Growing distrust of the market

12While the international comparison of economic inequality is now a field of study in itself, the comparative analysis of economic attitudes is underdeveloped given the importance of inequality in public debate and the interest the media and elites have in shifts in public opinion movements concerning it. Conflicts over meaning and the disputed nature of the data, which are sometimes dismissed as “soft” or “subjective,” in part explain this fact. But another factor should be taken into account. On the European scale, most of the important work on opinions of the state, wealth redistribution and social inequality is based on the results of large international surveys (“Eurobarometer,” “European Social Survey,” “European Values Study,” “World Values Survey” and the “International Social Survey Programme” [3]). And although these surveys have made significant sociological advances possible by systematizing the clash between theory and empiricism in the field of social change (Chenu, 2011), they have remained little used when it comes to examining changes in opinions towards the big economic and social issues.

13The “European Values Study” survey project [4] enables us to study the economic attitudes of Europeans, and their long-term evolution, in detail. Since the second wave, launched in 1990 in 29 countries, the questionnaire has invited respondents to position themselves on a series of bipolar scales asked “in series,” one after the other. [5] These ten-point scales present two opposing economic approaches, one characteristic of a liberal position and the other more typical of an interventionist position. Three questions can be said to relate to classical economic liberalism and concern its major doctrinal principles (competition, private property and free enterprise). They are formulated in the following way:

• “Competition is good. It stimulates people to work hard and develop new ideas” versus “Competition is harmful. It brings out the worst in people.”
• “Private ownership of business and industry should be increased” versus “Government ownership of business and industry should be increased.”
• “The state should give more freedom to firms” versus “The state should control firms more effectively.”

14Three other questions are about the role and involvement of the respondents in economic life. They better evaluate the degree of agreement with the individualistic dimension of economic liberalism.

• The theme of one question is responsibility for providing for individual needs: “Individuals should take more responsibility for providing for themselves” versus “The state should take more responsibility to ensure that everyone is provided for.”
• Another relates to individual effort and income inequality: “There should be greater incentives for individual effort” versus “Incomes should be made more equal.”
• The final one relates to social constraints and attitudes to the unemployed: “People who are unemployed should have to take any job available or lose their unemployment benefits” versus “People who are unemployed should have the right to refuse a job they do not want.”

15Figure 1 covers all 23 countries of the European Union present in the three survey waves. [6] It captures the overall development of economic attitudes between 1990 and 2008. It can be seen that the most liberal opinions in economic matters—represented here by responses between 7 and 10—have tended to decline. For the questions concerning the doctrinal pillars of economic liberalism, this decline is much more pronounced. It is as if the Europeans interviewed were expressing a growing distrust of the market economy as an optimal means of distributing wealth.

Figure 1

Evolution of economic attitudes in the European Union (1990–1999–2008)

Evolution of economic attitudes in the European Union (1990–1999–2008)

16Opinions in favour of free enterprise have declined from 40% to 36% between 1999 and 2008. Opinions in favour of private property have declined by about 15 points between 1990 and 2008. At this time, only 39% of Europeans thought there should be more private property. Complementary analysis shows that support for the private ownership of means of production has declined across the whole European Union, particularly in the Scandinavian and Eastern Europeans countries such as Hungary, Estonia, Finland and Austria. In Western European countries, such as Portugal, Italy and France, the decline is weaker.

17The trend is also downwards if we consider the opposition between individual efforts and income equality. A preference for more equal incomes has increased in most poor Eastern European countries (Romania, Czech Republic and Slovenia), but also in the rich Northern European countries (Sweden, Estonia and Finland). The same is true for competition (–9 points), the belief in whose virtues has declined above all in the East (Hungary, Estonia, Latvia, Lithuania and Finland) and the West (Spain, France and Belgium), but of which 58% of Europeans remained in favour in 2008 nonetheless.

18The last two items give a misleading impression of stability over time. They have in fact evolved very differently depending on the country. The question on the state taking responsibility for the needs of individuals reveals a division between social-democratic regimes (Denmark, Sweden, Finland, Austria, Italy, Belgium, and France) where it has tended to increase, and liberal regimes (Great Britain and Ireland) where, in contrast, it has tended to decline. The question on the rights of the unemployed seems to obey the same logic. The opinion that “the unemployed should have to take any job available” has declined in Denmark, Sweden and Finland. It is stable in what Esping-Andersen calls “corporatist-conservative” countries (Austria, Italy, Spain, Belgium, and France). But it has gained ground in liberal countries as well as in the East.

19Several other questions in the survey corroborate the upward trend of economic anti-liberalism in Europe. Trust in large firms, for example, has declined by 13 points between 1990 and 2008, from 48% to 35%. Conversely, faith in trade unions has increased continuously since 1990 to reach 40% in 2008 (+7 points). This evolution, however, hides significant differences between countries. Several countries from the former Soviet Bloc (Bulgaria, Hungary, Poland, and Czech Republic), for example, show very low levels of trust for both indicators; this is probably explained by a combination of low levels of unionization and weaker support for certain fundamentals of a market economy, such as private ownership of the means of production. In Spain and Sweden, where levels of unionization are very different (16% versus 71% in 2007 according to the ILO) trust in unions and large firms has declined in the same proportions since 1990.

20A similar observation can be made regarding confidence in the major social and economic functions of the state. Contradicting the received idea of a crisis in public institutions and the welfare state, nearly two-thirds of Europeans claim to have confidence in the “administration”, the “education system” and the “social security system.” [7] The general trend for these institutions is, moreover, an upward one, particularly between 1999 and 2008, but it is not generalizable for all European Union countries. In certain Eastern countries (Poland, Bulgaria, and Hungary) and in Northern Europe (Austria, Finland, and Sweden), confidence in the three institutions is instead declining.

