Introduction
Preferences for redistributive social spending play a central role in many of the most influential theories of democratic politics and political economy. From Downs’ (Reference Downs1957) median voter theorem, to Meltzer and Richard’s (Reference Meltzer and Richard1981) models of government redistribution, to Iversen and Soskice’s (Reference Iversen and Soskice2006) examination of the impact of electoral rules on levels of redistribution, models of political economy frequently begin with the simplifying assumption that democratic politics can be reduced to a single left–right continuum defined by preferred levels of redistributive taxes and spending. Traditionally, these models have assumed that preferences for redistribution are principally, if not completely, determined by individuals’ relative income. More recently, driven by increasingly high-quality cross-national survey data, researchers have sought to better understand the factors that explain variations in preferences for redistribution both among individuals and across different national contexts. This research has significantly expanded on the conventional view by describing a more nuanced and, by extension, complex calculation of economic self-interest, incorporating the role of unemployment risks (Blekesaune and Quadagno, Reference Blekesaune and Quadagno2003), skill-levels (Iversen and Soskice, Reference Iversen and Soskice2001; Rehm, Reference Rehm2011), and the structure of welfare institutions (Svallfors, Reference Svallfors1997; Bean and Papadakis, Reference Bean and Papadakis1998; Korpi and Palme, Reference Korpi and Palme1998). In addition to economic incentives, researchers have identified a range of non-economic motivations for support of redistributive social policies including social trust (Edlund, Reference Edlund1999), racism (Alesina and Glaeser, Reference Alesina and Glaeser2004), attitudes toward immigration (Finseraas, Reference Finseraas2008), and religion (Jordan, Reference Jordan2014).
This new body of work provides a more nuanced vision of redistributive preferences than the simplifying assumptions of political economy models, but it has also generated an image of citizens engaging in a complex process of preference formation based on balancing sophisticated calculations of economic self-interest with a variety of ethical and ideological concerns. This picture of preference formation runs counter to evidence from political psychology challenging the assumption that opinions expressed in surveys about policy issues reflect well-developed, or even coherent, preferences. Developed largely in the context of the study of American public opinion, this research demonstrates that many, if not most, citizens lack clearly formed opinions and attitudes concerning complex policy issues (Converse, Reference Converse1964, Reference Converse2000; Zaller Reference Zaller1992, Reference Zaller2012; Delli Carpini and Keeter, Reference Delli Carpini and Keeter1996). There exist significant differences in levels of political interest and knowledge across the population, which influence the extent to which individuals’ develop, hold, and express policy preferences that conform to their economic self-interests or ideological predispositions. From this perspective, the traditional model of individuals providing well-reasoned attitudes grounded in ideological concerns and calculated self-interest assumed by much of the cross-national literature on redistributive and social policy preferences is only appropriate for a relatively small share of the population at the highest levels of political sophistication.
The comparative study of redistributive and social policy preferences largely ignores the significant differences in levels of political interest and awareness across the population, raising a number of important questions. First, to what extent do individuals’ expressed opinions on surveys reflect well-reasoned and consistent preferences for redistribution and the welfare state? Second, does political awareness influence the extent to which individuals appear to hold and express social policy preferences that are internally consistent in a way assumed by traditional models? Third, does political awareness influence how well individuals connect their social preferences to their economic self-interests and expressed ideological predispositions?
This paper begins to answer these questions by exploring the relationship between political awareness and both the character and structure of social policy attitudes across 14 countries in Europe. The first section outlines existing theory and evidence on the role of political knowledge and interest in shaping preference formation before turning to an exploration of the extent to which individuals’ appear to hold and express coherent and consistent preferences about the welfare state. This discussion is followed by an examination of the influence of political awareness on individuals’ ability to connect their economic self-interests to particular policy preferences. The conclusion discusses the implications of these findings for cross-national research on social policy attitudes.
