analysed support for spending on social assistance (more or less) and attitudes towards the living standard that the public should provide claimants of social assistance. The chapter showed that perceptions of job opportunities,

144

The Institutional Logic of Welfare Attitudes

generosity, and cultural distinctiveness (measured as perceived work ethic among claimants of social assistance compared with work ethic among employed) had a strong and significant impact on support for social assistance. The three items were able to explain much of the individual support, and the connections did not disappear when we took each individual’s affiliation with political parties into account. Finally, Chapter 8 investigated to what extent the strong impact of the perception of the work ethic was rooted in negative stereotypes of immigrants. Such a replication of the American findings was not unlikely as (1) the number of immigrants from third world countries and their descendants has increased substantially in Denmark over the last two decades; (2) they are highly overrepresented in the social assistance system; and (3) the political mobilisation around the immigration issue has been substantial. However, the analyses showed that (1) no connection could be found between the perceptions of the share of immigrants among claimants of social assistance and support for social assistance; (2) a connection between the perception of work ethic among immigrants receiving social assistance and support for social assistance could be found (as in the American studies), but the effect was halved by taking the general assessment of work ethic among claimants of social assistance into account; and (3) that a decrease in public support for the entitlement of immigrants to social assistance (from 1998 to 2004) and to social support during unemployment in general (from 1993 to 2004) could not be found. This led again to the overall conclusion that the programmatic structure of the pursued welfare policy has a crucial influence on public support, which needs to be taken into account by the established theories of welfare state development when we explain the recent developments in labour market policies, and when we make predictions about the future public support for welfare policy. These perspectives will we discussed in the next four sections. Towards a better micro foundation for comparative welfare state theory In the previous chapters we have criticised mainstream comparative welfare state research, i.e. the power resource theory, the welfare regime theory, and the new politics theory, for paying too little attention to the micro-foundation. In all these mainstream theories the electorate plays a role, especially as a factor that can explain path-dependency, but expectations are rarely tested. As a consequence, the theories have a rather ‘mechanical’ perception of the welfare attitudes of the electorate. For example Pierson’s argument about a new politics of the welfare state rests on the assumption that citizens’ attitudes and behaviour are guided by self-interest. That 57 per cent of the Swedish electorate belong to a so-called welfare clientele is used to prove that the social democratic welfare regime is difficult to change. EspingAndersen uses the same kind of soft rational choice arguments when he stresses the importance of the coalition between working class and middle class in the social democratic regime (and the lack of coalition in the liberal regime). If the middle class is included, a welfare backlash is less likely, and an expansion of the welfare

Deservingness and Welfare State Development

145

state more likely. The arguments seem reasonable, but they have not been subjected to much empirical research. The previous studies reviewed in Chapter 3 sometimes showed a connection between ‘concentrated’ interest in the welfare state (status as unemployed showed the clearest pattern) and welfare attitudes, but it was not always the case. Furthermore, the differences in the size of the groups with vested concentrated interest do not seem able to explain the cross-national differences. The same holds for the power resource theory, which puts more emphasis on the longterm interest of the class. Indeed, class is still an important variable, but it apparently does not help us explain cross-national differences in attitudes. If these (admittedly rather fragile) empirical findings are taken into account in the theoretical reasoning, they often lead scholars to use ‘culture’ as a residual explanation. Typically, this only replaces one mechanical explanation with an even more mechanical explanation; at least if culture is believed to be reproduced through the internalisation of norms and values in childhood and youth. And again our analyses have shown that we do not find a direct connection between cross-national differences in egalitarian values and differences in support for welfare policy. Naturally, we should take into account that we have primarily dealt with attitudes towards welfare policy in the narrow sense. If we have had comprehensive data on attitudes towards e.g. pension reform, we might have been able to find a strong effect from short-term self-interest and long-term class interest. Nevertheless, in the case of policy that improves the living conditions of poor and unemployed, it seems promising to replace the ‘economic man’ and ‘sociological man’ found behind many previous study with a ‘political man’, whose policy attitudes are influenced by the experienced regime-dependent reality. Naturally, attitudes might also be influenced by societal norms and values, self-interest, and class interest, but in order to explain the cross-national differences it seems promising to equip the electorate with a more reflective mind. This line of reasoning follows the very pronounced shift in the election literature, where the old class theories are replaced by more dynamic theories of issue voting (e.g. Merill & Grofman 1999). It also follows the emphasis modern policy analyses put on the social construction of the target group of a given public policy (e.g. Schneider & Ingram 1993) and the emphasis modern media research puts on the way different framing of political issues influences public opinion (e.g. Gamson & Modigliani 1987; Kinder & Sanders 1996; Nelson, Clawson & Oxley 1997). Moreover, the theoretical construction of a more ‘moveable’ ‘political man’ does not mean that public support for welfare policy is simply an outcome of the way welfare issues are presented by the mass media or that the previous macro theories cannot be rooted. Actually, it is an old idea in comparative welfare state literature that the different institutional arrangements influence the way citizens perceive the world. The classic thesis (which was the one Titmuss had in mind) is the link between selectivism and the tendency to speak in terms of ‘us’ and ‘them’. We have tried to advance this line of reasoning by showing how three characteristics of Esping-Andersen’s welfare regimes are likely to influence the public perceptions of poor and unemployed. This led to a number of testable theses that specified the intervening variables between

146

The Institutional Logic of Welfare Attitudes

welfare regimes and support for welfare policy. It could be a future framework for analysing cross-national differences in welfare attitudes that avoid the simple ‘is Esping-Andersen right or wrong discussion’. To put it boldly, it is unfruitful simply to include a regime variable in a regression model and see whether it is significant or not. In order to advance welfare state theories, survey studies need to explain why there is a connection, and why not. Welfare attitudes, restructuring, and new labour market policies in the 1990s ‘Does it matter’ is one of the strongest objections against attitude research, and thereby also against our call for a micro-foundation for comparative welfare state theory informed by the findings from survey studies. Do attitudes actually influence public policy? Children in Western countries are brought up to believe that the sovereign people ‘gets what it wants’, but naturally the connection between public attitudes and public policy is far more complex. The issue has not been the object of much European research, but a number of American studies have been conducted. In Burstein’s review of 20 studies, mainly quantitative and mainly focusing on the US case, he found that ‘all but one reach the same conclusion: that public opinion influences public policy’ (1998, 41). Sometimes the effect is modest, sometimes stronger and sometimes clearly more important than any other variable. Furthermore, ‘all the studies examining the impact of public opinion on social welfare policies and defence find its impact significant’ (1998, 41). However, relatively few of these studies build on longitudinal data, which makes it difficult to determine whether public attitudes influence public policy or it is the other way around. Furthermore, we do not know to what extent conclusions derived from quantitative studies are valid for the other Western countries. However, in Jacobs’ comprehensive historical study on public opinion and the making of American and British health policy (1993), the conclusions basically correspond with the quantitative studies. Convincingly he shows that the different policies pursued in the otherwise rather similar countries (so Jacob argues) can largely be explained by differences in public attitudes. It led him to conclude that ‘systematic evidence that policymakers’ decisions corresponded with public opinion suggest that elites had relatively little discretion or leeway on issues that attracted strong, sustained, and focused public support’ (1993, 219). Thus, the argument is qualified by stressing that strong sustained and focused public support is necessary in order to establish a strong connection between public attitudes and policy. In Sharp’s qualitative study (1999) on the relationship between public attitudes and six social policy areas in the US, the term ‘the sometime connection’ is used. Nevertheless, these studies generally give us good reasons to believe that public support for welfare policy does influence the pursued welfare policy. Therefore our theoretical framework might contribute substantially to the current discussion of welfare state restructuring. In Pierson’s path-breaking work from 1994 he argues that due to a situation of permanent austerity, the current politics of the welfare state

