How do welfare regimes influence attitudes? A comparison of five European countries and the United States

Transcription

1 University of Bielefeld Department of Sociology Research project Wohlfahrtsstaatliche Maßnahmen und Einstellungen der Bürger (WME) Working paper no. 9 How do welfare regimes influence attitudes? A comparison of five European countries and the United States Paper presented at the ECSR-Workshop Causes and Consequences of Socio-Economic and Political Attitudes in Eastern and Western Europe, Mannheim, Germany (October 7-8, 1999) (Preliminary version, do not cite!) Thorsten Heien & Dirk Hofäcker Bielefeld September 1999

2 Thorsten Heien & Dirk Hofäcker: How do welfare regimes influence attitudes? 2 Contents 1. Introduction Theoretical concepts What are the dimensions of attitudes towards the welfare state? What affects attitudes towards the welfare state? Self interest Differential socialization Economic situation National welfare cultures Interaction of macro and micro level attitude determinants Causal model Data, measurement, and method Data Measurement Dependent variables Independent variables Method Results Measurement model Are there any cross-country attitude differences? What are the reasons of cross-country attitude differences? Country-specific interest and socialization effects Summary and discussion References Appendix I Appendix II Appendix III... 47

3 Thorsten Heien & Dirk Hofäcker: How do welfare regimes influence attitudes? 3 1. Introduction + The last two decades are marked by a grown interest of social science research in the description and explanation of attitudes towards the welfare state (HEIEN 1998). A central concern of this research is the comparison of attitudes between different countries resp. welfare states (see, e.g., PAPADAKIS & BEAN 1993, 1998, PEILLON, SVALLFORS 1997). Cross-country comparisons are today much more attractive and easier than twenty years ago because of the availability of large scale comparative data-sets like the International Social Justice Project (ISJP) or the International Social Survey Programme (ISSP). Research usually explains attitudinal variations between countries with different cultural, institutional, and structural characteristics of the countries considered, resulting in specific dominant welfare state ideologies (GELISSEN 1999, KLUEGEL & MIYANO 1995, SVALLFORS 1997). The most influential systematization of relevant characteristics was provided by ESPING-ANDERSEN in his study Three Worlds of Welfare Capitalism (ESPING-ANDERSEN ). With respect to the interplay of public and private actors in welfare provision, the protection of individuals against the market, and the kind of stratification that is promoted by social policy, ESPING-ANDERSEN distinguishes between social democratic, conservative and liberal welfare regimes. Attitudinal research assumes that each regime type is related with specific patterns of attitudes towards the welfare state: social democratic regimes should show the highest level of welfare state support, followed by conservative regimes and finally by liberal regimes with the lowest level of support. Many authors have reported cross-country differences confirming the hypothesized order of regime types (see, e.g., GUNDLACH 1994, MAU 1998, PEILLON, SVALLFORS 1997). However, others (GELISSEN 1999, PAPADAKIS & ΒEAN 1993) reject the hypothesis at least partially because their results show only small resp. no cross-country differences in welfare state attitudes or even an attitudinal rank order that contradicts the one derived from ESPING-ANDERSEN s regime typology. An alternative, but less popular macro level explanation of attitude differences between welfare states is the economic situation resp. level of the countries considered. In general, literature suggests that in times of economic hardship support for the welfare state declines because people are less willing to give parts of their income to the state (SIHVO & UUSITALO 1995b). When comparing countries with different economic levels at the same point of time, economic situation works as a competing hypothesis to the welfare regime hypothesis. Although it can be tested with cross-sectional data, the economic situation hypothesis is especially suited for trend data. Except for times of tremendous cultural, institutional, and structural changes, the alternative explanatory factor of a specific welfare regime type is held constant when focussing on the development of one country s welfare state attitudes. + The authors thank Hans-Jürgen Andreß (Bielefeld) for his comments on earlier versions of this paper.

