Identity priming and public opinion on income inequality: robustness testing of the micro-level mechanism of the paradox of redistribution

Identity priming and public opinion on income inequality: robustness testing of the micro-level... Abstract This study aims to contribute new insights into the ‘paradox of redistribution’ theory in light of identity priming. Korpi and Palme argued that low-income targeting leads to less redistribution and explained this trade-off as a result of coalition politics among different social strata. Indeed, recent empirical studies suggest that high-income earners tend to have more negative attitudes toward redistribution in the context of targeted spending (the polarization hypothesis). Contrary to most studies, which have relied on a single data source and measure, this study explores whether the polarization hypothesis is supported by multiple (both micro and macro) data sources. The findings of this study indicate that different question wordings produce different responses to inequality and redistribution. Some of the results confirmed the polarization hypothesis, whereas others did not. This article attempts to explain these discrepancies by comparing the wording patterns of inequality-attitude measures and arguing that high- and low-income earners polarize their views in the context of low-income targeting when their sense of ingroup favoritism and intergroup conflict is sufficiently triggered. 1. Introduction Several recent studies have focused on the role of target efficiency—that is, the extent to which public transfers are targeted to those at the bottom of the income scale—as a key cross-national institutional variation that explains the micro-level mechanism of redistribution politics (Brady and Bostic, 2015; Beramendi and Rehm, 2016; Sumino, 2016). The theoretical underpinnings for this insight come from the seminal work by Korpi and Palme (1998; ‘KP, 1998’ henceforth), in which they argued that the skewed distribution of welfare benefits among different income groups creates a cross-class polarization of interests and identities, hence less government redistribution—called the ‘paradox of redistribution’. Despite growing interest in the concentration of transfer benefits, relatively little attention has been paid to the role of identity/conflict awareness. Let me explain this point by turning to the original ‘paradox of redistribution’ theory (KP, 1998). The KP’s theory predicts that more targeted welfare systems result in less redistributive government: Targeted spending → redistribution. Here, ‘redistribution’ refers to the relative reduction in income inequality from market income to disposable (post-tax/transfer) income as a proportion of market income inequality (KP, 1998). The logic behind this process (targeted spending → redistribution) can be broken down into two components: the first in which given institutional structures (i.e. universal or targeted spending) mobilize coalitions (or conflicts) between different income groups, and the second in which formulated preferences are reflected in an actual redistributive policy. This micro-level mechanism can be expressed as: Targeted spending (a) → preference formation (b) → redistribution (c). Focusing on the former stage [(a) → (b)], this study hypothesizes that the relationship between income status and attitudes toward income inequality is conditioned by the degree of targeted spending. In other words, attitudinal cleavages between income groups grow wider in welfare states with low-income targeting, in which redistribution politics tends to be a group-interested and conflict-driven process. An empirical study by Beramendi and Rehm (2016) indicates that a more low-income targeting leads to a more contested welfare state (i.e. more negative income slopes). This study further specifies the above argument in light of identity awareness, arguing that the polarization process (Targeted spending → preference formation) occurs when individuals are exposed to a cue that elicits their sense of social identity (Tajfel, 1981) and self-categorization (Turner, 1985; Turner et al., 1987). In other words, targeted spending mobilizes coalitions (or conflicts) among income groups when members of high/low income groups articulate a sense of identity and conflict. This modifies the causal process as follows: Targeted spending (a) → identity priming (b1) → polarization in attitudes (b2) → redistribution (c), in which preference formation (b) is divided into (b1) and (b2). By ‘identity priming’, I mean an increased sensitivity to reducing inequality (e.g. redistributive policy) triggered by exposure to stimuli that evoke a sense of ingroup favoritism and intergroup conflict. In practice, this study argues that the polarization process [(a) through (b2)] occurs when a survey question accentuates the contrast/conflict between high- and low-income earners. To this end, unlike most previous studies that have replied on a single data source (e.g. ISSP) and single question item, this study attempts to provide a more comprehensive analysis using multiple data sources: the International Social Survey Program (ISSP, 2009), the European Social Survey (ESS, 2008) and the European Values Survey (EVS, 2008). From these surveys, I select a set of question items (inequality attitudes) with different question wordings, some of which emphasize the contrast/conflict between high- and low-income groups but others do not. The polarization process is expected to occur when respondents are exposed to a survey question that is designed to prime their sense of group identity/conflict. The findings of this study indicate that differences in the question wording, even slight ones, can yield different results, and that the validity of the hypothesis that targeted spending contextualizes the association between income and inequality attitudes depends on how the outcome variable (inequality attitude) is asked or worded/phrased. 2. Theoretical framework and hypotheses In this section, I briefly outline the ‘paradox of redistribution’ theory developed by KP (1998), and then identify the specific process of attitudinal polarization in the context of targeted spending. I then highlight the role of identity priming as a potentially important activator of ingroup favoritism and intergroup conflict that make the polarization process function. 2.1 Micro-level foundation of the paradox of redistribution The welfare regime theory argues that the differences in social policy arrangements generate distinct forms of social structuring, political discourse and attitudes (Korpi, 1980; Esping-Andersen, 1990; KP, 1998; Pierson, 2001; Rothstein and Steinmo, 2002; Larsen, 2008). The central idea is that institutional structures shape a collective understanding of redistributive politics—namely, whether and to what extent the well-being of the least well-off in society should be secured through statutory intervention or market-based solutions (Offe, 1992; Edlund, 1999). The political economy literature, on the other hand, highlights the conflict of material interest between income groups as the key motive for individualist behavior (Romer, 1975; Roberts, 1977; Meltzer and Richard, 1981). In this view, individuals are assumed to behave in a way that maximizes their personal utility. For instance, low-income earners are more likely to support redistribution than high-income earners because it benefits them financially more than it does in their high-income counterparts (Luttmer, 2001; Alesina and Ferrara, 2005; Rehm, 2009; Sumino, 2014). Synthesizing these two distinct theoretical traditions (i.e. institutionalist and political economy perspectives) prompts the following question: do different income groups respond uniformly to inequality across countries with different institutional structures? Political economists claim that income cleavages lead to differences in how people feel about inequality. But do high-income earners under a certain institutional context oppose egalitarian measures in the same manner as those under a different institutional context? If the answer is no, what might account for the differences? Revisiting the ‘paradox of redistribution’ theory proposed by KP (1998) gives us a good starting point. They argued that target efficiency—‘the proportion of program expenditures going exclusively to those below the official poverty line’ (p. 662)—plays a crucial role in the process of welfare state politics. They theorized that there is a reverse relationship between targeting and redistribution, namely that a welfare state institution based on means-tested programs targeted exclusively at the needy results in less redistributive policy outputs. According to them, this trade-off is explained as a product of socio-political dynamics. Marginal welfare states with targeted income transfers, in which social programs ‘unfairly’ benefit a small portion of the population (i.e. those below the poverty line), produce a preference gap between high/middle-income and low-income groups, and this, in turn, causes a potential revolt against a redistributive political agenda. On the other hand, welfare institutions characterized by universal programs, which benefit a broad range of people regardless of recipients’ market positions, mobilize cross-class alliances. Unlike means-tested transfer schemes, the comprehensive welfare state incorporates all citizens without accentuating disparities between net contributors and net beneficiaries (Pierson, 1996). Because social welfare programs in the universal welfare state are generously designed to cover the entire population instead of labelling or stigmatizing the target group, the expected costs and benefits of the welfare state are less visible or calculable to both manual and better-off workers (Skocpol, 1992; Rothstein, 1998). In other words, the universal welfare state blurs the boundaries between group identities and generates a seamless and integrated public demand for redistribution. In contrast, in selective welfare states, identities are constructed along a constellation of class interests, which lead to group-based conflicts between ‘us’ and ‘them’ (Edlund, 1999). Residual welfare systems explicitly divide the population into welfare recipients and non-recipients, and hence, tend to be associated with greater polarization in preferences. This is not to say that institutional structures (i.e. target efficiency) directly explain welfare stakeholders’ perceptions toward income inequality, but rather that the relationship between income and outcome attitudes is conditioned—strengthened or weakened—by the structure of social policy institutions, in particular, by the degree of targeting policy (Gingrich and Ansell, 2012). The thrust of the above argument is two-fold: (1) reactions to redistributive policies differ across income groups—while the poor demand redistribution, the rich oppose it, and (2) such reactions of different income groups are not uniform across welfare sates with different degrees of targeted spending. These theoretical propositions lead us to the following two hypotheses: Hypothesis 1. Higher-income earners are less supportive of reducing inequality. In other words, there exist attitudinal cleavages between income groups. Hypothesis 2. Attitudinal cleavages between income groups grow wider in the context of targeted spending. High-income earners’ opposition to reducing inequality becomes more intense in welfare states where transfer programs are ‘unequally’ targeted at low-income households. This study does not intend to test the whole process of the ‘paradox of redistribution’ theory developed by KP (1998), in which they noted that, (1)the institutional structures of welfare states are likely to affect the definitions of identity and interest among citizens. Thus, (2)an institutional welfare state model based on a universalistic strategy intended to maintain normal or accustomed standards of living is likely to result in greater redistribution than a marginal one based on targeting (p. 663; underlines added). The underlying logic of these statements can be summarized as follows: first, institutional structures shape identity/interest-based preferences toward inequality, and second, such formulated preferences are translated to redistributive policy outputs, most likely through political representation. The first stage concerns the role of institutional structures in triggering broad alliances or antagonisms among income groups, while the second stage refers, though not explicitly, to the impact of public opinion on redistributive policy. Thus, the whole causal process is expressed as: Targeted spending (a) → preference formation (b) → redistribution (c). The interest of this article is to explore the former part—the causal sequence from ‘targeted spending’ (a) to ‘preference formation’ (b). In this respect, while the present study draws upon the ‘paradox of redistribution’ theory, the main focus is on the micro-level mechanism behind the formation of preferences rather than on the causality from ‘preference’ (b) to ‘redistribution’ (c) or the aggregate-level relationship between ‘targeted spending’ (a) and ‘redistribution’ (c) (e.g. Kenworthy, 2011; Marx et al., 2013). 2.2 Identity priming in redistribution politics Central to understanding the mechanism underlying the polarization process (Targeted spending (a) → preference formation (b); Hypothesis 2) is the role of identity awareness. The issue of who, or which income group, gets transfer benefits and who bears the costs is politically sensitive, and attracts scrutiny from welfare stakeholders. The uneven distribution of public transfers, which is closely linked to subjective economic well-being, affects stakeholders’ reactions to the reallocation of income resources between ‘our group’ and ‘the others’. In the context of low-income targeting, high-income earners are expected to be more resistant to reducing inequality and redistribution because targeting, they perceive, ‘unfairly’ benefits ‘the others’ (i.e. low-income groups) at the expense of ‘our group’. This leads to the prediction that attitudinal differences between high- and low-income earners grow wider in countries with targeting policies. Redistributive preferences cannot be explained only by individual calculations of rational self-interest but are largely a function of social identity (Tajfel, 1981) and self-categorization (Turner, 1985; Turner et al., 1987). Individuals’ perceptions that they belong to a socio-economic group and their emotional attachment to the group are a powerful regulator of inequality attitudes. The process of self-categorization, in which individuals place themselves into a group that shares similar concerns and interests, draws boundaries between ‘us’ and ‘them’, and individuals ascribe positive attributes to their own group and negative traits and prejudice to the others. The self-categorization process helps ‘atomized’ individuals recognize their social self and elicits intragroup cohesion and intergroup conflict. Group polarization is particularly salient in the context of low-income targeting characterized by the reallocation of resources from high- to low-income groups. This implies that the polarization process occurs when individuals’ awareness of identities is stimulated, and when ingroup favoritism is triggered. In other words, attitudinal polarization (in the context of low-income targeting) is activated by an exposure to a stimulus that provokes a sense of belongingness and conflict. Hypothesis 2 predicts that targeted spending contextualizes the relationship between income position and attitudes toward inequality. Here, ‘attitudes toward inequality’ can be operationalized in a variety of ways (question wordings): some survey questions emphasize the contrast between those who benefit and those who contribute, but others do not. The polarization process (Hypothesis 2) is expected to occur when respondents are exposed to a question that is designed to prime their identification with a specific socioeconomic group (high/low-income status). Thus, the causal process described in the sub-section 2.1 will be modified as follows: Targeted spending (a) → identity priming (b1) → polarization in attitudes (b2) → redistribution (c) ‘Preference formation’ (b) can be divided into two sub-processes: identity priming (b1) and polarization in attitudes (b2). This causal sequence suggests that if respondents were not exposed to a cue that stimulates their sense of identification (b1), the process of attitudinal polarization (b2) would be less likely to happen. In the empirical analysis that follows, I examine how the validity of Hypothesis 2 is susceptible to the question wording of inequality-attitude measures. 3. Research methods 3.1 Data This study is based on individual-level data (aged 18–89 years) from three different sources: the International Social Survey Program (ISSP, 2009), the European Social Survey (ESS, 2008) and the European Values Survey (EVS, 2008). The three surveys include questions about attitudes toward inequality and redistribution, and cover (share) the same set of countries and time period (late-2000s). In this study, the analysis was restricted to capitalist democracies on which the KP theory is originally based, and macro-level data on target efficiency (OECD, 2008; Wang and Caminada, 2011) are fully available. To make the comparisons easier, this study used the same set of overlapping countries commonly covered by the three surveys. The analysis includes 20 countries: Austria, Belgium, the Czech Republic, Denmark, Estonia, Finland, France, Germany, Great Britain, Hungary, the Netherlands, Norway, Poland, Portugal, the Slovak Republic, Slovenia, Spain, Sweden, Switzerland and Turkey. The three datasets were analyzed in combination with macro-level data (OECD, 2008; Wang and Caminada, 2011). Due to data constraints, Portugal and Turkey were included only in the models estimated using the OECD data, while Estonia and Slovenia were included only in the models estimated using the LIS (Wang and Caminada, 2011 ) data (for details, see Appendix A, Table A3). Missing data were handled using multiple imputation (MI) techniques because the use of list-wise deletion results in loss of ∼20% of the cases—mainly due to missing information on the (country-specific) household income variables—thus, may lead to biased and inefficient estimates (King et al., 2001). Imputation was carried out with Amelia II for creating five (m = 5) complete case datasets (Honaker et al., 2011). Missing values were imputed separately within each country to account for the clustered data structure (Graham, 2009). Estimates from the five imputed datasets were pooled using Rubin’s rules (Rubin, 1987; 1996; White et al., 2011). 