Deservingness and Welfare State Development

147

differs qualitative from the politics of the golden age. As mentioned in chapter 3, Pierson argues that ‘retrenchment is generally an exercise in blame avoidance rather than credit claiming’ (1996, 145). The subsequent literature on this ‘new politics of the welfare state’ quickly discovered that the degree of retrenchment differed remarkably even in cases with similar fiscal pressure. In order to explain these crossnational differences – in what was given the broader label ‘restructuring’ (Pierson 2001) – scholars pointed at cross-national differences in the possibilities for blame avoidance.1 Following our the line of reasoning, we would rather argue that it can actually not be taken for granted that restructuring of the welfare state is always an exercise in blame avoidance. Whether politicians will be blamed or blessed differs significantly across regimes, especially in the case of restructuring of policies that concern poor and unemployed. The extent to which politicians could expect to be blamed or blessed for making ‘tougher’ policies towards poor and unemployed might help us explain why the Western countries have responded so differently to the experience of unemployment. After the unemployment experience of the 1970s and 1980s, the elite diagnosis of the nature of the labour market problems was actually quite similar across countries. OECD’s highly profiled job study from 1994 concludes that it was ‘an inability of OECD economies and societies to adapt rapidly and innovatively to a world of rapid structural change that is the principle cause of high and persistent unemployment’ (OECD 1994a, vii). Boldly speaking, the extraordinarily high unemployment, especially in Europe, was perceived to be caused by structural circumstances and not simply by lack of demand in times of recession.2 Therefore labour market policy turned into ‘high politics’ as reforms that could create structural improvement were seen as the only solution to the high level of unemployment in almost all Western countries 1 One of the major explanations given for harsh retrenchment in UK and USA (studied by Pierson in 1994) and modest retrenchment in the social democratic and conservative regimes was the differences in political and especially electoral systems. With reference to Lijphart (1984), scholars argue that the ‘majoritarian’ (‘Westminister’) model characterised by first-past-the-post voting, a two-party system, bare majority voting etc. makes it easier to make unpopular decisions. In contrast it should be more difficult to make unpopular decisions in the European ‘consensus’ models based on proportional representations, multi-party systems, coalition government etc., which lead to a fragmentation of power. As the social democratic welfare regimes have some of the ‘purest’ systems of proportional representation and strong labour unions – and a tradition for negotiating with the unions – this could be an explanation of a politic of cost containment (Pierson 2001). 2 The explanations about the causes of this structural unemployment can roughly be divided into external and internal. The external explanations focus on fundamental changes in the production structure caused by economic globalisation and technology, which leads to mismatch problems on the labour markets. The internal explanations focus on inflexible labour markets caused by corporatism and highly developed European welfare states (OECD 1994b, 51, 211, 265). This inflexibility turns into a huge problem when meeting the challenge of globalisation, and the old policy responses are no longer effective. The point is that due to mismatch problems on the European labour markets, traditional Keynesian demand side policy would only generate inflation.

148

The Institutional Logic of Welfare Attitudes

in the mid-1990s. In OECD’s so-called job strategy, the single most important piece of advice was to lower welfare benefits, especially unemployment benefits. In the short run, it would increase the incentives for unemployed to search for jobs, and in the long run it would decrease wage levels, which could prevent outsourcing to low-wage countries. Another piece of advice was to initiate an active labour market policy, which by means of activation measures, tighter eligibility criteria and more control could create some of the same structural improvements. Thus, policy makers spoke about a general shift from ‘passive’ to ‘active’ and academic scholars spoke about a general shift from ‘welfare’ to ‘workfare’ or from ‘decommodification’ to ‘recommodification’ or simply a process of retrenchment. However, it soon became evident that the reforms, which all were given the ‘active label’, came in many different varieties. Studying workfare reforms in France, Germany, the Netherlands, Norway, Denmark, UK, and USA, Lødemel & Trickey (2000) describe the different strategies on a continuum from ‘human resource development’ to ‘labour market attachment’. USA was the prime example of a labour market attachment strategy, which ‘promoted quick job placement, even at low wages, in the expectation that ‘work first’ is the best way to build (or forestall loss of) work habits and skills’ (2000, 227). One of the concrete policy outputs was the time limit (5 years over a life time) put on the successor of the AFDC (Aid to Families with Dependent Children). Denmark was the country that came closest to the ideal type of human resource development (2000, 297). Naturally, it is difficult to make this kind of judgment, and Lødemel & Trickey’s ranking is pretty inductive. It is striking, however, that the harshest reform of social assistance schemes took place in the country where the public has the most negative perception of poor and unemployed. The softest reforms took place in the countries where the public has the most positive perception. At the same time, the explanation of differences in electoral systems has a number of shortcomings.3 The point is that cross-national differences in recent labour market reforms might better be explained by differences in public opinion than by differences in political systems. The influence of public opinion is most observable in situations where governments pursue a policy that contradicts the ideological positions of the parties included in the government. In the US case, it is well known that the highly negative public opinion towards the welfare schemes (targeted towards poor and unemployed) forced the Democrats to pursue a harsh policy (Gilens 2000). Actually the public attention was 3 First of all, concentrated power also means concentrated accountability, i.e. the electorate knows who to blame (Pierson 1994), which should make it more difficult to make unpopular decisions. As Bonoli, George & Taylor-Gooby admit, ‘the formal institutions, however, seldom affect the course of policy per se’ (2000, 148). Secondly, it is clear that the political system in USA with two chambers and a federal structure creates many more vetopoints (Immergut 1992; Tsebelis 2002) than the system with one chamber and a centralised unitary government found in the social democratic welfare regimes. A detailed study of Danish welfare reforms in the last two decades actually reveals a political system with high decision-making capability and deliberate use of compromises and packages to conceal whom the electorate should blame (Albrekt Larsen & Goul Andersen 2004).

Deservingness and Welfare State Development

149

so strong in 1990s that Clinton made ‘ending welfare as we know it’ one of his major campaign issues in 1992. The reform package – which was implemented in a radically different version in 1996 – included both ‘human resource’ elements and tough-minded elements appealing to the conservatives. In 1996, the ‘Personal Responsibility and Work Opportunity Reconciliation Act’ signed by Clinton totally dismantled the federal AFDC program – under the outcry of many liberal party members – and replaced it with a block grant arrangement that empowered the states to constitute their own welfare programmes (Sharp 1999, 163-209; Wiseman 2001). Thus, in the US case, a harsh neo-liberal labour market policy was actually more a matter of credit-claming than one of blame avoidance. At the other end of Lødemel & Trickey’s continuum, Denmark is actually a good place to observe the impact of public opinion in social democratic regimes. Since the end of 1980, the policy elites have been convinced that Denmark was troubled by high structural unemployment, among other things caused by the generous welfare schemes (Albrekt Larsen 2002; Goul Andersen 2002ab; Albrekt Larsen & Goul Andersen 2004), and bourgeois parties have been in office more than in the other Nordic countries (from 1982 to 1993 and again from 2001). The Danish labour market policy did undergo a number of major reforms in the period from 1993 to 2003, but neither the bourgeois governments from 1982 to 1993, nor from 2001 and onwards dared to suggest policies that fundamentally reduced the generosity of the (comparatively) very generous schemes. Prior to the 2001 election, the bourgeois actually guaranteed that they would not make any cuts in unemployment benefits and social assistance. This was part of the general strategy of ‘not being too visibly against the social democratic welfare state’, which to a large extent explains the bourgeois victories in 2001 and again in 2005. The other social democratic regimes had more success avoiding the unemployment of the post-industrial phase, but in the beginning of the 1990s, both Finland and Sweden experienced a rather extreme fiscal crisis. This led to modest reductions in compensation rates (in Sweden in 1993 and 1996, in Finland in 1992 and 1994), but they were more guided by necessary cost containment (Pierson 2001, 423) than principle confrontation with previous policy. In electoral terms, this was a wise strategy as survey research reveals that the public in the social democratic regimes remained hostile towards these reductions (Albrekt Larsen 2004). Thus, in contrast to USA, the politicians’ cuts in unemployment benefits and social assistance in social democratic regimes have clearly been a matter of blame avoidance. Hence, in the cases of ‘extreme’ public opinions about poor and unemployed, we have evidence that seems to confirm the influence of regimedependent attitudes on welfare policy. Path-dependency and public support for welfare policy The beauty of Esping-Andersen’s The three worlds of welfare capitalism (1990) is that it is not only an attempt to cluster the Western countries into three different brackets, but also an attempt to understand the transition process from an industrial