4 Thorsten Heien & Dirk Hofäcker: How do welfare regimes influence attitudes? 4 Unfortunately, only some authors (COUGHLIN 1980, GELISSEN 1999, PAPADAKIS & BEAN 1993) have paid attention to the possibility that attitudinal variations at the aggregate level could result from country-specific variations at the individual level. Citizens of social democratic regimes (resp. citizens of countries in a positive economic situation) do not generally have to be more favourable towards the welfare state than citizens of conservative and liberal regimes (resp. citizens of countries in a negative economic situation). Instead of being the result of a dominant welfare state ideology, differences at the aggregate level could also occur because several groups of citizens of social democratic regimes (resp. citizens of countries in a positive economic situation), identified by certain individual characteristics, show a much higher (lower) degree of support than the equivalent groups of citizens of conservative and liberal regimes (resp. citizens of countries in a negative economic situation). This implies the interaction of macro level (welfare regime and/or economic situation) and micro level (individual characteristics) determinants of welfare state attitudes. A rationale for such interaction effects is given by FORMA: The welfare state and it s institutional design form one sort of the structure people are living in (1997: 237). But generally speaking, which individual characteristics influence welfare state attitudes? Current research refers either to specific economic interests connected with the structural position of individuals ( selfinterest ) or to different socialization patterns (see, e.g., ANDREß & HEIEN 1999, CNAAN et al. 1993, GELISSEN 1999, HASENFELD & RAFFERTY 1989). Concerning the micro-macrointeraction, we have to ask how economic interests and/or socialization experiences are shaped by the type of welfare regime resp. the economic situation of a country. To give just one example, it is plausible to assume that women in social democratic regimes are much more favourable towards the welfare state than women in conservative or liberal regimes because of their outstanding benefits from redistribution, taxation, and public services (SVALLFORS 1997). The aim of this paper is therefore to broaden the perspective of comparative research on welfare state attitudes by offering a comprehensive view on the determinants of attitudinal variations between countries. West and East Germany, Norway, Italy, Great Britain, and the United States were chosen because of their different cultural, structural, and institutional backgrounds, representing different welfare regimes. Though now being citizens of the conservative (West) German welfare regime, in the past East Germans were exposed to the fundamentally different welfare regime of the former GDR. 1 Since the typology of ESPING- ANDERSEN is originally limited to capitalist welfare states, we will extend it to former communist welfare states in the theoretical part of this paper. Our empirical analysis starts with the question whether the considered countries differ in aggregate levels of welfare state support and whether this support changes over time. 1 Therefore, in the following analysis, East Germany is treated as a separate country regardless of the (political) unification with West Germany.

5 Thorsten Heien & Dirk Hofäcker: How do welfare regimes influence attitudes? 5 Furthermore, we compare the rank-order of welfare state attitudes with the ones hypothesized by the two macro level explanations. First, do citizens of social democratic regimes indeed show a higher level of welfare state support than citizens of conservative and liberal regimes, and what about the citizens of former communist regimes? Second, does a bad economic situation result in a lower level of welfare state support? Our next research question is concerned with the explanation of possible differences in aggregate levels: do attitudinal differences between countries result from overall variations, or do they result from countryspecific effects in structuring the effects of interest and socialization variables at the individual level? Finally, we examine country-specific patterns of attitude determination. How do individual characteristics in detail influence people s welfare state attitudes in the countries considered? Our causal model of welfare state attitudes is explained in section 2, followed by a presentation of our data, measurement issues and the statistical method in section 3. The results of our analysis are discussed in section 4. The paper closes with a summary of the basic findings and raises some questions in section 5 that may form the basis of the discussion at the workshop and of future research in general. 2. Theoretical concepts 2.1. What are the dimensions of attitudes towards the welfare state? Without any doubt the welfare state is a complex phenomenon. Assuming homogenous attitudes towards the welfare state regardless of the aspects studied would be a rather unrealistic belief. Therefore, we developed a multidimensional model of welfare state attitudes. Based on similar typologies (e.g., CNAAN 1989, ROLLER 1992, SIHVO & UUSITALO 1995a) we differentiate between four main dimensions: a) the aims of the welfare state, b) the means (institutions, programs, actors) of the welfare state, c) the (intended and unintended) effects of the welfare state, and d) the financing of the welfare state (HEIEN 1998). In this paper we concentrate on the first dimension. In accordance with the definition of the welfare state by FLORA et al. (1977), the primary aim of the welfare state can be seen in ensuring socio-economic security and socio-economic equality. This can be done in several ways: for example, by guaranteeing a basic income, providing jobs for all, supporting children from poor families to go to university or reducing income differences. An important question in this respect is whether people see a governmental responsibility for doing this. Following ROLLER (1992), we call this subdimension of attitudes towards the welfare state the extent of the welfare state. In some ways, this dimension can be labelled fundamental compared to the other dimensions. Means, effects, and financing can be neglected if people see no governmental responsibility for ensuring socio-economic security and socio-economic equality at all (HEIEN 1998). Another important

6 Thorsten Heien & Dirk Hofäcker: How do welfare regimes influence attitudes? 6 advantage of using the extent dimension is its comparability across welfare states whereas questions concerning the other dimensions, especially questions concerning the means of the welfare state, are often nation-specific. Finally, there is a pragmatic reason to limit the analysis to the extent dimension. As GELISSEN (1999) points out, large scale comparative data-sets like the ISJP or the ISSP usually concentrate on this aspect of welfare state attitudes What affects attitudes towards the welfare state? The empirical literature that tries to explain attitudes towards the welfare state basically refers to four different determinants: a) self interest, b) different socialization patterns, c) the economic situation, and d) national welfare cultures. We want to define these four modes of explanation and their interrelationships very shortly Self interest The first mode of explanation focuses on the egoistic self interests of the citizens of the welfare state. Using rational choice arguments, it is hypothesized that individuals will evaluate those aspects of the welfare state positively from which they gain personally (i.e., which maximise utility ), and disapprove those aspects that do not appear advantageous for their own interests (COOK & BARRETT 1992, FORMA 1997, SANDERS 1988). The literature concentrates on three types of interests, which all are determined by the individual s position in the social structure. Firstly, consumers of services and transfers relying on the welfare state for material support, like women, old age pensioners, young families with children, persons with low income or low level of education, ethnic minorities, the unemployed or the disabled (the so-called transfer classes or underdogs ), will show a much higher degree of support than those who consume less (CNAAN et al. 1993, COOK & BARRETT 1992, ROBINSON & BELL 1978). Secondly, those contributing to the welfare state as taxpayers are likely to have negative attitudes towards the welfare state (GILENS 1995, SIHVO & UUSITALO 1995b). Difficulties arise because people do not only pay taxes, but also benefit from the welfare state. Some authors, e.g., assume that middle class people despite their tax burdens strongly support the welfare state because they benefit disproportionately as consumers of health and education services (PAPADAKIS & BEAN 1993: 258; see also PÖNTINEN & UUSITALO 1988). Others, especially supporters of the welfare backlash (WILENSKY 1975) and the working-class anger hypothesis (TAYLOR-GOOBY 1983), claim in contrast that members of the middle and the working class despite benefiting from programs will oppose to the welfare state because they resent having to pay taxes [...] and channel their anger against the poor (CNAAN et al. 1993: 125). Besides that, a subjective element has to be considered: the individual s perception of its own present and future social position possibly influences its views about the welfare state. Perceived risks of unemployment or other life events, for example, make it even