21The question on preference between equality and liberty is another indicator of the increasing strength of economic antiliberalism. [8] The attachment to equality greatly increased between 1990 and 2008: 42% of Europeans claimed to prefer equality to liberty in 2008, while the figure was only 36% in 1999 and 1990. This increase is particularly marked in certain countries that are unequal from the point view of the Gini coefficient, such as Hungary, Italy and northeastern Europe (Latvia and Estonia). But it also affects more egalitarian countries such as Finland and Belgium. The attachment to equality has tended, in contrast, to decline in very unequal countries, whether rich, such as Ireland and the Netherlands, or rather poor, such as Romania and Bulgaria.

22These results ultimately reflect a mixed picture of European economic attitudes. On the one hand, anti-liberalism, commitment to the welfare state and demands for equality have clearly increased. But, on the other hand, as Forsé and Parodi (2005) have shown using data from the 1999 EVS, opinions favourable to economic liberalism are most often in the majority and seem well established. In other words, while anti-liberalism is on the increase in most European Union countries, the dominant opinion nevertheless seems to be settled on a “social-liberal” model. It combines growing demands of the state with an overall high level of acceptance of the fundamentals of liberal economic doctrine with individual involvement in economic matters being highly valued. The non-systematic character of economic preferences appears even more clearly when considering a cross-distribution of opinions. Respondents who adopt a social position on one question do not necessarily adopt the same position on another question (Appendix, Table A1).

23In sum, an initial descriptive analysis indicates that Europeans are not “ideologues”: [9] their liberal and anti-liberal preferences in economic matters co-exist more than they are mutually exclusive. This analysis also illustrates the fact that individuals adhere to very varied and often contradictory principles of redistributive justice (Dubet et al., 2006), and that they are thus susceptible to accepting various forms of redistribution and state intervention. But these composite response rationales perhaps also relate to different aspects of economic liberalism.

Analysing the multidimensionality of economic attitudes

24One way to analyse public opinion and its variations over the long term is to construct a measure that aggregates questions from surveys conducted at different times but whose wording is sufficiently similar to allow comparison. James A. Stimson thus developed a longitudinal index—the Policy Mood—to account for the evolution of broad general attitudes in America public opinion and to measure their connection to the state of the economy or to electoral results (Stimson, 1999; Stimson, 2007). This approach gave rise to an extensive literature, sometimes seeking to better explain shifts in ‘policy mood’ (Erikson, MacKuen and Stimson, 2002), sometimes to refine the perspective by developing more specific indices (Enns and Kellstedt, 2008).

25The method was recently applied to a study of economic attitudes in France (Tiberj, 2012). It enabled the identification of a sharp increase in public support for state regulation of the economy since the mid-1990s. On the eve of the 2012 presidential election, social and economic expectations in public opinion had reached a pitch not seen since the end of the 1970s. By analysing shifts in opinion through large synthetic indices, this type of approach tends to make a reduction ad unum of collective preferences and to mask variations. [10] Other studies have highlighted the value of a multidimensional approach, showing, in particular, the intertwining of economic attitudes and moral and political attitudes (Dargent, 2006; Dargent and Gonthier, 2010; Degeorges and Gonthier, 2012).

26To assess the extent to which economic attitudes are organized around potentially distinct dimensions, principal component analysis (PCA) [11] was done based on cumulative data from the 1990, 1999 and 2008 EVS, which included the 23 European Union countries featuring in the three survey waves. Each national sample is of the same size (around 1,000 individuals per country per wave) so as not to bias the statistical analysis by giving greater weight to more densely populated countries (such as Germany, France and the United Kingdom). Principal component analysis was thus applied to a data set containing 68,472 individuals in total.

27PCA is a statistical method that aims to identify the underlying structure of a set of variables (Rouanet and Le Roux, 1993). It enables us to reduce a set of correlated variables to a smaller number of non-correlated dimensions. These dimensions summarize the most structuring oppositions of these variables in the form of linear combinations. They can be ranked by order of importance according to their explanatory power, i.e., depending on the variability of the point clouds they describe.

28Table 1 shows that European economic attitudes relate to three distinct dimensions. The first, represented by factor 1, can be described as “interventionism”: it mainly includes the questions concerning state control of the economy. This can be considered the key dimension at the European level since it explains 20% of the variability in point clouds. The questions on the dilemma between equality and freedom and the rights of the unemployed are also positively correlated to this dimension, although more weakly. This indicates that Europeans in favour of a greater role for the state tend to favour equality over freedom and to consider that the unemployed should be able to reject an offer of work that they do not want. Interventionism and equality thus seem, to some extent, to go together.

Table 1

The structure of economic attitudes in the European Union (1990–1999–2008)

Factor 1 Factor 2 Factor 3 Nationalization of firms and industries should be increased 0.66 –0.14 –0.19 The state should take more responsibility to ensure that 0.65 everyone is provided for Competition is harmful 0.64 The unemployed should have the right to refuse a job they 0.42 –0.58 do not want Index of confidence in the social functions of the state 0.80 A very great confidence in large firms –0.19 0.66 –0.15 A very great confidence in trade unions 0.22 0.69 Incomes should be more equal 0.28 0.70 Equality is more important than freedom 0.40 0.45 Percentage of variance explained 20 18 12

The structure of economic attitudes in the European Union (1990–1999–2008)

Note: The index of confidence in the social functions of the state summarizes, in the form of an additive scale, the questions on confidence in the administration, education system and social security system (Cronbach’s α = 0.65).
The correlations between components and active variables are also shown in the Appendix (Figure A1).