150

The Institutional Logic of Welfare Attitudes

golden age to a post-industrial era of restructuring. The technological development in the industrial production, the increased economic integration, and the Baumolleffect are mentioned as factors that initiate such a transition process, but the major contribution of Esping-Andersen (and other welfare state scholars following this lead) was to show that each of the three welfare regimes responded differently to a similar pressure. As mentioned, the exact mechanisms behind these different responses or path-dependent trajectories remain somewhat under-specified. Above we argued that public opinion probably does matters, and below we will discuss the findings in relation to the current discussion of path-dependency. First of all, the connection between welfare regimes and perceptions of poor and unemployed could suggest a system of reproduction. The line of reasoning would be that the liberal welfare regimes ‘reproduce’ negative perceptions of poor and unemployed, which ‘reproduce’ tight-fisted targeted welfare schemes and unregulated labour markets that uphold the cultural distinctiveness of poor and unemployed. Thereby liberal regimes have also reproduced the factors that make it difficult for unemployed and poor to fulfil the five deservingness criteria that for some reason seem to be rather stable both over time and place. Following the same line of reasoning, the social democratic regimes tend to reproduce the positive perceptions of poor and unemployed, which in turn ‘reproduce’ generous universal welfare schemes and regulated labour market that effectively repress potential cultural divisions between the majority and groups in risk of poverty. Following this logic, the conservative regimes reproduce ‘medium’ perceptions of poor and unemployed, which lead to policies towards poor and unemployed that in terms of generosity and targeting are somewhere in-between. Thus, in general the prediction would be that the cross-national differences in support for welfare policy is stable over time, as illustrated in the left-hand side of figure 9.1. However, the idea of reproduction carries an inappropriate deterministic understanding of society. The notion of path-dependency is somewhat better as it suggests that societies constantly develop (the path goes from one point to another), and it is more explicit about the many other variables outside our models (dependence is much more modest than reproduction). Following this logic, one could argue that the findings of this book might help uncover some of the pieces in the big puzzle of the regime-dependent transition process. As mentioned in chapter 2, EspingAndersen argued in 1996 that the welfare regimes tended to follow a ‘Scandinavian route’, a ‘neo-liberal route’ and a ‘labour reduction route’ in this transition process. Thus, instead of a static reproduction, we have a dynamic development, which could lead to the prediction that meeting the challenges of the post-industrial era the future support for welfare policy would decrease in the liberal regimes and increase in the social democratic regimes. In the former case, the argument would be that the neoliberal labour market trajectory tends to produce more and more negative perceptions of poor and unemployed, which leads to more and more targeted, tight-fisted policies and more and more unregulated labour markets. This again leads to a deeper cultural gap between the majorities and poor and unemployed, which altogether make it more and more difficult for the latter to fulfil the five deservingness criteria. Such

Deservingness and Welfare State Development

Figure 9.1

151

Illustration of reproduction of public support for welfare policy (left-hand side) and dynamic path-dependent public support for welfare policy (right-hand side)

a prediction of erosion of public support in the liberal regimes is illustrated in the right-hand side of figure 9.1. The argument goes nicely with the finding that social trust has eroded substantially in all the liberal regimes over the last decade (Albrekt Larsen 2006), and the idea was heavily discussed during the Thatcher and Reagan years. However, the dominant interpretation is that rather than eroding public support, the retrenchment in UK and USA increased support for the welfare state (Pierson 1994, 146-49; Mau 2003, 21). However, one has to distinguish between support for welfare policy in the broad and in the narrow sense, see figure 9.2. In general we do not have good time series of public support for welfare policy, but in the British case we can actually trace the proportion of voters that would prefer to ‘increase taxes and spend more on health, education and social benefits’ all the way back to 1983. Using this item as a measure for welfare policy in the broad sense, we can indeed observe an increase in public support for spending on welfare. In 1983, 32 per cent were in favour of spending more, which by 1999 had increased to 65 per cent, i.e. from constituting a minority those in favour of spending came to constitute a solid majority. From 1991 and onwards, we see a stability around 60 per cent except the drop in the 2000 survey. However, if we look at the share agreeing that ‘the government should spend more money on welfare benefits for the poor, even if it leads to higher taxes’, figure 9.2 shows a decrease from the late 1980s and onwards. In 1989, 61 per cent supported spending on the poor, compared to only 37 per cent in 2001. Naturally this is by no means bullet-proof evidence, but the best time series we have from the liberal regimes actually support the idea that the neoliberal labour market trajectory potentially makes it more and more difficult for poor and unemployed to fulfil the deservingness criteria.

152

Figure 9.2

The Institutional Logic of Welfare Attitudes

Public support for welfare policy in Britain from 1983 to 2002. Share in favour of increased spending on health, education, and social benefits (solid line) and share in favour of increased spending on welfare benefits for the poor (dotted line). Per cent

The other way around, one could predict that the ‘Scandinavian route’ of public sector employment and active labour market policy creates more and more positive perceptions of poor and unemployed, which leads to higher public support for welfare policy, see again right-hand side of figure 9.1. Such an argument goes nicely with those who argued that the intensified labour market policies in the Nordic countries during the 1990s is not a matter of workfare, but a matter of creating even better conditions for poor and unemployed (e.g. Jensen 1999), and the finding that social trust has become exceptionally high in the social democratic regimes (Albrekt Larsen 2006; Kumlin & Rothstein 2005). From the social democratic regimes, we cannot produce a time series that matches the one based on the British Social Attitude data. In Sweden, we are able to trace attitudes on a broad and narrow item back to 1986. As in Britain, figure 9.3 indicates an increase in support for the general welfare policy. In the period from 1986 and onwards, the proportion of voters that found reduction of the public sector to be a bad proposal increased from 35 per cent in 1986 to 52 per cent in 1996 and declined to 43 per cent in 1999 (solid line). The lowest share against the proposal was during the economic crisis in the beginning of the 1990s. In 1990, only 20 per cent found a reduction of the public sector to be a bad idea. Turning to welfare policy in the narrow sense, we only have tree observations. In 1986, 68 per cent were in favour of using the same amount of money or more money on social assistance. In 1992, the share was 65 per cent, and in 2002 the share was 70 per cent Thus, in contrast to Britain, these fragile Swedish data do not indicate any erosion in public support for welfare policy towards the

Deservingness and Welfare State Development

Figure 9.3

153

Public support for welfare policy in Sweden from 1986 to 2002. Share that finds reduction in public sector a bad proposal (solid line) and share in favour of maintaining or increasing public spending on social assistance (dotted line). Per cent

poor and unemployed in Sweden. Reproduction seems to be the best interpretation, i.e. as expected poor and unemployed in the social democratic regimes have not experienced larger difficulties in fulfilling the deservingness criteria, but at the same it has probably not become easier. The same is more or less the case in Denmark, where time series on public support the welfare state are also available. The proportion of voters who agreed most with a statement about maintaining social reforms at least at the present level (in contrast to the counter-argument that states that ‘social reforms has gone to far…’, see wording) has been more or less stable since 1979 and onwards, see figure 9.4. In 1979, 56 per cent agree, which increased to above 60 per cent in 1984, 1990, 1994, and 1998. From 1998 to 2001, we see a drop from 64 per cent to 59 per cent The proportion that wants to maintain or increase spending on social assistance has also been more or less stable in the covered period. In 1990, the percentage was 56, which increased to 64 per cent in 1994, 62 per cent in 1998, and again to 64 per cent in 2001. Thus, rather than erosion of support for the poor and unemployed, we actually see a slight increase in this social democratic regime. Finally, we have included the share would disagree in the statement that ‘many receive social benefits without needing them’ in figure 9.4. In 1973, only 7 per cent of the voters disagreed with this statement (naturally the persuasion effect should be taken into account), but in 1981 the proportion had increased to 38 per cent; ironically the bourgeois came to power the year after due to the severe macro-economic problems. After 1981, the figures show

154

Figure 9.4

The Institutional Logic of Welfare Attitudes

Public support for welfare policy in Denmark from 1973 to 2001. Proportion of voters that 1) want to maintain or increase spending on social assistance (dotted line), 2) want to maintain social reforms at least at present level, and 3) who disagree that too many receive social benefits without needing them

a decline to 1987 and thereafter stability. Thus, again we are able to conclude that in the social democratic regimes, it has not become more difficult for claimants of welfare benefits to fulfil the deservingness criteria. I the Danish case, we actually see a tendency in the opposite direction. The dynamics of the conservative regimes are more difficult to predict using this line of reasoning. Admittedly, the line of reasoning is also a bit speculative, but it illustrates how our theoretical framework might help to narrow down some of the mechanisms behind the path-dependent labour market trajectories. Nevertheless, in order to predict future support for welfare policy, it might be more relevant to discuss the impact foreseeable trends could have on the assessment of deservingness. Fiscal limitations, targeting, immigration, and future support for the welfare state In this final section, we will point at three concrete trends in most Western welfare states that are likely to influence future public support for welfare policy. Firstly, the fiscal limitation of the welfare state (or at least the perceived fiscal limitation) is likely to become a permanent condition for welfare state development in the foreseeable future. On the one hand, expenditures are likely to rise due to the ageing