7 Thorsten Heien & Dirk Hofäcker: How do welfare regimes influence attitudes? 7 for well-off citizens rational to support benefits not meeting their acute needs (FORMA 1997: 239; see also PÖNTINEN & UUSITALO 1988, ROBINSON & BELL 1978). Thirdly, apart from the question of benefits and burdens, some people have interests in the welfare state as its producers. Those employed in the public sector - according to PAPADAKIS & BEAN (1993), especially those employed in the welfare bureaucracies in reproduction and education as doctors, nurses, teachers or social workers - are more likely to be favourable towards the welfare state than others, because their employment, but also their careers, working conditions and economic rewards depend on the welfare state s prosperity (FORMA 1997, PÖNTINEN & UUSITALO 1988, SVALLFORS 1991, 1995) Differential socialization A second type of explanation assumes that individual socialization processes will lead to different attitudes towards the welfare state. Several individual characteristics are used as indicators of different socialization processes: among them age (i.e., generation), gender, education, and employment sector. A differentiation with respect to age can be found in INGLEHART s (1977) work on materialistic and post-materialistic values. According to this theory, the younger age cohorts should be characterized by rather post-materialistic values, i.e., they should favour values such as solidarity and community instead of pure self-interest. Therefore, we expect the younger age cohorts to show a comparatively stronger support for the welfare state than the older ones. Nevertheless, we may also expect different age effects in different countries due to their different levels of modernity. Similarly, a rich literature exists on different socialization patterns of men and women. A traditional view can be found in the following citation from KIDDER et al. (1981: 246): Women s behaviours are maximally tailored to the world of love, men s to the world of work. Although it may be questioned whether this traditional division of labour between men and women is still a good description of the living arrangements of both sexes today, it is obvious that care for others continues to be a central part of socialization for most women (SVALLFORS 1991, 1995). Therefore, it is assumed that caring and need have much higher value for women than for men. Hence, we expect women to show higher levels of support for the welfare state which provides those in need with services respectively financial benefits. Men are expected to show less support. The effects of education are not as clear as the other two variables. First of all, education is one of the ways in which the dominant welfare state ideology of a country is transferred to the individuals (HEIEN 1997). The longer a person spends in the educational system, the more she is assumed to adhere to this dominant ideology. This also implies, that the effect of education on welfare state attitudes varies between countries with different dominant ideologies. Secondly, education is supposed to enlighten people on general values of civilization. With enlightenment [...] comes a greater commitment to the idea of equality as a positive value

8 Thorsten Heien & Dirk Hofäcker: How do welfare regimes influence attitudes? 8 (ROBINSON & BELL 1978: 129). If this is true, the better educated will show greater preferences for an extensive welfare state. Finally, the longer a person remains successful in the educational system, the more she will be convinced that individual achievement is rewarding and should be rewarded. Therefore, one could also hypothesize that individual success ideologies prevail among the better educated; i.e. people with a higher education will tend to show lower levels of support for the welfare state. This hypothesis is based on the assumption that the welfare state is regarded as being distributive which to some extent contradicts the idea of individual achievement. Finally, compared to the private sector, employment in the public sector is connected with specific socialization experiences. As different authors argue working conditions in the public sector may create bonds of sympathy and solidarity with fellow public-sector employees and their clients, patients, and other welfare dependants (SVALLFORS 1995: 55; see also HOEL & KNUTSEN 1989, SVALLFORS 1991). Therefore, people working in the public sector should show higher levels of support for the welfare state than employees of the private sector Economic situation Turning to the macro level, it is plausible to assume that welfare attitudes are influenced by the present economic situation of a specific country. This is a very popular argument since it is compatible with current theories of modernization and social evolution. According to these theories, industrial and economic development is linked with a general change in values towards individualism and self-realization. Following GUNDELACH (1994), one can therefore expect those living in countries with a high degree of economic development to show comparably low levels of support for the welfare state. SIHVO and UUSITALO (1995b) support this hypothesis with a different argument. In times of economic recession, there can also be an increased need for social policies in order to protect citizens from the negative effects of a decreased economic growth and increasing unemployment. Therefore, we could theoretically expect a bad economic situation to result in a high support for welfare state actions. It is worth noting, that this is not the only possible direction of an effect of the economic level on welfare state attitudes. SIHVO and UUSITALO explain declining support for the welfare state in times of economic crises by the fact, that when one s own economy is on the decline, one is less willing to give part of one s own income to the state (1995b: 252). Their research results support this hypothesis because they found considerable evidence for a declining support for the welfare state following an economic backlash (SIHVO & UUSITALO 1995b: 260). Furthermore, our own results for East and West Germany provide evidence for reduced claims towards the welfare state in times of a declining economic growth (HOFÄCKER 1999). We therefore assume economic level to have a positive effect on welfare state attitudes: people in countries with a high economic level will show higher support for welfare state activities than those in a country with a lower economic level.

