European Societies ,» 7(2) 2005: 287-319 1^ ™?u,t'ed9^ © 2005 Taylor & Francis Group Ltd ISSN 1461-6696 print 1469-8307 online jj! % Taylor & Francis Group DOES TRANSITION MATTER? Bringing Welfare Attitudes into the Debate Steven Saxonberg Uppsala University, Department of Government and Dalarna University College, Sweden ABSTRACT: This article compares the socio-economic determinants of welfare attitudes in the Czech Republic to those in Sweden, using survey data from the 1996 ISSP survey 'The Role of Governments'. Many theorists of the transition have claimed that the post-communist countries have a different political dynamic that their west European neighbors. For example, David Ost claims that that citizens of post-communist countries during the 1990s were not sure of which class they would belong to and therefore, were not sure of their class interests. Similarly, Zagorski claims that in post-communist countries educational level becomes one of the most important determinants of welfare and socioeconomic attitudes, because the reform process is very complicated. Those with higher levels of education can better understand the complexities of the reforms and are more willing to accept short-term disadvantages for long-term gains. However, many experts have also claimed that the Czech Republic presents an exception to these trends. It is the one country in which party-competition is based on socioeconomic issues and voting is class-based. This study tests these three hypotheses, to determine whether the Czech Republic really does provide an exception to the general post-communist development. Key words: welfare attitudes; class; transition; post-communism; Czech Republic; Sweden Introduction One of the most important debates arising out of the post-communist transition is whether the political dynamics of the former Soviet bloc countries are approaching Western Europe or are different patterns of political behavior emerging. This article tests three of the most interesting hypotheses using empirical survey data. In addition, it fills an important gap in the transition literature by testing these hypotheses for the often-neglected field of welfare attitudes. Welfare attitudes are significant, because welfare issues play a central role in contemporary politics. In fact, DOI: 10.1080/14616690500083477 287 EUROPEAN SOCIETIES a number of social scientists claim that in the West class struggles have basically shifted from confrontations over ownership of the means of production to political battles over welfare policies. Since public ownership has been particularly discredited in the post-communist countries, we would expect this statement to also hold true for these countries. This study brings welfare attitudes into the debate by comparing the Czech Republic to Sweden. Thus, this article also represents one of the first attempts at testing these hypotheses by examining more closely whether the dynamics of a post-communist country resemble those of a relatively similar Western country (Sweden). The focus is on the Czech Republic, because it represents a 'critical case'. Much previous research on the post-communist countries has found that the Czech Republic is the one country, which has rather similar political dynamics to west European countries. For example, several studies have claimed that it is the only post-communist country in which class is a strong explainer of attitudes and voting. Moreover, party competition is based on socioeconomic issues (cf Cotta 1996; Machonin 1996; Markowski 1997; Antoš 1998; Večerník 1998: ch. 10; Kitchelt et al. 1999; Matějů 1999; Řeháková 2000a; Krause 2000). Thus, if this study refutes these claims of Czech exceptionahsm, it greatly increases the validity of those theories, which claim that the post-communist countries have different dynamics than west European countries. In particular, this study would then support David Ost's (1993, 1995) hypothesis that during the post-communist transformation, workers do not know their own class interests, because classes are in flux. Since Ost developed his theory mostly based on the Polish case, this article could give his theory greater validity by showing that classes were actually even in flow in the one country, for which many theorists have claimed otherwise. Even if this study were not to support the class-in-flux hypothesis, it could also question the claims of Czech exceptionahsm if it supports Zagórski's (1994) hypothesis about understanding complexity. According to this hypothesis, educational level matters more in post-communist countries than in others, because one must be well-educated to understand the complexity of the social and economic reforms. In contrast, less educated people living under stabile, established market economies are better able to understand the implications of socioeconomic policies. 1. See, for example, Esping-Andersen (1990), Korpi (1981) and Ginsburg (1992), and Svallfors (1997: 290), although Korpi originally believed that economic democracy would become a major issue in the next step of political development in Sweden. 2. As Prime Minister, Klaus talked about the Czech Republic being an 'exceptional' case. For a discussion, see Vachudová (2001) 288 Bringing Welfare Attitudes into the Debate SAXONBERG Theoretical Discourse So far it has been quite rare to compare welfare attitudes between western and post-communist countries. One exception is Renwick and Tóka (1998), who compare attitudes on many issues among EU countries and several of the central and east European countries planning to enter the Union. They basically look at aggregate attitudes and do not perform statistical tests of socioeconomic variables that could explain welfare attitudes. Thus, they do not test any of the hypotheses from this present study. In another study, Andreß et al. (2001) compare eastern Germany, to western Germany, Norway and the UK. They find that East Germans are the most positive toward welfare policies and that educational level is a significant explainer of welfare attitudes for all four countries. However, the former East Germany is a special case, because of the dynamics involved in unification with West Germany. Thus, it is not clear whether findings about the former East Germany are applicable to other post-communist countries. In addition, since they do not test for class, their study does not test the class-in-flux hypothesis, which is one of the main purposes of this present article. The class-in-flux hypothesis claims that at least in relatively homogenous societies, we would normally expect workers to have some idea about their class interests and to organize around them. Lipset (1963: 230) sums up this view in his classic Political Man: "In every modern democracy conflict among different groups is expressed through political parties which basically represent a 'democratic translation of the class struggle'". Similarly, Ingelhart (1997: 240) exclaims: 'for most of the twentieth century, the dominant axis of political cleavage was the Left-Right polarization based on economic issues, with the working class supporting the Left and the middle class supporting the Right'. To be sure Ingelhart (1997: 254-5) shows that class-based voting is on the decline in Western societies as post-materialist values increase. Nevertheless, his statistics show that the percentage of working-class voters supporting Leftist parties is 20—30 per cent higher than the percentage of middle-class voters supporting such parties among the countries, which Lijphart (1984: 43; 1977: 15, 110-11) labels 'non-plural', and 'culturally homogeneous' (such as Great Britain and Sweden). If the post-communist communists then have the same dynamics as western countries, we would expect class to be the most important political cleavage among the relatively homogenous countries. However, David Ost (1993, 1995) finds in his studies of Poland that during the transition to a market economy, people are not sure of their class status, since classes are flowing. Upward and downward mobility is quite high and most people 289 EUROPEAN SOCIETIES are not sure, who will succeed in joining the new and upcoming entrepreneurial class. Under these conditions, labor unions are also uncertain of what policies are in the workers' interests. During the first years of non-communist rule, for example, the Solidarity union was confused over its role in the transition to capitalism. Since union leaders believed that market reforms would raise living standards, they supported laissez-faire policies and opposed worker influence over the production process. This market liberal orientation confused workers and cost the unions much of their support. Thus, the traditional Western model of class interested has not yet emerged. Other theorists also note the inability for post-communist unions to help worker formulate their distinct class interests. Herod (1998: 206) argues that unions in the transitional period find themselves in 'contradictory positions' of claiming to defend workers' interests and calling for privatization and market reforms at the same time. These very market reforms, however, will likely cause loss of jobs and reduction of real incomes as prices rise sharply for basic needs, such as food and housing. Offe (1994: 120ff) adds that membership in the kinds of interest groups and political parties that could represent class interests are also in fluctuation during the transitional period, which makes Western-styled interest politics more difficult. Meanwhile, Zagórski (1994) posits another hypothesis about the uniqueness of the post-communist countries, by claiming that educational level matters more for citizens of post-communist countries than for comparable western countries. He claims that since economic and social-welfare reforms are complicated and difficult to understand, the ability to understand the complexity of these reforms becomes more important for voting and forming political attitudes than concerns of self-interest, such as class interests or income. If education turns out to be insignificant in the Czech case or a much weaker predictor of welfare attitudes than in Sweden, this study gives support to Zagórski's theory of the less highly educated not understanding the complexity of social and economic reforms. In contrast to these theories of the uniqueness of post-communist countries, many scholars have claimed that the Czech Republic comprises an exception to the post-communist trend. They claim