Deservingness and Welfare State Development

155

of Western populations, higher expectations to health services etc., and on the other hand, the possibilities to collect higher taxes are (at least perceived to be) modest. This situation creates a public awareness of the limited economic resources and the limited capacity of the state, which probably weaken the connection between fulfilment of deservingness criteria and support for public policy. The early AngloSaxon survey studies in the 1970s reported a strong connection between perception of causes of poverty and support for welfare policy. In chapter 5 we also found a strong connection between the cross-national differences in perceptions of causes of poverty and cross-national differences in support for welfare policy measured by the ISSP social inequality module in 1992. However, our analyses of the Eurobarometer survey from 2001 in appendix 5.2 suggest a less ‘automatic’ connection between deservingness and support for public welfare policy. It seems reasonable to except that ‘political men’ of the future will have both the degree of deservingness and the limited capacity of the state in mind when they form their preferences for welfare policy. The fiscal limitations could also lead to reductions in unemployment benefits and social assistance, which following chapter 7 (ceteris paribus) increase the cultural distinctiveness of poor and unemployed. This would then make it more difficult to fulfil the identity criterion. Such an identity effect is naturally most likely if we deal with severe cuts in benefits that create a new ‘underclass’, cf. the discussion on the impact of the Reagan and Thatcher era. However, also relative reduction (in many countries public benefits no longer follow wages) could potentially create a larger divide between ‘us’ and ‘them’ in the long run. Secondly, what might be more severe for future public support is the tendency to target public benefits at ‘those’ in need, which has been reported in recent studies (e.g. Goodin & Rein 2001, Bendix Jensen 2004). As discussed in chapter 4, selective welfare policies have a tendency to open discussions that question the deservingness of claimants in a number of ways. More specifically, targeting raises the questions, 1) are the claimants really needy, 2) are they to be blamed for their neediness, 3) do they receive the public support with the right attitude. Targeting also highlights the boundaries between those who give and those who receive, and often define the latter as a special group distinguished from the well-adjusted majority. Therefore it becomes harder for recipients to fulfil the deservingness criteria, which then tends to erode future public support. Thus, what in the short run seems to be an effective and socially acceptable way to save money could in the long run erode future public support for welfare policy. In that sense, our analyses have delivered an argument that helps to explain the old saying in comparative welfare studies; namely that ‘welfare for poor becomes poor welfare’. Finally, the increased immigration poses a threat to public support for welfare policy in the future; especially in Europe as USA, Canada, New Zealand and Australia have centuries of experience as cultural melting pots. The recent increases in the stock of immigrants in European countries combined with high unemployment rates and discrimination on the European labour markets could potentially generate the perception of a culturally distinct underclass, which is believed to play such a prominent role for support for welfare policy in the US. Whether the welfare states

156

The Institutional Logic of Welfare Attitudes

suppress or inflame these tensions is up for discussion. Economists often argue that the welfare state is an obstacle for the integration of immigrants into the labour markets as it limits job opportunities (by securing a high minimum wage) and limits the economic incentives of the immigrants. The bulk of labour market research conducted since the mid-1990s has shown that the impact of economic incentives on job search and actual integration into the labour market is modest; especially in the case of job search and job opportunities among immigrants. Therefore, we would rather argue that a comprehensive welfare state is needed more than ever. The integrative capability of the welfare state had a dominant place in the classic writings of Titmuss and Marshall. Marshall basically saw the welfare state as a way to overcome the class divisions imposed by capitalism and thereby to secure a stable democracy (1950). The next task of the welfare state could be to overcome the internal divisions between ‘them’ and ‘us’, which seems to be one of the major challenges globalisation imposes on the developed Western welfare states.

References

Aalberg, T. (2003), Achieving Justice: Comparative Public Opinion on Income Distribution (Leiden: Brill). Aalport, G. (1954/1979), The Nature of Prejudice (New York: Van NostrandReinhold). Albrekt Larsen, C. (2002), ‘Policy Paradigms and Cross-national Policy (mis)Learning from the Danish Employment Miracle’, European Journal of Public Policy 9: 5, 677–94. Albrekt Larsen, C. (2004), Blessing or Blame? Public Attitude to Nordic ‘Workfare’ in the 1990 (Aalborg:. CCWS working paper, 33). Albrekt Larsen, C. (2006), Welfare Regimes and Social Capital. The Impact of Living Conditions and Perceptions of Poor and Unemployed (Aalborg: CCWS working paper 38). Albrekt Larsen, C. & Goul Andersen, J. (2004), Magten på Borgen. En analyse af beslutningsprocesser ved større politiske reformer (Aarhus: Aarhus Universitetsforlag). Albæk, E., Eliason, L., Nørgaard, A. S. & Schwartz, H. (eds.) (forthcoming), Crisis, Miracles and Beyond: Negotiated Adaptation of the Danish Welfare State (Aarhus: Aarhus University Press). Alesina, A. et al. (2003), ‘Fractionalization’, Journal of Economic Growth 8: 2, 155–94. Alesina, A. & Glaeser, E. L. (2004), Fighting Poverty in the US and Europe. A World of Difference (Oxford: Oxford University Press). Alston, J. & Dean, K.I. (1972), ‘Socioeconomic factors associated with attitudes toward welfare recipients and the causes of poverty,’ Social Services Review 46, 13–23. Andress, H.-J. & Heien, T. (2001), ‘Four Worlds of Welfare State Attitudes? A Comparison of Germany, Norway, and the United States,’ European Sociological Review 17:4, 337–56. Anckar, D., Damgaard, E. & Valen, H. (eds) (1982), Partier, ideologier, väljere (Aabo: Aabo Akademy Press). Arts, W. & Gelissen, J. (2001), ‘Welfare States, Solidarity and Justice Principles: Does the Type Really Matter?’ Acta Sociologica 44: 4, 283–99. Baldwin, P. (1990), The Politics of Social Solidarity (Cambridge: Cambridge University Press). Bay, A. and Pedersen, A.W. (2003), The Limits of Social Solidarity: Popular Attitudes Towards Basic Income in a Nordic Welfare State. Paper presented at the annual

158

The Institutional Logic of Welfare Attitudes

ESPAnet conference, Copenhagen, 13–15 November. Bean, C. & Papadakis, E. (1998), ‘A Comparison of Mass Attitudes towards the Welfare State in different Institutional Regimes, 1985-1990’, International Journal of Public Opinion Research 10: 3, 211–36. Bendix Jensen, J. (2004), Targeting social security: Who gets what, why? An analysis of measures that target Danish and British unemployment benefits, social assistance, retirement pension and family allowance, 1990-2002. PhD Thesis, Department of Economics, Politics and Public Administration, Aalborg University. Blekesaune, M. & Quadagno J. (2003), ’Public Attitudes to Welfare State Policies: A comparative Analysis of 24 Nations’, European Sociological Review 19: 5, 415–27. Bommes, M. & Geddes, A. (eds) (2000), Immigration and Welfare: Challenging the Borders of the Welfare State (London: Routledge). Bonoli, G. (2000), ‘Public attitudes to social protection and political economy traditions in Western Europe’, European Societies 2: 4, 431–52. Bonoli, G., George V. & Taylor-Gooby, P. (2000), European Welfare Futures. Towards a Theory of Retrenchment (Cambridge: Polity Press). Borre, O. & Scarbourgh, E. (eds) (1995), The Scope of Government (Oxford: Oxford University Press). Burstein, P. (1998), ‘Bringing the Public Back In: Should Sociologists Consider the Impact of Public Opinion on Public Policy?’, Social Forces 77: 1, 27–62. Cameron, D. (1978), ‘The Expansion of the Public Economy: A Comparative Analysis’, American Political Science Review 72: 4, 1243–61. Castles, F. G. (2002), ‘Developing New Measures of Welfare State Change and Reform’, European Journal of Political Research 41: 5, 613–41. Castles, F. G. (1996), ‘Needs-based Strategies of Social Protection in Australia and New Zealand’, in Esping-Andersen (ed.). Castles, F. G. (ed.) (1993), Families of Nations (Dartmouth: Aldershot). Castles, F. G. & Mitchell D. (1993), ‘Worlds of Welfare and Families of Nations’, in: F.G. Castles (ed.) Families of Nations (Dartmouth: Aldershot). Checcchi, D., Ichino, A. & Rustichini, A. (1999), ‘More Equal and Less Mobile? Education, Financing and Intergenerational Mobility in Italy and in the US’, Journal of Public Economics 74: 3, 351–93. Christensen, L. R. (1987), Hver vore veje: Livsformer, familietyper og kvindeliv (Odense: Etnologisk forum). Cook, F. (1979), Who Should be Helped: Public Support for Social Services (Beverly Hills, CA: Sage Publications). Coughlin, R. M. (1980), Ideology, Public Opinion and Welfare Policy: Attitudes towards Taxes and Spending in Industrial Societies (Institute of international Studies, Research Series No. 42, Berkeley, CA: University of California). Crocker, J. & Major, B. (1989), ‘Social Stigma and Self-esteem: The Self-protective Properties of Stigma’, Psychological Review, 96: 4, 608–30. Danish Statistics (2000), Statistisk tiårsoversigt 2000 (Copenhagen: Danish