3.2 Measures The central aim of this study is to examine how question wordings affect the polarization process in the context of low-income targeting (Hypothesis 2). In essence, this study predicts that the polarization process is most likely to occur when survey questions elicit respondents’ identification and self-categorization. To test this, I employ and compare a set of four different indicators (with different question wordings) of inequality attitudes (Table 1). The first two items are standard ‘government redistribution’ questions from the ISSP and the ESS. The ISSP asked the respondents whether they agree or disagree with the statement, ‘it is the responsibility of the government to reduce the differences in income between people with high incomes and those with low incomes’, while the ESS has a similar but slightly different question item that asked whether or not respondents agree with the notion that ‘the government should take measures to reduce differences in income levels’. The ISSP question explicitly refers to the contrast between ‘people with high incomes and those with low incomes’, which is expected to invoke respondents’ awareness of group identity—e.g. high-income status (contributor) or low-income status (recipient). On the other hand, the ESS item is a less conflict-prone question in that it is vaguely worded as ‘differences in income levels’, with no reference to ‘high versus low’ income earners. The third item was taken from the EVS, which asked respondents for their views on the statement: ‘incomes should be made equal’ or ‘there should be greater incentives for individual effort’. Similar to the first and second items, this question asks about income inequality, but with a more explicit reference to the conflict between income equality and incentives on effort. The fourth inequality-attitude item also comes from the EVS, in which respondents were asked to choose which of the following statements came closer to their own view: ‘I would consider equality more important, that is, that nobody is underprivileged and that social class differences are not so strong’, or ‘I would consider personal freedom more important, that is, everyone can live in freedom and develop without hindrance’. This item is similar to the third one (i.e. equality versus freedom), but different in that the two opposite values were worded in a more abstract way: ‘equality’ rather than ‘income equality’ and ‘personal freedom’ rather than ‘incentives for effort’. In this sense, the fourth question is expected to send less explicit signals to respondents about the possibility that reducing income inequality might result in disincentives for effort and performance. Table 1. Classification of inequality-attitude items by question-wording patterns Identity/conflict priming words/phrases (a) Income inequality/ differences (b) High income versus low income (c) Incentives on effort Q1: Responsibility of the government to reduce the differences in income between people with high incomes and those with low incomes. (ISSP 2009) Yes Yes No Q2: Government should take measures to reduce differences in income levels. (ESS 2008) Yes No No Q3: Incomes should be made more equal. OR There should be greater incentives for individual effort. (EVS 2008) Yes No Yes Q4: I would consider equality more important, that is, that nobody is underprivileged and that social class differences are not so strong. OR I would consider personal freedom more important, that is, everyone can live in freedom and develop without hindrance. (EVS 2008) No No No Identity/conflict priming words/phrases (a) Income inequality/ differences (b) High income versus low income (c) Incentives on effort Q1: Responsibility of the government to reduce the differences in income between people with high incomes and those with low incomes. (ISSP 2009) Yes Yes No Q2: Government should take measures to reduce differences in income levels. (ESS 2008) Yes No No Q3: Incomes should be made more equal. OR There should be greater incentives for individual effort. (EVS 2008) Yes No Yes Q4: I would consider equality more important, that is, that nobody is underprivileged and that social class differences are not so strong. OR I would consider personal freedom more important, that is, everyone can live in freedom and develop without hindrance. (EVS 2008) No No No Sources: ISSP (2009); ESS (2008); EVS (2008). Table 1. Classification of inequality-attitude items by question-wording patterns Identity/conflict priming words/phrases (a) Income inequality/ differences (b) High income versus low income (c) Incentives on effort Q1: Responsibility of the government to reduce the differences in income between people with high incomes and those with low incomes. (ISSP 2009) Yes Yes No Q2: Government should take measures to reduce differences in income levels. (ESS 2008) Yes No No Q3: Incomes should be made more equal. OR There should be greater incentives for individual effort. (EVS 2008) Yes No Yes Q4: I would consider equality more important, that is, that nobody is underprivileged and that social class differences are not so strong. OR I would consider personal freedom more important, that is, everyone can live in freedom and develop without hindrance. (EVS 2008) No No No Identity/conflict priming words/phrases (a) Income inequality/ differences (b) High income versus low income (c) Incentives on effort Q1: Responsibility of the government to reduce the differences in income between people with high incomes and those with low incomes. (ISSP 2009) Yes Yes No Q2: Government should take measures to reduce differences in income levels. (ESS 2008) Yes No No Q3: Incomes should be made more equal. OR There should be greater incentives for individual effort. (EVS 2008) Yes No Yes Q4: I would consider equality more important, that is, that nobody is underprivileged and that social class differences are not so strong. OR I would consider personal freedom more important, that is, everyone can live in freedom and develop without hindrance. (EVS 2008) No No No Sources: ISSP (2009); ESS (2008); EVS (2008). Table 1 summarizes the wording patterns of the four inequality-attitude items. The items were classified by whether or not they include specific words/phrases that trigger a sense of ingroup favoritism and intergroup conflict—namely, explicit references to (a) income inequality, (b) high versus low incomes and (c) incentives on effort. Unlike the phrases such as ‘social class differences’ and ‘social inequality’, ‘income inequality/differences’ (a) is specific since it identifies a particular dimension of socio-economic disparities (i.e. income), which is expected to invoke a sense of identity as a high/low-income earner. This specification is especially important because this analysis focuses on attitudinal cleavages across income lines (Hypotheses 1–2). ‘High- versus low-income status’ (b) is also a potential trigger of identification and self-categorization. The phrase ‘high/low incomes’ elicits a sense of identity/conflict by making respondents aware, for example, that redistributive policy benefits low-income group but burdens high-income group. This would be less likely to happen if the question was asked in a more abstract fashion—for example, ‘differences in income levels’ (ESS). The phrase ‘incentives on effort’ (c) reminds high-income earners that they are economically successful because they worked hard for it, which leads to the justification of privileged status and existing income inequality. This process instills a sense of identity as a member of high-income group and triggers ingroup favoritism and outgroup prejudice (e.g. lazy low-income earners). As presented in Table 1, Questions 1–2 (ISSP/ESS) share all features except ‘high versus low incomes’. Question 3 (EVS) is similar to Questions 1–2 because it addresses income inequality, but differs because it addresses the potential trade-off between income equality and incentives on effort. Question 4 also deals with the conflict between equality and freedom but in a more abstract way: ‘equality’ not ‘income equality’; and ‘personal freedom’ not ‘incentives for individual effort’. As presented in Figure 1, respondents’ reactions to the four questions vary markedly both within and between countries. Figure 1. View largeDownload slide Attitudes toward inequality, %. Notes: Countries were sorted in descending order by the sum of ‘strongly agree’ and ‘agree’ of the ISSP item (top left). To facilitate comparison, countries were presented in the same order for all figures. N = 20. Sources: ISSP (2009); ESS (2008); EVS (2008). Figure 1. View largeDownload slide Attitudes toward inequality, %. Notes: Countries were sorted in descending order by the sum of ‘strongly agree’ and ‘agree’ of the ISSP item (top left). To facilitate comparison, countries were presented in the same order for all figures. N = 20. Sources: ISSP (2009); ESS (2008); EVS (2008). The primary independent variable of interest is household income. In the three surveys (ISSP, ESS, EVS), household income was reported in categories. The midpoint of each income interval was used as a proxy for actual income (Rueda et al., 2014; Rueda, 2014; Rueda and Stegmueller, 2015). The midpoint of the last open-ended category was extrapolated from the frequencies of the next-to-top and the top categories using a Pareto formula proposed by Hout (2004). To take into account the differences in price levels across countries, income values were adjusted by purchasing power parities (PPPs) in national currencies per US dollar (World Bank, 2008). To adjust for the differences in household compositions, household incomes were divided by the square root of household size (Buhmann et al., 1988). To reduce skewness, income values were transformed with a natural logarithm prior to analysis. The analysis includes the following demographic controls: gender (female =1, male = 0), age (in years/10), age2, educational attainment (primary, secondary, tertiary) and employment status (full-time employed, unemployed, student, retired, housekeeping, others). Although one may argue that party affiliation (or left-right ideological disposition) has a significant association with inequality attitudes (e.g. left partisanship leads to higher levels of support for egalitarian principles and practices), I prefer not to include this item in this analysis because of potential endogeneity bias problems (i.e. the possibility of reverse causation; see Cusack et al., 2006; but see Jæger, 2008; Margalit, 2013). Target efficiency. To evaluate the distribution of public transfers across income groups, I use the concentration coefficient for public cash benefits. The concentration index, calculated in the same way as the Gini index, measures the extent to which public transfers are concentrated or dispersed across different income groups (Fields, 1979; Kakwani, 1986). Theoretically, the index lies between −1.0 (the poorest gets all public transfers) and +1.0 (the richest gets all public transfers). A score of zero indicates that public cash benefits are evenly distributed across the income scale. Negative and lower scores represent higher levels of progressivity—public cash benefits are targeted efficiently (or ‘unevenly’) to reduce income inequality. For robustness purposes, this study uses two different sources of aggregate-level data on targeted spending: OECD (2008); Wang and Caminada (2011). Here, Wang and Caminada (2011) is an updated version of the concentration coefficient calculated by the Luxembourg Income Study (LIS), the data source which KP’s (1998) work was originally based on. As Figure 2 shows, OECD (2008) statistics indicate that the concentration coefficient varies greatly across countries, ranging from −0.32 in Denmark to 0.35 in Turkey (mean = −0.04; SD = 0.19; N = 18). Denmark, Great Britain and Finland have the highest levels of target efficiency with values < −0.20, while public transfers are highly regressive in Turkey and Portugal with values > 0.20. In Germany and Hungary, public cash benefits are (relatively) equally distributed across incomes. According to Wang and Caminada (2011), the concentration index ranges from −0.31 in Great Britain to 0.16 in Poland (mean = −0.08; SD = 0.14; N = 18). The target efficiency scores are highly negative in Great Britain and Denmark; close to zero in Slovenia and Hungary; positive and high in Poland and Austria. Both measures are largely consistent with each other (Pearson’s r = 0.87; N = 16). The variation in values is larger in the OECD data (SD = 0.19) than in the LIS data (SD = 0.14), partly due to the inclusion of Turkey and Portugal, two highly regressive countries, in the OECD data. Figure 2. View largeDownload slide Target efficiency across 20 countries. Note: Countries were sorted in descending order, with the same countries listed side by side to facilitate comparison. Sources: OECD (2008); Wang and Caminada (2011). Figure 2. View largeDownload slide Target efficiency across 20 countries. Note: Countries were sorted in descending order, with the same countries listed side by side to facilitate comparison. Sources: OECD (2008); Wang and Caminada (2011). In addition to the concentration index, I also include the size of public transfers, defined as the percentage share of public cash transfers in household income (OECD, 2008; Wang and Caminada, 2011). Higher score represents greater proportion of redistributive benefits in household income. For instance, public transfers constitute about 35% of household income in Hungary and Poland, while they account for <20% in Great Britain and Switzerland (Table 2). The size of public transfers is clearly distinguished from the concentration index, which is about who among different income groups receives cash benefits and not how much (KP, 1998; Barnes, 2015). For example, as shown in Table 2, although public transfers account for >25% (OECD) of household income in both Denmark and Portugal, the degree of low-income targeting is very strong in Denmark (−0.316), but rather weak in Portugal (0.247). Table 2. Size of public transfers and target efficiency, by welfare regime type Size of benefits Target efficiency Size of benefits Target efficiency Regimes/ countries OECD WC OECD WC Regimes/ countries OECD WC OECD WC Social democratic Liberal Austria 36.6 26.7 0.157 0.108 Great Britain 14.5 14.3 −0.275 −0.313 Sweden 32.7 24.6 −0.145 −0.128 Non-classified Belgium 30.5 7.9 −0.120 −0.244 Poland 35.8 32.5 0.185 0.157 Denmark 25.6 18.9 −0.316 −0.306 Hungary 35.1 35.7 −0.016 0.016 Norway 21.7 20.2 −0.183 −0.155 Slovak Republic 26.0 26.6 −0.056 −0.109 Netherlands 17.1 21.3 −0.198 −0.041 Portugal 25.5 – 0.247 – Conservative Czech Republic 24.3 20.8 −0.154 −0.218 France 32.9 26.2 0.136 0.077 Spain 21.3 20.7 0.063 0.068 Germany 28.2 21.2 0.013 −0.110 Turkey 16.9 – 0.347 – Switzerland 16.0 17.5 −0.170 −0.066 Slovenia – 27.5 – 0.011 Finland 14.4 23.2 −0.219 −0.127 Estonia – 17.9 – −0.099 Size of benefits Target efficiency Size of benefits Target efficiency Regimes/ countries OECD WC OECD WC Regimes/ countries OECD WC OECD WC Social democratic Liberal Austria 36.6 26.7 0.157 0.108 Great Britain 14.5 14.3 −0.275 −0.313 Sweden 32.7 24.6 −0.145 −0.128 Non-classified Belgium 30.5 7.9 −0.120 −0.244 Poland 35.8 32.5 0.185 0.157 Denmark 25.6 18.9 −0.316 −0.306 Hungary 35.1 35.7 −0.016 0.016 Norway 21.7 20.2 −0.183 −0.155 Slovak Republic 26.0 26.6 −0.056 −0.109 Netherlands 17.1 21.3 −0.198 −0.041 Portugal 25.5 – 0.247 – Conservative Czech Republic 24.3 20.8 −0.154 −0.218 France 32.9 26.2 0.136 0.077 Spain 21.3 20.7 0.063 0.068 Germany 28.2 21.2 0.013 −0.110 Turkey 16.9 – 0.347 – Switzerland 16.0 17.5 −0.170 −0.066 Slovenia – 27.5 – 0.011 Finland 14.4 23.2 −0.219 −0.127 Estonia – 17.9 – −0.099 Notes: Regime classification is based on Esping-Andersen (1990, pp. 51–52). Countries are sorted in descending order by the size of transfer benefits (OECD, 2008) within each welfare regime cluster. ‘WC’ = Wang and Caminada (2011). Sources: OECD (2008); Wang and Caminada (2011). Table 2. Size of public transfers and target efficiency, by welfare regime type Size of benefits Target efficiency Size of benefits Target efficiency Regimes/ countries OECD WC OECD WC Regimes/ countries OECD WC OECD WC Social democratic Liberal Austria 36.6 26.7 0.157 0.108 Great Britain 14.5 14.3 −0.275 −0.313 Sweden 32.7 24.6 −0.145 −0.128 Non-classified Belgium 30.5 7.9 −0.120 −0.244 Poland 35.8 32.5 0.185 0.157 Denmark 25.6 18.9 −0.316 −0.306 Hungary 35.1 35.7 −0.016 0.016 Norway 21.7 20.2 −0.183 −0.155 Slovak Republic 26.0 26.6 −0.056 −0.109 Netherlands 17.1 21.3 −0.198 −0.041 Portugal 25.5 – 0.247 – Conservative Czech Republic 24.3 20.8 −0.154 −0.218 France 32.9 26.2 0.136 0.077 Spain 21.3 20.7 0.063 0.068 Germany 28.2 21.2 0.013 −0.110 Turkey 16.9 – 0.347 – Switzerland 16.0 17.5 −0.170 −0.066 Slovenia – 27.5 – 0.011 Finland 14.4 23.2 −0.219 −0.127 Estonia – 17.9 – −0.099 Size of benefits Target efficiency Size of benefits Target efficiency Regimes/ countries OECD WC OECD WC Regimes/ countries OECD WC OECD WC Social democratic Liberal Austria 36.6 26.7 0.157 0.108 Great Britain 14.5 14.3 −0.275 −0.313 Sweden 32.7 24.6 −0.145 −0.128 Non-classified Belgium 30.5 7.9 −0.120 −0.244 Poland 35.8 32.5 0.185 0.157 Denmark 25.6 18.9 −0.316 −0.306 Hungary 35.1 35.7 −0.016 0.016 Norway 21.7 20.2 −0.183 −0.155 Slovak Republic 26.0 26.6 −0.056 −0.109 Netherlands 17.1 21.3 −0.198 −0.041 Portugal 25.5 – 0.247 – Conservative Czech Republic 24.3 20.8 −0.154 −0.218 France 32.9 26.2 0.136 0.077 Spain 21.3 20.7 0.063 0.068 Germany 28.2 21.2 0.013 −0.110 Turkey 16.9 – 0.347 – Switzerland 16.0 17.5 −0.170 −0.066 Slovenia – 27.5 – 0.011 Finland 14.4 23.2 −0.219 −0.127 Estonia – 17.9 – −0.099 Notes: Regime classification is based on Esping-Andersen (1990, pp. 51–52). Countries are sorted in descending order by the size of transfer benefits (OECD, 2008) within each welfare regime cluster. ‘WC’ = Wang and Caminada (2011). Sources: OECD (2008); Wang and Caminada (2011). Using the quantitative regime indicators—the concentration and size of public cash benefits—has at least two advantages over some other operationalizations of welfare regime. First, it seems obvious that cross-national variation in target efficiency cannot be captured within the framework of traditional regime typology. As shown in Table 2, while the regime classification roughly reflects the variation in the size of public transfers—the scores are comparatively high in the social democratic cluster and low in its liberal counterpart—the scores of target efficiency do not have clear between-regime differences or within-regime similarities. In previous empirical studies, capitalist democracies have often been clustered into three (or more) distinct regime groups. However, ideal types of welfare regimes (Esping-Andersen, 1990) cannot be interpreted as ‘real entities’; groups of (or individual) countries cannot necessarily be treated as effective representatives of different welfare regimes (Svallfors, 1997; Jæger, 2006). The notion that welfare states are qualitatively distinct relies on the implicit assumption that each regime cluster has internal homogeneity. If this assumption does not hold (as in the case of target efficiency), potential within-regime variations are to be ignored. In this respect, employing quantitative measures allows for carefully capturing institutional differences across individual countries. Second, the concentration and size of cash benefits seem to affect people’s preferences more straightforwardly than some other quantitative indicators. An indicator such as the size of social spending as a percentage of GDP appears less appropriate because welfare stakeholders might either not be aware of the level of government spending in their country, or it might not be evident to them as to what the ‘percentage of national GDP’ implies for the expected welfare benefits their households actually receive. In contrast, for welfare stakeholders, the proportion of public cash benefits in their household income (i.e. the size of public transfers) and the extent to which transfer benefits are distributed to their own income group relative to other ‘rival’ groups (i.e. the concentration of public transfers)—are more tangible and easier to calculate and thus to respond to. Finally, to hold constant the effects of other macro-level factors, the following two variables were included in the analysis: market income inequality and economic development. Market income inequality was assessed by the Gini coefficient before taxes and transfers (OECD, 2005; Wang and Caminada, 2011), while economic development was assessed by gross domestic product (GDP) per capita adjusted for PPP in US dollars (IMF, 2005). In particular, controlling for ‘market income inequality’ adds strength to the causal inferences because the degree of targeted spending at the bottom is expected to be greater in countries where the distribution of market income is highly concentrated at the top. All macro-level factors were standardized by subtracting their means and dividing by their standard deviations. For the four macro-level indicators, I used the values from a single year around 2005 (mid-2000s). Descriptive statistics are shown in Appendix Table A1. The bivariate correlations between macro-level measures are presented in appendix Table A2. 3.3 Analytical approach A cross-sectional multilevel modelling approach was applied to analyze hierarchically nested samples and simultaneously examine the interplay between macro-level factors (e.g. targeting) and short-term individual-level preferences (Goldstein, 1995; Kreft and de Leeuw, 1998; Snijders and Bosker, 1999). Another possible approach might be to analyze over-time trends in targeting policies and subsequent preferences within countries. Nevertheless, a multilevel modelling approach seems more justifiable given that the key motivation of this article is to examine how pre-existing regime differences across countries would affect the association between income position and outcome preferences. Moreover, sufficient multi-year data on target efficiency are currently unavailable to carry out a longitudinal assessment. A series of two-level ordered logit models was estimated using MLwiN version 2.30 (Rasbash et al., 2014) via the ‘runmlwin’ command in Stata (Leckie and Charlton, 2012). 