that the Czech Republic differs from its post-communist neighbors, because voting and political attitudes are in fact class-based and party competition is based on socioeconomic issues (cf Cotta 1996; Machonin 1996; Markowski 1997; Antoš 1998; Kitchelt et al. 1999; Matějů 1999; Krause 2000; Řeháková 2000a; Večerník 1998: ch. 10). Thus, Antoš (1998: 81-94), Machonin (1996) and Řeháková (2000a) all present statistics showing that workers are more likely than other groups to vote Left. Matějů (1999: 32) also admits 290 Bringing Welfare Attitudes into the Debate SAXONBERG that class influences voting in the Czech Republic, but not in Poland and Hungary. However, he adds that subjective socio-economic mobility provides a stronger predictor of voting in all three countries. Not only does class play an important role for voting in the Czech Republic, scholars have also claimed that the country is unique among the post-communist countries in that socioeconomic issues comprise the main political cleavage. For example, Kitschelt et al. (1999: 169) write in comparing the Czech Republic with Poland, Hungary and Bulgaria: Politicians in the Czech Republic signal the greatest capacity to engage in programmatic competition. They discriminate in the sharpest fashion among more or less salient issues and their judgements of party positions on the most salient economic issues indicate considerable programmatic crispness. Markowski (1997: 229) argues 'Czech politicians manage to divide the party spectrum clearly between pro-and anti-market forces'. Cotta (1996) and Krause (2000) agree, while Krause (2000: 36) adds that in contrast to Slovakia, where nationalist issues predominate, 'party choice in the Czech Republic involves primarily economic and socio-economic questions'. So far, very little has been written about the relationship between class and welfare attitudes in the Czech Republic. Nevertheless, Rabušic and Sirovátka (1999: 259) conclude that in the Czech Republic both subjective social class and party preferences are significantly correlated with attitudes toward welfare policies. Similarly, Matějů (2000: 175) shows that owners of their own firms (i.e. the class of self-employed) are more likely to oppose equality in the Czech Republic. These findings create a counter-hypothesis to Ost's (1993, 1995) and indicate that although classes might be flowing in most post-communist countries, the Czech Republic provides an exception. However, both of these studies have some limits in their measurement of class in a manner that prevents an adequate of the class-in-flow hypothesis. Rabušic and Sirovátka (1999) use the respondents' subjective class placement. This raises problems of tautological reasoning. If manual laborers are not sure of their class position, then they might answer that they belong to the 'middle class'. They would probably choose that categorization because they identify more with the middle class than their own class and thus have middle class values. Consequently, we could expect a higher correlation between subjective class-placement and attitudes than between objective class position and attitudes. In another study, Matějů (2000) examines attitudes toward distributional justice in the Czech Republic. However, among the explanatory variables, he limits class to the self-employed. This does not either provide a test of Ost's theory, because Ost (1993, 1995) never claims that the new 291 EUROPEAN SOCIETIES 'bourgeoisie' are unaware of their class interests, rather he emphasizes the case of workings being uncertain of their interests. Besides hypotheses about class, this study also tests a hypothesis about the special role of education for the post-communist countries. Zagórski (1994) argues in his study of Poland that educational level is particularly important for determining socioeconomic attitudes among post-communist countries. He claims (1994: 363): 'During the transformation period, the level of education that enhances the intellectual capacity to understand complex social and economic processes is a much more important determinant of support for the transformation than other dimensions of socio-economic status'. Thus, if Zagórski's hypothesis is correct then we should expect educational level to matter more for the post-communist Czech Republic than for western countries. So far, previous studies seem to indicate that educational level does matter for welfare attitudes in the Czech Republic. For example, Řeháková (2000b) finds that educational level influences attitudes toward inequality in the Czech Republic, while Matějů (2000) finds that Czechs with a college education are more skeptical of distributive justice than those without a college education. However, none of these studies have compared the influence of educational level between post-communist countries and west European countries, which means that still no serious test of the understanding-complexity hypothesis has been undertaken. If this study repeats these previous findings for the Czech Republic, then we can be more certain that educational level really matters more for post-communist countries, since the better educated are better able to understand the complexity of social and economic reforms. However, these previous studies all have the drawback that they do not compare the role of education in forming attitudes between post-communist and west European countries. Since previous studies of west European countries also show that educational level is important, we can provide a much more rigorous test of the understanding complexity hypothesis by comparing the Czech Republic to a comparable western country (Sweden). In summary, this article can either give greater validity to the claims of Czech exceptionalism, by showing that class continues to strongly influence attitudes in the Czech Republic even hold up when applied to the area of welfare attitudes and even in comparison to a relatively similar western country. Or, in contrast, this article can support the universal validity of general theories of post-communist transition, by showing that Ost's theory of class-in-flux and Zagórski's theory of understanding complexity even apply to the one country, which is least likely to support their claims. 292 Bringing Welfare Attitudes into the Debate SAXONBERG Variables in the Study In order to test these hypotheses about classes being in flow, Czech exceptionalism and understanding complexity, it is necessary to formulate the hypotheses in terms of measurable independent variables. The three variables are: CLASS, INCOME and EDUCATIONAL LEVEL. (1) CLASS If class turns out to either not be a significant explainer of attitudes in the Czech Republic or if it turns out to be a much weaker explainer of attitudes than in Sweden, this study will support the class-in-flux hypothesis, while refuting the hypothesis of Czech exceptionalism. This study follows Ahn (2000) in using a Marxian division of classes into three groups: workers ( = 1), professionals ( = 2) and bourgeoisie ( = 3). It does so for several reasons. First, this article tests Ost's hypothesis about classes being in flow and his hypothesis is based on the Marxian conception of class. Second, most authors claiming that Czech political dynamics are class-based have also used Marxian definitions of class or have at least used 'workers' or 'entrepreneurs' as dummy variables (for example, Matějů 2000; Řeháková 2000a, b). Since the main protagonists of the debate have used a Marxian definition of class, then it is best to use the same definition as them in order to test their hypotheses. This article follows Ahn (2000) in also including the spouse's class, because we can assume that spouses are also influenced by each other's class position. For example, we could expect a woman without a university education, who works part time at a manual job while taking care of the children, might still take on some of the values of the 'bourgeoisie' if her husband owns a large enterprise. Empirically, the results are again much better for each country if the spouse's class is included. An alternative would be to use the Erikson-Golthorpe division of classes, which social scientists often use in the Western discourse on welfare attitudes (i.e. Svallfors 1993, 1997, 1999). While the Marxian approach lumps all workers together into one class, the Erikson-Golthorpe approach divides workers, for example, into non-manual labor, skilled 3. The explained variance of CLASS on PRO-WELFARE ATTITUDES decreases from 20.3% to 5.3% is Sweden when the spouse's class is eliminated from the model. In the Czech case, the explained variance of CLASS remains the came when the spouse's occupation is eliminated. However, when CLASS is the only independent variable and both TAXPAYER and EDUCATION are eliminated, then CLASS explains 20.3% of the variance in PRO-WELFARE ATTITUDES, compared to only 1.2% when the class of the spouse is eliminated. 293 EUROPEAN SOCIETIES labor and routine non-manual labor. However, since neither Ost nor most of the authors writing about the class determinants of post-communist voting make these differentiations, a definition of class that separates these groups would not provide a test of their theories. Furthermore, when the Erikson-Goldthorpe definition is used, the results become much worse for the two countries. Another advantage of using the Marxian definition is that it allows the inclusion of the spouses' class. The most commonly used algorithm developed by Ganzeboom and Treiman (1996) excludes this possibility, because the spouse's occupation is used when the respondent's occupation is unknown. Finally, one could use a Bourdieuian definition of class that included such elements as social and cultural capital. However, since none of the social scientists engaged in the debate on welfare attitudes in general or post-communist welfare attitudes and voting in particular use such a definition of class, one cannot test their hypotheses using Bourdieu's definition. The breakup of classes according to the Marxian definition for the Czech Republic becomes 51.9 per cent workers, 27.1 per cent professionals and 21 per cent bourgeois. For Sweden the statistics are 34 per cent workers, 54.8 per cent professionals and 11.2 per cent bourgeois. The fact that the Czech Republic had a greater percentage of entrepreneurs and a much lower portion of professionals probably reflects the nature of the transitional economy, in which the public sector is rather poor and unable to pay professionals high salaries. Thus, many Czechs with professional backgrounds might have concluded that they could earn more money by starting small private businesses. (2) INCOME If classes are in flow and post-communist citizens are not sure which class they belong to or what their class interests were, they were at least likely to know what their incomes are. Furthermore, if they know their income, they will also likely to know approximately how much of their income will 4. In Sweden, in the full model the Marxian definition of class including the occupation of the spouse explains 20.2% of the variance of welfare attitudes, compared to 12.3% for the Erikson-Golthorpe definition. In a forthcoming article on Swedish welfare attitudes, I show that when other independent variables are used, the difference between the explained variable of Marxian and Erikson-Goldthorpe class definitions are even greater. As will be seen for the Czech case, class is such a poor explainer of welfare attitudes that its standardized coefficient would be extremely low, regardless of which definition of class one uses. The explained variance of class drops from .16% to 0.01%. 