References

159

Statistics). Danish Statistics (2003), Statistisk tiårsoversigt 2003 (Copenhagen: Danish Statistics). Dean, H. & Taylor-Gooby, P (1992), Dependency Culture: The Explosion of a Myth (New York: Harvester Wheatsheaf). DeMaio, T. J. (1984), ‘Social Desirability and Survey Measurement: A Review’, in: C.F. Turner & E. Martyin (eds) Surveying Subjective Phenomena (New York: Russell Sage Foundation). Department of Work and Pensions (2003), Income-Related Benefits: Estimates of Take-up in 2000/2001 (London: Department for Work and Pensions: Information and Analysis Directorate). De Swaan, A. (1988), In Care of the State: Health Care, Education and Welfare in Europe and the USA in the Modern Era (Oxford: Oxford University Press). Ditch, J. (1984), ‘The Perception of Poverty in Northern Ireland’, Policy and Politics 12: 2, 167–81. Elster, J. & Hylland, A. (eds) (1986), Foundations of Social Choice Theory (Cambridge, MA: Cambridge University Press). Esping-Andersen, G. (1980), Social Class, Social Democracy and State Policy. Party Policy and Party Decomposition in Denmark and Sweden (Copenhagen: Nyt fra Samfundsvidenskaberne). Esping-Andersen, G. (1990), The Three Worlds of Welfare Capitalism (Cambridge: Polity Press). Esping-Andersen, G. (ed.) (1996a), Welfare States in Transition. National Adaptions in Global Economies (London: Sage Publications). Esping-Andersen, G. (1996b), ‘After the Golden age? Welfare State Dilemmas in a Global Economcy’, in: G. Esping-Andersen (ed.) Welfare States in Transition. National Adaptions in Global Economies (London: Sage Publications). Esping-Andersen, G. (1999), Social Foundations of Postindustrial Economies (Oxford: Oxford University Press). Esping-Andersen, G. (2000), Social Foundations of Postindustrial Economies, paperback (Oxford: Oxford University Press). Evans, G. (1995), ‘Cross-national Differences in Support for Welfare and Redistribution: an Evaluation of Competing Theories’, in O.E. Kangas et al. (eds) In the Eye of the Beholder. Opinions on Welfare and Justice in Comparative Perspective (Umå: The Bank of Sweden Tercentenary Foundation). Feagin, J. (1972), ‘Poverty we Still Believe that God Helps who Help Themselves’, Psychology Today, 6, 101–29. Feather, N. (1974), ‘Explanation of Poverty in Australian and American Samples: the Person, Society and Fate’, Australian Journal of Psychology 26, 199–216. Ferge, Z. & Kolberg, J. (eds) (1992), Social Policy in a Changing Europe (Frankurt am Main: Campus Verlag). Ferrarini, T. & Nelson, K. (2002), The Impact of Taxation on the Equalising Effect of Social Insurance to Income Inequality: A Comparative Analysis of Ten Welfare

160

The Institutional Logic of Welfare Attitudes

States (Luxembourg Income Study Working Paper No. 327). Ferrera, M. (1996), ‘The ‘southern’ model of welfare in social Europe’, Journal of European Social Policy 6: 1, 17–37. Ferrera, M. & Rhodes, M. (2000), ‘Recasting European Welfare States: An Introduction’, West European Politics, 23: 2, 1–10. Förster, M. & Pearson, M. (2002), Income Distribution and Poverty in the OECD Area: Trends and Driving Forces (OECD Economic Studies 34, 2002/I, 7 –39). Furnham, A. (1982), ‘Why are the poor with us? Explanations for poverty in Britain’, Journal of Adolescence 5: 2, 135–47. Gallie, D. & Paugam, S. (eds) (2000a), Welfare Regimes and the Experience of Unemployment in Europe (Oxford: Oxford University Press). Gallie, D. & Paugam, S. (2000b), ‘The Experience of Unemployment in Europe: The Debate’, in: D. Gallie & S. Paugam (eds), Welfare Regimes and the Experience of Unemployment in Europe (Oxford: Oxford University Press). Gallie, D. & Paugam, S. (2002), Social Precarity and Social Integration. Report for the European Commission Based on Eurobarometer 56.1 (Brussels: The European Commission). Gallie, D., Jacobs, S. & Paugam, S. (2000), ‘Poverty and Financial Hardship among the unemployed’, in: D. Gallie & S. Paugam (eds) Welfare Regimes and the Experience of Unemployment in Europe (Oxford: Oxford University Press). Gamson, W.A. & Modigliani, A. (1987) ‘The Changing Culture of Affirmative Action’, Research in Political Sociology 3, 137–77. Gelissen, J. (2000), ‘Popular support for institutionalised solidarity: a comparison between European welfare states’, International Journal of Social Welfare 9: 4, 285–300. Gelissen, J. (2002), World of Welfare, Worlds of Consent? (Leiden: Brill). Gilens, M. (2000), Why Americans hate Welfare. Race, Media, and the Politics of Antipoverty Policy (Chicago: The University of Chicago Press). Goffman, E. (1963), Stigma. Notes on the Management of Spoiled Identity (Harmondsworth: Penguin). Golding, P. & Middleton, S. (1981), Images of Welfare (Oxford: Blackwell). Goodin, R. E. (1986), ‘Laundering preferences’, in: J. Elster & A. Hylland (eds) Foundations of Social Choice Theory (Cambridge, MA: Cambridge University Press). Goodin, R. E. & Rein, M. (2001), ‘Regimes on Pillars: Alternative Welfare State Logics and Dynamics’, Public Administration 79: 4, 769–801. Gottschalk, P. & Spolaore, E. (2002), ‘On the Evaluation of Economic Mobility’, Review of Economic Studies 69: 1, 191–208. Goul Andersen, J. (1982), ’Den folkelige tilslutning til socialpolitikken – en krise for velfærdsstaten?’ in: D. Anckar, E. Damgaard & H. Valen (eds) Partier, ideologier, väljere (Aabo: Aabo Akademy Press). Goul Andersen, J. (1998), ’Velfærdens veje i komparativt perspektiv’, Den jyske historiker 82, 114–38. Goul Andersen, J. (1999), ’Den universelle velfærdsstat er under pres – men hvad er