9 Thorsten Heien & Dirk Hofäcker: How do welfare regimes influence attitudes? 9 Figure 1: GDP growth rate % 8% 6% 4% 2% 0% -2% -4% Source: UNITED NATIONS 1994, USA GB W-GER E-GER ITA NOR An important methodological problem is how to measure the economic situation for it can be described by a great number of different indicators. We decide to rely on two approved indicators: the annual GDP growth rate and the rate of unemployment. Compared to the pure numerical value of the national GDP, the growth rate supplies us with a direct measure of the Figure 2: Unemployment rate % 16% 14% 12% 10% 8% 6% 4% 2% 0% Source: INTERNATIONAL LABOUR OFFICE 1993, USA GB W-GER E-GER ITA NOR 2 The following abbreviations are used: USA = United States; GB = Great Britain; W-GER = West Germany; E- GER = East Germany; ITA = Italy; NOR = Norway. Concerning East Germany, data prior to 1992 are not available resp. not comparable to the data for the other five countries (SACHVERSTÄNDIGENRAT 1991: 2). 3 Unemployment rates refer to the following age groups: United States, Great Britain, and Norway: persons aged 16 years and over; West and East Germany: persons aged 15 years and over; Italy: persons aged 15 years and over (prior to 1993: persons aged 14 years and over). Concerning East Germany, data prior to 1991 are not available.

10 Thorsten Heien & Dirk Hofäcker: How do welfare regimes influence attitudes? 10 economic development of a country compared to the preceding year and therefore provides us with a measure of the economic atmosphere of the country. The unemployment rate can additionally be seen as a measure of the fiscal pressure on the welfare state. Figure 1 and 2 report the GDP growth rate and the unemployment rate for the analyzed countries from 1983 to. Although the trends are generally very irregular, one can for example see, that especially Italy is characterized by a comparably low rate of economic growth combined with a fairly high level of unemployment. We should therefore expect Italians to show a relatively low support for welfare state activities. Further effects will be discussed in section National welfare cultures While models of differential socialization assume a non-uniform socialization according to individual experiences models of national welfare cultures deny such heterogeneity. It is assumed that certain cultural, structural, and institutional characteristics of a country crystallize into dominant welfare state ideologies. Relevant cultural aspects are, e.g., the religious and/or ideological background of a society. Some researchers distinguish between countries with catholic and protestant traditions (WEGENER & LIEBIG 1995a), others between countries with liberal, socialistic, or communistic traditions (HALLER et al. 1995). While Protestantism or liberalism stresses the responsibilities of the individual, leaving only residual functions for the state, Catholicism, socialism, or communism focus much more on notions of equality - in the communist extreme: the classless society - allowing more or less extensive interventions of the welfare state. Since welfare regimes always developed under the influence of the religious and ideological background of a society, the structural and institutional characteristics of a country are closely related to the aforementioned cultural aspects. A popular systematization of such structural and institutional aspects can be found in ESPING-ANDERSEN s () typology of welfare state regimes. 4 According to the author three worlds of welfare capitalism have to be differentiated: the liberal, the conservative (or corporatist) and the social democratic (or socialist) world. The indicators of his classification refer to specific configurations of institutions and distributive outcomes like the public-private interplay in welfare provision, the protection of individuals against market forces ( de-commodification ), and the kind of stratification that is promoted by social policy (see Table 1). It should be noted, that ESPING- ANDERSEN assumes no single pure cases of liberal, conservative, or social democratic welfare regimes in reality. Each existing welfare regime contains elements of the other regimes types: The degree to which clearly defined regime-clusters exist depends [...] on the extent to which regime-specific features are exclusively present only in one type (ESPING-ANDERSEN : 69). 4 For a discussion of alternative welfare state typologies see ARTS & GELISSEN (1998).