294 Bringing Welfare Attitudes into the Debate SAXONBERG be taken away from them in taxes. Thus, they still have interests as taxpayers. Even thought they might be able to find loopholes in the tax laws, they still might be afraid of potentially having to pay high taxes. Pettersen (2001: 29) terms this hypothesis the 'burdened taxpayers'. According to this theory, higher income groups feel burdened by high taxes, but have enough income security that they do not feel they will benefit much from welfare programs. Previous studies have found that income significantly influences welfare values or support for income inequality for Western countries (Svallfors 1993, 1995, 1997; Edlund 1999, 2000; Ahn 2000). For post-communist countries, Tworzecki (2000) shows in his study of Poland and Hungary during the years 1992-1994 that having a high income was statistically significant for both countries. (3) EDUCATION In order to measure Zagórski's hypothesis about education being important for understanding the complexity of the social and economic reforms, it is necessary to include the variable level of education. If the understanding complexity hypothesis is correct, then we would expect support for welfare policies to be statistically significant and negatively correlated with educational level. That is, those with higher education would be more likely to support the economic reforms that have often brought about cutbacks in welfare benefits (such as the cut-off of support for nursery schools, the introduction of private healthcare insurances, etc.). In addition, in order to confirm Zagórski's hypothesis, educational level must be a clearly greater explainer of welfare support in the Czech Republic than in Sweden. For if no large difference exists between the two countries or of the relationship between education and welfare support is higher in Sweden, then one cannot conclude that educational level plays a greater role in shaping public opinion for post-communist countries than for comparable western countries. Methodology This study uses confirmatory factor analysis and structural equation modeling (SEM). Confirmatory factor analysis has three main advantages. First, it allows one to test for multi-dimensionality Most previous studies of welfare attitudes have applied Cronbach's oc reliability test to decide which variables should be in the scale (i.e. Hansenfeld and Rafferty 1989; Svallfors 1993, 1997, 1999; Edlund 1999, 2000; Sundström 1999). Since 295 EUROPEAN SOCIETIES this method does not allow one to test for multi-dimensionality, they just assume that only one dimension exists. Exploratory factor analysis can partially solve this problem, because in contrast to Cronbach's a it can also test for multi-dimensionality. However, it is rarely used in the research on welfare attitudes, (for exceptions, see Svallfors 1996; and Rose and Makkai 1995). Although exploratory factor analysis presents an improvement over Cronbach's a, it is still problematical, because it is inductive and data driven, since it uses computer output as its starting point. In contrast, confirmatory factor analysis has the advantage of being deductive and theory driven, because it starts out by testing theoretical premises (Bollen 1989: ch. 7; Jöreskog and Sörbom 1993: 22; Kline 1998: ch. 3; Maruyama 1998: ch. 7). A third advantage of confirmatory factor analysis is that besides testing for multi-dimensionality and being theory driven, it can also allow for covariation among the factor indicators. However, in this particular study it is possible to create satisfactory measurement models of welfare attitudes without having any items covary. Next structural equation modeling (SEM) is used. SEM is the most suitable method for applying path analyis for models that include latent variables. This is especially true when one of the latent variables (PRO-WELFARE ATTITUDES) has several dimensions. Diagram 1 shows the basic theoretical model of this study, where INCOME, EDUCATIONAL LEAVE and CLASS explain PRO-WELFARE ATTITUDES and PRO- Figure 1. Theoretical model. 296 Bringing Welfare Attitudes into the Debate SAXONBERG WELFARE ATTITUDES might have several dimensions. SEM also allows us to bring in political behavior into the equation by making it possible to simultaneously test for the relationship between explanatory variables and welfare attitudes and the relationship between using these attitudes and voting for political parties that share these attitudes. In this study, VOTING is measured by the voting for leftwing parties (the social democrats and leftists in Sweden and the social democrats and communists in the Czech Republic). Voting is interesting, because previous studies have found that those, who support generous welfare policies, are more likely to vote for leftist parties, since such parties are more likely to implement generous welfare policies (Roller and Westle 1987; Edlund 1999, 2000; Forma 1999; Forma and Kangas 1999; Knudsen 2001). Thus, if Czech politics is really following western political dynamics, then according to the Czech exceptionalism hypothesis, we would also expect a strong relationship between welfare attitudes and voting in the Czech Republic. For this country, Sirovátka (2002) and Rabušic and Sirovátka (1999) find strong ties between having pro-welfare attitudes and voting for the social democratic and communist parties. These previous studies, however, suffer from the major drawback that they used voting as an independent variable, that explains welfare attitudes. Technically, this is understandable, because in normal regression analysis only one path is used, so voting becomes one of many independent variables that explain welfare attitudes. Theoretically and logically, however, this is very problematic, because it assumes that voters first decide which party they want to vote for and then decide what attitudes they should have. It is more logical and theoretically reasonable to assume that voters first develop certain attitudes and then vote for the parties that are most likely to implement the policies that correspond their preferences. Now that the theoretical SEM model has been presented, some words are necessary about the means of testing it. The vast majority of social scientists use various closeness-of-fit measures for determining the best model. Even though some controversy exists about what the minimum level of fit should be or which tests provide the best measures, one should remember that at least SEM modelers are forced to pay close attention to model fit. In contrast, the mainstream literature on welfare attitudes 5. It should be noted that in using a dichotomous variable, arguments could be made for using a tetrachoric correlations, which unfortunately, AMOS 4 is not capable of doing. However, it is not clear that using this method would improve the results. The simulation study by Yung and Bender (1994) indicates that that the sample size should be at least 2000, and possibly 5000, to obtain satisfactory results, while in the present study the sample size is 1238 for Sweden and 1100 for the Czech Republic. 297 EUROPEAN SOCIETIES normally uses traditional regression analysis and pays little or no interest overall model fit. Thus, it is common to publish articles where the model R is well below 0.05. This article will present models that fit the most basic closeness of fit criteria, in that both the goodness of fit index (gfi) and the adjusted goodness of fit (agfi) are > 0.9 and the root mean square error of approximation (RMSEA) <0.08. All the SEM-modeling in this article has been done using maximum likelihood calculations in AMOS 4. Although AMOS has a very advanced and highly acclaimed method for dealing with missing values, this method has the problem of only dealing with raw data, which means that the GFI and AGFI scores cannot be calculated. For this reason correlation matrixes were exported from SPSS using pairwise deletion for the missing variables. It should be noted, however, that all of the other model-fit results improved greatly when using raw data and the AMOS method for dealing with missing variables. For the confirmatory factor analysis, is does not matter much, since when calculating from correlation matrixes, the measurement models for both countries easily meet all closeness-of-fit criteria and even meet the much more stringent chi-square test (of P > 0.05). For the structural models, however, the difference is more important, because although the modified models for both countries meet the demand that rsmea < 0.08, Sweden barely misses the more stringent criteria of rsmea < 0.05, but it easily meets this criteria when using raw data and AMOS missing variable method. However, none of this affects the substantive results of this study. When conducting the chi-square difference tests, the raw data is used, since GFI and AGFI scores are not calculated for this test. Data Base This article uses the 1996 International Social Survey Programme (ISSP) survey 'The Role of Government'. Along with the World Value Surveys, the ISSP surveys are the most extensive in the world, with 26 countries participating. As one of the two most comprehensive and best known international surveys, its results are relatively reliable and little doubt exists about the methodology of its surveyors. Moreover, the participation of another 24 countries means that the results of this article can be tested 6. For example, in Svallfors' 1999 study (1999: 109) the explained variance (IT) for the six countries studied ranges from 4.2% to 19.3%. In a study only of Sweden, Svallfors (1995: 66-7) presents R s ranging from a mere 1.2% to 13.7%. In 16 of 24 cases his models explain less than 5% of the total variance and in 10 of 24 cases they explain less than 3%. 