Political awareness and preference formation
The growing body of research on social policy preferences has substantially improved our understanding of the causes of cross-national and individual-level variations in support for redistributive social policies. This research provides a significantly more nuanced picture of preference formation than that suggested by the simplifying assumptions of many common political economy models; however, this literature remains committed to a vision of individuals capable of easily linking their economic interests and ideological predispositions to preferences for social policy. This image of reasonably well-informed citizens developing and expressing sophisticated policy preferences is increasingly inconsistent with empirical evidence. Converse’s (Reference Converse1964, Reference Converse2000) seminal work on public opinion revealed significant deficiencies in political knowledge among the American public with large disparities in political interest and sophistication. Analysis of American public opinion consistently shows that large numbers of citizens cannot answer basic facts about the political system (Converse, Reference Converse1964; Delli Carpini and Keeter, Reference Delli Carpini and Keeter1996; Althaus, Reference Althaus2003; Bishop, Reference Bishop2005) and struggle to effectively connect their preferences to specific policy choices or party platforms (Lewis-Beck et al., Reference Lewis-Beck, Jacoby, Norpoth and Weisberg2008). Though reliable cross-national data are limited, recent evidence suggests similarly low levels of political knowledge in countries outside of the United States (Gronlund and Milner, Reference Gronlund and Milner2006) with potentially significant variation across countries (Fraile, Reference Fraile2013).
Recognizing the relatively low level of political awareness among the population has important consequences for understanding how individuals develop and express political preferences. Examining the psychological processes of preference formation and expression, Zaller (Reference Zaller1990, Reference Zaller1992) argues that when faced with questions concerning complex policy questions, those with limited political awareness formulate opinions on the spot by drawing upon a host of only partially consistent ideas and ‘considerations’ that come to mind at the moment the question is asked. The consequence of this ‘on the fly’ opinion formation is that the expressed opinions of those with low levels of political awareness are highly unstable both within a given survey (Schuman and Presser, Reference Schuman and Presser1981; Zaller, Reference Zaller1992) as well as over time (Converse, Reference Converse1964; Feldman, Reference Feldman1989; Delli Carpini and Keeter, Reference Delli Carpini and Keeter1996). Moreover, these low information individuals are highly influenced by seemingly minor changes to question wording, the order of questions, response option ordering, as well as the perceived social desirability of particular responses (Pasek and Krosnick, Reference Pasek and Krosnick2010). The combined effect of these problems is that the expressed preferences and attitudes of the least informed may appear largely indistinguishable from random variation (Converse, Reference Converse1964, Reference Converse2000).
The research on political knowledge and awareness thus undermines the assumption of comparative opinion research that individuals necessarily express, or even hold, well-reasoned preferences about social policy, grounded in economic self-interest and broader ideological considerations. Research in the American context suggests that only a small section of the public appears to exhibit ‘ideological constraint’ in the sense that their preferences are consistent both within and across issue areas (Delli Carpini and Keeter, Reference Delli Carpini and Keeter1996; Abramowitz, Reference Abramowitz2010; Frederico and Hunt, Reference Frederico and Hunt2013). Moreover, the ability of respondents to link their economic self-interest to specific policy preferences is determined by their level of political knowledge and interest. For example, data from the United States suggests that the expected link between income and support for redistributive social policy programs (Delli Carpini and Keeter, Reference Delli Carpini and Keeter1996; Kuklinski et al., Reference Kuklinski, Quirk, Jerit, Schwieder and Rich2000) and left-wing political parties (Andersen et al., Reference Andersen, Tilley and Heath2005; Arnold, Reference Arnold2012) is significantly higher among the politically aware than those with only limited knowledge and interest.
Though significant evidence exists concerning the effects of political awareness on survey responses, critics suggest that these effects may have limited practical effects due to the ability of voters to make effective choices using relatively limited amounts of information in real-world environments (Popkin, Reference Popkin1994; Lupia and McCubbins, Reference Lupia and McCubbins1998). Yet, the large body of literature on social policy preferences is primarily concerned with the determinants of support for redistribution and particular welfare policies. As such, it is difficult to ignore questions concerning the extent to which expressed preferences reflect well-reasoned attitudes. If individuals are shown to lack well-developed or reasoned opinions, this calls into question the underlying assumption of well-informed respondents underlying these models. It also raises important questions concerning how individuals with low political interest and knowledge gather the limited information needed to develop preferences and make voting decisions. This is particularly important in a cross-national perspective where the political learning environments may be quite different due to differences in levels of unionization, educational quality, and party systems (Iversen and Soskice, Reference Iversen and Soskice2015).