11 Thorsten Heien & Dirk Hofäcker: How do welfare regimes influence attitudes? 11 The liberal regime comprises Anglo-Saxon welfare states such as the United States, Canada, Great Britain, Australia, and New Zealand. 5 In these countries, the market plays the dominant role in the distribution of resources. Welfare provisions by the state are minimal and restricted to those who have proven need (means-tested programs), while the extent of social insurance programs and universal transfers is only modest. Consequently, the degree of decommodification is quite low and the order of stratification is characterized by a strict social dualism between the majority of citizens which are (market-differentiated) rewarded for their frugality, entrepreneurship and self-reliance and the stigmatized recipients of socialassistance. Table 1: Extended typology of welfare regimes Dimension Regime-type (example) Liberal (United States; Great Britain) Conservative (West Germany) Social democratic (Norway) Latin Rim (Italy) (former) Communist (East Germany) Dominant aspect of the welfare mix market subsidiarity state familialism state De-commodification low medium high low/medium perfect Logic of stratification exclusion segmentation inclusion segmentation inclusion The conservative regime is typical for most Continental European countries like Germany, Austria, or France. Although state activity is clearly higher than in the liberal type it is restricted to situations when the family fails to ensure social security (compare the subsidiarity principle in Catholicism). Welfare provisions depend on work performance resp. previous earnings and preserve former status with negligible redistributive impact. Therefore, the protection against the market depends primarily on the labour market position with (male) employees and their families being de-commodified to a high degree and others being insufficiently de-commodified. As a consequence, conservative welfare states are marked by a strict hierarchical segmentation. Finally, the social democratic regime, which is the smallest of the three clusters, can be found in Scandinavian countries like Sweden, Norway, or Denmark. In these countries the state plays a dominant role in the distribution of resources with the aim of promoting an equality of highest standards, not an equality of minimal needs (ESPING-ANDERSEN : 28). 5 Some authors have argued that the radical welfare states Australia and New Zealand must be distinguished from the truly liberal world because of their high benefit equality and their strong labour movements (see, e.g., CASTLES & MITCHELL 1992). Since we concentrate on the United States and Great Britain, we can neglect this differentiation.

12 Thorsten Heien & Dirk Hofäcker: How do welfare regimes influence attitudes? 12 Individuals are emancipated from market dependence by means of universal benefits resulting in a high degree of de-commodification. Because of universal entitlement rules society is not characterized by a dualism or segmentation like in the other regime-types. Generally speaking, ESPING-ANDERSEN s characterization of the liberal welfare regime is very much based on the prime example of the United States. It is not clear, if Great Britain can be regarded as a typical case for that type of regime. SCHMID (: 94) even regards Great Britain as a hybrid form of all the different ESPING-ANDERSEN regime types. Concerning the existing dominance of social insurance over social provision, it can be regarded as a conservative regime. Concerning its universal character, there are some parallels to the socialdemocratic logic of organizing welfare. On the other hand, the very low level of benefits suggests a classification as liberal. For the purpose of our research, we will - in accordance with almost any other study of welfare regimes - regard Great Britain as being liberal, but we will have to keep in mind that the British case is ambiguous. Another question of the current scientific discourse is whether the South European or Latin Rim countries (Italy, Spain, Portugal, and Greece) should be considered as a separate cluster in the universe of welfare states (FERRERA : 18). ESPING-ANDERSEN (, 1999) himself regards the South European countries as a special variant of the conservative regime and has fortified this evaluation in his latest publication. On the other hand, FERRERA () argues in favour of a southern model referring to the rudimentary and fragmented character of the South European welfare states due to their late industrial and socio-political development. Furthermore, the Italian welfare state is characterized by a particularisticclientelistic design donating special benefits to important electoral groups. Moreover, the important role of the family as a welfare provider and a central economic institution (MINGIONE 1994: 138) can be seen as a typical characteristic of South European welfare systems: The family is the explicit partner of social policies in Italy (SARACENO 1994: 65). For the purpose of our research, we will regard the case of Italy as part of a specific Southern Model of Welfare constituting an extreme case of the conservative model. Its typical characteristics are a weak state and a high degree of familialism. 6 As in conservative regimes, the level of de-commodification is comparably moderate while the logic of stratification is based on segmentation. The typology of ESPING-ANDERSEN is originally limited to capitalist welfare states. Since many (East) European countries have experienced a non-capitalist history the question remains whether these countries can be seen as a fourth regime-type. Without going into detail we argue that (former) communist countries have been those pure cases of the social democratic regime-type ESPING-ANDERSEN does not expect to occur for capitalist welfare 6 In accordance with POPENOE, familialism shall be understood as the belief in a strong sense of family identification and loyalty, mutual assistance among family members, and a concern for the perpetuation of the family unit (POPENOE 1988: 212).