4. Results The purpose of this study is to test how different indicators of inequality attitudes affect the micro-level process of the ‘paradox of redistribution’ (Hypothesis 2) in light of the theoretical assumption that attitudinal polarization is induced when respondents place themselves into a particular income group, and articulate a sense of ingroup favoritism and intergroup conflict. To explore this, I begin by comparing the results obtained from the ISSP and the ESS (Table 3). Estimated parameters in the left two columns are based on the ISSP 2009, while those in the right two columns are from the ESS 2008. As suggested in Hypothesis 1, higher income is indeed significantly associated with weaker support for redistribution (models 1–4). Model 1 shows that the coefficient of the interaction term between income and targeted spending (OECD) was positive and highly significant (P < 0.01), which lends support for Hypothesis 2 that low-income targeting contextualizes the relationship between income and inequality attitudes (the more targeted spending, the more attitudinal polarization). Model 2 shows that the interaction effect remains significant (but less marked; P < 0.10) when the same model was tested using the targeted-spending measure calculated by Wang and Caminada (2011). In models 3–4, Hypothesis 2 was re-examined using a slightly different inequality-attitude item that does not accentuate the contrast between high- and low-income earners (ESS). The results show that the interaction effect was in the expected direction but failed to reach statistical significance, regardless of which targeted-spending measure (OECD or WC) was used. Table 3. Government should reduce income differences: comparison of ISSP 2009 and ESS 2008 Q1. Govt. resp. to reduce income differences b/w high and low incomes (ISSP) Q2. Govt. should take measures to reduce differences in income levels (ESS) Model 1: OECD Model 2: WC Model 3: OECD Model 4: WC Variables coef. s.e. coef. s.e. coef. s.e. coef. s.e. Individual-level predictors Household income (equivalized) −0.502*** (0.076) −0.544*** (0.078) −0.395*** (0.064) −0.442*** (0.059) Level-1 controls Yes Yes Yes Yes Country-level factors Target efficiency (OECD) 0.246* (0.108) – – 0.288* (0.136) – – Size of public transfers (OECD) 0.009 (0.085) – – −0.071 (0.109) – – Target efficiency (WC) – – 0.199† (0.119) – – 0.221 (0.136) Size of public transfers (WC) – – 0.030 (0.121) – – 0.033 (0.139) Gini (market income) 0.151† (0.089) 0.133 (0.093) 0.180 (0.109) 0.176 (0.104) GDP(PPP)/capita −0.196† (0.106) −0.269* (0.120) −0.043 (0.128) −0.041 (0.133) Cross-level interactions Income × Target efficiency (OECD) 0.324** (0.102) – – 0.124 (0.088) – – Income × Size of public transfers (OECD) −0.226** (0.080) – – −0.082 (0.070) – – Income × Target efficiency (WC) – – 0.179† (0.103) – – 0.016 (0.079) Income × Size of public transfers (WC) – – −0.182† (0.104) – – 0.026 (0.083) Income × Gini (market income) 0.053 (0.083) 0.030 (0.080) 0.069 (0.070) 0.067 (0.062) Income × GDP(PPP)/capita 0.138 (0.099) 0.033 (0.102) 0.027 (0.084) 0.069 (0.081) Intercept variance 0.102 (0.035) 0.125 (0.043) 0.164 (0.055) 0.167 (0.056) Slope variance (income) 0.082 (0.031) 0.082 (0.031) 0.060 (0.022) 0.050 (0.019) Covariance b/w intercepts and slopes 0.009 (0.023) 0.028 (0.027) 0.033 (0.026) 0.040 (0.025) VPC (variance partition coefficient) 0.030 0.037 0.047 0.048 Nindividual 23,050 22,562 34,961 33,236 Ncountry 18 18 18 18 Q1. Govt. resp. to reduce income differences b/w high and low incomes (ISSP) Q2. Govt. should take measures to reduce differences in income levels (ESS) Model 1: OECD Model 2: WC Model 3: OECD Model 4: WC Variables coef. s.e. coef. s.e. coef. s.e. coef. s.e. Individual-level predictors Household income (equivalized) −0.502*** (0.076) −0.544*** (0.078) −0.395*** (0.064) −0.442*** (0.059) Level-1 controls Yes Yes Yes Yes Country-level factors Target efficiency (OECD) 0.246* (0.108) – – 0.288* (0.136) – – Size of public transfers (OECD) 0.009 (0.085) – – −0.071 (0.109) – – Target efficiency (WC) – – 0.199† (0.119) – – 0.221 (0.136) Size of public transfers (WC) – – 0.030 (0.121) – – 0.033 (0.139) Gini (market income) 0.151† (0.089) 0.133 (0.093) 0.180 (0.109) 0.176 (0.104) GDP(PPP)/capita −0.196† (0.106) −0.269* (0.120) −0.043 (0.128) −0.041 (0.133) Cross-level interactions Income × Target efficiency (OECD) 0.324** (0.102) – – 0.124 (0.088) – – Income × Size of public transfers (OECD) −0.226** (0.080) – – −0.082 (0.070) – – Income × Target efficiency (WC) – – 0.179† (0.103) – – 0.016 (0.079) Income × Size of public transfers (WC) – – −0.182† (0.104) – – 0.026 (0.083) Income × Gini (market income) 0.053 (0.083) 0.030 (0.080) 0.069 (0.070) 0.067 (0.062) Income × GDP(PPP)/capita 0.138 (0.099) 0.033 (0.102) 0.027 (0.084) 0.069 (0.081) Intercept variance 0.102 (0.035) 0.125 (0.043) 0.164 (0.055) 0.167 (0.056) Slope variance (income) 0.082 (0.031) 0.082 (0.031) 0.060 (0.022) 0.050 (0.019) Covariance b/w intercepts and slopes 0.009 (0.023) 0.028 (0.027) 0.033 (0.026) 0.040 (0.025) VPC (variance partition coefficient) 0.030 0.037 0.047 0.048 Nindividual 23,050 22,562 34,961 33,236 Ncountry 18 18 18 18 Notes: Random intercept and slope (income) models. RIGLS (reiterated generalized least squares) with PQL2 (second-order penalized quasi-likelihood). Intercept variances for the empty model (with no predictor) are: [Question 1] 0.343 with a VPC (variance partition coefficient) of 0.094 (OECD, 2008); 0.298 with a VPC of 0.083 (Wang and Caminada, 2011), [Question 2] 0.319 with a VPC of 0.089 (OECD, 2008); 0.259 with a VPC of 0.073 (Wang and Caminada, 2011). The countries included in each model are presented in Appendix Table A3. Missing data were imputed using multiple imputation methods (Rubin, 1987). Estimated cut-points not displayed. ‘WC’ = Wang and Caminada (2011). Sources: ISSP (2009); ESS (2008); OECD (2008); Wang and Caminada (2011). † P < 0.10; *P < 0.05; **P < 0.01; ***P < 0.001. Table 3. Government should reduce income differences: comparison of ISSP 2009 and ESS 2008 Q1. Govt. resp. to reduce income differences b/w high and low incomes (ISSP) Q2. Govt. should take measures to reduce differences in income levels (ESS) Model 1: OECD Model 2: WC Model 3: OECD Model 4: WC Variables coef. s.e. coef. s.e. coef. s.e. coef. s.e. Individual-level predictors Household income (equivalized) −0.502*** (0.076) −0.544*** (0.078) −0.395*** (0.064) −0.442*** (0.059) Level-1 controls Yes Yes Yes Yes Country-level factors Target efficiency (OECD) 0.246* (0.108) – – 0.288* (0.136) – – Size of public transfers (OECD) 0.009 (0.085) – – −0.071 (0.109) – – Target efficiency (WC) – – 0.199† (0.119) – – 0.221 (0.136) Size of public transfers (WC) – – 0.030 (0.121) – – 0.033 (0.139) Gini (market income) 0.151† (0.089) 0.133 (0.093) 0.180 (0.109) 0.176 (0.104) GDP(PPP)/capita −0.196† (0.106) −0.269* (0.120) −0.043 (0.128) −0.041 (0.133) Cross-level interactions Income × Target efficiency (OECD) 0.324** (0.102) – – 0.124 (0.088) – – Income × Size of public transfers (OECD) −0.226** (0.080) – – −0.082 (0.070) – – Income × Target efficiency (WC) – – 0.179† (0.103) – – 0.016 (0.079) Income × Size of public transfers (WC) – – −0.182† (0.104) – – 0.026 (0.083) Income × Gini (market income) 0.053 (0.083) 0.030 (0.080) 0.069 (0.070) 0.067 (0.062) Income × GDP(PPP)/capita 0.138 (0.099) 0.033 (0.102) 0.027 (0.084) 0.069 (0.081) Intercept variance 0.102 (0.035) 0.125 (0.043) 0.164 (0.055) 0.167 (0.056) Slope variance (income) 0.082 (0.031) 0.082 (0.031) 0.060 (0.022) 0.050 (0.019) Covariance b/w intercepts and slopes 0.009 (0.023) 0.028 (0.027) 0.033 (0.026) 0.040 (0.025) VPC (variance partition coefficient) 0.030 0.037 0.047 0.048 Nindividual 23,050 22,562 34,961 33,236 Ncountry 18 18 18 18 Q1. Govt. resp. to reduce income differences b/w high and low incomes (ISSP) Q2. Govt. should take measures to reduce differences in income levels (ESS) Model 1: OECD Model 2: WC Model 3: OECD Model 4: WC Variables coef. s.e. coef. s.e. coef. s.e. coef. s.e. Individual-level predictors Household income (equivalized) −0.502*** (0.076) −0.544*** (0.078) −0.395*** (0.064) −0.442*** (0.059) Level-1 controls Yes Yes Yes Yes Country-level factors Target efficiency (OECD) 0.246* (0.108) – – 0.288* (0.136) – – Size of public transfers (OECD) 0.009 (0.085) – – −0.071 (0.109) – – Target efficiency (WC) – – 0.199† (0.119) – – 0.221 (0.136) Size of public transfers (WC) – – 0.030 (0.121) – – 0.033 (0.139) Gini (market income) 0.151† (0.089) 0.133 (0.093) 0.180 (0.109) 0.176 (0.104) GDP(PPP)/capita −0.196† (0.106) −0.269* (0.120) −0.043 (0.128) −0.041 (0.133) Cross-level interactions Income × Target efficiency (OECD) 0.324** (0.102) – – 0.124 (0.088) – – Income × Size of public transfers (OECD) −0.226** (0.080) – – −0.082 (0.070) – – Income × Target efficiency (WC) – – 0.179† (0.103) – – 0.016 (0.079) Income × Size of public transfers (WC) – – −0.182† (0.104) – – 0.026 (0.083) Income × Gini (market income) 0.053 (0.083) 0.030 (0.080) 0.069 (0.070) 0.067 (0.062) Income × GDP(PPP)/capita 0.138 (0.099) 0.033 (0.102) 0.027 (0.084) 0.069 (0.081) Intercept variance 0.102 (0.035) 0.125 (0.043) 0.164 (0.055) 0.167 (0.056) Slope variance (income) 0.082 (0.031) 0.082 (0.031) 0.060 (0.022) 0.050 (0.019) Covariance b/w intercepts and slopes 0.009 (0.023) 0.028 (0.027) 0.033 (0.026) 0.040 (0.025) VPC (variance partition coefficient) 0.030 0.037 0.047 0.048 Nindividual 23,050 22,562 34,961 33,236 Ncountry 18 18 18 18 Notes: Random intercept and slope (income) models. RIGLS (reiterated generalized least squares) with PQL2 (second-order penalized quasi-likelihood). Intercept variances for the empty model (with no predictor) are: [Question 1] 0.343 with a VPC (variance partition coefficient) of 0.094 (OECD, 2008); 0.298 with a VPC of 0.083 (Wang and Caminada, 2011), [Question 2] 0.319 with a VPC of 0.089 (OECD, 2008); 0.259 with a VPC of 0.073 (Wang and Caminada, 2011). The countries included in each model are presented in Appendix Table A3. Missing data were imputed using multiple imputation methods (Rubin, 1987). Estimated cut-points not displayed. ‘WC’ = Wang and Caminada (2011). Sources: ISSP (2009); ESS (2008); OECD (2008); Wang and Caminada (2011). † P < 0.10; *P < 0.05; **P < 0.01; ***P < 0.001. Table 4 provides further results using alternative inequality-attitude items from the EVS 2008: ‘income equality versus incentives for effort’ (models 5–6) and ‘more important: equality versus freedom’ (models 7–8). Models 5–6 show that the coefficient of the interaction term (income and target efficiency) is positive and significant, indicating that higher income earners are more likely to favor incentives for effort rather than income equality in the context of targeted spending. On the other hand, models 7–8 show that the polarization process does not occur when the question item was phrased more vaguely (or more normatively) as ‘equality versus freedom’. Table 4. Conflict between equality and freedom/incentives for effort, EVS 2008 Q3. Income equality versus incentives on effort Q4. More important: equality versus freedom Model 5: OECD Model 6: WC Model 7: OECD Model 8: WC Variables coef. s.e. coef. s.e. coef. s.e. coef. s.e. Individual-level predictors Household income (equivalized) −0.265*** (0.040) −0.267*** (0.041) −0.190*** (0.050) −0.210*** (0.053) Level-1 controls Yes Yes Yes Yes Country-level factors Target efficiency (OECD) 0.272 (0.179) – – 0.178† (0.101) – – Size of public transfers (OECD) −0.019 (0.139) – – −0.004 (0.079) – – Target efficiency (WC) – – 0.190 (0.197) – – 0.132 (0.096) Size of public transfers (WC) – – 0.129 (0.212) – – −0.129 (0.103) Gini (market income) 0.041 (0.142) 0.050 (0.152) 0.036 (0.080) 0.076 (0.074) GDP(PPP)/capita 0.057 (0.171) 0.093 (0.200) 0.057 (0.096) 0.019 (0.097) Cross-level interactions Income × Target efficiency (OECD) 0.118* (0.054) – – 0.062 (0.071) – – Income × Size of public transfers (OECD) −0.026 (0.042) – – −0.118* (0.054) – – Income × Target efficiency (WC) – – 0.127* (0.051) – – 0.071 (0.073) Income × Size of public transfers (WC) – – −0.087 (0.055) – – −0.054 (0.077) Income × Gini (market income) 0.013 (0.043) 0.023 (0.040) 0.030 (0.056) 0.035 (0.054) Income × GDP(PPP)/capita 0.034 (0.055) −0.026 (0.055) −0.081 (0.070) −0.047 (0.073) Intercept variance 0.272 (0.091) 0.349 (0.117) 0.084 (0.029) 0.080 (0.028) Slope variance (income) 0.019 (0.008) 0.017 (0.008) 0.036 (0.016) 0.036 (0.015) Covariance b/w intercepts and slopes −0.025 (0.020) −0.040 (0.023) −0.019 (0.016) −0.019 (0.015) VPC (variance partition coefficient) 0.076 0.096 0.025 0.024 Nindividual 27,422 26,430 26,010 25,186 Ncountry 18 18 18 18 Q3. Income equality versus incentives on effort Q4. More important: equality versus freedom Model 5: OECD Model 6: WC Model 7: OECD Model 8: WC Variables coef. s.e. coef. s.e. coef. s.e. coef. s.e. Individual-level predictors Household income (equivalized) −0.265*** (0.040) −0.267*** (0.041) −0.190*** (0.050) −0.210*** (0.053) Level-1 controls Yes Yes Yes Yes Country-level factors Target efficiency (OECD) 0.272 (0.179) – – 0.178† (0.101) – – Size of public transfers (OECD) −0.019 (0.139) – – −0.004 (0.079) – – Target efficiency (WC) – – 0.190 (0.197) – – 0.132 (0.096) Size of public transfers (WC) – – 0.129 (0.212) – – −0.129 (0.103) Gini (market income) 0.041 (0.142) 0.050 (0.152) 0.036 (0.080) 0.076 (0.074) GDP(PPP)/capita 0.057 (0.171) 0.093 (0.200) 0.057 (0.096) 0.019 (0.097) Cross-level interactions Income × Target efficiency (OECD) 0.118* (0.054) – – 0.062 (0.071) – – Income × Size of public transfers (OECD) −0.026 (0.042) – – −0.118* (0.054) – – Income × Target efficiency (WC) – – 0.127* (0.051) – – 0.071 (0.073) Income × Size of public transfers (WC) – – −0.087 (0.055) – – −0.054 (0.077) Income × Gini (market income) 0.013 (0.043) 0.023 (0.040) 0.030 (0.056) 0.035 (0.054) Income × GDP(PPP)/capita 0.034 (0.055) −0.026 (0.055) −0.081 (0.070) −0.047 (0.073) Intercept variance 0.272 (0.091) 0.349 (0.117) 0.084 (0.029) 0.080 (0.028) Slope variance (income) 0.019 (0.008) 0.017 (0.008) 0.036 (0.016) 0.036 (0.015) Covariance b/w intercepts and slopes −0.025 (0.020) −0.040 (0.023) −0.019 (0.016) −0.019 (0.015) VPC (variance partition coefficient) 0.076 0.096 0.025 0.024 Nindividual 27,422 26,430 26,010 25,186 Ncountry 18 18 18 18 Notes: Random intercept and slope (income) models. RIGLS (reiterated generalized least squares) with PQL2 (second-order penalized quasi-likelihood). Intercept variances for the empty model (with no predictor) are: [Question 3] 0.257 with a VPC (variance partition coefficient) of 0.086 (OECD, 2008); 0.333 with a VPC of 0.092 (Wang and Caminada, 2011), [Question 4] 0.086 with a VPC of 0.026 (OECD, 2008); 0.068 with a VPC of 0.020 (Wang and Caminada, 2011). The countries included in each model are presented in Appendix Table A3. Missing data were imputed using multiple imputation methods (Rubin, 1987). Estimated cut-points not displayed. ‘WC’ = Wang and Caminada (2011). Sources: EVS (2008); OECD (2008); Wang and Caminada (2011). † P < 0.10; *P < 0.05; **P < 0.01; ***P < 0.001. Table 4. Conflict between equality and freedom/incentives for effort, EVS 2008 Q3. Income equality versus incentives on effort Q4. More important: equality versus freedom Model 5: OECD Model 6: WC Model 7: OECD Model 8: WC Variables coef. s.e. coef. s.e. coef. s.e. coef. s.e. Individual-level predictors Household income (equivalized) −0.265*** (0.040) −0.267*** (0.041) −0.190*** (0.050) −0.210*** (0.053) Level-1 controls Yes Yes Yes Yes Country-level factors Target efficiency (OECD) 0.272 (0.179) – – 0.178† (0.101) – – Size of public transfers (OECD) −0.019 (0.139) – – −0.004 (0.079) – – Target efficiency (WC) – – 0.190 (0.197) – – 0.132 (0.096) Size of public transfers (WC) – – 0.129 (0.212) – – −0.129 (0.103) Gini (market income) 0.041 (0.142) 0.050 (0.152) 0.036 (0.080) 0.076 (0.074) GDP(PPP)/capita 0.057 (0.171) 0.093 (0.200) 0.057 (0.096) 0.019 (0.097) Cross-level interactions Income × Target efficiency (OECD) 0.118* (0.054) – – 0.062 (0.071) – – Income × Size of public transfers (OECD) −0.026 (0.042) – – −0.118* (0.054) – – Income × Target efficiency (WC) – – 0.127* (0.051) – – 0.071 (0.073) Income × Size of public transfers (WC) – – −0.087 (0.055) – – −0.054 (0.077) Income × Gini (market income) 0.013 (0.043) 0.023 (0.040) 0.030 (0.056) 0.035 (0.054) Income × GDP(PPP)/capita 0.034 (0.055) −0.026 (0.055) −0.081 (0.070) −0.047 (0.073) Intercept variance 0.272 (0.091) 0.349 (0.117) 0.084 (0.029) 0.080 (0.028) Slope variance (income) 0.019 (0.008) 0.017 (0.008) 0.036 (0.016) 0.036 (0.015) Covariance b/w intercepts and slopes −0.025 (0.020) −0.040 (0.023) −0.019 (0.016) −0.019 (0.015) VPC (variance partition coefficient) 0.076 0.096 0.025 0.024 Nindividual 27,422 26,430 26,010 25,186 Ncountry 18 18 18 18 Q3. Income equality versus incentives on effort Q4. More important: equality versus freedom Model 5: OECD Model 6: WC Model 7: OECD Model 8: WC Variables coef. s.e. coef. s.e. coef. s.e. coef. s.e. Individual-level predictors Household income (equivalized) −0.265*** (0.040) −0.267*** (0.041) −0.190*** (0.050) −0.210*** (0.053) Level-1 controls Yes Yes Yes Yes Country-level factors Target efficiency (OECD) 0.272 (0.179) – – 0.178† (0.101) – – Size of public transfers (OECD) −0.019 (0.139) – – −0.004 (0.079) – – Target efficiency (WC) – – 0.190 (0.197) – – 0.132 (0.096) Size of public transfers (WC) – – 0.129 (0.212) – – −0.129 (0.103) Gini (market income) 0.041 (0.142) 0.050 (0.152) 0.036 (0.080) 0.076 (0.074) GDP(PPP)/capita 0.057 (0.171) 0.093 (0.200) 0.057 (0.096) 0.019 (0.097) Cross-level interactions Income × Target efficiency (OECD) 0.118* (0.054) – – 0.062 (0.071) – – Income × Size of public transfers (OECD) −0.026 (0.042) – – −0.118* (0.054) – – Income × Target efficiency (WC) – – 0.127* (0.051) – – 0.071 (0.073) Income × Size of public transfers (WC) – – −0.087 (0.055) – – −0.054 (0.077) Income × Gini (market income) 0.013 (0.043) 0.023 (0.040) 0.030 (0.056) 0.035 (0.054) Income × GDP(PPP)/capita 0.034 (0.055) −0.026 (0.055) −0.081 (0.070) −0.047 (0.073) Intercept variance 0.272 (0.091) 0.349 (0.117) 0.084 (0.029) 0.080 (0.028) Slope variance (income) 0.019 (0.008) 0.017 (0.008) 0.036 (0.016) 0.036 (0.015) Covariance b/w intercepts and slopes −0.025 (0.020) −0.040 (0.023) −0.019 (0.016) −0.019 (0.015) VPC (variance partition coefficient) 0.076 0.096 0.025 0.024 Nindividual 27,422 26,430 26,010 25,186 Ncountry 18 18 18 18 Notes: Random intercept and slope (income) models. RIGLS (reiterated generalized least squares) with PQL2 (second-order penalized quasi-likelihood). Intercept variances for the empty model (with no predictor) are: [Question 3] 0.257 with a VPC (variance partition coefficient) of 0.086 (OECD, 2008); 0.333 with a VPC of 0.092 (Wang and Caminada, 2011), [Question 4] 0.086 with a VPC of 0.026 (OECD, 2008); 0.068 with a VPC of 0.020 (Wang and Caminada, 2011). The countries included in each model are presented in Appendix Table A3. Missing data were imputed using multiple imputation methods (Rubin, 1987). Estimated cut-points not displayed. ‘WC’ = Wang and Caminada (2011). Sources: EVS (2008); OECD (2008); Wang and Caminada (2011). † P < 0.10; *P < 0.05; **P < 0.01; ***P < 0.001. To facilitate visualization, predicted probabilities were calculated based on the estimates from models 5–6 in Table 4. In Figure 3, the horizontal axis represents income position (in percentile); the vertical axis represents the predicted probability of strong support for income equality (v. incentives for effort). The dashed line illustrates the predicted probability for targeted spending, while the solid line corresponds to that for non-targeted spending. All other variables were held at their observed values, and then the estimated probabilities were averaged over all cases (Hanmer and Kalkan, 2013). Figure 3. View largeDownload slide Predicted probabilities: support for income equality (v. incentives for effort). Sources: EVS (2008); OECD (2008); Wang and Caminada (2011). Figure 3. View largeDownload slide Predicted probabilities: support for income equality (v. incentives for effort). Sources: EVS (2008); OECD (2008); Wang and Caminada (2011). 