In addition, the problems of ‘response instability’ may introduce significant statistical error into cross-national public opinion research. The high levels of measurement error produced by low levels of political awareness are not randomly distributed. Rather, the degree of measurement error varies systematically with levels of political awareness, which in turn are associated with many of the central predictors of social policy preference at the individual and national-level, including income, class, gender, education, inequality, and unionization rates (Fraile, Reference Fraile2013; Iversen and Soskice, Reference Iversen and Soskice2015). Because models of welfare preferences will systematically predict the preferences of the informed better than the uninformed, existing research may suffer from very high levels of unaccounted for heteroskedasticity, resulting in underestimation of standard errors and inappropriate inferences.
Despite the theoretical and empirical significance of political awareness for the study of social policy preferences, cross-national research largely ignores the issue while generating increasingly sophisticated models of preference formation. To rectify this situation, it is necessary to determine the extent to which political awareness influences the stability and internal coherence of welfare preferences. It is possible, given the high salience of social welfare politics in most Western societies, that even low levels of political knowledge may be sufficient to produce political preferences that are both internally consistent (i.e. exhibit ideological constraint) and meaningfully connected to individuals’ economic self-interests and ideological predispositions. The following two sections examine both the degrees of ideological constraint and its relationship to political awareness across a sample of European democracies.
Methodology
The availability and comparability of cross-national opinion data has significantly improved in recent years; however, few cross-national surveys contain the types of questions necessary for an analysis of political awareness and its implications for the structure and stability of social policy attitudes. An analysis of this type requires a large battery of different questions concerning social policy attitudes as well as a set of questions that allow for an assessment of individuals’ level of political awareness. Round four of the European Social Survey (ESS) conducted in 2008 provides the best opportunity for examining the effects of political awareness on redistribution and social policy attitudes in a multi-national context. The 2008 ESS interviewed individuals across 22 European countries with a special emphasis on social policy preferences. After considering data limitations and with an eye toward maintaining comparability with other research in the field, the sample was limited to 14 advanced capitalist democracies in Europe: Belgium, Denmark, Finland, France, Germany, Greece, Ireland, the Netherlands, Norway, Portugal, Spain, Sweden, Switzerland, and the United Kingdom.
The 2008 ESS contains a wide variety of questions on redistribution and the welfare state, providing a unique opportunity to investigate the structure and coherence of social policy preferences. The analysis focuses on support for redistribution because of its central theoretical importance both within the literature on public welfare attitudes and the broader political economy literature. Support for redistribution is measured through belief in the responsibility of government to reduce income inequalities. The question specifically asks whether ‘the government should take measures to reduce differences in income levels.’ Respondents were placed on a five-point scale varying from ‘strongly agree’ to ‘strongly disagree’ with higher values representing stronger support.
In order to examine the ideological consistency of support for redistribution with other expressed social policy preferences, the analysis includes a number of additional measures of support for the welfare state. First, preferences for overall social spending levels was gauged through a question asking individuals to place themselves on an 11-point scale between the two extreme policy positions ‘Government should decrease taxes a lot and spend much less on social benefits and services’ and ‘Government should increase taxes a lot and spend much more on social benefits and services.’ This question is commonly used as an overall measure of welfare support because it emphasizes the tradeoffs between social spending and taxation, which lie at the heart of many political economy models. The ‘welfare support index’ combines indicators of support for the government’s responsibility to address five social policy areas (unemployment, child care, sick leave, employment guarantees, and health care). The analysis also includes a ‘moral hazard’ index, which combines three measures that gauge the extent to which respondents believe that welfare policies reduce individual and social responsibility (e.g. Do ‘social benefits and services make people lazy?’). Respondents were also asked the extent to which they agreed or disagreed with two statements concerning their attitudes toward inequality. The first stated, ‘Large differences in people’s incomes are acceptable to properly reward differences in talents and efforts.’ The second statement, read ‘For a society to be fair, differences in people’s standard of living should be small.’ Together, these variables guage individuals tolerance for inequality, which should be associated with their overall support for government efforts to reduce income inequalities.