29The second component, which captures 18% of the inertia, is distinguished from the first by summarizing the questions relating to confidence in institutions. Here we find a point that has already been established empirically: confidence in social and political institutions is a cumulative phenomenon (Cahiers du CEVIPOF, 2011). An individual who claims to have confidence in one type of institution would also be inclined to claim to have faith in another institution and vice versa. This is particularly the case with confidence in trade unions and in large firms, which are strongly and positively correlated.

30The third dimension is principally defined by the question concerning incomes policy and equal pay. It is structured by the opposition between Europeans who think that incomes should be more equal and those that think that individual efforts should be encouraged more. The question on unemployment rights is also associated negatively with this dimension—although more weakly. This means that Europeans who value income equality mostly tend to think that the unemployed should accept a job they do not want. The demand for equality thus seems to relate to an expectation of social cohesion, illustrated here by the idea that the unemployed contribute to collective solidarity by accepting a job they do not want. As a consequence, the third PCA component relates more to a logic of reciprocity (in the sense of a balance of social contributions and benefits) rather than to a logic of theoretical equality (in the sense that everyone receives the same benefits whatever their contribution); which is confirmed elsewhere by the quite moderate correlation between this component and preferring equality over freedom.

31There are several specific methods for testing the comparability of scales between countries (Davidov, Schmidt and Billiet, 2010). The consistency of results was confirmed here using additional PCA on each survey wave and on each country studied. These analyses reveal how the structure of European economic attitudes is very stable over time. The hierarchy of dimensions barely changed between survey waves in the 23 countries examined. The same is true for the content of the components: interventionism, confidence and equality are the central dimensions of national economies, whatever country is considered.

32Some variations, however, are worth noting. In some Northern European countries (such as Denmark, Finland, the Netherlands, and Sweden), PCA, for example, only isolates two factors: interventionism and confidence. Items that also correspond to the third factor are in fact correlated to the first factor; which would tend to indicate that in these countries, equality is more directly associated with public intervention. In other countries, in particular in the east of Europe (Czech Republic, Hungary, Romania, Slovakia, Slovenia, and also Portugal), the first factor is slightly less explanatory than the second. The proportion of variance explained by interventionism can here fall to 16% and that explained by confidence reach 25%. This means that in Eastern Europe economic preferences perhaps relate more to issues of confidence in institutions than to issues of public intervention—we will come back to this point later.

“Parallel publics”

33The PCA conducted on data collected at different times has the advantage of being able to represent the evolution of European economic attitudes in the same factorial plane. We thus find that, between 1990 and 2008, interventionism has significantly increased at the European scale, while the overall level of confidence has remained quite static (Appendix, Figure A2). [12] This factorial representation nevertheless contains an ambiguity: it leads us to believe that “European public opinion” is a totality likely to change uniformly over time. Yet social groups have different interests, experiences, environments and abilities to influence that bring into play often-divergent representations, values and expectations. Preferences regarding the state are thus “embedded” (Brooks and Manza, 2007) in social contexts, positions and situations. The hypothesis to examine is therefore the following: To what is extent is the observed rise in interventionism at the aggregate level confirmed at the level of the sub-populations that make up the opinion? In other words, have social groups changed their preferences regarding the state in the same way since 1990?

34The idea that there are similar movements within public opinion was first formalized and empirically confirmed by two American political scientists, Benjamin Page and Robert Shapiro (1992). Examining the evolution of American opinions on a set of social issues between 1972 and 1990, they conclude that: “different social groups tend [authors’ emphasis] to change their preferences neither very often nor very differently. Among most of them, opinions tend to change (or not change) in roughly the same way: in the same direction and more or less to the same extent and at the same time.” This idea was later confirmed over a longer period by other authors (Erikson, MacKuen and Stimson, 2002; Uri and Ellis, 2008) and for other American data (Enns and Kellstedt, 2008).

35It is therefore interesting to observe that the “parallel publics” hypothesis is well supported by European data from the “Values” survey. Several of the factorial planes presented below and in the Appendix (Figure A2) indeed show that social groups became more interventionist between 1990 and 2008. This is particularly the case:

• by gender, although each survey wave shows that women make greater demands on that state than men;
• by age, although respondents aged between 45 and 59 are always more inclined towards interventionism;
• by household income: [13] whatever year, the degree of interventionism seems inversely proportional to income, but the different income groups have evolved in a parallel way on the horizontal axis (Figure 2a);
• by political orientation: [14] in 2008 as in 1999 and 1990, respondents on the left were much more interventionist than respondents on the right, the centre or those that refuse to position themselves on the left–right scale. All ideological groups, however, have moved in the same direction and have greater expectations of the state (Figure 2b).

Figure 2a

Evolution of interventionism by household income (1990–1999–2008)

Evolution of interventionism by household income (1990–1999–2008)

Figure 2b

Evolution of interventionism by political orientation (1990–1999–2008)

Evolution of interventionism by political orientation (1990–1999–2008)

Interpretation: Figures 2a–2b. The more a point is situated to the east of the factorial plane, the more it indicates an interventionist attitude. The more a point is situated to the north, the more it shows a confident attitude. For example, respondents living in a wealthy household in 2008 (Q4_08) are more interventionist and confident than their counterparts in 1999 (Q4_99).
Note: The modalities were introduced as illustrative modalities; i.e., they are not taken into account in building the dimensions. For gender, age, age at end of education and employment status, the results go in the same direction. The factorial planes showing countries are plotted in the Appendix (Figure A2).

Interests vs. values: a false opposition

36To clarify the effects of the characteristics of individuals and countries on interventionism, several multilevel models were evaluated. This type of model, which is appropriate for analysing hierarchialized data, enables us to answer two questions that can be stated thus: are demands on the state better explained by differences between European countries or by differences within these countries? How does this demand on the state vary from one country to another, once certain characteristics inherent to countries and the individuals that compose these countries are taken into account?