13 Thorsten Heien & Dirk Hofäcker: How do welfare regimes influence attitudes? 13 states, i.e. they were more or less free of liberal and conservative elements. The state did not just play the dominant role in the distribution of resources, it was even the one and only distributor (DELHEY 1998). Because of the absence of market elements, de-commodification was perfect in communist regimes. Finally, social inequality was minimal, because differences in income and living conditions were rather small. Although now being citizens of the conservative (West) German welfare regime, East Germans were in the past exposed to the communist regime of the former GDR. Therefore, at least right after the German reunification, we expect East Germans attitudes to be determined by the communist past. But in the following years, one can expect the East Germans to have adjusted their attitudes to the conservative regime. Therefore, they have probably become more similar to West Germans attitudes in the sense of a convergence between East and West Germany. The question remains how the level of people s support for an extensive welfare state differs between the six countries considered. We expect this support to be lowest in countries with a weak state activity in welfare provision and a small number of welfare state beneficiaries, i.e., where benefits are restricted to specific groups and only few universal transfers exist. In contrast to this, welfare state support should be highest in regimes where the state plays an important role in welfare provision and most of the population benefit from transfers because universalism serves as the guiding principle. Nonetheless, we expect the Italians to show higher support for welfare state activities than the Germans, since Italy is going through an earlier stage of welfare state development. People in these young welfare states will favour a higher governmental responsibility since they have much more to gain from increased welfare state efforts than the citizens of the other European states (GELISSEN 1999: 13). Therefore, concerning the people s support for an extensive welfare state we hypothesize the following rank order of the six countries considered in our analysis (from highest to lowest): 1. East Germany, 2. Norway, 3. Italy, 4. West Germany, 5. Great Britain, 6. United States Interaction of macro and micro level attitude determinants Following the discussion of the macro- and micro-determinants of welfare state attitudes, we have to think about possible interaction effects because we want to explain attitude differences between countries with country-specific interest and socialization effects. As SVALLFORS points out, there is an obvious link between macro and micro level determinants of welfare state attitudes: The identities and interests of social actors are not pre-determined from their structural positions. They are created in a process where the institutional framework within which people act, and the historical traditions through which events and processes are interpreted, have a decisive impact. The weight of exposure to different institutional regimes creates diverging world-views even between people in similar structural locations (SVALLFORS 1997: 291).

14 Thorsten Heien & Dirk Hofäcker: How do welfare regimes influence attitudes? 14 The work of ESPING-ANDERSEN gives some indications for country-specific interest effects. According to him, the different regime-types are connected with specific social cleavages and conflict structures in the transition from industrial to post-industrial societies. The liberal welfare state tends to create a class cleavage, especially within sexes and races: 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. As some women become yuppies and Blacks become bourgeois, the women and Blacks left behind will experience much more keenly the phenomenon of relative deprivation (ESPING- ANDERSEN : 228). For the conservative welfare state, we have to expect conflicts between the insiders and outsiders of the labour market because the highly productive workforce supports a growing but unproductive outsider population (ESPING-ANDERSEN : 227). Finally, in the social democratic regime gender and sector (private vs. public sector) will emerge as the most important cleavages because of their different welfare state benefits: In this sense, one might easily imagine a war between (largely) male workers in the private sector and (largely) female workers in the welfare state (ESPING-ANDERSEN : 227). Because ESPING-ANDERSEN concentrates on typical representatives of the liberal, conservative, and social democratic welfare regime, we have to think about country-specific interest effects for Great Britain, Italy, and East Germany. As SCHMID (1995: 95) points out, the British system of welfare can be labelled by three typical characteristics: First, it is centred on wagework as the central source of a person s income. Secondly, it is following a policy of full employment. And thirdly, full employment actually means male full employment, i.e. it presupposes a traditional understanding of family. Accordingly, we may expect two central cleavage lines for the British case. As in conservative regimes, we may be faced with an insider-outsider phenomenon, i.e., a cleavage between those inside and outside the official labour market. Due to the change in traditional family norms (that, for example, become visible in growing divorce rates), we can also expect an attitudinal cleavage between men and women where the latter are not as well cared for by the welfare system. For the Italian case of a Latin Rim welfare regime, we can expect the dominant cleavage to run between the core sectors of the labour force and those employed in the so-called irregular or non-institutional market. As MINGIONE (1994) points out, there is a high degree of protection for those working in large or medium-sized enterprises which is contrasted by a scarcity of protection for those unemployed. For example, there are only rudimentary security programs for the young unemployed. Furthermore, the labour market position of women is characterized by a low participation rate and a high and growing level of unemployment. These conditions force many people into the irregular economic sector which is of course not covered by welfare state activities. In accordance with a classification by FERRERA (), we therefore expect differences in welfare state attitudes between the hyper-protected beneficiaries (public employees, white-collar workers and private wage earners of medium

15 Thorsten Heien & Dirk Hofäcker: How do welfare regimes influence attitudes? 15 and large enterprises) and the under-protected workers and citizens (those working in weak economic sectors or the informal economy resp. those affected by youth and long-time unemployment). Concerning East Germany, we expect social cleavages resulting from the transformation of the communist welfare regime of the former GDR to the conservative welfare regime of (West) Germany to be more important than possible genuine conflict structures of the communist welfare regime itself. With women and the unemployed we can identify two main losers of the transformation process. Because in the former communist welfare regime of the GDR full-time employment (outside the home) was an element of both male and female ideal life biographies, labour force participation of women was much higher than it is in East Germany today (WEGENER & LIEBIG 1995b: 274). Additionally, due to the tremendous increase in East German unemployment rates (see Figure 2), larger parts of the population were exposed to the risk of being unemployed and therefore dependent on welfare state benefits. We therefore expect attitudinal cleavages between the sexes and between the employed and the unemployed. Furthermore, in East Germany, we can expect some specific socialization effects. Since they focussed on the capitalist experience, it can be questioned whether the hypotheses discussed in section 2.2.3, which assumed higher preferences for an extensive welfare state for women and younger people, make also sense for the East German case. The older age cohorts have built up the GDR, and because of selective migration processes to West Germany the older age cohorts remaining in the GDR can be assumed to be the ones that show the highest identification with the egalitarian socialistic model of society (HEIEN 1997, 1998, WEGENER & LIEBIG 1995b). On the other hand, the younger age cohorts born into the GDR took the socialistic achievements as granted, they realized their deficiencies, and were infected with Western capitalist ideologies by modern mass media. Therefore, if the hypothesized age effect for Western welfare regimes is true, it should work the other way round in East Germany. Similarly, it can be questioned whether the assumed socialization effects hold for East German women because of their higher labour force participation before the reunification Causal model The discussed attitude determinants are summarized in Figure 3. Firstly, cross-country variations in attitudes towards the extent of the welfare state are hypothesized as a result of the macro effects of different national welfare cultures and different economic situations of countries. For a comparison of both approaches, it makes sense to consider different points of time (t 1, t 2, t 3,...) instead of working with cross-sectional data only. Secondly, apart from these macro effects, we expect country-specific rational choice considerations (self interest) and values and norms acquired during socialization processes,