Each of these variables engages slightly different aspects of the welfare state, but all seek to determine the extent to which individuals support government intervention into the economy to support greater social equality. Mainstream political economy and public opinion models predict that these different attitudes should fit together into a reasonably coherent ‘pro’ or ‘anti’ welfare state position, which at least partially defines the left–right political spectrum. It is always possible to ‘thread the needle’ and imagine ways in which individuals could respond negatively to some of these questions and positively to others in an intellectually consistent way. For example, an individual might support social insurance (e.g. pensions and national health care) while simultaneously rejecting redistribution (Moene and Wallerstein, Reference Moene and Wallerstein2001). Yet, even in such an example, we would expect the variables to be correlated in the aggregate because, though it is possible to support social insurance and reject redistribution, the inverse position (i.e. supporting redistribution, but rejecting the social insurance programs used to produce it) is more difficult to see as intellectually consistent. In such a case, we would expect to see an overall correlation between these variables as supporters of redistribution should on balance have higher levels of support for the welfare state than those who are less supportive of redistribution. Moreover, taken separately, the theoretical expectations for each of these variables suggest effectively identical relationships with the most important causal variables (e.g. income is negatively correlated with all measures), which should produce an empirical correlation between the variables even if the questions are to some degree intellectually distinct. As a result, even if, as outside observers, we could imagine different ideological positions that could disentangle any two or three of these variables, as a battery of seven indicators, they should in practice be highly correlated with one another, if individuals give responses consistent with their economic self-interests and ideological predispositions.
The literature on political awareness is characterized by a significant diversity in measurement with scholars frequently relying upon whatever limited measures are available (Luskin, Reference Luskin1987; Zaller, Reference Zaller1992; Delli Carpini and Keeter, Reference Delli Carpini and Keeter1993). Though fact-based knowledge quizzes have become increasingly common in the literature on American public opinion (Zaller, Reference Zaller1992; Converse, Reference Converse2000), knowledge quizzes are rare in cross-national surveys and those that do exist (e.g. the European Election Studies or Comparative Study of Electoral Systems) do not have adequate questions concerning social policy preferences or income to allow for an examination of the effect of political awareness on welfare attitudes. An alternative approach measures political awareness through self-assessed interest in and knowledge of politics. From a theoretical perspective, political interest is essential for motivating individuals to invest the time and energy required to become politically informed and develop more stable and complex opinions about political and policy issues (Luskin, Reference Luskin1990; Frederico and Hunt, Reference Frederico and Hunt2013). In the American context, significant evidence exists that self-assessed political interest influences the internal consistency of policy preferences, reflecting greater political sophistication among those with higher levels of interest (Judd et al., Reference Judd, Krosnick and Milburn1981; Luskin, Reference Luskin1990; Baldassarri and Gelman, Reference Baldassarri and Gelman2008; Lewis-Beck et al., Reference Lewis-Beck, Jacoby, Norpoth and Weisberg2008).
Here, political awareness is measured through a standardized index of three questions concerning respondents’ self-assessed political knowledge and interest. The first question asks individuals’ their level of interest in politics with values ranging from ‘very interested’ to ‘not at all interested.’ The second question asks, ‘how often does politics seem so complicated that you can’t really understand what is going on?’ with respondents placing themselves on a five-point scale ranging from ‘never’ to ‘frequently.’ The third question asks, ‘how difficult or easy do you find it to make your mind up about political issues?’ with values ranging from ‘very difficult’ to ‘very easy.’ These three variables were combined into a standardized index with a fair level of internal consistency (α=0.68).