5. Discussion This study has attempted to provide a comprehensive picture of the role of targeted spending by using multiple data sources. Some evidence (models 1–2, Table 3) suggests that attitudinal cleavages grow wider in the context of targeted spending (Hypothesis 2), which lends some credence to the micro-level mechanism underlying the ‘paradox of redistribution’ (KP, 1998) and, more generally, to the institutionalist assumption that institutional structures have feedback effects on mass opinion (Pierson, 1993; Mettler and Soss, 2004). The result of model 1 (estimated using ISSP and OECD) was generally consistent with that of Beramendi and Rehm (2016) (i.e. the more low-income targeting, the more polarized attitudes), even though the analytical procedure was slightly different (e.g. level-2 sample, income measurement). This result (obtained from ISSP/OECD) was somewhat robust to the use of alternative targeted-spending measure (model 2, WC), though the significance level was less impressive (P < 0.10). However, the same hypothesis was not clearly evidenced when a slightly different question-wording was used (models 3–4, Table 3). Findings also indicate that the polarization process (Hypothesis 2) occurs when respondents were asked about their views on the conflict between income equality and incentives on effort (models 5–6, Table 4), but not when a similar question was asked in a more indirect manner (models 7–8). These results indicate that different question wordings provoke different reactions to inequality and redistribution. These inconsistent results could be explained by the function of identity priming. As presented in Table 1, Questions 1–2 (government redistribution) were worded in a similar manner, the only substantial difference being ‘high- versus low-income status’. The ISSP item, which clearly refers to the differences between high and low incomes, is expected to raise respondents’ self-awareness of their identity as a high-income group (welfare contributor) or low-income group (welfare recipient). In contrast, the ESS question is vaguely worded as ‘differences in income levels’, thus sends less explicit signals and implications regarding the conflict between the groups. This process makes it less likely that high- and low-income earners polarize their inequality attitudes in the context of targeted spending (Hypothesis 2). On the other hand, Question 3 addresses not only the issue of reducing income differences but also the potential trade-off between income equality and incentives on effort. The results (models 5–6) show that high-income earners prefer incentives for effort to income equality in welfare states with low-income targeting. As with Question 1, Question 3 is a conflict-prone question since it reminds respondents that reducing income inequality can be at odds with their striving for economic success. When asked in this manner, high-income earners may consider that the rich are rich because of their hard work, and that it is unfair to equalize income differences between hard-working high-income earners and ‘lazy’ low-income earners. This identification process might explain the polarization process (Hypothesis 2) observed in models 5–6 (Table 4). Finally, no significant interaction effect (Hypothesis 2) was observed for Question 4. In general, Question 4 was vaguely worded compared to the others: ‘underprivileged’ not ‘low-incomes’; ‘social class differences’ not ‘income inequality/differences’; and ‘personal freedom (without hindrance)’ not ‘incentives on effort’. Question 4 also lacks stimuli that evoke a sense of ingroup favoritism and intergroup conflict such as (a) income inequality, (b) high versus low incomes and (c) incentives on effort (Table 1). A few caveats are worth mentioning. First, the role of identity priming must be explored further using an experimental study design. The most basic form of inequality-attitude question would be: do you agree or disagree with the statement that inequality should be reduced. Respondents’ answers to this question would differ depending on whether their sense of identity/conflict is induced or not. In an experimental study, one can randomly assign respondents to experimental and control groups and then contrast the outcomes after the experimental group received an intervention that provokes their sense of identity/conflict. The treatment for the experimental group would include introductory/guiding sentences (before the basic question) such as ‘In [Country], you are in the top 20% of the income distribution’ (using demographic data collected); ‘By “inequality” we mean income differences’; ‘Reducing income inequality implies that some of the money you worked hard for will be used to help low-income earners who need assistance’; and/or ‘Please keep in mind that reducing inequality most likely entails tax increases on high-income earners (like yourself)’. Second, although this study sought to compare inequality-attitude items from different social surveys (ISSP, ESS, EVS), the comparison of different inequality-attitude measures should ideally be made using the same survey data. In this study, to harmonize potential inconsistences between the survey datasets, several efforts have been made: the same set of countries (level-2 sample), time-period, and explanatory variables and their measurements (Table A1). Several implications can be noted. First, the empirical results from this study direct our attention to the role of identity priming in redistributive politics. Hypothesis 2 predicted that the uneven distribution of public transfers serves as a contextual cue in defining expectations for the consequences of egalitarian practices. This notion is based on the premise that respondents are aware of their socio-economic status and their benefits/costs associated with an egalitarian goal. High-income earners become less willing to support egalitarian views in the context of low-income targeting because they consider that targeting unfairly benefits low-income group at the expense of their group. This mechanism does not function if a survey question fails to induce their sense of identity and conflict. Second, this study has provided a more detailed description of the ‘paradox of redistribution’ process (KP, 1998). The basic form of the ‘paradox of redistribution’ can be expressed as: targeted spending → redistribution. As argued in the theory section, the underlying (micro-level) logic of this process is that first, low-income targeting generates identity-based preferences toward inequality, and second, polarized preferences are translated to redistributive policy outputs. Thus, the ‘paradox of redistribution’ process can be illustrated as: Targeted spending (a) → identity priming (b1) → polarization in attitudes (b2) → redistribution (c). The whole ‘paradox of redistribution’ process occurs when each of these underlying events [from (a) through (c)] is activated. Interestingly, recent studies have shown that the aggregate-level relationship between target efficiency (a) and redistribution (c) can no longer be observed or at least has substantially weakened since the mid-1990s (Kenworthy, 2011; Marx et al., 2013). On the other hand, the results of this article (models 1–2, Table 3) have provided some evidence in favor of the causal sequence from (a) through (b2). These seemingly puzzling results could be explained by looking at the micro-level process—for instance, the link between attitudinal polarization (b2) and actual policy outputs (c) is by no means as self-evident as is often assumed. In conclusion, this study has presented suggestive evidence that the validity of the polarization hypothesis (Hypothesis 2) is susceptible to the question wording of inequality-attitude indicators, and has provided a theoretical rationale for why this might be the case by suggesting that identity priming plays a crucial role in the micro-level mechanism of the ‘paradox of redistribution’. A promising avenue for future research would be to expand on this line of interest and provide more rigorous (e.g. experimental) data to confirm its plausibility. Acknowledgements I wish to thank Professors Stephen Fisher, Nan Dirk de Graaf, Jane Gingrich and Tamas David-Barrett for their kind assistance during the study. I am also indebted to Professor David Rueda and anonymous reviewers for their valuable contributions to its development. References Alesina A. , Ferrara E. L. ( 2005 ) ‘ Preferences for Redistribution in the Land of Opportunities’, Journal of Public Economics , 89 , 897 – 931 . Google Scholar CrossRef Search ADS Barnes L. ( 2015 ) ‘ The Size and Shape of Government: Preferences over Redistributive Tax Policy’, Socio-Economic Review , 13 , 55 – 78 . Google Scholar CrossRef Search ADS Beramendi P. , Rehm P. ( 2016 ) ‘ Who Gives, Who Gains? Progressivity and Preferences’, Comparative Political Studies , 49 , 529 – 563 . Google Scholar CrossRef Search ADS Brady D. , Bostic A. ( 2015 ) ‘ Paradoxes of Social Policy: Welfare Transfers, Relative Poverty, and Redistribution Preferences’, American Sociological Review , 80 , 268 – 298 . Google Scholar CrossRef Search ADS Buhmann B. , Rainwater L. , Schmaus G. , Smeeding T. M. ( 1988 ) ‘ Equivalence Scales, Well-being, Inequality, and Poverty: Sensitivity Estimates across Ten Countries Using the Luxembourg Income Study (LIS) Database’, Review of Income and Wealth , 34 , 115 – 142 . Google Scholar CrossRef Search ADS Cusack T. , Iversen T. , Rehm P. ( 2006 ) ‘ Risks at Work: The Demand and Supply Sides of Government Redistribution’, Oxford Review of Economic Policy , 22 , 365 – 389 . Google Scholar CrossRef Search ADS Edlund J. ( 1999 ) ‘ Trust in Government and Welfare Regimes: Attitudes to Redistribution and Financial Cheating in the USA and Norway’, European Journal of Political Research , 35 , 341 – 370 . Esping-Andersen G. ( 1990 ) The Three Worlds of Welfare Capitalism , Cambridge , Polity Press . Fields G. S. ( 1979 ) ‘ Decomposing LDC Inequality’, Oxford Economic Papers , 31 , 437 – 459 . Google Scholar CrossRef Search ADS Gingrich J. , Ansell B. ( 2012 ) ‘ Preferences in Context: Micro Preferences, Macro Contexts, and the Demand for Social Policy’, Comparative Political Studies , 45 , 1624 – 1654 . Google Scholar CrossRef Search ADS Goldstein H. ( 1995 ) Multilevel Statistical Models , London , Edward Arnold . Graham J. W. ( 2009 ) ‘ Missing Data Analysis: Making it Work in the Real World’, Annual Review of Psychology , 60 , 549 – 576 . Google Scholar CrossRef Search ADS PubMed Hanmer M. J. , Kalkan K. O. ( 2013 ) ‘ Behind the Curve: Clarifying the Best Approach to Calculating Predicted Probabilities and Marginal Effects from Limited Dependent Variable Models’, America Journal of Political Science , 57 , 263 – 277 . Google Scholar CrossRef Search ADS Honaker J. , King G. , Blackwell M. ( 2011 ) ‘ Amelia II: A Program for Missing Data’, Journal of Statistical Software , 45 , 1 – 47 . Google Scholar CrossRef Search ADS Hout M. ( 2004 ) Getting the Most Out of the GSS Income Measures , GSS Methodological Report No. 101, Chicago , NORC . IMF ( 2005 ) World Economic Outlook Database (September) , Washington, DC , IMF . Jæger M. M. ( 2006 ) ‘ Welfare Regimes and Attitudes towards Redistribution: The Regime Hypothesis Revisited’, European Sociological Review , 22 , 157 – 170 . Google Scholar CrossRef Search ADS Jæger M. M. ( 2008 ) ‘ Does Left-Right Orientation have a Causal Effect on Support for Redistribution? Causal Analysis with Cross-sectional Data Using Instrumental Variables’, International Journal of Public Opinion Research , 20 , 363 – 374 . Google Scholar CrossRef Search ADS Kakwani N. ( 1986 ) Analyzing Redistribution Politics , Oxford , Oxford University Press . Kenworthy L. ( 2011 ) Progress for the Poor , Oxford , Oxford University Press . Google Scholar CrossRef Search ADS King G. , Honaker J. , Joseph A. , Scheve K. ( 2001 ) ‘Analyzing Incomplete Political Science Data: An Alternative Algorithm for Multiple Imputation, American Political Science Review , 95 , 49 – 69 . Korpi W. ( 1980 ) ‘ Social Policy and Distributional Conflict in the Capitalist Democracies: A Preliminary Comparative Framework’, West European Politics , 3 , 296 – 316 . Google Scholar CrossRef Search ADS Korpi W. , Palme J. ( 1998 ) ‘ The Paradox of Redistribution and Strategies of Equality: Welfare State Institutions, Inequality, and Poverty in the Western Countries’, American Sociological Review , 63 , 661 – 687 . Google Scholar CrossRef Search ADS Kreft I. , de Leeuw J. , ( 1998 ) Introducing Multilevel Modeling , London , Sage . Google Scholar CrossRef Search ADS Larsen C. A. ( 2008 ) ‘ The Institutional Logic of Welfare Attitudes: How Welfare Regimes Influence Public Support’, Comparative Political Studies , 41 , 145 – 168 . Google Scholar CrossRef Search ADS Leckie G. , Charlton C. ( 2012 ) ‘ runmlwin - A Program to Run the MLwiN Multilevel Modeling Software from within Stata’, Journal of Statistical Software , 52 , 1 – 40 . Google Scholar CrossRef Search ADS Luttmer E. F. P. ( 2001 ) ‘ Group Loyalty and the Taste for Redistribution’, Journal of Political Economy , 109 , 500 – 528 . Google Scholar CrossRef Search ADS Margalit Y. ( 2013 ) ‘ Explaining Social Policy Preferences: Evidence from the Great Recession’, American Political Science Review , 107 , 80 – 103 . Google Scholar CrossRef Search ADS Marx I. , Salanauskaite L. , Verbist G. ( 2013 ) The Paradox of Redistribution Revisited: And That It May Rest in Peace?, IZA Discussion Paper No. 7414, Bonn, Institute for the Study of Labor (IZA). Meltzer A. H. , Richard S. F. ( 1981 ) ‘ A Rational Theory of the Size of Government’, Journal of Political Economy , 89 , 914 – 927 . Google Scholar CrossRef Search ADS Mettler S. , Soss J. ( 2004 ) ‘ The Consequences of Public Policy for Democratic Citizenship: Bridging Policy Studies and Mass Politics’, Perspectives on Politics , 2 , 55 – 73 . Google Scholar CrossRef Search ADS OECD ( 2005 ) Income Distribution Database (IDD) , Paris , OECD . OECD ( 2008 ) ‘How Much Redistribution Do Governments Achieve? The Role of Cash Transfers and Household Taxes’. In OECD (ed.) Growing Unequal? Income Distribution and Poverty in OECD Countries , Paris , OECD , pp. 97 – 121 . Offe C. ( 1992 ) ‘A Non-Productivist Design for Social Policies’. In Parijs P. V. (ed.) Arguing for Basic Income , London , Verso . Pierson P. ( 1993 ) ‘ When Effect Becomes Cause: Policy Feedback and Political Change’, World Politics , 45 , 595 – 628 . Google Scholar CrossRef Search ADS Pierson P. ( 1996 ) ‘ The New Politics of the Welfare State’, World Politics , 48 , 143 – 179 . Google Scholar CrossRef Search ADS Pierson P. ( 2001 ) ‘Coping with Permanent Austerity: Welfare State Restructuring in Affluent Democracies’. In Pierson P. (ed.) The New Politics of the Welfare State , Oxford , Oxford University Press , pp. 410 – 456 . Google Scholar CrossRef Search ADS Rasbash J. , Steele F. , Browne W. J. , Goldstein H. , ( 2014 ) A User's Guide to MLwiN version 2.31, Centre for Multilevel Modelling, University of Bristol. Rehm P. ( 2009 ) ‘ Risks and Redistribution: An Individual-level Analysis’, Comparative Political Studies , 42 , 855 – 881 . Google Scholar CrossRef Search ADS Roberts K. W. S. ( 1977 ) ‘ Voting over Income Tax Schedules’, Journal of Public Economics , 8 , 329 – 340 . Google Scholar CrossRef Search ADS Romer T. ( 1975 ) ‘ Individual Welfare, Majority Voting, and the Properties of a Linear Income Tax’, Journal of Public Economics , 4 , 163 – 185 . Google Scholar CrossRef Search ADS Rothstein B. ( 1998 ) Just Institutions Matter: The Moral and Political Logic of the Universal Welfare State , Cambridge , Cambridge University Press . Google Scholar CrossRef Search ADS Rothstein B. , Steinmo S. , ( 2002 ) ‘Restructuring Politics: Institutional Analysis and the Challenges of Modern Welfare States’. In Rothstein B. , Steinmo S. (eds) Restructuring the Welfare State: Political Institutions and Policy Change , New York , Palgrave Macmillan , pp. 1 – 19 . Google Scholar CrossRef Search ADS Rubin D. B. ( 1987 ) Multiple Imputation for Nonresponse in Surveys , New York , John Wiley & Sons, Inc . Google Scholar CrossRef Search ADS Rubin D. B. ( 1996 ) ‘ Multiple Imputation After 18+ Years’, Journal of the American Statistical Association , 91 , 473 – 489 . Google Scholar CrossRef Search ADS Rueda D. ( 2014 ) Food Comes First, Then Morals: Redistribution Preferences, Altruism and Group Heterogeneity in Western Europe, CAGE Working Paper Series No. 200, Competitive Advantage in the Global Economy (CAGE). Rueda D. , Stegmueller D. ( 2015 ) ‘ The Externalities of Inequality: Fear of Crime and Preferences for Redistribution in Western Europe’, American Journal of Political Science , 60 , 472 – 489 . Google Scholar CrossRef Search ADS Rueda D. , Segmueller D. , Idema T. , ( 2014 ) ‘The Effects of Income Expectations on Redistribution Preferences in Western Europe’, paper presented at the Annual Meeting of the American Political Science Association (APSA), 2014. Skocpol T. ( 1992 ) Protecting Soldiers and Mothers , Cambridge, MA , Harvard University Press . Snijders T. A. B. , Bosker R. J. ( 1999 ) Multilevel Analysis: An Introduction to Basic and Advanced Multilevel Modeling , London , Sage . Sumino T. ( 2014 ) ‘ Escaping the Curse of Economic Self-interest: An Individual-level Analysis of Public Support for the Welfare State in Japan’, Journal of Social Policy , 43 , 109 – 133 . Google Scholar CrossRef Search ADS Sumino T. ( 2016 ) ‘ Level or Concentration? A Cross-national Analysis of Public Attitudes towards Taxation Policies’, Social Indicators Research , 129 , 1115 – 1134 . Google Scholar CrossRef Search ADS Svallfors S. ( 1997 ) ‘ Worlds of Welfare and Attitudes to Redistribution: A Comparison of Eight Western Nations’, European Sociological Review , 13 , 283 – 304 . Google Scholar CrossRef Search ADS Tajfel H. ( 1981 ) Human Groups and Social Categories . Cambridge , Cambridge University Press . Turner J. C. ( 1985 ) ‘Social Categorization and the Self-concept: A Social Cognitive Theory of Group Behaviour’. In Lawler E. J. (eds) Advances in Group Processes , Greenwich, CT , JAI Press , pp. 77 – 122 . Turner J. C. , Hogg M. A. , Oakes P. J. , Reicher S. D. , Wetherell M. S. , ( 1987 ) Rediscovering the Social Group: A Self-categorization Theory . Oxford and New York , Basil Blackwell . Wang C. , Caminada K. ( 2011 ) Disentangling Income Inequality and the Redistributive Effect of Social Transfers and Taxes in 36 LIS Countries , LIS Working Paper Series No. 567, Luxembourg , Luxembourg Income Study (LIS ). White I. R. , Royston P. , Wood A. M. ( 2011 ) ‘ Multiple Imputation Using Chained Equations: Issues and Guidance for Practice’, Statistics in Medicine , 30 , 377 – 399 . Google Scholar CrossRef Search ADS PubMed World Bank ( 2008 ) International Comparison Program Database . Washington, DC , World Bank . Appendix A Table A1. Descriptive statistics ISSP 2009 ESS 2008 EVS 2008 Variables Mean SD Min. Max. Mean SD Min. Max. Mean SD Min. Max. Inequality attitude (5-pt. scale) 2.12 1.09 1 5 2.19 1.05 1 5 2.76 1.30 1 5 Inequality attitude (dummy) – – – – – – – – 0.46 0.50 0 1 Ln(income) 7.47 0.85 5.39 9.93 7.14 0.75 4.26 9.18 7.07 0.87 3.39 9.90 Male 0.46 0.50 0 1 0.46 0.50 0 1 0.45 0.50 0 1 Female 0.54 0.50 0 1 0.54 0.50 0 1 0.55 0.50 0 1 Age/10 4.88 1.71 1.8 8.9 4.84 1.78 1.8 8.9 4.84 1.76 1.8 8.9 Lower secondary or less 0.29 0.45 0 1 0.32 0.47 0 1 0.34 0.47 0 1 Upper secondary 0.46 0.50 0 1 0.44 0.50 0 1 0.44 0.50 0 1 Tertiary 0.25 0.43 0 1 0.24 0.43 0 1 0.23 0.42 0 1 Full-time employed 0.46 0.50 0 1 0.45 0.50 0 1 0.47 0.50 0 1 Unemployed 0.06 0.23 0 1 0.05 0.22 0 1 0.06 0.24 0 1 Student 0.05 0.23 0 1 0.06 0.23 0 1 0.05 0.22 0 1 Retired 0.25 0.43 0 1 0.24 0.43 0 1 0.25 0.43 0 1 Housekeeping 0.06 0.24 0 1 0.10 0.30 0 1 0.08 0.27 0 1 Other 0.12 0.32 0 1 0.10 0.30 0 1 0.09 0.29 0 1 ISSP 2009 ESS 2008 EVS 2008 Variables Mean SD Min. Max. Mean SD Min. Max. Mean SD Min. Max. Inequality attitude (5-pt. scale) 2.12 1.09 1 5 2.19 1.05 1 5 2.76 1.30 1 5 Inequality attitude (dummy) – – – – – – – – 0.46 0.50 0 1 Ln(income) 7.47 0.85 5.39 9.93 7.14 0.75 4.26 9.18 7.07 0.87 3.39 9.90 Male 0.46 0.50 0 1 0.46 0.50 0 1 0.45 0.50 0 1 Female 0.54 0.50 0 1 0.54 0.50 0 1 0.55 0.50 0 1 Age/10 4.88 1.71 1.8 8.9 4.84 1.78 1.8 8.9 4.84 1.76 1.8 8.9 Lower secondary or less 0.29 0.45 0 1 0.32 0.47 0 1 0.34 0.47 0 1 Upper secondary 0.46 0.50 0 1 0.44 0.50 0 1 0.44 0.50 0 1 Tertiary 0.25 0.43 0 1 0.24 0.43 0 1 0.23 0.42 0 1 Full-time employed 0.46 0.50 0 1 0.45 0.50 0 1 0.47 0.50 0 1 Unemployed 0.06 0.23 0 1 0.05 0.22 0 1 0.06 0.24 0 1 Student 0.05 0.23 0 1 0.06 0.23 0 1 0.05 0.22 0 1 Retired 0.25 0.43 0 1 0.24 0.43 0 1 0.25 0.43 0 1 Housekeeping 0.06 0.24 0 1 0.10 0.30 0 1 0.08 0.27 0 1 Other 0.12 0.32 0 1 0.10 0.30 0 1 0.09 0.29 0 1 Sources: ISSP (2009); ESS (2008); EVS (2008). View Large Table A1. Descriptive statistics ISSP 2009 ESS 2008 EVS 2008 Variables Mean SD Min. Max. Mean SD Min. Max. Mean SD Min. Max. Inequality attitude (5-pt. scale) 2.12 1.09 1 5 2.19 1.05 1 5 2.76 1.30 1 5 Inequality attitude (dummy) – – – – – – – – 0.46 0.50 0 1 Ln(income) 7.47 0.85 5.39 9.93 7.14 0.75 4.26 9.18 7.07 0.87 3.39 9.90 Male 0.46 0.50 0 1 0.46 0.50 0 1 0.45 0.50 0 1 Female 0.54 0.50 0 1 0.54 0.50 0 1 0.55 0.50 0 1 Age/10 4.88 1.71 1.8 8.9 4.84 1.78 1.8 8.9 4.84 1.76 1.8 8.9 Lower secondary or less 0.29 0.45 0 1 0.32 0.47 0 1 0.34 0.47 0 1 Upper secondary 0.46 0.50 0 1 0.44 0.50 0 1 0.44 0.50 0 1 Tertiary 0.25 0.43 0 1 0.24 0.43 0 1 0.23 0.42 0 1 Full-time employed 0.46 0.50 0 1 0.45 0.50 0 1 0.47 0.50 0 1 Unemployed 0.06 0.23 0 1 0.05 0.22 0 1 0.06 0.24 0 1 Student 0.05 0.23 0 1 0.06 0.23 0 1 0.05 0.22 0 1 Retired 0.25 0.43 0 1 0.24 0.43 0 1 0.25 0.43 0 1 Housekeeping 0.06 0.24 0 1 0.10 0.30 0 1 0.08 0.27 0 1 Other 0.12 0.32 0 1 0.10 0.30 0 1 0.09 0.29 0 1 ISSP 2009 ESS 2008 EVS 2008 Variables Mean SD Min. Max. Mean SD Min. Max. Mean SD Min. Max. Inequality attitude (5-pt. scale) 2.12 1.09 1 5 2.19 1.05 1 5 2.76 1.30 1 5 Inequality attitude (dummy) – – – – – – – – 0.46 0.50 0 1 Ln(income) 7.47 0.85 5.39 9.93 7.14 0.75 4.26 9.18 7.07 0.87 3.39 9.90 Male 0.46 0.50 0 1 0.46 0.50 0 1 0.45 0.50 0 1 Female 0.54 0.50 0 1 0.54 0.50 0 1 0.55 0.50 0 1 Age/10 4.88 1.71 1.8 8.9 4.84 1.78 1.8 8.9 4.84 1.76 1.8 8.9 Lower secondary or less 0.29 0.45 0 1 0.32 0.47 0 1 0.34 0.47 0 1 Upper secondary 0.46 0.50 0 1 0.44 0.50 0 1 0.44 0.50 0 1 Tertiary 0.25 0.43 0 1 0.24 0.43 0 1 0.23 0.42 0 1 Full-time employed 0.46 0.50 0 1 0.45 0.50 0 1 0.47 0.50 0 1 Unemployed 0.06 0.23 0 1 0.05 0.22 0 1 0.06 0.24 0 1 Student 0.05 0.23 0 1 0.06 0.23 0 1 0.05 0.22 0 1 Retired 0.25 0.43 0 1 0.24 0.43 0 1 0.25 0.43 0 1 Housekeeping 0.06 0.24 0 1 0.10 0.30 0 1 0.08 0.27 0 1 Other 0.12 0.32 0 1 0.10 0.30 0 1 0.09 0.29 0 1 Sources: ISSP (2009); ESS (2008); EVS (2008). View Large Table A2. Pearson correlations between macro-level variables Variables 1a 1b 2a 2b 3 4 1a. Target efficiency (OECD) 1.00 1b. Target efficiency (WC) 0.87*** 1.00 2a. Size of public transfers (OECD) 0.37 0.47* 1.00 2b. Size of public transfers (WC) 0.57* 0.70** 0.52* 1.00 3. Gini (market income) 0.47* 0.21 0.39 0.32 1.00 4. GDP(PPP)/capita −0.63** −0.36 −0.14 −0.50* −0.47* 1.00 Variables 1a 1b 2a 2b 3 4 1a. Target efficiency (OECD) 1.00 1b. Target efficiency (WC) 0.87*** 1.00 2a. Size of public transfers (OECD) 0.37 0.47* 1.00 2b. Size of public transfers (WC) 0.57* 0.70** 0.52* 1.00 3. Gini (market income) 0.47* 0.21 0.39 0.32 1.00 4. GDP(PPP)/capita −0.63** −0.36 −0.14 −0.50* −0.47* 1.00 Note: ‘WC’ = Wang and Caminada (2011). * P < 0.05; **P < 0.01; ***P < 0.001. View Large Table A2. Pearson correlations between macro-level variables Variables 1a 1b 2a 2b 3 4 1a. Target efficiency (OECD) 1.00 1b. Target efficiency (WC) 0.87*** 1.00 2a. Size of public transfers (OECD) 0.37 0.47* 1.00 2b. Size of public transfers (WC) 0.57* 0.70** 0.52* 1.00 3. Gini (market income) 0.47* 0.21 0.39 0.32 1.00 4. GDP(PPP)/capita −0.63** −0.36 −0.14 −0.50* −0.47* 1.00 Variables 1a 1b 2a 2b 3 4 1a. Target efficiency (OECD) 1.00 1b. Target efficiency (WC) 0.87*** 1.00 2a. Size of public transfers (OECD) 0.37 0.47* 1.00 2b. Size of public transfers (WC) 0.57* 0.70** 0.52* 1.00 3. Gini (market income) 0.47* 0.21 0.39 0.32 1.00 4. GDP(PPP)/capita −0.63** −0.36 −0.14 −0.50* −0.47* 1.00 Note: ‘WC’ = Wang and Caminada (2011). * P < 0.05; **P < 0.01; ***P < 0.001. View Large Table A3. List of countries included in the analysis (Tables 3–4) OECD (2008) Wang and Caminada (2011) ISSP 09 ESS 08 EVS 08 ISSP 09 ESS 08 EVS 08 Countries M1 M3 M5/7 M2 M4 M6/8 Austria x x x x x x Belgium x x x x x x Czech Rep. x x x x x x Denmark x x x x x x Estonia n/a n/a n/a x x x Finland x x x x x x France x x x x x x Germany x x x x x x Great Britain x x x x x x Hungary x x x x x x Netherlands x x x x x x Norway x x x x x x Poland x x x x x x Portugal x x x n/a n/a n/a Slovak Rep. x x x x x x Slovenia n/a n/a n/a x x x Spain x x x x x x Sweden x x x x x x Switzerland x x x x x x Turkey x x x n/a n/a n/a OECD (2008) Wang and Caminada (2011) ISSP 09 ESS 08 EVS 08 ISSP 09 ESS 08 EVS 08 Countries M1 M3 M5/7 M2 M4 M6/8 Austria x x x x x x Belgium x x x x x x Czech Rep. x x x x x x Denmark x x x x x x Estonia n/a n/a n/a x x x Finland x x x x x x France x x x x x x Germany x x x x x x Great Britain x x x x x x Hungary x x x x x x Netherlands x x x x x x Norway x x x x x x Poland x x x x x x Portugal x x x n/a n/a n/a Slovak Rep. x x x x x x Slovenia n/a n/a n/a x x x Spain x x x x x x Sweden x x x x x x Switzerland x x x x x x Turkey x x x n/a n/a n/a Sources: ISSP (2009); ESS (2008); EVS (2008); OECD (2008); Wang and Caminada (2011). View Large Table A3. List of countries included in the analysis (Tables 3–4) OECD (2008) Wang and Caminada (2011) ISSP 09 ESS 08 EVS 08 ISSP 09 ESS 08 EVS 08 Countries M1 M3 M5/7 M2 M4 M6/8 Austria x x x x x x Belgium x x x x x x Czech Rep. x x x x x x Denmark x x x x x x Estonia n/a n/a n/a x x x Finland x x x x x x France x x x x x x Germany x x x x x x Great Britain x x x x x x Hungary x x x x x x Netherlands x x x x x x Norway x x x x x x Poland x x x x x x Portugal x x x n/a n/a n/a Slovak Rep. x x x x x x Slovenia n/a n/a n/a x x x Spain x x x x x x Sweden x x x x x x Switzerland x x x x x x Turkey x x x n/a n/a n/a OECD (2008) Wang and Caminada (2011) ISSP 09 ESS 08 EVS 08 ISSP 09 ESS 08 EVS 08 Countries M1 M3 M5/7 M2 M4 M6/8 Austria x x x x x x Belgium x x x x x x Czech Rep. x x x x x x Denmark x x x x x x Estonia n/a n/a n/a x x x Finland x x x x x x France x x x x x x Germany x x x x x x Great Britain x x x x x x Hungary x x x x x x Netherlands x x x x x x Norway x x x x x x Poland x x x x x x Portugal x x x n/a n/a n/a Slovak Rep. x x x x x x Slovenia n/a n/a n/a x x x Spain x x x x x x Sweden x x x x x x Switzerland x x x x x x Turkey x x x n/a n/a n/a Sources: ISSP (2009); ESS (2008); EVS (2008); OECD (2008); Wang and Caminada (2011). View Large © The Author 2017. Published by Oxford University Press and the Society for the Advancement of Socio-Economics. All rights reserved. For Permissions, please email: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Socio-Economic Review Oxford University Press

Identity priming and public opinion on income inequality: robustness testing of the micro-level mechanism of the paradox of redistribution

Socio-Economic Review , Volume 16 (3) – Jul 1, 2018

Loading next page...
 
/lp/ou_press/identity-priming-and-public-opinion-on-income-inequality-robustness-jZzVFYNJZU
Publisher
Oxford University Press
Copyright
© The Author 2017. Published by Oxford University Press and the Society for the Advancement of Socio-Economics. All rights reserved. For Permissions, please email: journals.permissions@oup.com
ISSN
1475-1461
eISSN
1475-147X
D.O.I.
10.1093/ser/mwx038
Publisher site
See Article on Publisher Site

Abstract

Abstract This study aims to contribute new insights into the ‘paradox of redistribution’ theory in light of identity priming. Korpi and Palme argued that low-income targeting leads to less redistribution and explained this trade-off as a result of coalition politics among different social strata. Indeed, recent empirical studies suggest that high-income earners tend to have more negative attitudes toward redistribution in the context of targeted spending (the polarization hypothesis). Contrary to most studies, which have relied on a single data source and measure, this study explores whether the polarization hypothesis is supported by multiple (both micro and macro) data sources. The findings of this study indicate that different question wordings produce different responses to inequality and redistribution. Some of the results confirmed the polarization hypothesis, whereas others did not. This article attempts to explain these discrepancies by comparing the wording patterns of inequality-attitude measures and arguing that high- and low-income earners polarize their views in the context of low-income targeting when their sense of ingroup favoritism and intergroup conflict is sufficiently triggered. 1. Introduction Several recent studies have focused on the role of target efficiency—that is, the extent to which public transfers are targeted to those at the bottom of the income scale—as a key cross-national institutional variation that explains the micro-level mechanism of redistribution politics (Brady and Bostic, 2015; Beramendi and Rehm, 2016; Sumino, 2016). The theoretical underpinnings for this insight come from the seminal work by Korpi and Palme (1998; ‘KP, 1998’ henceforth), in which they argued that the skewed distribution of welfare benefits among different income groups creates a cross-class polarization of interests and identities, hence less government redistribution—called the ‘paradox of redistribution’. Despite growing interest in the concentration of transfer benefits, relatively little attention has been paid to the role of identity/conflict awareness. Let me explain this point by turning to the original ‘paradox of redistribution’ theory (KP, 1998). The KP’s theory predicts that more targeted welfare systems result in less redistributive government: Targeted spending → redistribution. Here, ‘redistribution’ refers to the relative reduction in income inequality from market income to disposable (post-tax/transfer) income as a proportion of market income inequality (KP, 1998). The logic behind this process (targeted spending → redistribution) can be broken down into two components: the first in which given institutional structures (i.e. universal or targeted spending) mobilize coalitions (or conflicts) between different income groups, and the second in which formulated preferences are reflected in an actual redistributive policy. This micro-level mechanism can be expressed as: Targeted spending (a) → preference formation (b) → redistribution (c). Focusing on the former stage [(a) → (b)], this study hypothesizes that the relationship between income status and attitudes toward income inequality is conditioned by the degree of targeted spending. In other words, attitudinal cleavages between income groups grow wider in welfare states with low-income targeting, in which redistribution politics tends to be a group-interested and conflict-driven process. An empirical study by Beramendi and Rehm (2016) indicates that a more low-income targeting leads to a more contested welfare state (i.e. more negative income slopes). This study further specifies the above argument in light of identity awareness, arguing that the polarization process (Targeted spending → preference formation) occurs when individuals are exposed to a cue that elicits their sense of social identity (Tajfel, 1981) and self-categorization (Turner, 1985; Turner et al., 1987). In other words, targeted spending mobilizes coalitions (or conflicts) among income groups when members of high/low income groups articulate a sense of identity and conflict. This modifies the causal process as follows: Targeted spending (a) → identity priming (b1) → polarization in attitudes (b2) → redistribution (c), in which preference formation (b) is divided into (b1) and (b2). By ‘identity priming’, I mean an increased sensitivity to reducing inequality (e.g. redistributive policy) triggered by exposure to stimuli that evoke a sense of ingroup favoritism and intergroup conflict. In practice, this study argues that the polarization process [(a) through (b2)] occurs when a survey question accentuates the contrast/conflict between high- and low-income earners. To this end, unlike most previous studies that have replied on a single data source (e.g. ISSP) and single question item, this study attempts to provide a more comprehensive analysis using multiple data sources: the International Social Survey Program (ISSP, 2009), the European Social Survey (ESS, 2008) and the European Values Survey (EVS, 2008). From these surveys, I select a set of question items (inequality attitudes) with different question wordings, some of which emphasize the contrast/conflict between high- and low-income groups but others do not. The polarization process is expected to occur when respondents are exposed to a survey question that is designed to prime their sense of group identity/conflict. The findings of this study indicate that differences in the question wording, even slight ones, can yield different results, and that the validity of the hypothesis that targeted spending contextualizes the association between income and inequality attitudes depends on how the outcome variable (inequality attitude) is asked or worded/phrased. 2. Theoretical framework and hypotheses In this section, I briefly outline the ‘paradox of redistribution’ theory developed by KP (1998), and then identify the specific process of attitudinal polarization in the context of targeted spending. I then highlight the role of identity priming as a potentially important activator of ingroup favoritism and intergroup conflict that make the polarization process function. 2.1 Micro-level foundation of the paradox of redistribution The welfare regime theory argues that the differences in social policy arrangements generate distinct forms of social structuring, political discourse and attitudes (Korpi, 1980; Esping-Andersen, 1990; KP, 1998; Pierson, 2001; Rothstein and Steinmo, 2002; Larsen, 2008). The central idea is that institutional structures shape a collective understanding of redistributive politics—namely, whether and to what extent the well-being of the least well-off in society should be secured through statutory intervention or market-based solutions (Offe, 1992; Edlund, 1999). The political economy literature, on the other hand, highlights the conflict of material interest between income groups as the key motive for individualist behavior (Romer, 1975; Roberts, 1977; Meltzer and Richard, 1981). In this view, individuals are assumed to behave in a way that maximizes their personal utility. For instance, low-income earners are more likely to support redistribution than high-income earners because it benefits them financially more than it does in their high-income counterparts (Luttmer, 2001; Alesina and Ferrara, 2005; Rehm, 2009; Sumino, 2014). Synthesizing these two distinct theoretical traditions (i.e. institutionalist and political economy perspectives) prompts the following question: do different income groups respond uniformly to inequality across countries with different institutional structures? Political economists claim that income cleavages lead to differences in how people feel about inequality. But do high-income earners under a certain institutional context oppose egalitarian measures in the same manner as those under a different institutional context? If the answer is no, what might account for the differences? Revisiting the ‘paradox of redistribution’ theory proposed by KP (1998) gives us a good starting point. They argued that target efficiency—‘the proportion of program expenditures going exclusively to those below the official poverty line’ (p. 662)—plays a crucial role in the process of welfare state politics. They theorized that there is a reverse relationship between targeting and redistribution, namely that a welfare state institution based on means-tested programs targeted exclusively at the needy results in less redistributive policy outputs. According to them, this trade-off is explained as a product of socio-political dynamics. Marginal welfare states with targeted income transfers, in which social programs ‘unfairly’ benefit a small portion of the population (i.e. those below the poverty line), produce a preference gap between high/middle-income and low-income groups, and this, in turn, causes a potential revolt against a redistributive political agenda. On the other hand, welfare institutions characterized by universal programs, which benefit a broad range of people regardless of recipients’ market positions, mobilize cross-class alliances. Unlike means-tested transfer schemes, the comprehensive welfare state incorporates all citizens without accentuating disparities between net contributors and net beneficiaries (Pierson, 1996). Because social welfare programs in the universal welfare state are generously designed to cover the entire population instead of labelling or stigmatizing the target group, the expected costs and benefits of the welfare state are less visible or calculable to both manual and better-off workers (Skocpol, 1992; Rothstein, 1998). In other words, the universal welfare state blurs the boundaries between group identities and generates a seamless and integrated public demand for redistribution. In contrast, in selective welfare states, identities are constructed along a constellation of class interests, which lead to group-based conflicts between ‘us’ and ‘them’ (Edlund, 1999). Residual welfare systems explicitly divide the population into welfare recipients and non-recipients, and hence, tend to be associated with greater polarization in preferences. This is not to say that institutional structures (i.e. target efficiency) directly explain welfare stakeholders’ perceptions toward income inequality, but rather that the relationship between income and outcome attitudes is conditioned—strengthened or weakened—by the structure of social policy institutions, in particular, by the degree of targeting policy (Gingrich and Ansell, 2012). The thrust of the above argument is two-fold: (1) reactions to redistributive policies differ across income groups—while the poor demand redistribution, the rich oppose it, and (2) such reactions of different income groups are not uniform across welfare sates with different degrees of targeted spending. These theoretical propositions lead us to the following two hypotheses: Hypothesis 1. Higher-income earners are less supportive of reducing inequality. In other words, there exist attitudinal cleavages between income groups. Hypothesis 2. Attitudinal cleavages between income groups grow wider in the context of targeted spending. High-income earners’ opposition to reducing inequality becomes more intense in welfare states where transfer programs are ‘unequally’ targeted at low-income households. This study does not intend to test the whole process of the ‘paradox of redistribution’ theory developed by KP (1998), in which they noted that, (1)the institutional structures of welfare states are likely to affect the definitions of identity and interest among citizens. Thus, (2)an institutional welfare state model based on a universalistic strategy intended to maintain normal or accustomed standards of living is likely to result in greater redistribution than a marginal one based on targeting (p. 663; underlines added). The underlying logic of these statements can be summarized as follows: first, institutional structures shape identity/interest-based preferences toward inequality, and second, such formulated preferences are translated to redistributive policy outputs, most likely through political representation. The first stage concerns the role of institutional structures in triggering broad alliances or antagonisms among income groups, while the second stage refers, though not explicitly, to the impact of public opinion on redistributive policy. Thus, the whole causal process is expressed as: Targeted spending (a) → preference formation (b) → redistribution (c). The interest of this article is to explore the former part—the causal sequence from ‘targeted spending’ (a) to ‘preference formation’ (b). In this respect, while the present study draws upon the ‘paradox of redistribution’ theory, the main focus is on the micro-level mechanism behind the formation of preferences rather than on the causality from ‘preference’ (b) to ‘redistribution’ (c) or the aggregate-level relationship between ‘targeted spending’ (a) and ‘redistribution’ (c) (e.g. Kenworthy, 2011; Marx et al., 2013). 2.2 Identity priming in redistribution politics Central to understanding the mechanism underlying the polarization process (Targeted spending (a) → preference formation (b); Hypothesis 2) is the role of identity awareness. The issue of who, or which income group, gets transfer benefits and who bears the costs is politically sensitive, and attracts scrutiny from welfare stakeholders. The uneven distribution of public transfers, which is closely linked to subjective economic well-being, affects stakeholders’ reactions to the reallocation of income resources between ‘our group’ and ‘the others’. In the context of low-income targeting, high-income earners are expected to be more resistant to reducing inequality and redistribution because targeting, they perceive, ‘unfairly’ benefits ‘the others’ (i.e. low-income groups) at the expense of ‘our group’. This leads to the prediction that attitudinal differences between high- and low-income earners grow wider in countries with targeting policies. Redistributive preferences cannot be explained only by individual calculations of rational self-interest but are largely a function of social identity (Tajfel, 1981) and self-categorization (Turner, 1985; Turner et al., 1987). Individuals’ perceptions that they belong to a socio-economic group and their emotional attachment to the group are a powerful regulator of inequality attitudes. The process of self-categorization, in which individuals place themselves into a group that shares similar concerns and interests, draws boundaries between ‘us’ and ‘them’, and individuals ascribe positive attributes to their own group and negative traits and prejudice to the others. The self-categorization process helps ‘atomized’ individuals recognize their social self and elicits intragroup cohesion and intergroup conflict. Group polarization is particularly salient in the context of low-income targeting characterized by the reallocation of resources from high- to low-income groups. This implies that the polarization process occurs when individuals’ awareness of identities is stimulated, and when ingroup favoritism is triggered. In other words, attitudinal polarization (in the context of low-income targeting) is activated by an exposure to a stimulus that provokes a sense of belongingness and conflict. Hypothesis 2 predicts that targeted spending contextualizes the relationship between income position and attitudes toward inequality. Here, ‘attitudes toward inequality’ can be operationalized in a variety of ways (question wordings): some survey questions emphasize the contrast between those who benefit and those who contribute, but others do not. The polarization process (Hypothesis 2) is expected to occur when respondents are exposed to a question that is designed to prime their identification with a specific socioeconomic group (high/low-income status). Thus, the causal process described in the sub-section 2.1 will be modified as follows: Targeted spending (a) → identity priming (b1) → polarization in attitudes (b2) → redistribution (c) ‘Preference formation’ (b) can be divided into two sub-processes: identity priming (b1) and polarization in attitudes (b2). This causal sequence suggests that if respondents were not exposed to a cue that stimulates their sense of identification (b1), the process of attitudinal polarization (b2) would be less likely to happen. In the empirical analysis that follows, I examine how the validity of Hypothesis 2 is susceptible to the question wording of inequality-attitude measures. 