37The principles of multilevel analysis can be broken down into different stages (Bressoux, 2008). The first stage is to estimate the so-called “empty” model. This model compares the percentage of variance of a dependent variable, [15] explained by the individual level and the percentage of variance explained by one (or several) aggregate levels (Table 2, model 1). Here we can calculate that 97% of variation in interventionism is attributable to individuals while only 3% is attributable to countries. [16] This initial finding indicates that interventionist attitudes are much more dependent on differences between Europeans than those between European Union countries. It allows us to put culturalist theses into perspective, which often presuppose that collective preferences are very homogenous and vary more between one country and another than from one individual and another.

Table 2

Multilevel models explaining interventionism by different characteristics of European Union individuals and countries

Model 1 Model 2 Model 3 Model 4 Model 5 Fixed effects Coef p-value Coef p-value Coef p-value Coef p-value Coef p-value Constant 0.001 –1.07*** –1.39*** –2.40*** –2.62*** Individual variables Woman (ref. man) 0.21*** 0.20*** 0.20*** 0.20*** 18–29 years (ref. 60 +) 0.05** 0.02 0.02 0.02 30–44 years 0.11*** 0.07*** 0.07*** 0.07*** 45–59 years 0.17*** 0.14*** 0.14*** 0.14*** Less than 30 h 0.45*** 0.39*** 0.39*** 0.39*** (ref. self-employed) More than 30 h 0.38*** 0.34*** 0.35*** 0.35*** Unemployed 0.78*** 0.72*** 0.73*** 0.73*** Inactive 0.40*** 0.36*** 0.37*** 0.37*** Retired 0.50*** 0.45*** 0.45*** 0.45*** Q1 (ref. Q4) 0.42*** 0.38*** 0.43*** 0.72*** Q2 0.35*** 0.32*** 0.33*** 0.65*** Q3 0.22*** 0.20*** 0.19*** 0.39*** IncDKNA 0.15*** 0.14*** 0.17*** 0.45*** Left (ref. Right) 0.87*** 0.87*** 1.32*** Centre 0.40*** 0.40*** 0.63*** IdeoDK 0.52*** 0.53*** 0.82*** IdeoNA 0.43*** 0.47*** 0.57*** Left*Q1 (ref. Left*Q4) –0.62*** Left*Q2 –0.61*** Left*Q3 –0.41*** Left*IncDKNA –0.62*** Centre*Q1 (ref. –0.29*** Centre*Q4) Centre*Q2 –0.33*** Centre*Q3 –0.20*** Centre*IncDKNA –0.30*** IdeoDK*Q1 –0.36*** (ref. IdeoDK*Q4) IdeoDK*Q2 –0.41*** IdeoDK*Q3 –0.30*** IdeoDK*IncDKNA –0.40*** IdeoNA*Q1 –0.22** (ref. IdeoNA*Q4) IdeoNA*Q2 –0.22** IdeoNA*Q3 –0.08 IdeoNA*IncDKNA –0.08 Aggregate variables EVS 1999 0.22*** 0.20*** 0.22*** 0.21*** (ref. EVS 1990) EVS 2008 0.36*** 0.34*** 0.35*** 0.34***

Model 1 Model 2 Model 3 Model 4 Model 5 GDP2008+ 0.82*** 0.82*** (ref. GDP+++ in 2008) GDP2008++ 0.47*** 0.47*** GDPEvol++ 0.00 0.00 (ref. GDPEvol+ – ∆ 1990–2008) GDPEvol+++ 0.68*** 0.70*** UnemployRate2008+++ 0.23*** 0.23*** (ref. UnemployRate+ and ++ in 2008) WelfareSpend2008+++ 0.32** 0.32** (ref. WelfareSpend+ and ++ in 2008) WelfareSpendEvol+ –0.30*** –0.30*** (ref. WelfareSpendEvol++ – ∆ 2000–2008) WelfareSpendEvol+++ 0.46*** 0.47*** GINI2008++ –0.27* –0.27* (ref. GINI+ in 2008) GINI2008+++ 0.48*** 0.48*** GINI2008++++ 0.23* 0.23* Random effects Level 1 (individuals): 1.733 (0.009) 1.658 (0.009) 1.581 (0.009) 1.559 (0.008) 1.552 (0.008) Intracountry variance (error type) Level 2 (countries): Variance of constants 0.046 (0.014) 0.041 (0.012) 0.039 (0.012) 0.216 (0.067) 0.212 (0.067) (error type) Variance of slopes 0.034 (0.010) 0.034 (0.010) (error type) Constants–slopes –0.084 –0.083 covariance (error type) (0.026) (0.026) N (individuals) 68,472 67,121 67,121 67,121 67,121 N (countries) 23 23 23 23 23 Deviance 232,075.2 224,536.6 221,312.4 220,445.0 220,159.0 Akaike information 232,081.2 224,572.7 221,356.5 220,514.9 220,261.0 criterion (AIC) Bayesian information 232,108.6 224,736.7 221,557.0 220,833.9 220,725.8 criterion (BIC)

Multilevel models explaining interventionism by different characteristics of European Union individuals and countries

Interpretation: According to model 3, a man responding to the survey in 1990 and corresponding to the reference situation (aged 60 or more, self-employed, living in a high income household and positioning themselves on the right politically) would have a score of –1.39 on the interventionism scale. A woman (+0.2) aged between 45 and 59 years (+0.14), unemployed (+0.72), living in a low income household (+0.38), positioning themselves on the left politically (+0.87) and responding in 2008 (+0.34) would have an interventionism score 2.65 points higher, reaching 1.26.
According to model 4, if a woman with the same characteristics lives in a country with a low GDP (+0.82) that grew moderately between 1999 and 2008 (0.00), but where there is high welfare spending (+0.32) that greatly increased between 2000 and 2008 (+0.46), where the unemployment rate in 2008 (+0.23) and income inequality are greater (+0.23), she would have an interventionism score that is 2.06 points higher.
Note: *** p < 0,001; ** p < 0.01; * p < 0,05.