Income is measured by placing individuals into income decile brackets for each country, allowing for comparisons of the relative income positions of individuals controlling for the significant cross-national variations in average incomes.Footnote 1 To further capture the effects of economic self-interest, the analysis includes a measure of occupational class status. The measure places individuals into three categories (upper, middle, and working class) based on the European Socio-Economic Classification’s (ESEC) three-part class schema (Erikson and Goldthorpe, Reference Erikson and Goldthorpe1992; Rose and Harrison, Reference Rose and Harrison2007).
In addition to these central variables, the analysis includes a variety of standard controls. Individuals’ assessment of their own ideological position were measured through a question asking respondents to place themselves on an 11-point left–right scale with higher values representing a more leftist political position. Other standard controls include sex, age, union membership, and education (Svallfors, Reference Svallfors1997; Jæger, Reference Jæger2006). In addition, controls were introduced to capture those who might be members of so called ‘transfer classes,’ such as the retired, students, the disabled, and the unemployed. The analysis includes a control for level of religiosity, which, at the individual level, has been shown to be negatively associated with support for redistribution (Scheve and Stasavage, Reference Scheve and Stasavage2006; Jordan, Reference Jordan2014).
Assessing ideological constraint in welfare preferences
To what extent does political awareness influence the consistency and coherence of expressed preferences for social policy and redistribution? As a first step, I constructed a standardized index of support for redistribution, left–right self-positioning, and the five other variables concerning preferences for the welfare state (preferences for welfare spending levels, the welfare support index, the moral hazard index, and the two measures of tolerance for inequality). Chronbach’s α expresses the average inter-item correlation between the different elements of the index, indicating the extent to which these different questions fit together as a measure of a central concept, social policy preferences.
Table 1 presents the Chronbach’s α of the social policy attitudes index for the sample as a whole and then for individuals divided into five quintiles of political awareness. The results conform to the expectations described above. For the sample as a whole, the seven indicators have an α value of 0.61, but there exists significant variation across levels of political awareness. The Chronbach’s α value rises from 0.48 among the least politically aware to 0.75 among those in the highest category. This demonstrates that the consistency of social policy preferences significantly increases with levels of political awareness. Further, the analysis reveals that the index reaches traditional standards of index reliability and consistency only among those with the highest levels of political awareness (Nunnally and Bernstein, Reference Nunnally and Bernstein1994). This conforms with the expectation that for large sections of the population, expressed attitudes on a host of related questions do not appear to cohere into a consistent set of social policy preferences as assumed by traditional models of welfare opinion.
Table 1 Political awareness and ideological consistency
a Chronbach’s α for a standardized index, including Left–Right Self-Position, Support for Redistribution, the Moral Hazard Index, the Welfare Support Index, Social Spending Preferences, and two variables gauging tolerance for inequality: Inequality Rewards Talent and Large Income Differences Unjust. See the Appendix for a full description of all variables.
The analysis of inter-item correlations is limited by its exclusive focus on the bivariate relationship between political awareness and attitudinal consistency. It is possible that the findings above result from other individual and national-level variables not addressed in simple correlational analysis. To deepen our understanding of the impact of political awareness on the consistency of social policy preferences, it is necessary to incorporate a multivariate analysis that includes the controls for the individual-level variables described above as well as the national context. Table 2 shows the results from a series of ordered probit regressions examining the relationship between support for redistribution and other expressed social policy preferences controlling for other common predictors. The analysis employs cluster-robust standard errors as a simple and straightforward solution to the common statistical problems associated with cross-national public opinion data (Steenbergen and Jones, Reference Steenbergen and Jones2002; Franzese, Reference Franzese2005). All models include national-level fixed effects to control for differences in mean levels of support across countries that result from unmodeled differences in the institutional or cultural context.
Table 2 Political awareness and attitudinal consistency: predictors of support for redistribution
Results from ordered probit regression using cluster-robust standard errors and fixed effects by country. P-values in parentheses.
*P<0.1, **P<0.05, ***P<0.001.