298 Bringing Welfare Attitudes into the Debate SAXONBERG on many more cases. Many questions in this survey concern attitudes toward social policy, making it the most comprehensive cross-national survey available. However, this article restricts itself to two cases, in order to focus on the Czech Republic as a critical case. A study involving many countries would likely become a more inductive study that focuses on unexpected patterns that emerge, rather than trying to clearly discuss whether the Czech Republic presents an exception to previous studies of post-communist studies or whether it confirms the more universalistic claims of Ost and Zagórski about the unique nature of the post-communist transition. The ISSP survey from 1996 is the only data base so far that includes questions on welfare policy, in which both the Czech Republic and West European countries participate. Since data only exists for one year, it is not possible to run any time-series analysis of welfare attitudes for the Czech Republic. However, the ISSP is planning to replicate this survey in 2006, so eventually time series data will be available. The fact that data is only available for these two countries for 1996 should not provide any problem for our present purposes. First, virtually all of the literature on welfare attitudes uses static analyses to determine the socioeconomic variables that influence welfare attitudes. When data over time has been used, it has almost always been limited to descriptive statistics. As a result no testable models of welfare attitudes have been developed that could use time-series data. Secondly, in a sense 1996 provides an optimal period for examining the Czech Republic. Although Prime Minister Klaus was starting to lose some popularity, he was still one of the country's most popular politicians and was re-elected that year. Unemployment was still low at under 4 per cent; Klaus had declared that the country had already completed its transformation to a market economy; and the country was still widely considered the model of successful post-communist transformation. One year later, much of this was to change. The country entered into an economic downturn; unemployment grew rapidly; and the mass media began criticizing Klaus and his party ODS for corruption scandals. By the end of 1997, he was forced to resign. The following summer a minority social democratic government came to power, which failed to build a coalition and thus signed a controversial cooperation agreement with its greatest enemy: the ODS (Saxonberg 1999a, 1999b, 2003). Thus, 1996 represents the year in which many Czechs probably thought that they had come closest to becoming a 'Western' type of democracy. If transition does not matter, then in 1996 Czechs probably had attitudes and voting behavior that was closer to similar homogenous, non-fragmented West European societies than in the following years, which were marked more by an atmosphere of crisis and uncertainty about the transformation. 299 EUROPEAN SOCIETIES In the Swedish case, 1996 represents a period in which political life began to return to 'normalcy'. Although unemployment had ranged from 1 to 3 per cent during the 1980s, the 1990s began with an economic crisis, which brought open unemployment levels up to over 8 per cent (SCB: 2003). As the economic crisis began the social democrats lost the 1991 elections, which was only their third electoral defeat since coming to power in 1932. Despite three years of center-right rule from 1991 to 1994, the cutbacks in social insurances during the 1990s were comparatively mild (Bonoli et al. 2000). From a gender perspective, Sweden even made some clear advancements in the 1990s, such as the introduction of a 'pappa month' to the parental leave insurance as early as 1992 and increased support for daycare (cf Bergkvist 1999). By 1996 the social democrats had already returned to power and the economic crisis had already reached its height and was soon to radically decline. Thus, citizens had reason to become more optimistic about the future and the crisis atmosphere of previous years had already begun to disappear. Studies of welfare attitudes indicate that throughout the 1990s the relationship between voting and welfare attitudes remained stable in Sweden, as has the relationship between socioeconomic variables and welfare attitudes (i.e. Edlund 2000). Choice of Countries Since this article tests the hypothesis as to whether a homogeneous, post-communist country has similar political dynamics as a more established Western European democracy, it is best to choose the Western country that is as similar as possible to the post-communist country. This method corresponds basically to 'the method of difference' (see, for example, Skocpol and Somers 1980). That is, the two countries are as similar as possible, but they have different outcomes if one independent variable is much different. In the present case, the independent variable is the post-communist transition. If being a transitional country matters, then either Ost's hypothesis of classes-in-flux or Zagorski's hypothesis of the importance of educational level will be confirmed. That is, if Ost's hypothesis is correct, then class will be a better explainer of welfare attitudes in Sweden than in the Czech Republic. Meanwhile, if Zagorski is correct, then educational level will be a better explanation of welfare attitudes in the Czech Republic than in Sweden. This would also give further evidence again the argument of Czech exceptionalism. Sweden is the West European country that is the most comparable to the Czech Republic for several reasons. First, these are both rather homogeneous and non-pluralist countries, with class as the main cleavage. Thus, they are more likely to have class-based politics than pluralist 300 Bringing Welfare Attitudes into the Debate SAXONBERG societies. For example, Lijphart (1977: 3-4) lists six 'segmental cleavages' that can cause a country to become plural: (1) religious, (2) linguistic, (3) regional, (4) cultural, (5) racial or (6) ethnic. Since none of these cleavages exist to any large extent in these two countries, then class is clearly the main social cleavage and these societies are homogeneous rather than plural. Some minor exceptions do exist. For example, Sweden has a growing immigrant population Sweden and in the Czech Republic, some tensions have arisen with the Roma minority. However, in neither country have any racial or ethnic cleavages emerged in the political system, although small xenophobic parties in each country came into parliament for a few years, before disappearing into obscurity. Thus, as in the Czech Republic, previous studies have found that in Sweden class still provides a robust explanation for both voting (Gillijam and Homberg 1995; Holmberg 2000; Holmberg 2001; Knutsen and Kumlin 2003) and for welfare attitudes (Ahn 2000; Edlund 2000; Svallfors 1993, 1996, 1997, 1999). Consequently, class clearly represents the main cleavage for both countries. Second, Sweden is clearly more similar to the Czech Republic than any other west European countries. For example, Great Britain is also basically dominated by a class/socioeonomic cleavage. Nevertheless, Sweden still compares better to the Czech Republic, because both Sweden and the Czech Republic have relatively small populations (about 9 million inhabitants in Sweden and approximately 10 million in the Czech Republic, compared to around 60 million in the UK). Furthermore, while Great Britain has a strong liberal cultural legacy that is common for Anglo-Saxon countries, both the Czech Republic and Sweden have relatively egalitarian cultures (see Holy 1996; Wolchik 1991 for the Czech Republic; and Heclo and Madsen 1987 for Sweden). Finally, Great Britain has a majoritarian electoral system, while both Sweden and the Czech Republic have proportional systems, which means that this difference in electoral systems might conceivably cause a different relationship between attitudes and voting in Great Britain than in Sweden and the Czech Republic. Even though Sweden and the Czech Republic both comprise rather homogeneous societies, with similar electoral systems and comparable population sizes, it has been more popular to compare the Czech Republic with pluralist societies in Austria (for example, Raven 1976) and the Netherlands (Sirovátka 2000; Sirovátka et al. 2002). However, as Lijphart (1984: 42ff and 1977: 73-74) observes, the pluralist Netherlands has a strong cleavage between Protestants and Catholics, while Austria has a strong cleavage between Catholics and non-practicing Catholics, which mean that class is a less important political factor in these pluralist countries than in homogeneous Sweden. Recently, an ethnic/national 301 EUROPEAN SOCIETIES cleavage has also emerged in Austria, as Haider has transformed the Freedom Party into a large xenophobic, populist party. Thus, homogenous Sweden provides much more comparable to the Czech Republic than pluralist Austria and the Netherlands. Of course, in the real world two countries are never completely similar and every country has its special characteristics, the question then becomes whether these differences are important enough to influence the results. The two major criticisms for choosing Sweden could be that the Social Democratic Party has dominated the country so greatly that it makes Sweden a special, incomparable case. The second criticism could be that Sweden is not comparable to the Czech Republic because it is much wealthier. Sweden is unique in that the social democrats have continuously been in power since 1932 for all but 9 years (1976-82 and 1991-94). This could conceivably influence welfare attitudes, as it is generally assumed that universalistic, social democratic welfare policies increase support for welfare programs. However, this might actually make Sweden more comparable than other Western countries to the Czech Republic, since the communist regimes also provided extensive benefits. Not surprisingly then, surveys of welfare