3. Research methods 3.1 Data This study is based on individual-level data (aged 18–89 years) from three different sources: the International Social Survey Program (ISSP, 2009), the European Social Survey (ESS, 2008) and the European Values Survey (EVS, 2008). The three surveys include questions about attitudes toward inequality and redistribution, and cover (share) the same set of countries and time period (late-2000s). In this study, the analysis was restricted to capitalist democracies on which the KP theory is originally based, and macro-level data on target efficiency (OECD, 2008; Wang and Caminada, 2011) are fully available. To make the comparisons easier, this study used the same set of overlapping countries commonly covered by the three surveys. The analysis includes 20 countries: Austria, Belgium, the Czech Republic, Denmark, Estonia, Finland, France, Germany, Great Britain, Hungary, the Netherlands, Norway, Poland, Portugal, the Slovak Republic, Slovenia, Spain, Sweden, Switzerland and Turkey. The three datasets were analyzed in combination with macro-level data (OECD, 2008; Wang and Caminada, 2011). Due to data constraints, Portugal and Turkey were included only in the models estimated using the OECD data, while Estonia and Slovenia were included only in the models estimated using the LIS (Wang and Caminada, 2011 ) data (for details, see Appendix A, Table A3). Missing data were handled using multiple imputation (MI) techniques because the use of list-wise deletion results in loss of ∼20% of the cases—mainly due to missing information on the (country-specific) household income variables—thus, may lead to biased and inefficient estimates (King et al., 2001). Imputation was carried out with Amelia II for creating five (m = 5) complete case datasets (Honaker et al., 2011). Missing values were imputed separately within each country to account for the clustered data structure (Graham, 2009). Estimates from the five imputed datasets were pooled using Rubin’s rules (Rubin, 1987; 1996; White et al., 2011). 3.2 Measures The central aim of this study is to examine how question wordings affect the polarization process in the context of low-income targeting (Hypothesis 2). In essence, this study predicts that the polarization process is most likely to occur when survey questions elicit respondents’ identification and self-categorization. To test this, I employ and compare a set of four different indicators (with different question wordings) of inequality attitudes (Table 1). The first two items are standard ‘government redistribution’ questions from the ISSP and the ESS. The ISSP asked the respondents whether they agree or disagree with the statement, ‘it is the responsibility of the government to reduce the differences in income between people with high incomes and those with low incomes’, while the ESS has a similar but slightly different question item that asked whether or not respondents agree with the notion that ‘the government should take measures to reduce differences in income levels’. The ISSP question explicitly refers to the contrast between ‘people with high incomes and those with low incomes’, which is expected to invoke respondents’ awareness of group identity—e.g. high-income status (contributor) or low-income status (recipient). On the other hand, the ESS item is a less conflict-prone question in that it is vaguely worded as ‘differences in income levels’, with no reference to ‘high versus low’ income earners. The third item was taken from the EVS, which asked respondents for their views on the statement: ‘incomes should be made equal’ or ‘there should be greater incentives for individual effort’. Similar to the first and second items, this question asks about income inequality, but with a more explicit reference to the conflict between income equality and incentives on effort. The fourth inequality-attitude item also comes from the EVS, in which respondents were asked to choose which of the following statements came closer to their own view: ‘I would consider equality more important, that is, that nobody is underprivileged and that social class differences are not so strong’, or ‘I would consider personal freedom more important, that is, everyone can live in freedom and develop without hindrance’. This item is similar to the third one (i.e. equality versus freedom), but different in that the two opposite values were worded in a more abstract way: ‘equality’ rather than ‘income equality’ and ‘personal freedom’ rather than ‘incentives for effort’. In this sense, the fourth question is expected to send less explicit signals to respondents about the possibility that reducing income inequality might result in disincentives for effort and performance. Table 1. Classification of inequality-attitude items by question-wording patterns Identity/conflict priming words/phrases (a) Income inequality/ differences (b) High income versus low income (c) Incentives on effort Q1: Responsibility of the government to reduce the differences in income between people with high incomes and those with low incomes. (ISSP 2009) Yes Yes No Q2: Government should take measures to reduce differences in income levels. (ESS 2008) Yes No No Q3: Incomes should be made more equal. OR There should be greater incentives for individual effort. (EVS 2008) Yes No Yes Q4: I would consider equality more important, that is, that nobody is underprivileged and that social class differences are not so strong. OR I would consider personal freedom more important, that is, everyone can live in freedom and develop without hindrance. (EVS 2008) No No No Identity/conflict priming words/phrases (a) Income inequality/ differences (b) High income versus low income (c) Incentives on effort Q1: Responsibility of the government to reduce the differences in income between people with high incomes and those with low incomes. (ISSP 2009) Yes Yes No Q2: Government should take measures to reduce differences in income levels. (ESS 2008) Yes No No Q3: Incomes should be made more equal. OR There should be greater incentives for individual effort. (EVS 2008) Yes No Yes Q4: I would consider equality more important, that is, that nobody is underprivileged and that social class differences are not so strong. OR I would consider personal freedom more important, that is, everyone can live in freedom and develop without hindrance. (EVS 2008) No No No Sources: ISSP (2009); ESS (2008); EVS (2008). Table 1. Classification of inequality-attitude items by question-wording patterns Identity/conflict priming words/phrases (a) Income inequality/ differences (b) High income versus low income (c) Incentives on effort Q1: Responsibility of the government to reduce the differences in income between people with high incomes and those with low incomes. (ISSP 2009) Yes Yes No Q2: Government should take measures to reduce differences in income levels. (ESS 2008) Yes No No Q3: Incomes should be made more equal. OR There should be greater incentives for individual effort. (EVS 2008) Yes No Yes Q4: I would consider equality more important, that is, that nobody is underprivileged and that social class differences are not so strong. OR I would consider personal freedom more important, that is, everyone can live in freedom and develop without hindrance. (EVS 2008) No No No Identity/conflict priming words/phrases (a) Income inequality/ differences (b) High income versus low income (c) Incentives on effort Q1: Responsibility of the government to reduce the differences in income between people with high incomes and those with low incomes. (ISSP 2009) Yes Yes No Q2: Government should take measures to reduce differences in income levels. (ESS 2008) Yes No No Q3: Incomes should be made more equal. OR There should be greater incentives for individual effort. (EVS 2008) Yes No Yes Q4: I would consider equality more important, that is, that nobody is underprivileged and that social class differences are not so strong. OR I would consider personal freedom more important, that is, everyone can live in freedom and develop without hindrance. (EVS 2008) No No No Sources: ISSP (2009); ESS (2008); EVS (2008). Table 1 summarizes the wording patterns of the four inequality-attitude items. The items were classified by whether or not they include specific words/phrases that trigger a sense of ingroup favoritism and intergroup conflict—namely, explicit references to (a) income inequality, (b) high versus low incomes and (c) incentives on effort. Unlike the phrases such as ‘social class differences’ and ‘social inequality’, ‘income inequality/differences’ (a) is specific since it identifies a particular dimension of socio-economic disparities (i.e. income), which is expected to invoke a sense of identity as a high/low-income earner. This specification is especially important because this analysis focuses on attitudinal cleavages across income lines (Hypotheses 1–2). ‘High- versus low-income status’ (b) is also a potential trigger of identification and self-categorization. The phrase ‘high/low incomes’ elicits a sense of identity/conflict by making respondents aware, for example, that redistributive policy benefits low-income group but burdens high-income group. This would be less likely to happen if the question was asked in a more abstract fashion—for example, ‘differences in income levels’ (ESS). The phrase ‘incentives on effort’ (c) reminds high-income earners that they are economically successful because they worked hard for it, which leads to the justification of privileged status and existing income inequality. This process instills a sense of identity as a member of high-income group and triggers ingroup favoritism and outgroup prejudice (e.g. lazy low-income earners). As presented in Table 1, Questions 1–2 (ISSP/ESS) share all features except ‘high versus low incomes’. Question 3 (EVS) is similar to Questions 1–2 because it addresses income inequality, but differs because it addresses the potential trade-off between income equality and incentives on effort. Question 4 also deals with the conflict between equality and freedom but in a more abstract way: ‘equality’ not ‘income equality’; and ‘personal freedom’ not ‘incentives for individual effort’. As presented in Figure 1, respondents’ reactions to the four questions vary markedly both within and between countries. Figure 1. View largeDownload slide Attitudes toward inequality, %. Notes: Countries were sorted in descending order by the sum of ‘strongly agree’ and ‘agree’ of the ISSP item (top left). To facilitate comparison, countries were presented in the same order for all figures. N = 20. Sources: ISSP (2009); ESS (2008); EVS (2008). Figure 1. View largeDownload slide Attitudes toward inequality, %. Notes: Countries were sorted in descending order by the sum of ‘strongly agree’ and ‘agree’ of the ISSP item (top left). To facilitate comparison, countries were presented in the same order for all figures. N = 20. Sources: ISSP (2009); ESS (2008); EVS (2008). The primary independent variable of interest is household income. In the three surveys (ISSP, ESS, EVS), household income was reported in categories. The midpoint of each income interval was used as a proxy for actual income (Rueda et al., 2014; Rueda, 2014; Rueda and Stegmueller, 2015). The midpoint of the last open-ended category was extrapolated from the frequencies of the next-to-top and the top categories using a Pareto formula proposed by Hout (2004). To take into account the differences in price levels across countries, income values were adjusted by purchasing power parities (PPPs) in national currencies per US dollar (World Bank, 2008). To adjust for the differences in household compositions, household incomes were divided by the square root of household size (Buhmann et al., 1988). To reduce skewness, income values were transformed with a natural logarithm prior to analysis. The analysis includes the following demographic controls: gender (female =1, male = 0), age (in years/10), age2, educational attainment (primary, secondary, tertiary) and employment status (full-time employed, unemployed, student, retired, housekeeping, others). Although one may argue that party affiliation (or left-right ideological disposition) has a significant association with inequality attitudes (e.g. left partisanship leads to higher levels of support for egalitarian principles and practices), I prefer not to include this item in this analysis because of potential endogeneity bias problems (i.e. the possibility of reverse causation; see Cusack et al., 2006; but see Jæger, 2008; Margalit, 2013). Target efficiency. To evaluate the distribution of public transfers across income groups, I use the concentration coefficient for public cash benefits. The concentration index, calculated in the same way as the Gini index, measures the extent to which public transfers are concentrated or dispersed across different income groups (Fields, 1979; Kakwani, 1986). Theoretically, the index lies between −1.0 (the poorest gets all public transfers) and +1.0 (the richest gets all public transfers). A score of zero indicates that public cash benefits are evenly distributed across the income scale. Negative and lower scores represent higher levels of progressivity—public cash benefits are targeted efficiently (or ‘unevenly’) to reduce income inequality. For robustness purposes, this study uses two different sources of aggregate-level data on targeted spending: OECD (2008); Wang and Caminada (2011). Here, Wang and Caminada (2011) is an updated version of the concentration coefficient calculated by the Luxembourg Income Study (LIS), the data source which KP’s (1998) work was originally based on. As Figure 2 shows, OECD (2008) statistics indicate that the concentration coefficient varies greatly across countries, ranging from −0.32 in Denmark to 0.35 in Turkey (mean = −0.04; SD = 0.19; N = 18). Denmark, Great Britain and Finland have the highest levels of target efficiency with values < −0.20, while public transfers are highly regressive in Turkey and Portugal with values > 0.20. In Germany and Hungary, public cash benefits are (relatively) equally distributed across incomes. According to Wang and Caminada (2011), the concentration index ranges from −0.31 in Great Britain to 0.16 in Poland (mean = −0.08; SD = 0.14; N = 18). The target efficiency scores are highly negative in Great Britain and Denmark; close to zero in Slovenia and Hungary; positive and high in Poland and Austria. Both measures are largely consistent with each other (Pearson’s r = 0.87; N = 16). The variation in values is larger in the OECD data (SD = 0.19) than in the LIS data (SD = 0.14), partly due to the inclusion of Turkey and Portugal, two highly regressive countries, in the OECD data. Figure 2. View largeDownload slide Target efficiency across 20 countries. Note: Countries were sorted in descending order, with the same countries listed side by side to facilitate comparison. Sources: OECD (2008); Wang and Caminada (2011). Figure 2. View largeDownload slide Target efficiency across 20 countries. Note: Countries were sorted in descending order, with the same countries listed side by side to facilitate comparison. Sources: OECD (2008); Wang and Caminada (2011). In addition to the concentration index, I also include the size of public transfers, defined as the percentage share of public cash transfers in household income (OECD, 2008; Wang and Caminada, 2011). Higher score represents greater proportion of redistributive benefits in household income. For instance, public transfers constitute about 35% of household income in Hungary and Poland, while they account for <20% in Great Britain and Switzerland (Table 2). The size of public transfers is clearly distinguished from the concentration index, which is about who among different income groups receives cash benefits and not how much (KP, 1998; Barnes, 2015). For example, as shown in Table 2, although public transfers account for >25% (OECD) of household income in both Denmark and Portugal, the degree of low-income targeting is very strong in Denmark (−0.316), but rather weak in Portugal (0.247). Table 2. Size of public transfers and target efficiency, by welfare regime type Size of benefits Target efficiency Size of benefits Target efficiency Regimes/ countries OECD WC OECD WC Regimes/ countries OECD WC OECD WC Social democratic Liberal Austria 36.6 26.7 0.157 0.108 Great Britain 14.5 14.3 −0.275 −0.313 Sweden 32.7 24.6 −0.145 −0.128 Non-classified Belgium 30.5 7.9 −0.120 −0.244 Poland 35.8 32.5 0.185 0.157 Denmark 25.6 18.9 −0.316 −0.306 Hungary 35.1 35.7 −0.016 0.016 Norway 21.7 20.2 −0.183 −0.155 Slovak Republic 26.0 26.6 −0.056 −0.109 Netherlands 17.1 21.3 −0.198 −0.041 Portugal 25.5 – 0.247 – Conservative Czech Republic 24.3 20.8 −0.154 −0.218 France 32.9 26.2 0.136 0.077 Spain 21.3 20.7 0.063 0.068 Germany 28.2 21.2 0.013 −0.110 Turkey 16.9 – 0.347 – Switzerland 16.0 17.5 −0.170 −0.066 Slovenia – 27.5 – 0.011 Finland 14.4 23.2 −0.219 −0.127 Estonia – 17.9 – −0.099 Size of benefits Target efficiency Size of benefits Target efficiency Regimes/ countries OECD WC OECD WC Regimes/ countries OECD WC OECD WC Social democratic Liberal Austria 36.6 26.7 0.157 0.108 Great Britain 14.5 14.3 −0.275 −0.313 Sweden 32.7 24.6 −0.145 −0.128 Non-classified Belgium 30.5 7.9 −0.120 −0.244 Poland 35.8 32.5 0.185 0.157 Denmark 25.6 18.9 −0.316 −0.306 Hungary 35.1 35.7 −0.016 0.016 Norway 21.7 20.2 −0.183 −0.155 Slovak Republic 26.0 26.6 −0.056 −0.109 Netherlands 17.1 21.3 −0.198 −0.041 Portugal 25.5 – 0.247 – Conservative Czech Republic 24.3 20.8 −0.154 −0.218 France 32.9 26.2 0.136 0.077 Spain 21.3 20.7 0.063 0.068 Germany 28.2 21.2 0.013 −0.110 Turkey 16.9 – 0.347 – Switzerland 16.0 17.5 −0.170 −0.066 Slovenia – 27.5 – 0.011 Finland 14.4 23.2 −0.219 −0.127 Estonia – 17.9 – −0.099 Notes: Regime classification is based on Esping-Andersen (1990, pp. 51–52). Countries are sorted in descending order by the size of transfer benefits (OECD, 2008) within each welfare regime cluster. ‘WC’ = Wang and Caminada (2011). Sources: OECD (2008); Wang and Caminada (2011). Table 2. Size of public transfers and target efficiency, by welfare regime type Size of benefits Target efficiency Size of benefits Target efficiency Regimes/ countries OECD WC OECD WC Regimes/ countries OECD WC OECD WC Social democratic Liberal Austria 36.6 26.7 0.157 0.108 Great Britain 14.5 14.3 −0.275 −0.313 Sweden 32.7 24.6 −0.145 −0.128 Non-classified Belgium 30.5 7.9 −0.120 −0.244 Poland 35.8 32.5 0.185 0.157 Denmark 25.6 18.9 −0.316 −0.306 Hungary 35.1 35.7 −0.016 0.016 Norway 21.7 20.2 −0.183 −0.155 Slovak Republic 26.0 26.6 −0.056 −0.109 Netherlands 17.1 21.3 −0.198 −0.041 Portugal 25.5 – 0.247 – Conservative Czech Republic 24.3 20.8 −0.154 −0.218 France 32.9 26.2 0.136 0.077 Spain 21.3 20.7 0.063 0.068 Germany 28.2 21.2 0.013 −0.110 Turkey 16.9 – 0.347 – Switzerland 16.0 17.5 −0.170 −0.066 Slovenia – 27.5 – 0.011 Finland 14.4 23.2 −0.219 −0.127 Estonia – 17.9 – −0.099 Size of benefits Target efficiency Size of benefits Target efficiency Regimes/ countries OECD WC OECD WC Regimes/ countries OECD WC OECD WC Social democratic Liberal Austria 36.6 26.7 0.157 0.108 Great Britain 14.5 14.3 −0.275 −0.313 Sweden 32.7 24.6 −0.145 −0.128 Non-classified Belgium 30.5 7.9 −0.120 −0.244 Poland 35.8 32.5 0.185 0.157 Denmark 25.6 18.9 −0.316 −0.306 Hungary 35.1 35.7 −0.016 0.016 Norway 21.7 20.2 −0.183 −0.155 Slovak Republic 26.0 26.6 −0.056 −0.109 Netherlands 17.1 21.3 −0.198 −0.041 Portugal 25.5 – 0.247 – Conservative Czech Republic 24.3 20.8 −0.154 −0.218 France 32.9 26.2 0.136 0.077 Spain 21.3 20.7 0.063 0.068 Germany 28.2 21.2 0.013 −0.110 Turkey 16.9 – 0.347 – Switzerland 16.0 17.5 −0.170 −0.066 Slovenia – 27.5 – 0.011 Finland 14.4 23.2 −0.219 −0.127 Estonia – 17.9 – −0.099 Notes: Regime classification is based on Esping-Andersen (1990, pp. 51–52). Countries are sorted in descending order by the size of transfer benefits (OECD, 2008) within each welfare regime cluster. ‘WC’ = Wang and Caminada (2011). Sources: OECD (2008); Wang and Caminada (2011). Using the quantitative regime indicators—the concentration and size of public cash benefits—has at least two advantages over some other operationalizations of welfare regime. First, it seems obvious that cross-national variation in target efficiency cannot be captured within the framework of traditional regime typology. As shown in Table 2, while the regime classification roughly reflects the variation in the size of public transfers—the scores are comparatively high in the social democratic cluster and low in its liberal counterpart—the scores of target efficiency do not have clear between-regime differences or within-regime similarities. In previous empirical studies, capitalist democracies have often been clustered into three (or more) distinct regime groups. However, ideal types of welfare regimes (Esping-Andersen, 1990) cannot be interpreted as ‘real entities’; groups of (or individual) countries cannot necessarily be treated as effective representatives of different welfare regimes (Svallfors, 1997; Jæger, 2006). The notion that welfare states are qualitatively distinct relies on the implicit assumption that each regime cluster has internal homogeneity. If this assumption does not hold (as in the case of target efficiency), potential within-regime variations are to be ignored. In this respect, employing quantitative measures allows for carefully capturing institutional differences across individual countries. Second, the concentration and size of cash benefits seem to affect people’s preferences more straightforwardly than some other quantitative indicators. An indicator such as the size of social spending as a percentage of GDP appears less appropriate because welfare stakeholders might either not be aware of the level of government spending in their country, or it might not be evident to them as to what the ‘percentage of national GDP’ implies for the expected welfare benefits their households actually receive. In contrast, for welfare stakeholders, the proportion of public cash benefits in their household income (i.e. the size of public transfers) and the extent to which transfer benefits are distributed to their own income group relative to other ‘rival’ groups (i.e. the concentration of public transfers)—are more tangible and easier to calculate and thus to respond to. Finally, to hold constant the effects of other macro-level factors, the following two variables were included in the analysis: market income inequality and economic development. Market income inequality was assessed by the Gini coefficient before taxes and transfers (OECD, 2005; Wang and Caminada, 2011), while economic development was assessed by gross domestic product (GDP) per capita adjusted for PPP in US dollars (IMF, 2005). In particular, controlling for ‘market income inequality’ adds strength to the causal inferences because the degree of targeted spending at the bottom is expected to be greater in countries where the distribution of market income is highly concentrated at the top. All macro-level factors were standardized by subtracting their means and dividing by their standard deviations. For the four macro-level indicators, I used the values from a single year around 2005 (mid-2000s). Descriptive statistics are shown in Appendix Table A1. The bivariate correlations between macro-level measures are presented in appendix Table A2. 