38The next stage of the multilevel analysis is to introduce several independent variables into the model in order to measure different fixed effects at the individual level. These fixed effects correspond here to the normal parameters of a linear regression. For example, model 3 aims to assess the impact of gender, age, household income levels, employment status, political orientation, and survey wave on interventionism. Model 3 is more powerful than the empty model. The six independent variables explain around a tenth of the intracountry variance in interventionism scores, and 15% of intercountry variance of the same scores. [17] In other words, for analysis of this type, it accounts for a modest, but not negligible, portion of the differences between individuals and countries.

39Models 2, 3, 4 and 5 provide the non-standardized coefficients (β) of sociodemographic characteristics introduced in the form of “dummy” variables (the absence or presence of each modality is coded 0 or 1; the latter modality, which is excluded from the model, thus serves as a reference). We can thus easily reconstruct the equation of the regression line. The ordinate at the origin is the β value of the table’s constant, while the slope is indicated by the β value for an independent variable. The effect on interventionism of each independent variable is then read according to the following equation: predicted value of interventionism = constant + β.

40In sociological literature, economic attitudes are often understood in terms of ex post rationalization. Above all, attitudes are seen to enable individuals to justify their objective interests. These justifications are more or less complex depending on the context, the actors involved and the material and symbolic resources the individuals possess. They are also more or less explicit depending on the individuals’ intentions, the degree of awareness of their interests and the means they can use to defend them. The relationship between interests and attitudes is therefore thought of as circular or “practical” (Bourdieu, 1980): interests engender attitudes, which in turn legitimize the interests that inform them.

41Internationally, empirical studies on the perception of the role of the state confirm that individuals tend to develop economic preferences that conform to their personal interests or their representations of these interests (Svallfors, 1997; Edlund, 1999). But they also highlight that, alongside interests, elements relating to values can enter the equation (Sears et al., 1980; Sears and Funk, 1990). How is this translated here?

42Figure 3 presents the most significant coefficients for the variables estimated in model 3. It is a statistically robust illustration (i.e., “all things being equal”) of trends emerging from the PCA mappings. Firstly we observe that gender has a very significant effect on interventionism, men being less in favour of a growing role for the state in economic life than women. This result confirms studies on the greater propensity of women to support the welfare state. In the literature, this propensity is generally explained by the gender division in moral sentiments, which leads women to develop empathetic dispositions and show greater solidarity towards others and thus to be more inclined to equality and redistribution (Arts and Gelissen, 2001). But women’s interventionism can also be explained in some cases by their objective dependence on the state: more often employed by the state or beneficiaries of certain social welfare assistance, they more willingly support public policies they are likely to benefit from (Sainsbury, 1996).

Figure 3

The effects of different individual and aggregated characteristics on interventionism (with 95% confidence intervals)

The effects of different individual and aggregated characteristics on interventionism (with 95% confidence intervals)

43The relationship between age and level of interventionism takes the form of a “bell” curve: being aged between 30 and 44 years or between 45 and 59 years, rather than being 60 years or over, tends to increase the interventionism score (by a tenth of a point, for example, for the 45–59-year-olds). 18–29-year-olds are not significantly different from those aged 60 and over. This result is consistent with young people’s greater economic liberalism observed in France, and undoubtedly reflects their comparative distance from the economic world and its constraints (Gonthier, 2012). The size of these coefficients, however, remains quite moderate and invites more detailed study on the impact of age on interventionism.

44The effect of employment status, however, is much clearer. Compared with the self-employed, all Europeans are more likely to endorse an interventionist attitude. This is particularly the case for the unemployed: the fact of being without employment thus increases the interventionist score by 0.72 points, against half this amount (+0.36 points) for the inactive (housewives, students, etc.). In this trend we can see an illustration of the dependency of the unemployed on the state; their support for state intervention thus expresses support for public policies on unemployment from which they personally benefit. But it can also be interpreted as showing a more profound support for the institutional programme underpinning the welfare state.

45The relationship between household income and level of interventionism is a declining linear relationship: the higher the disposable income, the lower the demands on the state. The Europeans who live in vulnerable households are thus more in favour of the state (+0.38 points); which might either reflect agreement with the values conveyed by the state or a belief that growing state interventionism would lead to an improvement in their economic situation through redistributive policies.

46Having performed a multilevel analysis on a dataset aggregating several surveys, binary variables corresponding to the survey waves were estimated. These variables allow us to control for the effects of the sociopolitical and socioeconomic contexts of each of the surveys. It also provides us with a measure of the evolution of economic preferences over time that is more detailed and statistically sound than the factorial planes previously presented. Figure 3 shows that the fact of being interviewed in 2008 or 1999 rather than 1990 significantly increases the probability of developing strong expectations of the state. Keeping all independent variables constant, Europeans were thus more interventionist in 1999 than in 1990 and also more interventionist in 2008 than in 1990.

47We know that political affiliation is likely to have a significant influence on economic attitudes (Hasenfeld and Rafferty, 1989; Jacobi, 1994; Gelissen, 2000). People with different political orientations but with the same incomes or employment situations can, for example, develop very different economic preferences. Unlike the circular schema set forth above, symbolic orientations can thus modulate personal interests and dampen their influence without necessarily legitimizing them.