References

161

universalisme?’, GRUS 56, 40–62. Goul Andersen, J. (2002a), ‘Coping with Long-Term Unemployment: Economic Security, Labour Market Integration and Well-being. Results from a Danish Panel Study, 1994–1999’, International Journal of Social Welfare 11: 2, 178–90. Goul Andersen, J. (2002b), ‘Work and Citizenship: Unemployment and Unemployment Policies in Denmark, 1980–2000’ in: J. Goul Andersen & P.H. Jensen (eds) Changing Labour Markets, Welfare Policies and Citizenship (Bristol: Policy Press). Goul Andersen, J. (2003), ’Vælgernes nye politiske dagsorden’, in: J. Goul Andersen & O. Borre (eds) Politisk forandring. Værdipolitik og nye skillelinjer ved folketingsvalget 2001 (Aarhus: Systime Academic). Goul Andersen, J. (forthcoming), ‘Public Support for the Danish Welfare State. Interests and Values, Institutions and Performance’, in: E. Albæk, L. Eliason, A.S. Nørgaard & H. Schwartz (eds) Crisis, Miracles and Beyond: Negotiated Adaptation of the Danish Welfare State (Aarhus: Aarhus University Press). Goul Andersen, J. & Albrekt Larsen, C. (2002), Pension Politics and Policy in Denmark and Sweden: Path Dependency, Policy Style, and Policy Outcome (Aalborg: CCWS working paper nr. 27). Goul Andersen, J. & Borre, O. (eds) (2003), Politisk forandring. Værdipolitik og nye skillelinjer ved folketingsvalget 2001 (Aarhus: Systime Academic). Goul Andersen, J. & Jensen, P.H. (eds) (2002), Changing Labour Markets, Welfare Policies and Citizenship (Bristol: Policy Press). Graubard, S.R. (ed.) (1986), Norden: the Passion for Equality (Oslo: Norwegian University Press). Halvorsen, K. (1999), Arbeitløshet som sosialt problem (Oslo: HiO-rapport nr. 13). Hancock, R. et al. (2003), The Take-up of Multiple Means-tested Benefits by British Pensioners. Evidence from the Family Resources Survey. (Discussion Papers in Economics, Department of Economics, University of Leicester). Hauser, R. et al. (2000), ‘Unemployment and Poverty: Change over Time’, in D. Gallie & S. Paugam (eds) Welfare Regimes and the Experience of Unemployment in Europe (Oxford: Oxford University Press). Heien, T. & Hofäcker, D. (1999), How Do Welfare Regimes Influence Attitudes? A Comparison of Fife European Countries and the United States 1985-1996. (Working Paper No. 9, Bielefeld). Hinrichs, K. (2000), Elephants on the Move: Patterns of Public Pension Reform in OECD Countries (Bremen: Zentrum für Sozialpolitik, Universität Bremen). Hochschild, J.L. (1981), What’s Fair? American Beliefs about Distributive Justice. (Harvard: Harvard University Press). Homans, G.C. (1974), Social Behaviour: Its Elementary Forms (New York: Harcourt Brace). Hsieh, C.-C. & Pugh, M.D. (1993), ‘Poverty, Income Equality, and Violent Crime: A Meta-analysis of Recent Aggregate Data Studies’, Criminal Justice Review 18: 2, 182–202. Immergut, E. (1992), Health Politics: Interests and Institutions in Western Europe

162

The Institutional Logic of Welfare Attitudes

(Cambridge: Cambridge University Press). Inglehart, R. (1997), Modernization and Postmodernization: Cultural, Economic, and Political Change in 43 Societies (Princeton: Princeton University Press). Jacobs, L.R. (1993), The Health of Nations: Public Opinion and the Making of American and British Health Policy (Cornell: Cornell University Press). Jensen, P.H. (1999), Activation of the Unemployed in Denmark since the early 1990s. Welfare or Workfare? (Aalborg: CCWS working paper nr. 7). Jordan, B. (1996), A Theory of Poverty and Social Exclusion (Cambridge: Polity Press). Jæger, M.M. (forthcoming), ‘Welfare Regimes and Attitudes Towards Redistribution: The Regime Hypothesis Revisited’, European Sociological Review 22: 2. Kangas, O.E. (1997), ‘Self-Interest and the Common Good: The Impact of Norms, Selfishness and Context in Social Policy Opinions’, Journal of Socio-Economics 26: 5, 475–94. Kangas, O.E. et al. (eds) (1995) In the Eye of the Beholder. Opinions on Welfare and Justice in Comparative Perspective (Umå: The Bank of Sweden Tercentenary Foundation). Kelley, J. & Evans, M.D.R. (1993), ‘The Legitimation of Inequality: Occupational Earnings in Nine Nations’, The American Journal of Sociology 99: 1, 75–125. Kersbergen, K. van (1995), Social Capitalism. A Study of Christian Democracy and the Welfare State (London: Routledge). Kersbergen, K. van (2000), ‘The Declining Resistance of Welfare States to Change?’ in: Kuhnle, S. (ed.) Survival of the European Welfare State (London: Routledge). Kim, M. & Mergoupis, T. (1997), ‘The Working Poor and Welfare Reipiency: Participation, Evidence, and Policy Directions’, Journal of Economic Issues 31: 3, 707–28. Kinder, D.R. & Sanders, L.M. (1996), Divided by Color: Racial Politics and Democratic Ideals (Chicago: University of Chicago Press). Kluegel, J. et al. (1995), Social Justice and Political Change: Public Opinion in Capitalist and Post-communist States (Hawthorne: Aldine de Gruyter). Kluegel, J. & Smith, E. (1986), Beliefs about Inequality: Americans’ View of What is and What Ought to Be (New York: Aldine De Gruyter). Kluegel, J. & Miyano, M. (1995), ‘Justice Beliefs and Support for Welfare State in Advanced. Capitalism’, in: Kluegel, J. et al. (1995), Social Justice and Political Change: Public Opinion in Capitalist and Post-communist States (Hawthorne: Aldine de Gruyter). Korpi, W. (1983), The Democratic Class Struggle (London: Routledge and Kegan Paul). Korpi, W. & Palme, J. (1998), ‘The Paradox of Redistribution and Strategies of Equality: Welfare State Institutions, Inequality, and Poverty in the Western Countries’, American Sociological Review 63: 5, 661–87. Kristensen, O.P. (1980), ‘The Logic of Political–Bureaucratic Decision-making as a Cause of Governmental Growth: Or why Expansion of Public Programs is a ”Private Good” and their Restrictions is a ”Public Good” ’, European Journal of

References

163

Political Research 8, 249–64. Kuhnle, S. (ed.) (2000), Survival of the European Welfare State (London: Routledge). Kumlin, S. & Rothstein, B. (2005), ‘Making and Breaking Social Capital. The Impact of Welfare-state Institutions’, Comparative Political Studies 38: 4, 339–65. Lee, B.A. & Jones, S.H. et al. (1990), ‘Public Beliefs about the Causes of Homelessness’, Social Forces 69:1, 253–65. Lee, B.A. & Lewis, D.W. et al. (1992), ‘Are the Homeless to Blame? A Test of Two Theories’, The Sociological Quarterly 33: 4, 535–52. Leibfried, S. (1992), ‘Towards a European Welfare State? On Integrating Poverty Regimes into the European Community’, in: Z. Ferge & J. Kolberg (eds) Social Policy in a Changing Europe (Frankurt am Main: Campus Verlag). Lemert, C. (2002), Social Things: an Introduction to the Sociological Life (Lanham, MD: Rowman & Littlefield Publishers). Lepianka, D. (2004), Towards the Understanding of Popular Poverty Explanations. Paper presented at CCWS/ESPAnet Young Researcher Conference, 1–3 October, Gammel Vraa Castle, Denmark. Lijphart, A. (1984), Democracies: Pattern of Majoritarian and Consensus Government in Twenty-one Countries (New Haven and London: Yale University Press). Lindbeck, A. & Snower, D. (1988), The Insider-Outsider Theory of Unemployment (Cambridge, MA: MIT Press). Lipset, S. M. (1996), American Exceptionalism (New York: Norton). Lødemel, I. & Trickey, H. (eds.) (2000), An Offer you can’t Refuse. Workfare in International Perspective (Bristol: The Policy Press). March, J.G. & Olsen, J.P. (1984), ‘The New Institutionalism: Organizational Factors in Political Life’, American Political Science Review 78: 3, 734–49. March, J.G. & Olsen, J.P. (1989), Rediscovering Institutions (New York: Free Press). Marshall, T.H. (1950), Citizenship and Social Class (Cambridge: Cambridge University Press). Mau, S. (2003), The Moral Economy of Welfare States. Britain and Germany Compared (London: Routledge). Merrill, S. & Grofman, B. (1999), A Unified Theory of Voting. Directional and Proximity Spatial Models (Cambridge: Cambridge University Press). Meyer, N.M. (1999), Kærlighed og opdragelse i socialaktiveringen (Copenhagen: Gyldendal). Miller, D. (1992), ‘Distributive Justice: What the People Think’, Ethics 102: 3, 555– 93. Ministry of Refugee, Immigration and Integration Affairs (2004), Årbog om udlændinge i Danmark. Status of Udvikling (Copenhagen: Ministry of Refugee, Immigration and Integration Affairs). Ministry of Social Affairs (2000), Socialsektoren i tal 2000 (Copenhagen: Ministry