attitudes generally show very high support for generous welfare polices among post communist-countries (see, for example, Nadas 1993; Tworzecki 2000), even though Czechs are not always the most supportive of welfare benefits among post-communist citizens. For example, Verčerník (1996: 211) writes that in 1994 Slovaks were slightly more dissatisfied than Czechs with social benefits (83 compared to 79 per cent). Nevertheless, this generally high level of support for welfare policies in the Czech Republic implies that just as we would expect Swedes to be supportive of generous welfare policies after decades of social democratic rule, we can also expect Czechs to be supportive of generous welfare policies after decades of communist rule. Even though this article argues that Sweden provides the best comparison to the Czech Republic previous studies indicate that it does not really matter which west European country one compares to the Czech Republic. Those studies show that the same socioeconomic variables used in this present study are generally good predictors of welfare attitudes among all or most western countries (Coughlin 1980; Svallfors 1993, 1999; Blomberg et al. 1996; Bean and Papadakis 1998; Edlund 1999; Andreß and Thorsten 2001; Pettersen 2001) and that a strong connection exists between pro-welfare attitudes and leftist voting (Roller 1987; Edlund 1999; Forma and Kangas 1999; Listhaug and Aalberg 1999). In other words, even though Sweden provides the best comparison with the Czech 302 Bringing Welfare Attitudes into the Debate SAXONBERG Republic, the studies cited above indicate that the results would not change significantly even if another Western country were chosen. The Measurement Model of Welfare Attitudes Now that the theoretical model has been presented, as have the independent variables, the methodology and choice of countries, the empirical study can begin by stipulating the measurement model for welfare attitudes. Once we know how welfare attitudes will be measured, we can develop the full structural model. As already noted, most studies of welfare attitudes have used one-dimensional scales based on Cronbach's reliability test. This article tests whether welfare attitudes really only have one-dimension or whether they might have two dimensions: one for support for a BIG PUBLIC SECTOR and one for supporting EQUALITY. Esping-Andersen (1990) argues that welfare regimes have two-dimensions. Liberal regimes generally spend less on welfare than social democratic or conservative-corporatist welfare regimes. However, a second dimension arises, because conservative-corporatist welfare regimes can be as generous as social democratic welfare regimes concerning spending levels and governmental responsibility for social programs. In contrast to social democratic regimes, though, they aim to maintain or even strengthen inequalities. For example, conservative welfare states traditionally support gender inequality by encouraging women to stay at home. Even though Esping-Andersen writes about welfare regimes rather than welfare attitudes, empirical studies of welfare attitudes also indicate that an EQUALITY dimension exists, along side support for a BIG PUBLIC SECTOR. For example, Aalberg (1998) finds that support for income equality is much higher than average for social democratic Sweden and Norway, while support for welfare programs and increased spending on these programs is not higher than other countries. For measuring support for BIG PUBLIC SECTOR, questions are chosen about government responsibility for providing housing and healthcare, as well as providing jobs. In addition, questions are included about attitudes toward increased spending on unemployment benefits, pensions, education, healthcare and job creation. EQUALITY is measured in terms of support for income redistribution, wage control and price control and supporting higher taxes for those with high and medium incomes. All the questions are based on five-point scales except the ones on governmental responsibility, which have four-point scales. These differences in scales affect the unstandardized coefficients, but not the 303 EUROPEAN SOCIETIES standardized ones. All the questions are uniformly recoded, so that the highest score denotes support for welfare policies. The aggregated responses to these questions are summarized in Table 1. As Table 1 shows, although Simkus and Robert (1995)—96) find that Czechs are less supportive of equality than Westerners, in 1996 Czechs were generally more favorable toward income equality and a big public sector than social democratic Swedes. Unemployment benefits and housing provide exceptions. However, the actual levels of support for welfare policies are not central to testing the three hypotheses of this article. Instead, the hypotheses are about the relationship between explanatory variables and welfare attitudes and the relationship between welfare attitudes and voting. Nevertheless, even at this descriptive level we find some support for the class-in-flux hypothesis. Table 1 shows that although Czechs are generally much more favorable than Swedes toward welfare programs and welfare spending, they are much less willing to raise taxes to support these programs. Whereas 62.4 per cent of the Swedish respondents think that taxes are too low for those with high incomes, only 44.8 per cent of Czechs believe so. If Czechs were more conscious of their interests, they would be more likely to see the connection between BIG PUBLIC SECTOR and the need to raise taxes to support increased spending on such programs. Now that the descriptive data has been examined, it is necessary to create a measurement model for measuring welfare attitudes. Confirmatory factor analysis confirms the hypothesis that PRO-WELFARE ATTITUDES has two dimensions: EQUALITY and BIG PUBLIC SECTOR. To create a good-fitting measurement model with the same indicators for both countries, it is necessary to eliminate several of the survey questions. Thus, EQUALITY is measured by support for higher taxes for those with high incomes (TAXHI) and support for redistribution (REDISTR). BIG PUBLIC SECTOR is measured by the three variables: government responsibility for jobs (RJOBS), government responsibility for healthcare (RHEALTH) and government spending on pensions (PENSION). Table 2 shows that this measurement model easily meets all of the model fit criteria. It should also be noted that these results somewhat confirm Rose and Makkai's (209-210) findings that two dimensions of welfare attitudes exist among post-communist countries. However, they use exploratory factor analysis and different types of questions. The validity of their study is further weakened by the fact that they only use five questions in their original calculations. In their study, egalitarianism and collectivism become different dimensions of welfare attitudes, which is not all that different from the present finding that support for BIG GOVERNMENT and EQUALITY provide different dimensions for measuring welfare 304 Bringing Welfare Attitudes into the Debate SAXONBERG TABLE 1. Support for pro-welfare polices (country with the highest score in bold) Sweden Czech (n = 1238) Republic (n = 1100) Factor 1: BIG PUBLIC SECTOR (standardized correlation with BIG PUBLIC SECTOR in parenthesis) v44: On the whole, do you think it should be or should not be the government's responsibility to provide decent housing for those who can't afford it? % answering 'definitely should be' or 'probably should be', (rhouse) v38: On the whole, do you think it should be or should not be the government's responsibility to provide health care for the sick?% answering 'definitely should be' or 'probably should be', (rhealth) v36: On the whole, do you think it should be or should not be the government's responsibility to provide a job for everyone who wants one?% answering 'definitely should be' or 'probably should be', (rjobs) v31: % agreeing that: the government should spend 'much more' or 'more' on unemployment benefits. (unemp) v30: % agreeing that: the government should spend 'much more' or 'more' on old age pensions, (pension) v28: % agreeing that: the government should spend 'much more' or 'more' on education, (edu) v26: % agreeing that: the government should spend 'much more' or 'more' on health, (health) v20: % 'strongly in favor of or 'in favor of government financing of projects to create new jobs, (job) Factor 2: EQUALITY (standardized correlation with EQUALITY in parenthesis) vl6: % 'Agree strongly' or 'Agree' that it is the responsibility of the government to reduce the differences in income between people with high incomes and those with low incomes, (redistr) vl7: % 'strongly in favor of or 'in favor of control of wages by law. (wagecon) vl8: % 'strongly in favor of or 'in favor of control of prices by law. (pricecon) vl9: % 'strongly against' or 'against' cuts in government spending (cuts) v57: % describing taxes in respondent's country as generally being 'too low or 'much too low' for those with high incomes, (taxhi) v58: % describing taxes in respondent's country as generally being 'too low or 'much too low' for those with medium incomes, (taxme) 81.8% 79.7% (0.61) (0.55) 96.2% 96.8% (0.42) (0.58) 65.1% 76.3% (0.75) (0.71) 42.7% 19.7 (0.69) (0.21) 56.9 66.7 (0.56) (0.47) 58.8% 66.7 (0.11) (0.30) 76.6% 82.7% (0.53) (0.45) 69.3% 74.1% (0.54) (0.26) 59.6 60.3% (0.77) (0.63) 28.3% 47.1% (0.60) (0.58) 58% 72% (0.64) (0.72) 20.4% 16.0% (0.30) (0.04) 62.4% 44.8% (0.88) (-1.10) 3.2% 6.6% (0.52) (0.43) EUROPEAN SOCIETIES TABLE 2. Results of confirmatory factor analysis (standardized coefficients given) Sweden Czech Republic (n = 1129) (n = 916) Indicators Factor I Factor II: Factor I: Factor II: BIG PUBLIC EQUALITY BIG PUBLIC EQUALITY SECTOR SECTOR pension 0.49 0.39 rhealth 0.41 0.77 rjobs 0.73 0.54 taxhi 0.68 0.33 red i st r 0.79 0.76 fit indices chi-square 5.048 8.641 df 4 4 P value 0.282 0.071 GFI 0.998 0.996 AGFI 0.993 0.986 RMSEA 0.015 0.036 P-close 0.952 0.725 attitudes. Strangely, in their study support for higher taxation provides the only measure of egalitarianism, while support for equal incomes becomes an indicator of support for collectivism rather than egalitarianism. This is a puzzling result and shows the kinds of problems that can arise when one uses data driven methods, such as exploratory factor analysis, rather than theory driven methods, such as confirmatory factor analysis. The Full Structural Model: Explaining Attitudes toward Welfare Now that it has been seen that PRO-WELFARE ATTITUDES have two dimensions, the full structural model is shown in Figure 2. The three independent variables CLASS, EDUCATIONAL LEVEL and INCOME explain PRO-WELFARE ATTITUDES. PRO-WELFARE ATTITUDES, in turn, has two dimensions: BIG-PUBLIC SECTOR (indicated by governmental responsibility for healthcare and jobs and increased spending on pensions) and EQUALITY (indicated by support for redistribution and higher taxes for the wealthy). Finally, PRO-WELFARE ATTITUDES explains voting. Those who support generous welfare policies and income equality are more likely than others to vote for Leftist parties. The small circles labled el—el4 denote the error terms of the variables. 