48It is interesting to consider model 3 in this perspective. It enables us to test the hypothesis of an effect of political values, and to compare the importance of this impact with the impact of personal interests (measured here by income levels and socio-occupational situation). It is immediately obvious, by comparing models 2 and 3, that the introduction of the political orientation model does not greatly modify the significance or the magnitude of the coefficients of other explanatory variables; which tends to confirm that ideology and interests have specific influences on attitudes towards the state.

49Note then that political orientation is as good a predictor of interventionism as income or employment status—a better one even. In other words, the most vulnerable Europeans are more interventionist than the less vulnerable Europeans; but the fact of being on the left rather than the right is strongly linked to marked support for the state. Specifically, the fact of being a left sympathizer increases the interventionism score by 0.87 points in comparison to the reference situation. The Europeans who do not position themselves politically, moreover, resemble those on the left and centre: they are much more inclined toward interventionism than those on the right.

50The direction of the relationship observed here between interventionism and political orientation merits discussion. If we are statistically rigorous, the coefficients presented do not allow us to interpret this relationship in terms of causality. We can, however, rely on political psychology studies to clarify its meaning. It is also acknowledged that individual attitudes are fashioned by symbolic predispositions transmitted by parents and peer groups (Almond and Verba, [1963] 1989; Inglehart, 1990). Of these symbolic predispositions, inherited political orientations are considered to be very stable and produce persistent effects, particularly on attitudes connected to political attitudes, such as, for example, attitudes towards minority groups (Sears and Funk, 1999). The fact that support for the state is strongly correlated with political preferences thus illustrates the influence of ideology on opinions in economic matters. In other words, the greater or lesser propensity of individuals to interventionism can be considered an attitudinal reflection of deep-rooted political values acquired during primary socialization.

51Ultimately we observe that the oft-cited opposition between ideology and interests does not stand up to empirical examination. Echoing David A. Sears and Carolyn L. Funk’s (1991) findings on the combined influence of economic interests and political values, our results show that being unemployed and being on the left both have significant but distinct impacts on interventionist attitudes, i.e., that their respective effects are of a strong and comparable intensity when we control for gender, age, respondents’ incomes and survey wave.

Similarities between micro- and macro-inequalities

52The last stage of a multilevel analysis is generally to estimate the fixed effects of independent variables documented at the aggregate level (for example, demographic and economic data by country). This means seeing to what extent these explain the variance between aggregates and whether their predictive power is more significant than the independent variables reported at the individual level.

53One of the principal challenges for the comparative analysis of economic attitudes is to be able to link changes at the micro- and macrosocial levels. This difficulty is due in part to opinion data, which are not compiled to be checked against major societal trends. From this point of view, although the questions in the “Values” survey enable us to understand the role of the state in general, relating them directly to developments in public policies and national economic aggregates is problematic. But the difficulty is also due to the availability of appropriate and exhaustively reported indicators. Large international organizations and their statistical agencies today provide several global indicators of wealth and inequality. These indicators are, however, ill-adapted for understanding the degree of economic liberalism of European societies (Gadrey and Jany-Catrice, 2005; Lebaron, 2011). And they remain poorly documented in the long term. There are similar problems with more academic sources. [18]

54Comparing changes at the micro- and macrosocial level is an important issue for this analysis. Are attitudes towards the state linked to levels of wealth and inequality in different European Union countries? To what extent do they depend on the degree of economic liberalism of these countries? To attempt to respond to these questions I used two traditional indicators of wealth (GDP per capita from 1990 to 2008, source: UNDP) and inequality (the Gini coefficient in 2008, source: Eurostat- SILC). Two other macro-indicators were used. Per capita welfare spending in 2008 (source Eurostat) enables us to understand national differences in terms of state regulation. The unemployment rate in 2008 (source: UNDP) enables us to examine, in a complementary way, expectations in terms of state protection. These indicators were encoded as terciles or quartiles, ranging from the lowest to the highest value for all 23 countries analysed. The encoding is designed to test the hypothesis of a non-linear relationship with interventionist attitudes. Macro-indicators were introduced in the form of binary variables to make a comparison with other explanatory variables.

55Model 4 shows the macro-indicators that have the most significant impact on interventionism once the same sociodemographic characteristics as in model 3 are controlled for. According to current statistical criteria (AIC and BIC), model 4 fits the data better than the two previous models. The introduction of macro-indicators also raises the inter-country variance explained by the model by 12% (0.216/(0.216 + 1.559)). This means that these macro-indicators account for more than 10% of the observable differences between countries. In other words, a significant proportion of variation in interventionist attitudes at the country level depends on overall inequalities between these countries. [19]

56Unlike model 3 with random constants (i.e., countries can take different values for interventionism but the lines that represent them are “constrained” and remain parallel), model 4 is a comprehensive multilevel model. It estimates regression lines with constant and random slopes. Specifically, the coefficients of the slopes and regression lines are allowed to vary here from one country to another by survey wave. In other words, it tested a differential hypothesis by allowing the relationship between interventionism and the countries to vary depending on survey context, which meant heeding the fact that interventionist attitudes can have different meanings, not only from one country to another, but at one time or another in these countries.

57The first result to pick out is that Europeans living in a country where GDP in 2008 is low (GDP2008+) or moderately high (GDP2008++) are more interventionist than those who live in a country with a very high GDP. This result can be interpreted in the following way: inhabitants of a country whose macro-economic performance is poor seek compensation for this poor performance and favour a socially generous state (Brooks and Manza, 2007).