164

The Institutional Logic of Welfare Attitudes

of Social Affairs). Murray, C. (1984), Losing Ground: American Social Policy, 1950-1980 (New York: Basic Books). Nelson, T.E., Clawson, R.A. & Oxley, Z.M. (1997), ‘Media Framing of a Civil Liberties Conflict and its Effect on Tolerance’, American Political Science Review 91: 3, 567–83. Nolan, B. & Hauser, R. (2000), ‘Unemployment and Poverty’, in: D. Gallie & S. Paugam, S. (eds) Welfare Regimes and the Experience of Unemployment in Europe (Oxford: Oxford University Press). Nososco (2002), Social tryghed I de nordiske lande 2000 (København: Nordisk Socialstatistisk Komité). Nososco (2004), Social tryghed I de nordiske lande 2002 (København: Nordisk Socialstatistisk Komité). O’Connor, J. (1973), The Fiscal Crisis of the State (New York: St. James Press). OECD (1994a), The OECD Jobs Study. Evidence and Explanations. Part I. Labour Market Trends and Underlying Forces of Change (Paris: OECD). OECD (1994b), The OECD Jobs Study. Evidence and Explanations. Part II. The Adjustment Potential of the Labour Market (Paris: OECD). Offe, C. (1972), ‘Advanced capitalism and the welfare state’, Politics and Society no 4. Oorschot, W. van (1991), ‘Non-Take-Up of Social Security Benefits in Europe’, Journal of European Social Policy 1: 1, 15–30. Oorschot, W. van (2000), ‘Who Should Get What, and Why? On Deservingness Criteria and the Conditionality of Solidarity among the Public’, Policy & Politics 28:1, 33–48. Oorschot, W. van (2005), Immigrants, Welfare and Deservingness Opinions in European Welfare States (Aalborg: CCWS Working Paper no. 2005-34). Oorschot, W. van & Halman, L. (2000), ‘Blame or fate? Individual or Social? An International Comparison of Popular Explanations of Poverty’, European Societies 2: 1, 1–28. Oorschot, W. van. & Arts, W. (2005), ‘The Social Capital of European Welfare States: the Crowding Out Hypothesis Revisited’, Journal of European Social Policy 15: 1, 5–26. Overbye, E. (1996), ‘How do Organizational Designs Influence Welfare Politics? A Study of the Nordic Pension Systems’, Scandinavian Political Studies 19: 3, 227–55. Overbye, E. & Eia, H. (1995), Oppfatninger om velferd. Rapport fra en intervjuundersøkelse, INAS-notat, (Oslo: Institutt for sosialforskning). Park, A. et al. (eds) (2003), British Social Attitudes: Continuity and Change over Two Decades (London: Sage Publications). Peters, B.G. (1999), Institutional Theory in Political Science. The ‘New Institutionalism’ (London: Printer). Petterson, P.A. (1995), ‘The Welfare State: The Security Dimension’, in: O. Borre & E. Scarbourgh (eds) The Scope of Government (Oxford: Oxford University

References

165

Press). Pierson, P. (1994), Dismantling the Welfare State? Reagan, Thatcher and the Politics of Retrenchment (Cambridge: Cambridge University Press). Pierson, P. (1996), ‘The New Politics of the Welfare State’, World Politics 48: 2, 143–79. Pierson, P. (2000), ‘Increasing Returns, Path Dependence, and the Study of Politics’, American Political Science Review, 94: 2, 251–67. Pierson, P. (ed.) (2001a), The New Politics of the Welfare State (Oxford: Oxford University Press). Pierson, P. (2001b), ‘Coping with Permanent Austerity: Welfare State Restructuring in Affluent Democracies’, in: P. Pierson (ed.) The New Politics of the Welfare State (Oxford: Oxford University Press). Piven F. F. & Cloward R. A. (1971), Regulating the Poor (New York: Vintage). Quadagno, J. (1994). The Color of Welfare. How Racism Undermined the War on Poverty (Oxford: Oxford University Press). Rasbach, J. et al. (2000), A User’s Guide to Mlwin (version 2.1d) (Centre for Multilevel Modelling, Institute for Education, University of London). Riphahn, R.T. (2001), ‘Rational Poverty or Poor Rationality? The Take-up of Social Assistance Benefits’, Review of Income and Wealth 47: 3, 379–98. Rodrik, D. (1998), ‘Why Do More Open Economies Have Bigger Governments?’ Journal of Political Economy, 106, 997–1032. Rothstein, B. (1994), Vad bör staten göra? (Stockholm: SNS Förlag). Rothstein, B. (1998), Just Institutions Matter: The Moral and Political Logic of the Universal Welfare State (Cambridge: Cambridge University Press). Rosenberg, M. (1979), Conceiving the Self (New York: Basic Books). Saurama, L. (2001). Description of Finnish Data Collection for Research Program: Unemployment, Early Retirement and Citizenship: Marginalisation and Integration in the Nordic Countries (Department of Social Policy, University of Turku). Sayers (stavemåden?), L.W. (1989), Pooled Time Series Analysis. (Sage University Paper, Quantitative Applications in the Social Sciences, no. 70). Scherer, K.R. (1992a), ‘Issues in the Study of Justice’, in: K.R. Scherer (ed.) Justice: Interdisciplinary Perspectives (Cambridge: Cambridge University Press). Scherer, K.R. (ed.) (1992b), Justice: Interdisciplinary Perspectives (Cambridge: Cambridge University Press). Schneider, A. & Ingram, H. (1993), ‘Social Construction of Target Populations: Implications for Politics and Policy’, The American Political Science Review 87: 2, 334–47. Schmidt, V.A. (2002), ‘Does Discourse Matter in the Politics of Welfare State Adjustment?’, Comparative Political Studies 35: 2, 168–93. Sefton, T. (2003), ‘What we Want from the Welfare State’, in Park et al. (eds) British Social Attitudes: Continuity and Change over Two Decades (London: Sage

166

The Institutional Logic of Welfare Attitudes

Publications). Sharp, E. B. (1999), The Sometime Connection. Public Opinion and Social Policy (Albany: State University of New York). Sharpf, F.W. & Schmidt, V.A. (eds) (2000), Welfare and Work in the Open Economy, vols 1–2 (Oxford: Oxford University Press). Svallfors, S. (1995), ‘Institutions and the Comparative Study of Beliefs about Justice’, in: O.E. Kangas et al. (eds) In the Eye of the Beholder. Opinions on Welfare and Justice in Comparative Perspective (Umå: The Bank of Sweden Tercentenary Foundation). Svallfors, S. (1997), ‘Worlds of Welfare and Attitudes to Redistribution: A Comparison of Eight Western Nations’, European Sociological Review 13: 3, 283–304. Svallfors, S. (1999), ‘Välfärdsregimer och välfärdsopinioner: en jämförelse mellan åtta västländer’, Sociologisk Forskning, 1, 93–119. Svallfors, S. (2004), Klassamhällets kollektiva medvetande (Umeå: Borëa). Taylor-Gooby, P. (1995), ‘Who Wants the Welfare State? Support for State Welfare in European Countries’, in: O.E. Kangas et al. (eds) In the Eye of the Beholder. Opinions on Welfare and Justice in Comparative Perspective (Umå: The Bank of Sweden Tercentenary Foundation). Titmuss, R.M. (1974), Social Policy: An Introduction (London: Allen and Unwin). Togeby, L. & Øystein, G. (2000). Holdninger til flygtninge og indvandrere, 1993 (Odense: Dansk Data Arkiv). Tsebelis, G. (2002), Veto Players: How Political Institutions Work (Princeton, NJ: Princeton University Press). Tullock, G. (1984), Economics of Income Redistribution (Boston: Kluwer Nijhoff). Turner, C.F. & Martyin, E. (eds) (1984), Surveying Subjective Phenomena. (New York: Russell Sage Foundation). Wadel, C. (1973), ‘Now, Whose Fault is That?’, Social and Economic Studies no. 11 (Toronto: Memorial University of New Foundland). Wilensky, H. L. (1975), The Welfare State and Equality (Berkeley: University of California Press). Will, J. (1993), ‘The Dimensions of Poverty: Public Perceptions of the Deserving Poor’, Social Science Research 22: 3, 312–32. Wilson, G. (1996), ‘Toward a Revised Framework for Examining Beliefs about the Causes of Poverty’, The Sociological Quarterly 37: 3, 413–28. Wilson, J.Q. (ed.) (1980a), The Politics of Regulation (New York: Basic Books). Wilson, J.Q. (1980b), ‘The Politics of Regulation’, in: J.Q. Wilson (ed.) The Politics of Regulation (New York: Basic Books). Wiseman, M. (2000), ‘Making Work for Welfare in the United States’, in: I. Lødemel & H. Trickey (eds) An Offer You Can’t Refuse. Workfare in International Perspective (Bristol: The Policy Press).