Model 1 examines the relationship between redistribution and preferred levels of social spending, two of the most common dependent variables used in comparative analyses of welfare preferences. To be clear, these models are not meant to suggest causal relationships between attitudinal variables, but rather to examine the influence of political awareness on the internal consistency of social policy preferences controlling for individual-level factors associated with both levels of political awareness and support for the welfare state (e.g. income, gender, etc.). Model 1 reveals that support for increased social spending is positively correlated with support for redistribution, as would typically be expected. Model 2 introduces an interaction term between political awareness and social spending preferences to examine whether the relationship between preferences for redistribution and social spending is influenced by levels of political awareness. The interaction term is positive and statistically significant, suggesting that rising levels of political awareness increase the strength of the correlation between preferences for social spending and redistribution. Substantively, this means that as political awareness increases attitudes about redistribution and social spending become increasingly consistent with one another.
In order to ease interpretation of the model, Figures 1 and 2 presents the predicted probabilities of different levels of support for redistribution for four ‘typical’ individuals in the sample who differ only in their attitudes toward social spending and levels of political awareness.Footnote 2 Among those with limited political awareness (Figure 1), preferences for redistribution are not statistically significantly different from those who support ‘decreasing taxes and social spending a lot’ and those who support ‘increasing taxes and social spending a lot.’ Moreover, the relationship is the opposite of theoretical expectations, with individuals who support significant social spending cuts more likely to support redistribution than those who support higher spending. At higher levels of political awareness (Figure 2), the results conform with traditional expectations with a predicted probability of support for redistribution (‘agree’ and ‘strongly agree’) among those who support significant increases in social spending of 0.91 compared to only 0.36 for those who support significant cuts to social spending. These results demonstrate that at low levels of political awareness attitudes about redistribution and social spending are almost entirely unrelated; only at reasonably high levels of awareness do the two attitudes appear to conform to a consistent set of preferences.
Figure 1 The government should reduce differences in income levels (low political awareness).
Figure 2 The government should reduce differences in income levels (high political awareness).
Models 3 and 4 (Table 2) repeat the analysis comparing support for redistribution to attitudes about the fairness of income inequalities. Though less dramatic than those for social spending cuts, the overarching pattern is similar. The correlation between ideas about the fairness of inequality and redistribution significantly increases at higher levels of political awareness. Looking again to the predicted probabilities, for individuals who strongly disagree that ‘differences in standard of living’ should be small, the predicted probability of support for redistribution (‘agree’ and ‘strongly agree’) is 0.46 among those with low political awareness compared to only 0.13 for those at the highest level of political awareness. This provides further evidence that the correspondence between related attitudes toward the welfare state, in this case, the unfairness of inequality and the responsibility of the government to do something about it, significantly increases with levels of political awareness.
Taken together, the results of the ordered probit analysis conform to the patterns suggested by the bivariate analyses above. After controlling for a number of individual-level factors as well as national context, the results suggest that attitudes toward social welfare are highly inconsistent among those with low levels of political awareness. As political awareness increases, responses to related questions concerning social welfare policy begin to cohere into more consistent preferences. In other words, only in those with high levels of political awareness do expressed attitudes for related social policy questions appear to be a part of a broader set of welfare preferences that we might describe as ideological or coherent. This provides strong evidence that the consistency of expressed attitudes toward redistribution and social policy depend heavily upon the levels of political interest and awareness among respondents. This poses a significant challenge to traditional models by suggesting that our assumptions about how individuals develop and express policy preferences in surveys may poorly reflect the actual behavior of individuals with low levels of political awareness.