306 Bringing Welfare Attitudes into the Debate SAXONBERG First this article presents the results for Sweden to create a yardstick, which can be used for comparing the Czech Republic. In each case the full structural model is presented as well as the modified model that meets closeness-of-fit criteria. As Table 3 shows, in Sweden CLASS explains much more of the variance in PRO-WELFARE ATTITUDES than the other independent variables. In the full structural model, CLASS explains 20.3 per cent of the variance in PRO-WELFARE ATTITUDES (i.e. the square of the standardized coefficient - 0.45), while EDUCATIONAL LEVEL only explains 9.6 per cent ( —0.31 ) of the variance and INCOME only explains 7.8 per cent (i.e. the square of — 0.28). In the full model, AGFI is unacceptable at 0.891 and so is RMSEA at 0.097. However, by removing EDUCATIONAL LEVEL and INCOME, the AGFI becomes > 0.9 and RMSEA < 0.08. In addition, although the RMSEA score is slightly above the more stringent level of 0.05, which some statisticians now recommend, it easily meets this more stringent standard by decreasing to 0.045 once the AMOS algorithm for dealing with missing values is used; although, as Figure 2. Full structural model of welfare attitudes and voting. 307 EUROPEAN SOCIETIES TABLE 3. Support for welfare (listing of the latent variables) Sweden Czech Republic Full Model Modified Full Model Modified Model (n = 1098) Model (n = 916) Influence of CLASS on PRO-WELFARE ATTITUDES (CLASS is measured by own occupation and spouse's occupation) standardized coefficient -0.45 -0.64 -0.04 unstandardized coefficient -1.44* -2.13* -0.01 std. Error 0.22 0.27 0.01 Influence of INCOME on PRO-WELFARE ATTITUDES standardized coefficient —0.28 unstandardized coefficient —0.14* std. Error 0.02 Influence of EDUCATIONAL LEVEL on PRO-WELFARE ATTITUDES standardized coefficient —0.31 —0.27 unstandardized coefficient —0.19* —0.06* std. Error 0.02 0.01 Influence of PRO-WELFARE ATTITUDES on VOTING (Left-voting = 1) -0.33 -0.43 -0.04* -0.07* 0.01 0.01 standardized coefficient 0.53 0.58 0.28 0.30 unstandardized coefficient 0.30* 0.31* 0.50* 0.39* std. Error 0.02 0.02 0.07 0.07 Model fit indices chi-square 397.969 75.236 425.509 36.067 df 32 17 34 12 P value 0.000 0.000 0.000 0.000 GFI 0.936 0.985 0.930 0.989 AGFI 0.891 0.968 0.886 0.975 RMSEA 0.097 0.053a 0.102 0.047 AIC 443.969 72.000 467.509 68.067 P-close 0.000 0.308a 0.000 0.586 * = significant at .001 level. For the CR, EQUALITY was set to equal 1 as well as BIG PUBLIC SECTOR in order to not have standard coefficients greater than 1. aNote: if AMOS' missing variable function is used, then RMSEA becomes 0.045 and P-close becomes 0.744. However, the GFI AND AGFI cannot be calculated. already noted, when using this method, one cannot calculate the AGFI and GFI scores. It should also be noted that when using the AMOS missing data algorithm, the P-close score nearly doubles to 0.704, which is well above the recommended level of 0.5. Furthermore, when eliminating INCOME and EDUCATION, the AIC score becomes much lower, which also indicates that the modified model performs much better. Thus, CLASS is clearly the best explainer of welfare attitudes, since it both explains the greatest variance in PRO-WELFARE ATTITUDES and it is the only independent variable left in the model that meets the closeness-of-fit criteria. 308 Bringing Welfare Attitudes into the Debate SAXONBERG The results are much different for the Czech Republic than for Sweden. Most importantly, the Czech case confirms the class-in-flux hypotheses and contradicts the hypothesis of Czech exceptionalism. First, while CLASS only explains 0.2 per cent of the variance in PRO-WELFARE ATTITUDES in the Czech Republic it explains 20.3 per cent of the variance in Sweden. Second, while CLASS is a more powerful explainer of PRO-WELFARE ATTITUDES than INCOME in Sweden, in the Czech Republic, INCOME becomes a much more powerful explainer of PRO-WELFARE ATTITUDES than CLASS. In the full structural model in the Czech Republic, INCOME has a standard coefficient of — 0.33 compared to a mere — 0.04 for CLASS. Thus, INCOME explains 10.9 per cent of the variance in PRO-WELFARE ATTITUDES in the Czech Republic compared to 0.2 per cent for CLASS. Third, in the Czech Republic CLASS is not even significant at the 0.05 level, while it is significant at the 0.001 level for Sweden. The data also shows that Zagórski's (1994) hypothesis of understanding complexity does not hold up. To be sure, EDUCATIONAL LEVEL explains much more variance of PRO-WELFARE ATTITUDES in the Czech Republic than does CLASS (7.3 compared to 0.2 per cent). EDUCATIONAL LEVEL is also significant at the 0.001 level, while CLASS is insignificant. However, EDUCATIONAL LEVEL actually explains less variance of PRO-WELFARE ATTITUDES for the Czech Republic than for Sweden (7.3 compared to 9.6 per cent). Thus, these findings give no indication that educational level is especially important for post-communist countries. In addition, EDUCATIONAL LEVEL must be eliminated in the modified model in order to meet closeness-of-fit criteria. Thus, among the explanatory variables, Ost's hypothesis holds up well: Czech exceptionalism seems to be a myth, as CLASS actually does a very poor job of explaining Czech welfare attitudes. Zagorski's hypothesis of understanding complexity holds up better than the one about Czech exceptionalism, since it is statistically significant. Nevertheless, this article suggests that his hypothesis should be rejected. First, it must be eliminated from the modified model in order to meet closeness-of-fit criteria. Second, educational level actually explains more variance of PRO-WELFARE ATTITUDES in Sweden than in the post-communist Czech Republic. 7. In the Czech case, it is necessary to set both EQUALITY = 1.0 and BIG PUBLIC SECTOR = 1.0, because when BIG PUBLIC SECTOR was set free, the standardized coefficient for EQUALITY becomes 1.04, which is not persmissable. However, this restriction only marginal influences on the result and has absolutely no bearings on the substantive results. 309 EUROPEAN SOCIETIES Further evidence that classes are in flow in Czech Republic comes from the relatively weak relationship between PRO-WELFARE ATTITUDES and VOTING. In the full model the standardized coefficient between these two variables is 0.28, explaining 7.8 per cent of the variance. While statistically significant, it is an extremely low score compared to Sweden, where the explained variance is above 28.1 per cent. In the modified models that meet the closeness-of-fit criteria, the difference remains approximately the same. Nevertheless, this relationship is a bit ambivalent, because although the standardized coefficient for PRO-WELFARE ATTITUDES is much higher for Sweden than for the Czech Republic, the unstandardized coefficient is actually higher for the Czech Republic. One possible explanation for this difference is that perhaps the link is a bit stronger for the Czech Republic, but the standard error is also much higher (0.07 compared to 0.02 for Sweden). Since the standard error influences the standardized coefficient one possible conclusion is that welfare policies on the average in fact, do play a larger role for voting in the Czech Republic than in Sweden, which could reflect the situation in which post-communist countries have not yet been influence as much by post-materialist values as comparable western countries. At the same time, the greater standard error and lower explained variance of PRO-WELFARE ATTITUDES reflects the lack of 'class consciousness' that comes about when classes are in flow Since Czechs are less certain about the class interests, a large number of Czechs do not make the link between welfare attitudes and voting, even though on the average they are more concerned about welfare programs than Swedes. Finally, it should be noted that although space does not allow for an analysis of the details, for the Czech Republic the relationship between PRO-WELFARE ATTITUDES and VOTING is slightly stronger when it comes to voting for rightwing parties (explaining more than 5 per cent more of the variance of voting), indicating that leftwing voters (which tend more often to be workers) are less sure of their class interests than professionals and the bourgeoisie, who are more likely to vote for the rightwing parties that support their attitudes toward taxes and welfare policies. Next the Chi-square difference test was used to ascertain whether the differences in coefficients between Swedish and Czech respondents are statistically significant. This was done using multi-group analysis. For each country (group) the coefficient of a particular variable was set equal to the coefficient of the same variable in the other country. If the chi-square difference has a probability <0.05 we can conclude that no significant difference exists between the coefficients for each country. That is, the difference in the unstandardized coefficients of a particular variable 310 Bringing Welfare Attitudes into the Debate SAXONBERG between the two countries is not really significant. We cannot reject the null hypothesis that the two coefficients could both be set equal to each other. If the probability >0.05 then we can conclude that the differences in coefficients are statistically significant. Since AFGI and GFI scores are not used for this test the raw data was used rather than coefficient matrixes in order to utilize AMOS' missing-data program. As Table 4 shows, when the three independent variables are constrained to have the same unstandardized coefficients for each country, they all fail the chi-square difference test, with P >0.05 for all three cases. In other words, we can reject the null hypothesis that no difference exists between the strength of CLASS in