16 Thorsten Heien & Dirk Hofäcker: How do welfare regimes influence attitudes? 16 Figure 3: Determination of attitudes towards the welfare state National welfare cultures & economic situation East Germany Norway East Germany Norway East Germany Norway Italy Italy Italy West Germany West Germany West Germany Great Britain Great Britain Great Britain United States United States United States Differential socialization Extent of the welfare state Differential socialization Extent of the welfare state Differential socialization Extent of the welfare state Self-interest Self-interest Self-interest t 1 t 2 Economic situation t 3 symbolized by different strata resp. models in Figure 3 (one for each country and point of time considered in the analysis), to be responsible for cross-country attitude variations. 3. Data, measurement, and method 3.1. Data The data used in this study are from the International Social Survey Programme (ISSP), which in 1985,, and launched its three Role of government surveys. 22,975 interviews from people from West and East Germany, Norway, Italy, Great Britain, and the United States of age 18 to 74 were gathered altogether. 7 The sixteen sample sizes, East Germany and Norway did not take part in the 1985 survey, vary between 677 (United States 1985) and 2,812 (West Germany ) Measurement Dependent variables The Role of government surveys 1985,, and offer a broad range of indicators of our dependent construct extent of the welfare state. Unfortunately, some of them were not asked at every time and in every country. Therefore, we have to restrict our analysis to six items. These items measure in various ways whether the state should be responsible for ensuring socio-economic security and/or socio-economic equality. More specifically, the respondent s opinion about the statements shown in Table 2 was asked for (abbreviations of items in parentheses). 7 We confine our analysis to this age group because of corresponding restrictions of the Italian and Norwegian data. The amount of non-response ranged from 26.0% (Great Britain 1985) to 58.3% (West Germany 1985) with an average of 42.5%. For further details see ZENTRALARCHIV FÜR EMPIRISCHE SOZIALFORSCHUNG 1985,,.

17 Thorsten Heien & Dirk Hofäcker: How do welfare regimes influence attitudes? 17 Table 2: ISSP items It is the responsibility of the government to reduce the differences in income between people with high incomes and those with low incomes. (EX_INC1) The government should provide a job for everyone who wants one. (EX_JOBS) The government should provide health care for the sick. (EX_HEALT) The government should provide a decent standard of living for the old. (EX_RETIR) The government should provide a decent standard of living for the unemployed. (EX_UNEMP) The government should reduce income differences between the rich and poor. (EX_INC2) Source: ZENTRALARCHIV FÜR EMPIRISCHE SOZIALFORSCHUNG 1985,,. The question remains, whether these six items really measure the same thing. In other words, is there only one single underlying dimension, namely the extent of the welfare state, or do we have to differentiate some attitudinal sub-dimensions? Figure 4 illustrates four theoretically plausible measurement models. Figure 4: Competing measurement models for extent of the welfare state a) EX_RETIR b) EX_RETIR EX_HEALT EX_HEALT Social security EX_UNEMP EX_JOBS Extent of the welfare state EX_UNEMP EX_JOBS EX_INC1 EX_INC1 Social equality EX_INC2 EX_INC2 c) EX_RETIR EX_HEALT Social security 1 d) EX_RETIR EX_HEALT Social security EX_UNEMP EX_JOBS Social security 2 Social equality 1 EX_UNEMP EX_JOBS Labour market EX_INC1 EX_INC2 Social equality 2 EX_INC1 EX_INC2 Social equality 8 A detailed description of all items used, including the answer categories, is shown in Appendix I.