3.3 Analytical approach A cross-sectional multilevel modelling approach was applied to analyze hierarchically nested samples and simultaneously examine the interplay between macro-level factors (e.g. targeting) and short-term individual-level preferences (Goldstein, 1995; Kreft and de Leeuw, 1998; Snijders and Bosker, 1999). Another possible approach might be to analyze over-time trends in targeting policies and subsequent preferences within countries. Nevertheless, a multilevel modelling approach seems more justifiable given that the key motivation of this article is to examine how pre-existing regime differences across countries would affect the association between income position and outcome preferences. Moreover, sufficient multi-year data on target efficiency are currently unavailable to carry out a longitudinal assessment. A series of two-level ordered logit models was estimated using MLwiN version 2.30 (Rasbash et al., 2014) via the ‘runmlwin’ command in Stata (Leckie and Charlton, 2012). 4. Results The purpose of this study is to test how different indicators of inequality attitudes affect the micro-level process of the ‘paradox of redistribution’ (Hypothesis 2) in light of the theoretical assumption that attitudinal polarization is induced when respondents place themselves into a particular income group, and articulate a sense of ingroup favoritism and intergroup conflict. To explore this, I begin by comparing the results obtained from the ISSP and the ESS (Table 3). Estimated parameters in the left two columns are based on the ISSP 2009, while those in the right two columns are from the ESS 2008. As suggested in Hypothesis 1, higher income is indeed significantly associated with weaker support for redistribution (models 1–4). Model 1 shows that the coefficient of the interaction term between income and targeted spending (OECD) was positive and highly significant (P < 0.01), which lends support for Hypothesis 2 that low-income targeting contextualizes the relationship between income and inequality attitudes (the more targeted spending, the more attitudinal polarization). Model 2 shows that the interaction effect remains significant (but less marked; P < 0.10) when the same model was tested using the targeted-spending measure calculated by Wang and Caminada (2011). In models 3–4, Hypothesis 2 was re-examined using a slightly different inequality-attitude item that does not accentuate the contrast between high- and low-income earners (ESS). The results show that the interaction effect was in the expected direction but failed to reach statistical significance, regardless of which targeted-spending measure (OECD or WC) was used. Table 3. Government should reduce income differences: comparison of ISSP 2009 and ESS 2008 Q1. Govt. resp. to reduce income differences b/w high and low incomes (ISSP) Q2. Govt. should take measures to reduce differences in income levels (ESS) Model 1: OECD Model 2: WC Model 3: OECD Model 4: WC Variables coef. s.e. coef. s.e. coef. s.e. coef. s.e. Individual-level predictors Household income (equivalized) −0.502*** (0.076) −0.544*** (0.078) −0.395*** (0.064) −0.442*** (0.059) Level-1 controls Yes Yes Yes Yes Country-level factors Target efficiency (OECD) 0.246* (0.108) – – 0.288* (0.136) – – Size of public transfers (OECD) 0.009 (0.085) – – −0.071 (0.109) – – Target efficiency (WC) – – 0.199† (0.119) – – 0.221 (0.136) Size of public transfers (WC) – – 0.030 (0.121) – – 0.033 (0.139) Gini (market income) 0.151† (0.089) 0.133 (0.093) 0.180 (0.109) 0.176 (0.104) GDP(PPP)/capita −0.196† (0.106) −0.269* (0.120) −0.043 (0.128) −0.041 (0.133) Cross-level interactions Income × Target efficiency (OECD) 0.324** (0.102) – – 0.124 (0.088) – – Income × Size of public transfers (OECD) −0.226** (0.080) – – −0.082 (0.070) – – Income × Target efficiency (WC) – – 0.179† (0.103) – – 0.016 (0.079) Income × Size of public transfers (WC) – – −0.182† (0.104) – – 0.026 (0.083) Income × Gini (market income) 0.053 (0.083) 0.030 (0.080) 0.069 (0.070) 0.067 (0.062) Income × GDP(PPP)/capita 0.138 (0.099) 0.033 (0.102) 0.027 (0.084) 0.069 (0.081) Intercept variance 0.102 (0.035) 0.125 (0.043) 0.164 (0.055) 0.167 (0.056) Slope variance (income) 0.082 (0.031) 0.082 (0.031) 0.060 (0.022) 0.050 (0.019) Covariance b/w intercepts and slopes 0.009 (0.023) 0.028 (0.027) 0.033 (0.026) 0.040 (0.025) VPC (variance partition coefficient) 0.030 0.037 0.047 0.048 Nindividual 23,050 22,562 34,961 33,236 Ncountry 18 18 18 18 Q1. Govt. resp. to reduce income differences b/w high and low incomes (ISSP) Q2. Govt. should take measures to reduce differences in income levels (ESS) Model 1: OECD Model 2: WC Model 3: OECD Model 4: WC Variables coef. s.e. coef. s.e. coef. s.e. coef. s.e. Individual-level predictors Household income (equivalized) −0.502*** (0.076) −0.544*** (0.078) −0.395*** (0.064) −0.442*** (0.059) Level-1 controls Yes Yes Yes Yes Country-level factors Target efficiency (OECD) 0.246* (0.108) – – 0.288* (0.136) – – Size of public transfers (OECD) 0.009 (0.085) – – −0.071 (0.109) – – Target efficiency (WC) – – 0.199† (0.119) – – 0.221 (0.136) Size of public transfers (WC) – – 0.030 (0.121) – – 0.033 (0.139) Gini (market income) 0.151† (0.089) 0.133 (0.093) 0.180 (0.109) 0.176 (0.104) GDP(PPP)/capita −0.196† (0.106) −0.269* (0.120) −0.043 (0.128) −0.041 (0.133) Cross-level interactions Income × Target efficiency (OECD) 0.324** (0.102) – – 0.124 (0.088) – – Income × Size of public transfers (OECD) −0.226** (0.080) – – −0.082 (0.070) – – Income × Target efficiency (WC) – – 0.179† (0.103) – – 0.016 (0.079) Income × Size of public transfers (WC) – – −0.182† (0.104) – – 0.026 (0.083) Income × Gini (market income) 0.053 (0.083) 0.030 (0.080) 0.069 (0.070) 0.067 (0.062) Income × GDP(PPP)/capita 0.138 (0.099) 0.033 (0.102) 0.027 (0.084) 0.069 (0.081) Intercept variance 0.102 (0.035) 0.125 (0.043) 0.164 (0.055) 0.167 (0.056) Slope variance (income) 0.082 (0.031) 0.082 (0.031) 0.060 (0.022) 0.050 (0.019) Covariance b/w intercepts and slopes 0.009 (0.023) 0.028 (0.027) 0.033 (0.026) 0.040 (0.025) VPC (variance partition coefficient) 0.030 0.037 0.047 0.048 Nindividual 23,050 22,562 34,961 33,236 Ncountry 18 18 18 18 Notes: Random intercept and slope (income) models. RIGLS (reiterated generalized least squares) with PQL2 (second-order penalized quasi-likelihood). Intercept variances for the empty model (with no predictor) are: [Question 1] 0.343 with a VPC (variance partition coefficient) of 0.094 (OECD, 2008); 0.298 with a VPC of 0.083 (Wang and Caminada, 2011), [Question 2] 0.319 with a VPC of 0.089 (OECD, 2008); 0.259 with a VPC of 0.073 (Wang and Caminada, 2011). The countries included in each model are presented in Appendix Table A3. Missing data were imputed using multiple imputation methods (Rubin, 1987). Estimated cut-points not displayed. ‘WC’ = Wang and Caminada (2011). Sources: ISSP (2009); ESS (2008); OECD (2008); Wang and Caminada (2011). † P < 0.10; *P < 0.05; **P < 0.01; ***P < 0.001. Table 3. Government should reduce income differences: comparison of ISSP 2009 and ESS 2008 Q1. Govt. resp. to reduce income differences b/w high and low incomes (ISSP) Q2. Govt. should take measures to reduce differences in income levels (ESS) Model 1: OECD Model 2: WC Model 3: OECD Model 4: WC Variables coef. s.e. coef. s.e. coef. s.e. coef. s.e. Individual-level predictors Household income (equivalized) −0.502*** (0.076) −0.544*** (0.078) −0.395*** (0.064) −0.442*** (0.059) Level-1 controls Yes Yes Yes Yes Country-level factors Target efficiency (OECD) 0.246* (0.108) – – 0.288* (0.136) – – Size of public transfers (OECD) 0.009 (0.085) – – −0.071 (0.109) – – Target efficiency (WC) – – 0.199† (0.119) – – 0.221 (0.136) Size of public transfers (WC) – – 0.030 (0.121) – – 0.033 (0.139) Gini (market income) 0.151† (0.089) 0.133 (0.093) 0.180 (0.109) 0.176 (0.104) GDP(PPP)/capita −0.196† (0.106) −0.269* (0.120) −0.043 (0.128) −0.041 (0.133) Cross-level interactions Income × Target efficiency (OECD) 0.324** (0.102) – – 0.124 (0.088) – – Income × Size of public transfers (OECD) −0.226** (0.080) – – −0.082 (0.070) – – Income × Target efficiency (WC) – – 0.179† (0.103) – – 0.016 (0.079) Income × Size of public transfers (WC) – – −0.182† (0.104) – – 0.026 (0.083) Income × Gini (market income) 0.053 (0.083) 0.030 (0.080) 0.069 (0.070) 0.067 (0.062) Income × GDP(PPP)/capita 0.138 (0.099) 0.033 (0.102) 0.027 (0.084) 0.069 (0.081) Intercept variance 0.102 (0.035) 0.125 (0.043) 0.164 (0.055) 0.167 (0.056) Slope variance (income) 0.082 (0.031) 0.082 (0.031) 0.060 (0.022) 0.050 (0.019) Covariance b/w intercepts and slopes 0.009 (0.023) 0.028 (0.027) 0.033 (0.026) 0.040 (0.025) VPC (variance partition coefficient) 0.030 0.037 0.047 0.048 Nindividual 23,050 22,562 34,961 33,236 Ncountry 18 18 18 18 Q1. Govt. resp. to reduce income differences b/w high and low incomes (ISSP) Q2. Govt. should take measures to reduce differences in income levels (ESS) Model 1: OECD Model 2: WC Model 3: OECD Model 4: WC Variables coef. s.e. coef. s.e. coef. s.e. coef. s.e. Individual-level predictors Household income (equivalized) −0.502*** (0.076) −0.544*** (0.078) −0.395*** (0.064) −0.442*** (0.059) Level-1 controls Yes Yes Yes Yes Country-level factors Target efficiency (OECD) 0.246* (0.108) – – 0.288* (0.136) – – Size of public transfers (OECD) 0.009 (0.085) – – −0.071 (0.109) – – Target efficiency (WC) – – 0.199† (0.119) – – 0.221 (0.136) Size of public transfers (WC) – – 0.030 (0.121) – – 0.033 (0.139) Gini (market income) 0.151† (0.089) 0.133 (0.093) 0.180 (0.109) 0.176 (0.104) GDP(PPP)/capita −0.196† (0.106) −0.269* (0.120) −0.043 (0.128) −0.041 (0.133) Cross-level interactions Income × Target efficiency (OECD) 0.324** (0.102) – – 0.124 (0.088) – – Income × Size of public transfers (OECD) −0.226** (0.080) – – −0.082 (0.070) – – Income × Target efficiency (WC) – – 0.179† (0.103) – – 0.016 (0.079) Income × Size of public transfers (WC) – – −0.182† (0.104) – – 0.026 (0.083) Income × Gini (market income) 0.053 (0.083) 0.030 (0.080) 0.069 (0.070) 0.067 (0.062) Income × GDP(PPP)/capita 0.138 (0.099) 0.033 (0.102) 0.027 (0.084) 0.069 (0.081) Intercept variance 0.102 (0.035) 0.125 (0.043) 0.164 (0.055) 0.167 (0.056) Slope variance (income) 0.082 (0.031) 0.082 (0.031) 0.060 (0.022) 0.050 (0.019) Covariance b/w intercepts and slopes 0.009 (0.023) 0.028 (0.027) 0.033 (0.026) 0.040 (0.025) VPC (variance partition coefficient) 0.030 0.037 0.047 0.048 Nindividual 23,050 22,562 34,961 33,236 Ncountry 18 18 18 18 Notes: Random intercept and slope (income) models. RIGLS (reiterated generalized least squares) with PQL2 (second-order penalized quasi-likelihood). Intercept variances for the empty model (with no predictor) are: [Question 1] 0.343 with a VPC (variance partition coefficient) of 0.094 (OECD, 2008); 0.298 with a VPC of 0.083 (Wang and Caminada, 2011), [Question 2] 0.319 with a VPC of 0.089 (OECD, 2008); 0.259 with a VPC of 0.073 (Wang and Caminada, 2011). The countries included in each model are presented in Appendix Table A3. Missing data were imputed using multiple imputation methods (Rubin, 1987). Estimated cut-points not displayed. ‘WC’ = Wang and Caminada (2011). Sources: ISSP (2009); ESS (2008); OECD (2008); Wang and Caminada (2011). † P < 0.10; *P < 0.05; **P < 0.01; ***P < 0.001. Table 4 provides further results using alternative inequality-attitude items from the EVS 2008: ‘income equality versus incentives for effort’ (models 5–6) and ‘more important: equality versus freedom’ (models 7–8). Models 5–6 show that the coefficient of the interaction term (income and target efficiency) is positive and significant, indicating that higher income earners are more likely to favor incentives for effort rather than income equality in the context of targeted spending. On the other hand, models 7–8 show that the polarization process does not occur when the question item was phrased more vaguely (or more normatively) as ‘equality versus freedom’. Table 4. Conflict between equality and freedom/incentives for effort, EVS 2008 Q3. Income equality versus incentives on effort Q4. More important: equality versus freedom Model 5: OECD Model 6: WC Model 7: OECD Model 8: WC Variables coef. s.e. coef. s.e. coef. s.e. coef. s.e. Individual-level predictors Household income (equivalized) −0.265*** (0.040) −0.267*** (0.041) −0.190*** (0.050) −0.210*** (0.053) Level-1 controls Yes Yes Yes Yes Country-level factors Target efficiency (OECD) 0.272 (0.179) – – 0.178† (0.101) – – Size of public transfers (OECD) −0.019 (0.139) – – −0.004 (0.079) – – Target efficiency (WC) – – 0.190 (0.197) – – 0.132 (0.096) Size of public transfers (WC) – – 0.129 (0.212) – – −0.129 (0.103) Gini (market income) 0.041 (0.142) 0.050 (0.152) 0.036 (0.080) 0.076 (0.074) GDP(PPP)/capita 0.057 (0.171) 0.093 (0.200) 0.057 (0.096) 0.019 (0.097) Cross-level interactions Income × Target efficiency (OECD) 0.118* (0.054) – – 0.062 (0.071) – – Income × Size of public transfers (OECD) −0.026 (0.042) – – −0.118* (0.054) – – Income × Target efficiency (WC) – – 0.127* (0.051) – – 0.071 (0.073) Income × Size of public transfers (WC) – – −0.087 (0.055) – – −0.054 (0.077) Income × Gini (market income) 0.013 (0.043) 0.023 (0.040) 0.030 (0.056) 0.035 (0.054) Income × GDP(PPP)/capita 0.034 (0.055) −0.026 (0.055) −0.081 (0.070) −0.047 (0.073) Intercept variance 0.272 (0.091) 0.349 (0.117) 0.084 (0.029) 0.080 (0.028) Slope variance (income) 0.019 (0.008) 0.017 (0.008) 0.036 (0.016) 0.036 (0.015) Covariance b/w intercepts and slopes −0.025 (0.020) −0.040 (0.023) −0.019 (0.016) −0.019 (0.015) VPC (variance partition coefficient) 0.076 0.096 0.025 0.024 Nindividual 27,422 26,430 26,010 25,186 Ncountry 18 18 18 18 Q3. Income equality versus incentives on effort Q4. More important: equality versus freedom Model 5: OECD Model 6: WC Model 7: OECD Model 8: WC Variables coef. s.e. coef. s.e. coef. s.e. coef. s.e. Individual-level predictors Household income (equivalized) −0.265*** (0.040) −0.267*** (0.041) −0.190*** (0.050) −0.210*** (0.053) Level-1 controls Yes Yes Yes Yes Country-level factors Target efficiency (OECD) 0.272 (0.179) – – 0.178† (0.101) – – Size of public transfers (OECD) −0.019 (0.139) – – −0.004 (0.079) – – Target efficiency (WC) – – 0.190 (0.197) – – 0.132 (0.096) Size of public transfers (WC) – – 0.129 (0.212) – – −0.129 (0.103) Gini (market income) 0.041 (0.142) 0.050 (0.152) 0.036 (0.080) 0.076 (0.074) GDP(PPP)/capita 0.057 (0.171) 0.093 (0.200) 0.057 (0.096) 0.019 (0.097) Cross-level interactions Income × Target efficiency (OECD) 0.118* (0.054) – – 0.062 (0.071) – – Income × Size of public transfers (OECD) −0.026 (0.042) – – −0.118* (0.054) – – Income × Target efficiency (WC) – – 0.127* (0.051) – – 0.071 (0.073) Income × Size of public transfers (WC) – – −0.087 (0.055) – – −0.054 (0.077) Income × Gini (market income) 0.013 (0.043) 0.023 (0.040) 0.030 (0.056) 0.035 (0.054) Income × GDP(PPP)/capita 0.034 (0.055) −0.026 (0.055) −0.081 (0.070) −0.047 (0.073) Intercept variance 0.272 (0.091) 0.349 (0.117) 0.084 (0.029) 0.080 (0.028) Slope variance (income) 0.019 (0.008) 0.017 (0.008) 0.036 (0.016) 0.036 (0.015) Covariance b/w intercepts and slopes −0.025 (0.020) −0.040 (0.023) −0.019 (0.016) −0.019 (0.015) VPC (variance partition coefficient) 0.076 0.096 0.025 0.024 Nindividual 27,422 26,430 26,010 25,186 Ncountry 18 18 18 18 Notes: Random intercept and slope (income) models. RIGLS (reiterated generalized least squares) with PQL2 (second-order penalized quasi-likelihood). Intercept variances for the empty model (with no predictor) are: [Question 3] 0.257 with a VPC (variance partition coefficient) of 0.086 (OECD, 2008); 0.333 with a VPC of 0.092 (Wang and Caminada, 2011), [Question 4] 0.086 with a VPC of 0.026 (OECD, 2008); 0.068 with a VPC of 0.020 (Wang and Caminada, 2011). The countries included in each model are presented in Appendix Table A3. Missing data were imputed using multiple imputation methods (Rubin, 1987). Estimated cut-points not displayed. ‘WC’ = Wang and Caminada (2011). Sources: EVS (2008); OECD (2008); Wang and Caminada (2011). † P < 0.10; *P < 0.05; **P < 0.01; ***P < 0.001. Table 4. Conflict between equality and freedom/incentives for effort, EVS 2008 Q3. Income equality versus incentives on effort Q4. More important: equality versus freedom Model 5: OECD Model 6: WC Model 7: OECD Model 8: WC Variables coef. s.e. coef. s.e. coef. s.e. coef. s.e. Individual-level predictors Household income (equivalized) −0.265*** (0.040) −0.267*** (0.041) −0.190*** (0.050) −0.210*** (0.053) Level-1 controls Yes Yes Yes Yes Country-level factors Target efficiency (OECD) 0.272 (0.179) – – 0.178† (0.101) – – Size of public transfers (OECD) −0.019 (0.139) – – −0.004 (0.079) – – Target efficiency (WC) – – 0.190 (0.197) – – 0.132 (0.096) Size of public transfers (WC) – – 0.129 (0.212) – – −0.129 (0.103) Gini (market income) 0.041 (0.142) 0.050 (0.152) 0.036 (0.080) 0.076 (0.074) GDP(PPP)/capita 0.057 (0.171) 0.093 (0.200) 0.057 (0.096) 0.019 (0.097) Cross-level interactions Income × Target efficiency (OECD) 0.118* (0.054) – – 0.062 (0.071) – – Income × Size of public transfers (OECD) −0.026 (0.042) – – −0.118* (0.054) – – Income × Target efficiency (WC) – – 0.127* (0.051) – – 0.071 (0.073) Income × Size of public transfers (WC) – – −0.087 (0.055) – – −0.054 (0.077) Income × Gini (market income) 0.013 (0.043) 0.023 (0.040) 0.030 (0.056) 0.035 (0.054) Income × GDP(PPP)/capita 0.034 (0.055) −0.026 (0.055) −0.081 (0.070) −0.047 (0.073) Intercept variance 0.272 (0.091) 0.349 (0.117) 0.084 (0.029) 0.080 (0.028) Slope variance (income) 0.019 (0.008) 0.017 (0.008) 0.036 (0.016) 0.036 (0.015) Covariance b/w intercepts and slopes −0.025 (0.020) −0.040 (0.023) −0.019 (0.016) −0.019 (0.015) VPC (variance partition coefficient) 0.076 0.096 0.025 0.024 Nindividual 27,422 26,430 26,010 25,186 Ncountry 18 18 18 18 Q3. Income equality versus incentives on effort Q4. More important: equality versus freedom Model 5: OECD Model 6: WC Model 7: OECD Model 8: WC Variables coef. s.e. coef. s.e. coef. s.e. coef. s.e. Individual-level predictors Household income (equivalized) −0.265*** (0.040) −0.267*** (0.041) −0.190*** (0.050) −0.210*** (0.053) Level-1 controls Yes Yes Yes Yes Country-level factors Target efficiency (OECD) 0.272 (0.179) – – 0.178† (0.101) – – Size of public transfers (OECD) −0.019 (0.139) – – −0.004 (0.079) – – Target efficiency (WC) – – 0.190 (0.197) – – 0.132 (0.096) Size of public transfers (WC) – – 0.129 (0.212) – – −0.129 (0.103) Gini (market income) 0.041 (0.142) 0.050 (0.152) 0.036 (0.080) 0.076 (0.074) GDP(PPP)/capita 0.057 (0.171) 0.093 (0.200) 0.057 (0.096) 0.019 (0.097) Cross-level interactions Income × Target efficiency (OECD) 0.118* (0.054) – – 0.062 (0.071) – – Income × Size of public transfers (OECD) −0.026 (0.042) – – −0.118* (0.054) – – Income × Target efficiency (WC) – – 0.127* (0.051) – – 0.071 (0.073) Income × Size of public transfers (WC) – – −0.087 (0.055) – – −0.054 (0.077) Income × Gini (market income) 0.013 (0.043) 0.023 (0.040) 0.030 (0.056) 0.035 (0.054) Income × GDP(PPP)/capita 0.034 (0.055) −0.026 (0.055) −0.081 (0.070) −0.047 (0.073) Intercept variance 0.272 (0.091) 0.349 (0.117) 0.084 (0.029) 0.080 (0.028) Slope variance (income) 0.019 (0.008) 0.017 (0.008) 0.036 (0.016) 0.036 (0.015) Covariance b/w intercepts and slopes −0.025 (0.020) −0.040 (0.023) −0.019 (0.016) −0.019 (0.015) VPC (variance partition coefficient) 0.076 0.096 0.025 0.024 Nindividual 27,422 26,430 26,010 25,186 Ncountry 18 18 18 18 Notes: Random intercept and slope (income) models. RIGLS (reiterated generalized least squares) with PQL2 (second-order penalized quasi-likelihood). Intercept variances for the empty model (with no predictor) are: [Question 3] 0.257 with a VPC (variance partition coefficient) of 0.086 (OECD, 2008); 0.333 with a VPC of 0.092 (Wang and Caminada, 2011), [Question 4] 0.086 with a VPC of 0.026 (OECD, 2008); 0.068 with a VPC of 0.020 (Wang and Caminada, 2011). The countries included in each model are presented in Appendix Table A3. Missing data were imputed using multiple imputation methods (Rubin, 1987). Estimated cut-points not displayed. ‘WC’ = Wang and Caminada (2011). Sources: EVS (2008); OECD (2008); Wang and Caminada (2011). † P < 0.10; *P < 0.05; **P < 0.01; ***P < 0.001. To facilitate visualization, predicted probabilities were calculated based on the estimates from models 5–6 in Table 4. In Figure 3, the horizontal axis represents income position (in percentile); the vertical axis represents the predicted probability of strong support for income equality (v. incentives for effort). The dashed line illustrates the predicted probability for targeted spending, while the solid line corresponds to that for non-targeted spending. All other variables were held at their observed values, and then the estimated probabilities were averaged over all cases (Hanmer and Kalkan, 2013). Figure 3. View largeDownload slide Predicted probabilities: support for income equality (v. incentives for effort). Sources: EVS (2008); OECD (2008); Wang and Caminada (2011). Figure 3. View largeDownload slide Predicted probabilities: support for income equality (v. incentives for effort). Sources: EVS (2008); OECD (2008); Wang and Caminada (2011). 