Sweden and the Czech Republic and the same is true for INCOME and EDUCATIONAL LEVEL, as the probability of the chi-square difference is well above 0.05 for all three cases. This confirms the previous analysis supporting the class-in-flux hypothesis, while rejecting both the understanding-complexity and Czech exception-alism hypotheses, as CLASS and EDUCATIONAL LEVEL are more important for explaining welfare attitudes in Sweden than in the Czech Republic. Finally, we have the relationship between PRO-WELFARE ATTITUDES and VOTING. Here, the results are ambiguous, since the standardized coefficient of PRO-WELFARE ATTITUDES is much greater for Sweden than for the Czech Republic, but the unstandardized coefficient is greater for the Czech Republic than for Sweden. The chi-square difference test shows that we cannot reject the null hypothesis that PRO-WELFARE ATTITUDES has the same unstandardized coefficient for each country. In other words, the difference in unstandardized coefficients is not statistically significant. At the same time the standardized coefficient is much stronger for Sweden than for the Czech Republic, since as already noted the standard error is greater in the Czech Republic. Thus, even though the results are ambivalent here, reason exists TABLE 4. Chi-square difference tests: Are coefficients for explanatory variables the same in Sweden and the Czech Republic? df CMIN P Taxpayer equal 1 0.213 0.644 Class equal 1 0.128 0.720 Understanding complexity equal 1 1.572 0.210 Pro-welfare attitudes equal 1 33.080 0.000 Note: in the Czech case, the coefficient for TAXHI on the latent variable EQUALITY had to be set at 0.34 in order to achieve a positive definite matrix. The standardized coefficient of EQUALITY was also above 1 before the constraint was added. This slight decline from the freely estimated score of 0.40 to 0.34 for TAXHI did not change the substantive results at all. 311 EUROPEAN SOCIETIES to question the claim that party competition in the Czech Republic revolves around socioeconomic issues. To some extent welfare attitudes do influence voting in the Czech Republic, but for a large number of persons this is not the case, which causes a high standard error and a relatively low standardized coefficient. Conclusion The Czech Republic represents a critical case, since many social scientists have singled out the country as the one post-communist case where class is important and party competition centers on socio-economic issues. However, the result of this study refute the claims of Czech exception-alism and support Ost's hypothesis that classes are in flux during the transition, which hinders workers from understanding their class interests. CLASS cannot significantly explain PRO-WELFARE ATTITUDES in the Czech Republic. Meanwhile, in a comparable West European country, Sweden, CLASS actually provides the strongest explanation of welfare attitudes and this difference in the explanatory value of CLASS is statistically significant. Moreover, the hypothesis that citizens in post-communist countries are not aware of their class interests gains support from the fact that Czechs are generally much more positive toward welfare programs and increased welfare spending than Swedes, but less willing to increase taxes in order to finance such measures. If class-based organizations had succeeded in making Czech more conscious of their class interests they would have been more likely to see the connection between the need to raise taxes in order to finance the programs that they support. This study also refutes Zagórski's (1994) hypothesis about understanding complexity. He claims that that educational level is especially important for post-communist countries, because of the nature of the complicated economic and socioeconomic reforms, which post-communist regimes must carry out during the transition to a market economy. According to this logic, those with higher education are more able to understand the need for reforms that will lower living standards in the short-run in order to raise them in the long run. It turns out, however, that educational level explains much more variance in PRO-WELFARE ATTITUDES in Sweden than in the Czech Republic and that this difference is statistically significant. The most ambivalent result is the relationship between PRO-WELFARE ATTITUDES and VOTING. On the one hand, in Sweden PRO-WELFARE ATTITUDES explain a much greater portion of the variance in VOTING than in the Czech Republic. On the other hand, the unstandardized coefficient is actually stronger for the Czech Republic. A 312 Bringing Welfare Attitudes into the Debate SAXONBERG reasonable interpretation is that welfare attitudes influence the voting of average Czechs more than average Swedes, although this difference is not statistically significant. However, since the country is going through a transition, in which classes are in flow, many Czechs are uncertain of their class interests. Thus, a large group exists that behaves differently than in comparable western countries, which accounts for the larger standard error among Czech respondents. Consequently, PRO-WELFARE ATTITUDES explains a greater portion of the variance in VOTING for Sweden than the Czech Republic. Finally, this study makes a good comparison for the upcoming ISSP survey on the Role of Goverment that will be repeated in the year 2006. Since respondents in both countries will have to answer the same questions, it will become possible to see whether the differences still remain or whether the Czech Republic has become nearer the 'classical model' of homogenous countries, in which class becomes the main cleavage and party competition centers on socioeconomic issues. It just may turn out that the differences will remain once class relations crystallize, which would indicate that the post-communist countries will continue to have different political dynamics than western European countries. However, we might just find out that the new members of the EU will develop similarly to their western counterparts. Some evidence is already arising that the Czech political system is coming closer to western countries, such as Sweden. For example, six years of social democratic rule have largely alleviated fears that social democratic rule means the return of communism. Thus, voters, who support generous social policies are more likely to vote according to their socioeconomic preferences and less likely to vote for rightest parties for anti-communist reasons. Similarly, party politics is coming closer to Western countries, such as Sweden, as pragmatic political compromises between parties along ideological lines are starting to replace personal conflicts as the motor of daily politics. For example, when the social democrats won the 1998 elections, they were unable to form a majority government, as the leader of the center-right Freedom Union, Jan Ruml, absolutely refused to consider forming a coalition with either the social democratic leader Miloš Zeman or the market-liberal leader of ODS, Václav Klaus. By the 2002 elections, however, all three party leaders had resigned from their positions (although Klaus later made a comeback and was elected president), which allowed the Freedom Union to join in a majority government together with a new generation of social democrats and the Christian democrats. In a situation in which the political parties are behaving more pragmatically and class relations are becoming more crystallized, we would expect workers today to follow their class interests more than during the 1990s, when the previous ISSP survey was taken. 313 EUROPEAN SOCIETIES Acknowledgements I would like to thank David Ost, Stefan Svallfors and Tomas Sirovatka as well as the anonymous reviewers who have given me helpful comments and suggestions on my drafts. References Ahn, S.-H. (2000) Pro-Welfare Politics: A Model for Changes in European Welfare States, Uppsala: Uppsala University. Aalberg, T. (1998) 'Nórske likhetsverdier i et kmparativt perspektiv', Tidskrift for Samfunnsforskning 39 (4): 490—515. Andreß, H.-J. and Heien, T. (2001) 'Four worlds of welfare state atttitudes? A comparison of Germany, Norway, and the United States', European Sociological Review 17 (4): 337—56. Antoš, P. (1998) Volby 98'... a co dal, Prague: Duel. Bean, C. and Papadakis, E. (1993) 'Popular support for the welfare state: A comparison between institutional regimes', Journal of Public Policy 13 (3): 227-54. Blomberg, H., Kroll, C, Suominen, S. and Helenius, H. (1996) 'Socialkass och attityderna till nedskärningar i välfärdssystemet i Finland', Sociologisk Forskning 4: 57-78, reprinted in Helena Blom-berg-Kroll (1999) Kosta vad det kosta vili? — Attitydmô'nster och attitydfô'rändringar hos befolkning och eliter betráffande välfärdsservicen i nedskärningarnas tid (doctoral dissertation, Institutionen för socialpoli-tik, Abo Akademi). Bollen, K. A. (1989) Structural Equations with Latent Variables, New York: John Wiley & Sons. Bergqvist, C. (1999) 'Modeller for barnomsorg och föräldraledighet', in C. Berqvist et al. (eds), Likestilte demokratier?, Oslo: Universitetsforlaget, pp. 113-28. Cotta, M. (1996) 'Structuring the new party systems after the dictatorship: Coalitions, alliances, fusions and splits during the transition and post-transition stages', in G. Pridham and P. G. Lewis (eds), Stabilising Fragile Democracies. Comparing New Party Systems in Southern and Eastern Europe, London & New York: Routledge, pp. 69-99. Coughlin, R. M. (1980) Ideology, Public Opinion and Welfare Policy, Berkeley, CA: Institute for International Studies. Csepeli, G., Kolosi, T. and Örkeny, A. (1993) 'Our futureless values: The forms of justice and injustice perception in Hungary in 1991', Social Research 60 (4): 865-893. 