Political awareness and ‘enlightened preferences’
The previous section reveals dramatic differences in the coherence of welfare preferences, with extremely limited internal consistency among those with low political awareness. The finding that welfare preferences are highly inconsistent for large sections of the population is not necessarily surprising; however, these findings have important implications beyond the statistical problems described above. An important starting point for many of the most influential political economy models is the claim that preferences for redistribution are driven by income, with the poor more likely to support higher taxes and social spending (Downs, Reference Downs1957; Meltzer and Richard, Reference Meltzer and Richard1981; Iversen and Soskice, Reference Iversen and Soskice2006). The findings above challenge this common assumption by suggesting that less informed individuals do not hold coherent or consistent social policy preferences. In addition, existing research suggests that individuals with low political awareness may struggle to connect their economic self-interest to particular policy preferences (Delli Carpini and Keeter, Reference Delli Carpini and Keeter1996; Kuklinski et al., Reference Kuklinski, Quirk, Jerit, Schwieder and Rich2000). These information effects are not politically neutral. Due to both the financial and time costs of becoming politically informed, political awareness is inversely correlated with income. As a result, the poor and working classes are more likely to exhibit preferences for redistribution that are disconnected from their economic self-interests. Because the ‘natural constituency’ of redistributive policies may suffer most from low interest and information, the effect of information effects may be to undermine overall support for redistributive social policies (Andersen et al., Reference Andersen, Tilley and Heath2005; Arnold, Reference Arnold2012). Such an expectation is consistent with Iversen and Soskice (Reference Iversen and Soskice2015), who find that low-levels of political awareness produces a centrist bias among uninformed voters as they err on the side of caution by supporting more centrist policies and parties.
Table 3 investigates these claims by evaluating the impact of income on preferences for redistribution at varying levels of political awareness. Model 5 is a baseline model, revealing the expected negative correlation between income and redistribution preferences and no statistically significant effect of political awareness on attitudes toward redistribution. Model 6 introduces an interaction term between income and political awareness to determine the effect of income on preferences for redistribution at different levels of political knowledge and interest. Consistent with the expectations from above, the results demonstrate a statistically significant negative interaction between income and political awareness, meaning that the negative effect of income on support for redistribution increases as political awareness increases. As above, the calculation of predicted probabilities ease the interpretation of these results. Figures 3 and 4 provide a comparison of individuals in the highest and lowest income deciles at the highest and lowest levels of political awareness. Figure 3 illustrates the effect of income on redistributive preferences when political awareness is low. The results demonstrate that income does not affect redistributive preferences when political awareness is low. At the lowest levels of political awareness, the poor are more likely to support redistribution than their high-income counterparts; however, the difference is substantively small and statistically insignificant. These results change dramatically when political awareness is high. Figure 4 demonstrates a significant difference between the predicted probabilities of support for redistribution among the rich and poor at high levels of political awareness. The rich are now significantly more likely to oppose redistribution than the poor. These results conform to the expectations above by revealing that income does not influence preferences for redistribution among those with low levels of political awareness, but does have a powerful affect at higher levels of awareness.
Figure 3 The government should reduce differences in income levels (low political awareness).
Figure 4 The government should reduce differences in income levels (high political awareness).
Table 3 Political awareness, economic interest and support for redistribution
Results from ordered probit regression using cluster-robust standard errors and fixed effects by country. P-values in parentheses.
*P<0.1, **P<0.05, ***P<0.001.
Interestingly, the principle cause of the interaction effect lies not in changes in the support for redistribution among the poor, but rather from the effect of political awareness on the attitudes of the rich. Among the poor, the probability of support (‘agree’ and ‘strongly agree’) for redistribution remains strong at both high (prob.=0.76) and low (prob.=82) levels of political awareness; however, for the rich, the predicted probability of support falls from 0.75 among the least politically aware to 0.44 for the most politically aware. Perhaps counter intuitively, this suggests that increasing the levels of political awareness might actually lower overall support for redistribution. Because the expressed opinions of the poor are largely unchanged by increasing information, higher levels of information in the electorate would have the most significant effects on the wealthy by turning them increasingly against redistribution.
Why does political awareness lower support for redistribution among the wealthy, while having little discernible impact for the poor? One explanation for these results is that the saliency of questions concerning redistribution may be significantly higher for the poor compared with the rich, which would increase the likelihood that poorer individuals have more concrete opinions about these issues even at lower levels of political awareness. An alternative explanation is that at low levels of political awareness, individuals at all income levels are more likely to place themselves into the ‘agree’ category due to ‘acquiescence response bias’ (Pasek and Krosnick, Reference Pasek and Krosnick2010). If this was the case, then informed and uninformed responses to questions about redistribution would be largely the same for the poor, both of which would choose to support redistribution. Only among higher income individuals would a more informed opinion differ substantially from the default response of ‘agree’ among the uninformed.