58A similar observation can also be made regarding the unemployment rate (although the magnitude of the coefficient is smaller): respondents from countries with a high unemployment rate in 2008 (UnemployRate2008+++) develop greater expectations of the state than other Europeans. This relationship is consistent with Morten Blekesaune and Jill Quadagno’s (2003) results. Like them, we might hypothesize that a high unemployment rate creates a certain sympathy within public opinion towards the unemployed. Individuals think they themselves are at risk of becoming unemployed; are more likely to be in contact with someone affected by unemployment; and political elites are more willing to put this issue on the agenda.

59The trends at the aggregate level on these two first macro-indicators are consistent with those at the individual level: just as the most economically disadvantaged individuals are the most inclined towards interventionism, the least advantaged countries (in terms of GDP per capita and unemployment rate) are also those where interventionism is most marked.

60The relationship between interventionism and the amount of welfare spending per capita in 2008 is more complex. One might expect that interventionism and welfare spending vary inversely; demands on the state being lower when a country allocates a significant proportion of its resources to social protection, and thus responds to the expectations of individuals in terms of interventionism. The results show the reverse, that intervention reaches the highest levels in countries where welfare spending is highest (Welfspend2008+++). This result suggests that expectations of the state do not directly depend on the quantity of public resources, but that they can be more sensitive to other factors, such as, for example, the way in which these resources are used.

61The differences between countries are even clearer if one examines the changes to macro-indicators over time. One might imagine that if not linearly related to the amount of welfare spending, interventionism would be tempered by its upward trend. Countries investing in social protection should thus be able to progressively respond to the social expectations of individuals. Here again, the results from model 3 are counterintuitive: it is among the respondents from countries where welfare spending has increased the most since 2000 (WelfSpendEvol+++) that interventionism is the most pronounced. In fact, it is weaker in countries where welfare spending has increased the least [20] (WelfSpendEvol+) compared to those where it grew moderately. It is as if a large increase in welfare spending leads to increased demands on the state, and a small increase in welfare spending to fewer demands.

62The impact of changes in GDP seems to obey the same logic. While one might expect that demands on the state would be weaker in countries where the level of wealth has grown the most between 1999 and 2008, on the contrary it is in these countries that interventionism is the most marked (GDPEvol+++) in comparison to countries where GDP has increased the least. This result only confirms that social expectations are relatively elastic and may combine with the level of country wealth in a variety of ways. On the one hand, it is indeed the respondents in the poorest European countries who are the most interventionist (GDP+). But, on the other, respondents are more interventionist when they live in a country becoming wealthier (GDPEvol+++).

63In sum, we cannot observe a simple and perfectly linear relationship between the degree of interventionism and degree of wealth or inequality of European countries. The Gini coefficient, which measures the inequality in incomes in a country, provides a very good illustration of the complexity of the relationship between inequality and demands on the state. In comparison to the most egalitarian countries (i.e., where the Gini is the lowest in 2008—GINI2008+), it is in the most unequal countries (i.e., where the Gini is the highest—GINI2008+++ and GINI2008++++) that Europeans have developed the greatest expectations of the state.

64This result helps to confirm the hypothesis that social expectations are the most strongly asserted where inequality is greater. Once again, this fits with the correlation identified at the macrosociological level between interventionism and individual interests. Just as the least advantaged respondents are the most inclined to interventionism, respondents living in the most unequal countries are also more inclined towards interventionism. However, in countries where the Gini is low but not the lowest (GINI2008++), respondents tend to be less interventionist than in the most egalitarian countries (GINI2008+). Thus we are far from confirming the homologies between inequality at the macrosociological and microsociological levels.

The rise in interventionism across virtually the entire European Union

65How have interventionist attitudes evolved? The multilevel models presented provide evidence of these changes over time all other things being equal. We have thus seen that demands on the state were greater in 2008 and 1999 than in 1990 when we control certain characteristics of individuals and countries. To better understand these changes, it is possible to graphically present the interventionist score regression lines depending on survey wave for each of the 23 countries of the European Union.

66Figure 4 shows that almost all of these countries have become more interventionist (with the exception of the United Kingdom, Portugal, Bulgaria and Romania which will be discussed below). But is this also the case for all publics within these countries? The evolution of interventionism country by country by employment status groups, income groups and ideological groups is presented in the Appendix (Figures A3). The trend to increasing interventionism observable at the global (European Union) and local (country) levels is found at the level of these different social groups, the movements of which are highly parallel within each of these countries. These figures also confirm the results of multilevel models for countries taken individually. Whether interventionism is rising or falling in each of these countries, it is notable that the unemployed are systematically more interventionist than the employed and self-employed; people from low-income households are more interventionist than those in wealthy ones; and left-wing individuals are more interventionist than those on the right wing.

Figure 4

Evolution of interventionism in the European Union by country

Evolution of interventionism in the European Union by country

Interpretation: The vertical axis corresponds to the interventionism scale, the horizontal axis to the survey wave. A positive slope indicates that a country becomes more interventionist between 1990 and 2008; a negative slope that it becomes less so. The inclination of the slope represents the size of this change. For example, the Czech Republic is much more interventionist in 2008 than in 1990. Romania is only slightly less so.
Note: Here the regression lines represent the “adjusted” values for interventionism for model 4. These values combine the values predicted by the fixed effects part of the model with the specific contributions based on the random effects predicted by the model.

67These graphs do not however show the way in which economic and ideological characteristics can interact. Political positioning can indeed act as a moderating variable for the relationship between interventionism and individual interests. It has been shown that, in France, left-wing individuals, even with very different incomes, have much more uniform opinions of the role of the state than those on the right (Degeorges and Gonthier, 2012). Model 5 puts this observation to the test by estimating not just the main effects of political orientation and incomes on demands on the state, but also an interaction effect between them.