Index

absolute versus relative support 26, 58, 123-6 AFDC 13, 100, 148-9 Alesina, A, & Glaeser, E. L, 3-4, 43, 60, 70-6, 85 American exceptionalism 3, 36-7, 43-5, 95, 140 Anglo-Saxon poverty tradition 45-8 anti-egalitarianism 33, 41, see also egalitarianism bad risks 5 13, 21, 55 binary logistic model 80-83, 86 blame avoidance 1, 17, 147, 149 see also Pierson and concentrated versus dispersed interests child allowances 102-8, 142 class effects 84, 93, see also class-interest class-interest 1, 9, 13-25, 29-34, 53, 84, 140, 145 compensation rates 112, 149 compositional effects 32-3, 36, 80 concentrated versus dispersed interest 1, 17, 21-2, 44, 145 coughlin, R. M. 24, 26, 47 cultural distinctiveness 80, 85, 111-23, 128-55, 166, see also ethnic fractionalisation and selectivism culture thesis 3, 18-9, 22, 25, 34, 37, 43-4, 53, 57, 135, 140, 145 de-commodification 9-10, 12, 19, 29 dependency culture 57

deservingness criteria 4, 21, 45-7, 4952, 55, 60-2, 140, 150-1, 153-5 deviant cases 20, 31, 67, 69-71, 89, 93, 141 disability pension 17, 53, 103-6, distributive justice 19 early retirement 11, 55-9, 103-06, 1089, 142 economic man 4, 21, 22, 135, 145 economic resources 52, 57, 61, 111-2, 155 egalitarianism 4, 6, 25, 41 Esping-Andersen, G., see welfare regime ethnic fractionalisation 4, 6, 65, 71, 7381, 93, 99, 109, 121, 123, 129, 141-2 ethnic homogeneity 6, 62, 78 ethnic stereotypes 123, 129, 132, 138 eurobarometer 6, 25, 27, 35, 59, 62, 668, 70-1, 76, 78, 88-9, 91-3, 95-8, 141, 155 feed back 15-20, 44, 57-8, 72, 88, 95, 139-140, see also path-dependency financial hardship 112-4, 116, 118, 120, 143 fiscal limitations 155 focused comparisons 6, 62-3, see also most-similar design foodstamps 100 framing 5, 22, 35, 145 future public support 7, 144, 155 Gelissen, J. 27, 29-33, 35 gender 33, 66, 88, 93, 117, 141

168

The Institutional Logic of Welfare Attitudes

generosity 46, 53, 56, 58, 62-3, 65, 71, 75, 80, 85, 88, 93, 109, 111-5, 119-21, 123, 126-9, 135-7, 1404, 149-50 gilens, M. 3, 22, 37, 43, 129, 149 globalisation 147, 156 golden-age 1, 11, 56, 76, 129, 147, 150 Goul Andersen, J. 11, 15, 21-2, 26, 34, 54, 102, 111, 133, 148-9

large-n-logic 69, 92, 141 level of unemployment 32, 59, 70-9, 81, 91, 102-5, 108-9, 111-2, 119-21, 141, 148 Lipset, S. M. 3, 18 see also American exceptionalism living standard 7, 18, 123-6, 128-9, 132, 135-7, 143

high status occupations 37, 39-40 Homans, G. C. 19, 41, 43 housing allowances 103-4, 106-9, 142

Marshall, T. H. 15, 156 means-testing 4, 100, 102-3 media research 145 Medicaid 13, 100 micro-foundation 3, 7, 25, 144, 146 middle class 1-2, 16, 21, 32-3, 44, 66, 85, 93, 144 missing link 2, 20, 45, 49, 89 most similar design 6, 62-3, 108-9, 111, 117, 119-21, 123, 143 multilevel model 32, 80

identity effects 111, 155 immigration 3, 7, 49, 73, 124, 129-30, 132, 134, 142-4, 155-6 income distribution 5, 19, 35-6, 38-9, 46, 56, 97 individual level variables 32, 65, 81, 83, 93 insider outsider divide 12, 87 interaction effects 33 international social survey project (ISSP) Social inequality module 28-30, 345, 37, 41, 45, 89-90, 92-3, 106, 140, 155 Role of government module 26, 29 30, 35, 59 job opportunities 4, 6, 56, 56, 59, 61-3, 65, 70-2, 80, 85, 93, 117, 123, 126-9, 135-7, 140-1, 143, 156 job protection 55, 59, 60, 105 justice belief 4, 34, 35 Kangas, O. 4, 34-5 Korpi, W. 1-2, 15-6, 21, 25, 29, 57, 139-40, see also power resource theory

national sample 6, 28, 63, 133-5, 143 new politics 1-4, 17, 21, 144, 147, see also P. Pierson non-take-up 99-100 OLS regression 76, 79, 118, 129 Oorschot , W. V. 4, 47-51, 67, 100, 140 parties 51, 130-7, 144, 148-51 passion for equality 3-4, 18, 37 path-dependency 2, 7, 139, 144, 14950, see also feedback pension 10-1, 13, 17, 21, 30, 43, 46, 48, 53-5, 100, 102-8, 117-9, 139, 142, 145, 158-9, 161, 164 Pierson, P. 1, 15, 17, 21, 76, 139, 144, 146-9, 151, see also New politics and blame avoidance political man 5, 20-2, 43, 46, 135, 137, 142, 145

Index

political mobilisation 32, 132-3, 138, 144 political orientation 81-4, 86-7, 132, 137 pooled data 78-9 post-industrial 59, 139, 149-50 poverty lines 20, 46, 52 power resource theory 1-3, 15-6, 21-2, 32-3, 37, 53, 57, 66, 93, 139, 144-5, see also W. Korpi public expenditures 11, 34 Quadagno, J. 3, 59, 75 rational choice 4, 17, 88, 144 reciprocity 48-9, 52-5, 61, 140 restructuring 1, 2, 10-1, 14-5, 17, 146-7, 150 Rothstein, B. 16, 22, 52-4, 152 selectivism 6, 56, 58, 62, 65, 71, 73-6, 79-80, 85, 93, 99, 101, 103, 105, 107-9, 119, 121, 140, 142, 145 self-esteem 111-2, 118, 120-1, 143 self-interest 1, 4-5, 13, 18-9, 21-3, 29, 32-4, 44, 66, 75, 88, 135, 140, 144-5 skilled workers 25, 36-7, 39, 41, 43-4, 84-5, 87, 93, 140-1 small-n problem 62, 72, 93, 123 social assistance 26, 46-9, 55-6, 102-9, 111, 114, 117, 119, 123-38, 1424, 148-9, 153-5 social trust 6, 62, 151-2 sociological man 5, 22, 135, 137, 145 SOCX 72 solidarity 14-6, 18-9, 22, 50, 143 stigma 16, 57, 99-105, 107-9, 111-21, 142-3, 158 stratification 15, 139

169

structural unemployment 147, 149 structural equation models 27, 33 Svallfors. S. 14, 31-4, 37-9, 80, 90-1 take-up rates 100 targeting 109, 150, 155 the potential poor 46, 56-7, 61, 111, 141 Titmuss, R. 139, 145, 156 trajectories 14, 43, 58-9, 139, 150, 154, see also path-dependency and feedback unemployment benefits 17, 46, 53-6, 103-5, 107-9, 112, 114, 117-9, 121, 124, 142-3, 148-9, 155 universalism 4, 16, 53, 99, 102 unskilled worker 36-7, 39, 41, 43-4, 87, 93, 141 validity 25-6, 95, 105, 124 vicious circle 53-4 visibility argument 21, 53 wage setting 59-60 welfare dependency 49 welfare regime theory 3-4, 6, 9-25, 45, 76, 137, 139-40, 144 definition 9-10 ideal type 5, 11-2, 16, 20, 36, 40-1, 44, 49, 51, 53-7, 60-2, 69, 71, 76, 85, 95-6, 148 work ethic 123, 126-9, 131-2, 135-8, 144 working class 16, 32-3, 43, 50, 66, 845, 144 World Values Study first wave 27, 73, 94, 96 second wave 27, 66, 68, 94-5, 141 third wave 27, 66