18 Thorsten Heien & Dirk Hofäcker: How do welfare regimes influence attitudes? 18 The first model (a) assumes only one latent construct behind the six items. This implies that people do not differentiate between any attitudinal sub-dimensions. Contrarily, the second model (b) differentiates between social security (EX_UNEMP, EX_RETIR, EX_JOBS) and social equality (EX_INC1, EX_INC2, EX_JOBS) which were discussed in section 2.1 as aspects of the extent of the welfare state. The problem with this solution is that the resulting latent constructs are still measured by indicators of different aspects of the welfare state. The indicators of the social equality construct, for example, refer to questions concerning income (EX_INC1, EX_INC2) and employment (EX_JOBS), although people might perceive them as distinct attitudinal dimensions. Therefore, our third model (c) differentiates between four latent constructs, two concerned with social security and two with social equality. Unfortunately, with only six items altogether this model contains two single-indicator constructs. Because of the necessary technical restrictions (see section 3.3) this implies no advantage compared to an analysis of single items. The fourth model (d) takes into account that the items EX_UNEMP and EX_JOBS are both concerned with aspects of the labour market. Therefore, it is plausible to use them as indicators of a common latent construct labour market (policy). Table 3: Fit of competing measurement models L² (df) p RMSEA SRMR GFI 1-factor model (a) 6, (144) factor model (b) 3, (128) factor model (c) (80) factor model (d) (96) Source: ISSP 1985,, (own calculations). To evaluate the four different models we used confirmatory factor analysis. 9 The results of our analysis shown in Table 3 indicate that the first two models have to be clearly rejected. 10 Both models L² values are extraordinarily high compared to degrees of freedom. Furthermore, the other goodness-of-fit indices reported are far away from their optimum values of 0 (RMSEA, SRMR) and 1 (GFI). 11 In contrast, the overall model fit of the 4-factor model (c) is 9 Data were analyzed with LISREL 8.12 (JÖRESKOG & SÖRBOM 1993) using maximum likelihood estimation. Listwise deletion was used for missing values to get consistent estimators. Because of possible differences in factor loadings and error variances between countries (see section 3.3), we did a sixteen-group analysis without any between-group restrictions except for the same factor structure, i.e., items load on the same factors across groups. 10 Because LISREL was not able to estimate the fit of the 1-factor model, L² values and corresponding degrees of freedom shown in Table 3 are calculated as the sum of 16 separate confirmatory factor analyses. Therefore, other fit indices are not calculated. 11 The L² test statistic tests the null hypothesis that the population covariance matrix equals the covariance matrix implied by our model. The alternative fit indices (RMSEA, SRMR, GFI) adjust some deficiencies of the L² test statistic. In large samples, for example, even trivial deviations of the hypothesized model from the observed covariance matrix will be detected and lead to rejection of the null hypothesis (BAUMGARTNER & HOMBURG, BOLLEN 1989, JÖRESKOG & SÖRBOM 1993).

19 Thorsten Heien & Dirk Hofäcker: How do welfare regimes influence attitudes? 19 satisfactory, although the L² value is statistically significant (L² = ; df = 80; p <.001). But in view of the large sample size (N Total = 17,657), the sensitivity of the L² value to the number of observations should be taken into consideration. The other fit indices, however, are excellent. Nearly the same is true for our last model with only three factors (d). Although its absolute fit is slightly worse than the fit of the previous model and the chi-square estimate of the conditional likelihood test is statistically significant (L² dif = ; df = 16), indicating a better fit of the 4-factor model, we will use the 3-factor model in the following analysis. The reason for this decision is that the only slightly better fitting 4-factor model is very restricted compared to the 3-factor model since working with single-indicator factors, as necessary in the case of the 4-factor model, implies no dimensional reduction Independent variables Compared to our dependent construct extent of the welfare state, the selection of indicators for our independent constructs was even more difficult with the ISSP data. Some aspects were measured with different indicators across countries and some indicators were not consistently collected. 12 Our independent variables include ten interest and socialization indicators. To evaluate the impact of national welfare cultures and the economic situation of the countries we estimate separate models for West and East Germany, Norway, Italy, Great Britain, and the United States for any possible point of time (see section 3.1). Implicitly, these models estimate the effect of the dominant welfare state ideology the respondent has lived with (see section 3.3). Different interest groups of welfare state provisions are identified by two age dummies (18 to 34 years; 35 to 54 years; reference group: 55 to 74 years), which were used to control for nonlinear age effects, a dummy for women, a standardized education score 13, a standardized household equivalent income score 14, three employment dummies (unemployed or short-time workers; old age pensioners; others not in labour force, i.e., helping family members, housewives resp. housemen and students reference group: full-time and part-time employed) and by a dummy for self employment. 12 In view of the attitude determinants discussed in section 2.2, especially the absence of indicators of the individual s perception of its own social position and of employment in the public sector is deplorable. 13 To get an international comparable measure of education, in a first step country specific educational classifications were transformed to formal length of schooling in years (ZENTRALARCHIV FÜR EMPIRISCHE SOZIALFORSCHUNG 1995). In a second step, the ratio of a respondent s schooling in years to the mean schooling in years for a given country (and point of time) was calculated. 14 The income score results from a multi-stage procedure. Because the extent of missing data for household income was in some cases very high (especially in West Germany), we used hot-deck-imputation methods, with the respondent s individual income, age, sex, and household size as predictors of household income, to impute missing values (LITTLE & SCHENKER 1995). To calculate household equivalent income, the following weights were used: 1 for the household s first adult person and 0.7 for any other person. Finally, to get an international comparable measure of income, the ratio of a respondent s household equivalent income to the mean household equivalent income for a given country (and point of time) was calculated.