68Figure 5 represents the linear predictions for these two variables. It shows that right-wing Europeans are even less interventionist when they are economically advantaged, while support for the state from Europeans on the left is much more homogeneous and only varies slightly depending on income level. For the wealthiest left-wing Europeans, political orientation seems to blunt the influence of interests: their ideology leads them to support opinions in favour of the state even when their income level “should” incite them to be less interventionist. [21] As for right-wing Europeans, they develop preferences that conform more to their ideology (with the same incomes, they are much less interventionist than their left-wing equivalents) and to their interests (they are even less interventionist when they have high incomes; anti-interventionism being even stronger among the richest of them).

Figure 5

Interventionism by ideological group and by income level (with 95% confidence intervals—23 countries, 1990–1999–2008)

Interventionism by ideological group and by income level (with 95% confidence intervals—23 countries, 1990–1999–2008)

Interpretation: A positive value indicates an interventionist orientation, a negative value an anti-interventionist orientation.
Note: The linear predictions were calculated based on the fixed effects component of model 5.
The results are similar to predictions adjusted for fixed and random effects. Additional graphs, included in the Appendix (Figures A4), show that these results can be found in the majority of European Union countries and for each survey wave.
The predicted values are the average of the individual marginal effects for a given sub-group and can be interpreted as average deviation from the mean.

69It is also interesting to note that the Europeans who refuse to position themselves on the left–right scale (IdeoDK) and those who do not answer this question (IdeoNA) have a level of interventionism that is similar to Europeans close to the centre. The former, however, more closely resemble those on the left (their level of interventionism varies little depending on income level), while the latter are closer to right-wing individuals less in favour of the state.

70These results suggest that social vulnerability is linked to interventionism, but it is not a sufficient condition. As Yeheskel Hasenfeld and Jane A. Rafferty (1989) suggested based on a smaller sample (i.e., 682 inhabitants of Detroit interviewed in 1983), political values mediate individuals’ objective interests. They can thus strongly modulate expectations of the state. The results presented here also invite a reconsideration of the literature on the decline of the “class vote” (Cautrès and Mayer, 2010). Contrary to the received idea that the role of the state has not been a strong ideological marker since the liberal turn on the European left during the 1980s, we find that the divide between left- and right-wing Europeans in terms of interventionism is far from being effaced; and that ideological groups are in fact as polarized in 2008 as they may have been in 1999 and 1990.

71Finally, we can examine four countries (United Kingdom, Portugal, Bulgaria, and Romania) in which public opinion contradicts the European trend and is becoming less interventionist. Some explanation can be found in the second factor of the PCA, which summarizes the level of confidence in the social functions of the state and major actors in economic life. Figure 6a shows the evolution in “adjusted” interventionism values and confidence in the four countries in which interventionism has declined, comparing them to other European Union countries. We see that confidence in institutions in the United Kingdom, Portugal, Bulgaria and Romania, already very low in 1990, plummeted between 1990 and 2008; while, in other countries, confidence that was comparatively higher in 1990 instead increased between 1990 and 2008.

72Figure 6b clarifies the comparison by representing the change in confidence in institutions for different social groups. In the United Kingdom, Portugal, Bulgaria and Romania, the confidence deficit seems to particularly affect the groups that in the rest of the European Union support the state the most. Thus the unemployed exhibit much lower confidence scores than their equivalents in countries where interventionism has increased. As for disadvantaged households (Q1), they completely “plummet” and fall to the bottom [22] of the confidence scale in 2008.

73The confidence deficit among these publics of course has an impact on their respective interventionism. While their equivalents in other countries became more interventionist between 1990 and 2008, the unemployed and most disadvantaged households tended to become much less so (Figure 6c). We could summarize these trends by saying that in the United Kingdom, Portugal, Bulgaria and Romania, the rise in interventionism is somehow “blocked” by a deficit in confidence in institutions, and that the support of the most vulnerable publics, which represent the driving force of the rise in interventionism at the European scale, is thus lacking in these countries. [23]

Figure 6

Evolution of interventionism and confidence in institutions

a

By group of countries

By group of countries

b

Evolution of confidence in institutions depending on different social group and by group of countries

Evolution of confidence in institutions depending on different social group and by group of countries

c

Evolution of interventionism depending on different social group and by group of countries

Evolution of interventionism depending on different social group and by group of countries

Evolution of interventionism and confidence in institutions

Note: Figures 6a-6b-6c. Interventionism declines in the United Kingdom, Portugal, Bulgaria, and Romania. The regression lines for overall confidence represent “adjusted” values of a model identical to model 4, but in which the dependent variable corresponds to the factorial coordinates/correlations of individuals on the second PCA factor.

74We can draw two general conclusions from this analysis. The first is a methodological one and calls for a multidimensional approach to economic attitudes. Far from being reducible to a single indicator, economic preferences encompass multiple dimensions and combine in a complex way. This can be seen here with confidence in institutions, where changes over time directly affect changes in interventionism. In most European Union countries confidence and interventionism rose in unison between 1990 and 2008. But in other countries, the ebb in confidence seems to have given rise to a decline in expectations of the state (and vice versa). The second conclusion is more theoretical and extends classical sociological studies that, from Simmel to Putnam via Goffman and Luhmann, see interpersonal confidence as the foundation of social life. Our results provide superficial confirmation that confidence in institutions constitutes the foundations of demands on the state. But they also show that, in most European Union countries, confidence and interventionism evolved concomitantly between 1990 and 2008. This is anything but trivial at a time when there is no shortage of theses concerning the alleged rise in mistrust in institutions and the welfare state (Algan, Cahuc and Zylberberg, 2012).

75* * *