Models 7 and 8 repeat the analysis focusing on preferences for levels of social spending, rather than redistribution. The results of these models do not reveal a statistically meaningful interaction effect between awareness and income. Model 8 uncovers no statistically significant interaction effect between knowledge and income when considering preferences for social spending. Interestingly, there also appears to be no statistically significant relationship between income and preferences for social spending, independent of political awareness (Model 7). It is unclear why these results are inconsistent with those from the previous analysis of redistribution preferences. One explanation may lie in the structure of the social spending variable. The variable asks individuals to place themselves on an 11-point scale between the extremes of ‘decrease taxes and social spending a lot’ to ‘increase taxes and social spending a lot.’ A response of ‘5’ would thus imply a satisfaction with status quo levels of social spending and taxation. Unfortunately, research has also consistently demonstrated that the middle response in these questions frequently acts as a hidden ‘don’t know’ category. The result is that the variable confounds absence of an opinion with support for the status quo.
Conclusions
How does the wide variation in levels of political awareness influence the structure and nature of public support for redistributive social policy? The findings here suggest that at low levels of political interest and knowledge, little discernible connection exists between preferences for redistribution and those for social spending levels, left–right positioning, tolerance for inequality, or support for specific social policies. Only at the highest levels of political awareness, do we begin to see the consistency in expressed attitudes that might suggest the types of reasoned and strongly held preferences assumed by research on comparative social policy attitudes. The results also challenge the assumption that individuals can clearly and easily identify their economic self-interests in questions of social policy. Differences in support for redistribution between high and low-income individuals only appear at higher levels of political awareness with a seemingly broad consensus in support of redistribution for those at the lowest levels of knowledge and interest.
Recognition of the significant variations in political awareness and attitudinal consistency among citizens has a number of important consequences for the study of social policy preferences. First, if political awareness influences the ideological constraint and consistency of respondents’ opinions, then the correlations between attitudes toward redistribution and other attitudes concerning the welfare state, inequality, general economic policy, and left–right positioning may be extremely low for many, if not most, individuals. This disconnect between theoretically related attitudes may cause public opinion models to be highly sensitive to the choice of dependent variable and may also produce disparate results across different surveys as a the result of minor differences in question wording or order. A second implication of these results is that much of the variance across countries and individuals in support for the welfare state may largely be noise, rather than real substantive variation. Importantly, framing and question order effects combined with the tendency for acquiescence bias suggests that this error is not random, but is rather systematically related to political awareness, which is in turn correlated with many of the most theoretically important explanatory variables. This systematic error could produce significant heteroskedasticity as public opinion models will systematically perform better for some groups than for others. Third, these effects may distort aggregate measures of public support for the welfare state. If large sections of the population are expressing uninformed preferences, then much of the strong support we see for welfare states across countries may simply reflect high levels of acquiescence response bias, rather than broad based support for social policy (Althaus, Reference Althaus2003). As demonstrated here, support for redistribution is uniformly high among the politically uninformed, with significant opposition between rich and poor only arising at higher levels of political awareness. The result is that raising political awareness might actually reduce overall levels of expressed support for redistribution.
Recognizing the role of political awareness and knowledge raises important new questions concerning how public support for the welfare state may be influenced by the information environment. In particular, this research highlights the need to investigate more carefully how individuals in different societies come to know and understand how the welfare state functions and how it relates to their specific economic self-interests and normative values. Following on the recent evidence from Iversen and Soskice (Reference Iversen and Soskice2015) that the welfare state itself might affect overall levels of political awareness through its influence on inequality and unionization rates, the results here suggest significant new avenues of research on how differences in the availability and quality of political information across countries may influence and be influenced by levels of public support for the welfare state.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/S1755773917000017