314 Bringing Welfare Attitudes into the Debate SAXONBERG Dalton, R. J. (1996) Citizen Politics (2nd edition), Chatham, NJ: Chatham House. Edlund, J. (1999) 'Trust in government and welfare regimes: Attitudes to redistribution and financial cheating in the USA and Norway', European Journal of Political Research 35: 341—70. Edlund, J. (2000) 'Public attitudes towards taxation: Sweden 1981-1997', Scandinavian Political Studies 23 (1): 37—64. Esping-Andersen, G. (1990) The Three Worlds of Welfare Capitalism, Cambridge: Polity Press. Forma, P. (1999) 'Welfare state opinions among citizens and local government elites in Finland, 1992—6', in S. Svallfors and P. Taylor- Gooby (eds), The End of the Welfare State? Responses to State Retrenchment, London: Routledge, pp. 87-105. Forma, P. and Kangas, O. (1999) 'Need, citizenship or merit: Public opinion on pension policy in Australia, Finland and Poland', in S. Svallfors and P. Taylor-Gooby (eds), The End of the Welfare State? Responses to State Retrenchment, London: Routledge, pp. 61-189. Gilljam, M. and Holmberg, S. (1995) Väljarnas Val, Stockholm: Norstedts juridik. Ginsburg, N. (1992) Divisions of Welfare: A Critical Introduction to Comparative Social Policy, London: Sage. Hasenfeld, Y. and Rafferty, J. A. (1989) 'The determinants of pulibc attitudes toward the welfare state', Social Forces 67 (4): 1027-48. Herod, A. (1997) Theorizing Trade Unions in Transition, inj. Pickles and A. Smith (eds), London: Routledge, pp. 197-217. Holmberg, S. (2000) Välja Parti, Stockholm: Norstedts juridik. Holmberg, S. (2001) Swedish Voting Behavior, Göeborg: Göteborg University, Swedish Elections Studies Program. Inglehart, R. (1997) Modernization and Postmodernization: Cultural, Economic, And Political Change in 43 Societies, Princeton, NJ: Princeton University Press. Jöreskog, K. and Sörbom, D. (1993) LISREL 8: Structural Equation Modeling with the SIMPLIS Command Language, Chicago, IL: Scientific Software International. Kitschelt, H., Mansfeldova, Z., Markowski, R. and Tóka, G. (1999) Post-Communist Party Systems: Competition, Representation, and Inter-Party Cooperation, Cambridge: Cambridge University Press. Kline, R. B. (1998) Principles and Practice of Structural Equation Modeling, New York: Guilford Press. Knudsen, K. (2001) 'Hvor stor bor ulikheten v sere? Nórske meinger om legitim lonn 1992-1999', Tidskrift for samfunnsforskning 42 (4): 507-36. 315 EUROPEAN SOCIETIES Knutsen, O. and Kumlin, S. (2003) 'Value orientations and party choice - A comparative longitutdinal study of five countries', Paper presented at the 2nd ECPR Conference in Marburg, 18—21 September 2003. Korpi, W. (1981) Den Demokratiska Klasskampen. Svensk Politik i Jämförande Perspektiv, Stockholm: Tidens fórlag. Krause, K. D. (2000) 'Public opinion and party choice in Slovakia and the Czech Republic', Party Politics 6 (1): 23-46. Lijphart, A. (1977) Democracy in Plural Societies: A Comparative Exploration, New Haven, CT: Yale University Press. Lijphart, A. (1984) Democracies: Patterns of Majoritarian and Consensus Governmet in Twenty-One Countries, New Haven, CT: Yale University Press. Lipset, S. M. (1963) Political Man: The Social Bases of Politics, Baltimore, MD: John Hopkins University Press. Machonin, P. et al. (1996) Strategie sociálni transformace Ceske Společnosti? Anosti, Brno: Nakladatelství Doplnki. Markowski, R. (1997) 'Political Parties and Ideological Spaces in East Central Europe', Communist and Post-Communist Studies 30 (3): 221 — 254. Maruyama, G. M. (1998) Basics of Structural Equation Modeling, London: Sage. Matějů, P. (1999) 'Who votes left after the fall of Communism? The Czech Republic in comparative perspective', International Journal of Comparative Sociology XL(1): 13—39. Matějů, P. (2000) 'Krystalizace norem distributivni spravedlnosti', in P. Matijt and K. Vlachová (eds), Nerovnost, Spravedlnost, Politika: Česka' Republika 1991-1998, Prague: Slon, pp. 145-90. Offe, C. (1994) Der Tunnel am Ende des Lichts: Erkundungen der Politischen Transformation im Neuen Osten, Frankfukrt: Campus Verlag. Ost, D. (1993) 'The politics of interest in post-communist East Europe', Theory and Society 22: 453—86. Ost, D. (1995) 'Labor, class, and democracy: Shaping political antagonisms in post-communist society', in B. Crawford (ed.), Markets, States and Democracy: The Political Economy of Post-Communist Transformation, Boulder, CO: Westview Press, pp. 177-203. Rabušic, L. and Sirovátka, T. (1999) 'Czech welfare state and its legitimacy', Polish Sociological Review 126 (2): 239-253. Raven, E (1876) 'La situation economique en Tschéchoslovaquie', Est et v Ouest, 15 June (574): 206-16. Řeháková, B. (2000a) 'The early parliamentary elections of 1998: An analysis of voting behaviour in comparison to the parliamentary election of 1996', Czech Sociological Review 8 (1): 2000. 316 Bringing Welfare Attitudes into the Debate SAXONBERG Řeháková, B. (2000b) 'Jak vnímáme příjmové nerovnosti a jaké bychom je chtěli mít', Sociologický Časopis 37 (1): 23-40. Renwick, A. and Tóka, G. (1998) 'East meets West', in R. Jowell et al. (eds), International Social Attitudes: The 15th British Social Attitudes Report, Aldershot: Dartmouth, pp. 149-71. Roller, E. and Westle, B. (1987) 'Zur Politisierung und depolitisierung von wohlfahrtsansprüchen', Soziale Welt 38: 227-51. Rose, R. and Makkai, T. (1995) 'Consensus or dissensus about welfare in post-communist societies?', European Journal of Political Research 28: 203-24. Saxonberg, S. (1999a) 'A new phase in Czech polities', Journal of Democracy 10 (1): 96—111. Saxonberg, S. (1999b) 'Václav Klaus: the rise and fall and re-emergence of a charismatic leader', East European Politics and Society 13 (2). Saxonberg, S. (2003) Czech Republic Before the New Millenium: Politics, Parties and Gender, New York/Boulder, CO: Columbia University Press, East European Monographs. Simkus, A. and Robert, P. (1995-96) 'Attitudes toward inequality: Cross-national comparison and class division in Hungary', International Journal of Sociology 25 (Winter): 85-102. SCB (Statistiska centralbyrán) (2003) http://www.scb.se/statistik/ AM0401 /am0401 _sammanstallning.asp. Sirovátka, T. (2002) 'Opinions of Czechs citizens about the welfare state', Czech Sociological Review 38 (3): 327—44. Sirovátka, T., van Oorschot, W and Rabušic, L. (2002) 'Solidarity and welfare state support. Czech Republic compared to the Netherlands', in P. Towsend and D. Gordon (eds), New Policies to Defeat an Old Enemy, Bristol: The Policy Press, pp. 147—69. Skocpol, T. and Somers, M. (1980) 'The uses of comparative history in macrosocial theory', Comparative Studies in Society and History 22: 174— 97. Sundström, E. (1999) 'Should mothers work? Age and attitudes in Germany, Italy and Sweden', International Journal of Social Welfare 8: 193-205. Svallfors, S. (1993) 'Dimensions of inequality: A comparison of attitudes in Sweden and Britain', European Sociological Review 9 (3): 267-87. Svallfors, S. (1996) Välfärdsstatens Moraliska Ekonomi: Välfärdsopinionen i 90-tals Sverige, Umeá: Boréa. 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', Sociologisk forskning (1): 93—119. 317 EUROPEAN SOCIETIES Taylor-Gooby, P. (1995) 'Who wants the welfare state? Support for welfare in European countries', in S. Svallfors (ed.), In the Eye of the Beholder: Opinions on Welfare and Justice in Comparative Perspective, Umeá: Sälj support, pp. 11-51. Tworzeciki, H. (2000) 'Welfare-state attitudes and electoral outcomes in Poland and Hungary', Problems of Post-Communismm 47 (6): 17-28. Vachudová, M. A. (2001) 'The Czech Republic: The unexpected force of institutional constraints', in A. Pravda and J. Zielonka (eds), Democratic Consolidation in Eastern Europe: International and Transnational Factors, Oxford: Oxford University Press, pp. 325-62. Večerník, J. (1996) Markets and People: The Czech Reform Experience in a Comparative Perspective, Aldershot: Avebury Večerník, J. (ed.) (1998) Zprava o Vývoji Ceske Společnosti, Prague: Academia. Yuan, K. H. and Bentier, P. M. (1994) 'Bootstrap-corrected ADF test statistics in covariance structure analysis', British Journal of Mathematical and Statistical Psychology 47: 63—84. Zagórski, K. (1994) 'Hope factor, inequality, and legitimacy of systemic transformations: The case of Poland', Communist and Post-Communist Studies 27 (4): 357-76. Appendix A: Recoding for class The ISSP survey only lists occupations, which make it necessary to recode occupations into class. Sometimes members of several professions could conceivably either be salaried employees (and thus 'professionals') or owners of private enterprises (and thus 'bourgeois'). However, this can be clarified, since another question in the survey (V213) asks whether the respondent is self-employed or is employed by others. Since this question is not asked about the respondent's spouse, it is not possible to be as accurate about the classification of the spouse's class. Nevertheless, it turns out that there was hardly any difference between the R of class of respondents and the class of their spouses, which indicates that this inexact measurement of the class of the spouse functions relatively well. Classes were scaled as: 3 = bourgeois, 2 = professional ('white collar'), 1 = worker. The Czech Republic uses the ISCO system. V208 was used to measure class and transformed as follows: professional = 1-20, 31, 40, 42-45, 60, 70; while bourgeois = 21, 30, 41, 51, 61; and worker = 32-38, 46-49, 52-59 and 71-99. In Sweden, var 209 was used, because Sweden abides by the NSCO system. The class of the respondant was coded in the following steps: (1) 318 Bringing Welfare Attitudes into the Debate SAXONBERG create file, bourg, for var213 (self-employed). Recoded, so that if v213 = 1 (self-employed), then now = 3. Otherwise, = 0. (2) create file profswe, for var209. Use filter for profswe, if v = 13 (i.e. Sweden). (3) recode profswe, so that < 405 = 2 (that is professional), 406-980 = 1 (manual worker) and 981 = 2 (officer), 990 and 999 are SYSMIS. (4) create file sweclass, so that sweclass = profswe. (5) recode sweclass so that x El 3 = 3. Coding spouse's profession for Sweden: 3 = 400, 297, 331, 210-219; 2 = 981, 609, 621, 1-399, 402-404, 602-603; 0 = SYSMIS = 990-999; all others = 1 Recoding of income for the Czech Republic: 1 =0-3500 crowns, 2 = 3501-4600, 3=4601-5700, 4 = 5701-6800, 5=6801-7900, 6 = 7901-9000, 7 = 9001-10100, 8 = everything above. Steven Saxonberg is an associate professor of political science. He has a research position at Uppsala University (Department of Government) and has a lectureship in sociology and social science at Dalarna University College in Sweden. His books include The Fall: A Comparative Study of the End of Communism in Czechoslovakia, East Germany, Hungary and Poland (Amsterdam: Harwood Academic, 2001), and The Czech Republic Before the New Millennium (East European Monographs/Columbia University Press, 2003). He has written extensively on the collapse of communism, the post-communist transition, welfare policies and attitudes, gender issues and gender attitudes in Central Europe. Address for Correspondence: Steven Saxonberg, Uppsala University, Department of Government, Box 514, 751 20 Uppsala, Sweden. E-mail: sax@post.utfors.se 319