5. Discussion This study has attempted to provide a comprehensive picture of the role of targeted spending by using multiple data sources. Some evidence (models 1–2, Table 3) suggests that attitudinal cleavages grow wider in the context of targeted spending (Hypothesis 2), which lends some credence to the micro-level mechanism underlying the ‘paradox of redistribution’ (KP, 1998) and, more generally, to the institutionalist assumption that institutional structures have feedback effects on mass opinion (Pierson, 1993; Mettler and Soss, 2004). The result of model 1 (estimated using ISSP and OECD) was generally consistent with that of Beramendi and Rehm (2016) (i.e. the more low-income targeting, the more polarized attitudes), even though the analytical procedure was slightly different (e.g. level-2 sample, income measurement). This result (obtained from ISSP/OECD) was somewhat robust to the use of alternative targeted-spending measure (model 2, WC), though the significance level was less impressive (P < 0.10). However, the same hypothesis was not clearly evidenced when a slightly different question-wording was used (models 3–4, Table 3). Findings also indicate that the polarization process (Hypothesis 2) occurs when respondents were asked about their views on the conflict between income equality and incentives on effort (models 5–6, Table 4), but not when a similar question was asked in a more indirect manner (models 7–8). These results indicate that different question wordings provoke different reactions to inequality and redistribution. These inconsistent results could be explained by the function of identity priming. As presented in Table 1, Questions 1–2 (government redistribution) were worded in a similar manner, the only substantial difference being ‘high- versus low-income status’. The ISSP item, which clearly refers to the differences between high and low incomes, is expected to raise respondents’ self-awareness of their identity as a high-income group (welfare contributor) or low-income group (welfare recipient). In contrast, the ESS question is vaguely worded as ‘differences in income levels’, thus sends less explicit signals and implications regarding the conflict between the groups. This process makes it less likely that high- and low-income earners polarize their inequality attitudes in the context of targeted spending (Hypothesis 2). On the other hand, Question 3 addresses not only the issue of reducing income differences but also the potential trade-off between income equality and incentives on effort. The results (models 5–6) show that high-income earners prefer incentives for effort to income equality in welfare states with low-income targeting. As with Question 1, Question 3 is a conflict-prone question since it reminds respondents that reducing income inequality can be at odds with their striving for economic success. When asked in this manner, high-income earners may consider that the rich are rich because of their hard work, and that it is unfair to equalize income differences between hard-working high-income earners and ‘lazy’ low-income earners. This identification process might explain the polarization process (Hypothesis 2) observed in models 5–6 (Table 4). Finally, no significant interaction effect (Hypothesis 2) was observed for Question 4. In general, Question 4 was vaguely worded compared to the others: ‘underprivileged’ not ‘low-incomes’; ‘social class differences’ not ‘income inequality/differences’; and ‘personal freedom (without hindrance)’ not ‘incentives on effort’. Question 4 also lacks stimuli that evoke a sense of ingroup favoritism and intergroup conflict such as (a) income inequality, (b) high versus low incomes and (c) incentives on effort (Table 1). A few caveats are worth mentioning. First, the role of identity priming must be explored further using an experimental study design. The most basic form of inequality-attitude question would be: do you agree or disagree with the statement that inequality should be reduced. Respondents’ answers to this question would differ depending on whether their sense of identity/conflict is induced or not. In an experimental study, one can randomly assign respondents to experimental and control groups and then contrast the outcomes after the experimental group received an intervention that provokes their sense of identity/conflict. The treatment for the experimental group would include introductory/guiding sentences (before the basic question) such as ‘In [Country], you are in the top 20% of the income distribution’ (using demographic data collected); ‘By “inequality” we mean income differences’; ‘Reducing income inequality implies that some of the money you worked hard for will be used to help low-income earners who need assistance’; and/or ‘Please keep in mind that reducing inequality most likely entails tax increases on high-income earners (like yourself)’. Second, although this study sought to compare inequality-attitude items from different social surveys (ISSP, ESS, EVS), the comparison of different inequality-attitude measures should ideally be made using the same survey data. In this study, to harmonize potential inconsistences between the survey datasets, several efforts have been made: the same set of countries (level-2 sample), time-period, and explanatory variables and their measurements (Table A1). Several implications can be noted. First, the empirical results from this study direct our attention to the role of identity priming in redistributive politics. Hypothesis 2 predicted that the uneven distribution of public transfers serves as a contextual cue in defining expectations for the consequences of egalitarian practices. This notion is based on the premise that respondents are aware of their socio-economic status and their benefits/costs associated with an egalitarian goal. High-income earners become less willing to support egalitarian views in the context of low-income targeting because they consider that targeting unfairly benefits low-income group at the expense of their group. This mechanism does not function if a survey question fails to induce their sense of identity and conflict. Second, this study has provided a more detailed description of the ‘paradox of redistribution’ process (KP, 1998). The basic form of the ‘paradox of redistribution’ can be expressed as: targeted spending → redistribution. As argued in the theory section, the underlying (micro-level) logic of this process is that first, low-income targeting generates identity-based preferences toward inequality, and second, polarized preferences are translated to redistributive policy outputs. Thus, the ‘paradox of redistribution’ process can be illustrated as: Targeted spending (a) → identity priming (b1) → polarization in attitudes (b2) → redistribution (c). The whole ‘paradox of redistribution’ process occurs when each of these underlying events [from (a) through (c)] is activated. Interestingly, recent studies have shown that the aggregate-level relationship between target efficiency (a) and redistribution (c) can no longer be observed or at least has substantially weakened since the mid-1990s (Kenworthy, 2011; Marx et al., 2013). On the other hand, the results of this article (models 1–2, Table 3) have provided some evidence in favor of the causal sequence from (a) through (b2). These seemingly puzzling results could be explained by looking at the micro-level process—for instance, the link between attitudinal polarization (b2) and actual policy outputs (c) is by no means as self-evident as is often assumed. In conclusion, this study has presented suggestive evidence that the validity of the polarization hypothesis (Hypothesis 2) is susceptible to the question wording of inequality-attitude indicators, and has provided a theoretical rationale for why this might be the case by suggesting that identity priming plays a crucial role in the micro-level mechanism of the ‘paradox of redistribution’. A promising avenue for future research would be to expand on this line of interest and provide more rigorous (e.g. experimental) data to confirm its plausibility. Acknowledgements I wish to thank Professors Stephen Fisher, Nan Dirk de Graaf, Jane Gingrich and Tamas David-Barrett for their kind assistance during the study. I am also indebted to Professor David Rueda and anonymous reviewers for their valuable contributions to its development. References Alesina A. , Ferrara E. L. ( 2005 ) ‘ Preferences for Redistribution in the Land of Opportunities’, Journal of Public Economics , 89 , 897 – 931 . Google Scholar CrossRef Search ADS Barnes L. ( 2015 ) ‘ The Size and Shape of Government: Preferences over Redistributive Tax Policy’, Socio-Economic Review , 13 , 55 – 78 . Google Scholar CrossRef Search ADS Beramendi P. , Rehm P. ( 2016 ) ‘ Who Gives, Who Gains? Progressivity and Preferences’, Comparative Political Studies , 49 , 529 – 563 . Google Scholar CrossRef Search ADS Brady D. , Bostic A. ( 2015 ) ‘ Paradoxes of Social Policy: Welfare Transfers, Relative Poverty, and Redistribution Preferences’, American Sociological Review , 80 , 268 – 298 . Google Scholar CrossRef Search ADS Buhmann B. , Rainwater L. , Schmaus G. , Smeeding T. M. ( 1988 ) ‘ Equivalence Scales, Well-being, Inequality, and Poverty: Sensitivity Estimates across Ten Countries Using the Luxembourg Income Study (LIS) Database’, Review of Income and Wealth , 34 , 115 – 142 . Google Scholar CrossRef Search ADS Cusack T. , Iversen T. , Rehm P. ( 2006 ) ‘ Risks at Work: The Demand and Supply Sides of Government Redistribution’, Oxford Review of Economic Policy , 22 , 365 – 389 . Google Scholar CrossRef Search ADS Edlund J. ( 1999 ) ‘ Trust in Government and Welfare Regimes: Attitudes to Redistribution and Financial Cheating in the USA and Norway’, European Journal of Political Research , 35 , 341 – 370 . Esping-Andersen G. ( 1990 ) The Three Worlds of Welfare Capitalism , Cambridge , Polity Press . Fields G. S. ( 1979 ) ‘ Decomposing LDC Inequality’, Oxford Economic Papers , 31 , 437 – 459 . Google Scholar CrossRef Search ADS Gingrich J. , Ansell B. ( 2012 ) ‘ Preferences in Context: Micro Preferences, Macro Contexts, and the Demand for Social Policy’, Comparative Political Studies , 45 , 1624 – 1654 . Google Scholar CrossRef Search ADS Goldstein H. ( 1995 ) Multilevel Statistical Models , London , Edward Arnold . Graham J. W. ( 2009 ) ‘ Missing Data Analysis: Making it Work in the Real World’, Annual Review of Psychology , 60 , 549 – 576 . Google Scholar CrossRef Search ADS PubMed Hanmer M. J. , Kalkan K. O. ( 2013 ) ‘ Behind the Curve: Clarifying the Best Approach to Calculating Predicted Probabilities and Marginal Effects from Limited Dependent Variable Models’, America Journal of Political Science , 57 , 263 – 277 . Google Scholar CrossRef Search ADS Honaker J. , King G. , Blackwell M. ( 2011 ) ‘ Amelia II: A Program for Missing Data’, Journal of Statistical Software , 45 , 1 – 47 . Google Scholar CrossRef Search ADS Hout M. ( 2004 ) Getting the Most Out of the GSS Income Measures , GSS Methodological Report No. 101, Chicago , NORC . IMF ( 2005 ) World Economic Outlook Database (September) , Washington, DC , IMF . Jæger M. M. ( 2006 ) ‘ Welfare Regimes and Attitudes towards Redistribution: The Regime Hypothesis Revisited’, European Sociological Review , 22 , 157 – 170 . Google Scholar CrossRef Search ADS Jæger M. M. ( 2008 ) ‘ Does Left-Right Orientation have a Causal Effect on Support for Redistribution? Causal Analysis with Cross-sectional Data Using Instrumental Variables’, International Journal of Public Opinion Research , 20 , 363 – 374 . Google Scholar CrossRef Search ADS Kakwani N. ( 1986 ) Analyzing Redistribution Politics , Oxford , Oxford University Press . Kenworthy L. ( 2011 ) Progress for the Poor , Oxford , Oxford University Press . Google Scholar CrossRef Search ADS King G. , Honaker J. , Joseph A. , Scheve K. ( 2001 ) ‘Analyzing Incomplete Political Science Data: An Alternative Algorithm for Multiple Imputation, American Political Science Review , 95 , 49 – 69 . Korpi W. ( 1980 ) ‘ Social Policy and Distributional Conflict in the Capitalist Democracies: A Preliminary Comparative Framework’, West European Politics , 3 , 296 – 316 . Google Scholar CrossRef Search ADS Korpi W. , Palme J. ( 1998 ) ‘ The Paradox of Redistribution and Strategies of Equality: Welfare State Institutions, Inequality, and Poverty in the Western Countries’, American Sociological Review , 63 , 661 – 687 . Google Scholar CrossRef Search ADS Kreft I. , de Leeuw J. , ( 1998 ) Introducing Multilevel Modeling , London , Sage . Google Scholar CrossRef Search ADS Larsen C. A. ( 2008 ) ‘ The Institutional Logic of Welfare Attitudes: How Welfare Regimes Influence Public Support’, Comparative Political Studies , 41 , 145 – 168 . Google Scholar CrossRef Search ADS Leckie G. , Charlton C. ( 2012 ) ‘ runmlwin - A Program to Run the MLwiN Multilevel Modeling Software from within Stata’, Journal of Statistical Software , 52 , 1 – 40 . Google Scholar CrossRef Search ADS Luttmer E. F. P. ( 2001 ) ‘ Group Loyalty and the Taste for Redistribution’, Journal of Political Economy , 109 , 500 – 528 . Google Scholar CrossRef Search ADS Margalit Y. ( 2013 ) ‘ Explaining Social Policy Preferences: Evidence from the Great Recession’, American Political Science Review , 107 , 80 – 103 . Google Scholar CrossRef Search ADS Marx I. , Salanauskaite L. , Verbist G. ( 2013 ) The Paradox of Redistribution Revisited: And That It May Rest in Peace?, IZA Discussion Paper No. 7414, Bonn, Institute for the Study of Labor (IZA). Meltzer A. H. , Richard S. F. ( 1981 ) ‘ A Rational Theory of the Size of Government’, Journal of Political Economy , 89 , 914 – 927 . Google Scholar CrossRef Search ADS Mettler S. , Soss J. ( 2004 ) ‘ The Consequences of Public Policy for Democratic Citizenship: Bridging Policy Studies and Mass Politics’, Perspectives on Politics , 2 , 55 – 73 . Google Scholar CrossRef Search ADS OECD ( 2005 ) Income Distribution Database (IDD) , Paris , OECD . OECD ( 2008 ) ‘How Much Redistribution Do Governments Achieve? The Role of Cash Transfers and Household Taxes’. In OECD (ed.) Growing Unequal? Income Distribution and Poverty in OECD Countries , Paris , OECD , pp. 97 – 121 . Offe C. ( 1992 ) ‘A Non-Productivist Design for Social Policies’. In Parijs P. V. (ed.) Arguing for Basic Income , London , Verso . Pierson P. ( 1993 ) ‘ When Effect Becomes Cause: Policy Feedback and Political Change’, World Politics , 45 , 595 – 628 . Google Scholar CrossRef Search ADS Pierson P. ( 1996 ) ‘ The New Politics of the Welfare State’, World Politics , 48 , 143 – 179 . Google Scholar CrossRef Search ADS Pierson P. ( 2001 ) ‘Coping with Permanent Austerity: Welfare State Restructuring in Affluent Democracies’. In Pierson P. (ed.) The New Politics of the Welfare State , Oxford , Oxford University Press , pp. 410 – 456 . Google Scholar CrossRef Search ADS Rasbash J. , Steele F. , Browne W. J. , Goldstein H. , ( 2014 ) A User's Guide to MLwiN version 2.31, Centre for Multilevel Modelling, University of Bristol. Rehm P. ( 2009 ) ‘ Risks and Redistribution: An Individual-level Analysis’, Comparative Political Studies , 42 , 855 – 881 . Google Scholar CrossRef Search ADS Roberts K. W. S. ( 1977 ) ‘ Voting over Income Tax Schedules’, Journal of Public Economics , 8 , 329 – 340 . Google Scholar CrossRef Search ADS Romer T. ( 1975 ) ‘ Individual Welfare, Majority Voting, and the Properties of a Linear Income Tax’, Journal of Public Economics , 4 , 163 – 185 . Google Scholar CrossRef Search ADS Rothstein B. ( 1998 ) Just Institutions Matter: The Moral and Political Logic of the Universal Welfare State , Cambridge , Cambridge University Press . Google Scholar CrossRef Search ADS Rothstein B. , Steinmo S. , ( 2002 ) ‘Restructuring Politics: Institutional Analysis and the Challenges of Modern Welfare States’. In Rothstein B. , Steinmo S. (eds) Restructuring the Welfare State: Political Institutions and Policy Change , New York , Palgrave Macmillan , pp. 1 – 19 . Google Scholar CrossRef Search ADS Rubin D. B. ( 1987 ) Multiple Imputation for Nonresponse in Surveys , New York , John Wiley & Sons, Inc . Google Scholar CrossRef Search ADS Rubin D. B. ( 1996 ) ‘ Multiple Imputation After 18+ Years’, Journal of the American Statistical Association , 91 , 473 – 489 . Google Scholar CrossRef Search ADS Rueda D. ( 2014 ) Food Comes First, Then Morals: Redistribution Preferences, Altruism and Group Heterogeneity in Western Europe, CAGE Working Paper Series No. 200, Competitive Advantage in the Global Economy (CAGE). Rueda D. , Stegmueller D. ( 2015 ) ‘ The Externalities of Inequality: Fear of Crime and Preferences for Redistribution in Western Europe’, American Journal of Political Science , 60 , 472 – 489 . Google Scholar CrossRef Search ADS Rueda D. , Segmueller D. , Idema T. , ( 2014 ) ‘The Effects of Income Expectations on Redistribution Preferences in Western Europe’, paper presented at the Annual Meeting of the American Political Science Association (APSA), 2014. Skocpol T. ( 1992 ) Protecting Soldiers and Mothers , Cambridge, MA , Harvard University Press . Snijders T. A. B. , Bosker R. J. ( 1999 ) Multilevel Analysis: An Introduction to Basic and Advanced Multilevel Modeling , London , Sage . Sumino T. ( 2014 ) ‘ Escaping the Curse of Economic Self-interest: An Individual-level Analysis of Public Support for the Welfare State in Japan’, Journal of Social Policy , 43 , 109 – 133 . Google Scholar CrossRef Search ADS Sumino T. ( 2016 ) ‘ Level or Concentration? A Cross-national Analysis of Public Attitudes towards Taxation Policies’, Social Indicators Research , 129 , 1115 – 1134 . Google Scholar CrossRef Search ADS Svallfors S. ( 1997 ) ‘ Worlds of Welfare and Attitudes to Redistribution: A Comparison of Eight Western Nations’, European Sociological Review , 13 , 283 – 304 . Google Scholar CrossRef Search ADS Tajfel H. ( 1981 ) Human Groups and Social Categories . Cambridge , Cambridge University Press . Turner J. C. ( 1985 ) ‘Social Categorization and the Self-concept: A Social Cognitive Theory of Group Behaviour’. In Lawler E. J. (eds) Advances in Group Processes , Greenwich, CT , JAI Press , pp. 77 – 122 . Turner J. C. , Hogg M. A. , Oakes P. J. , Reicher S. D. , Wetherell M. S. , ( 1987 ) Rediscovering the Social Group: A Self-categorization Theory . Oxford and New York , Basil Blackwell . Wang C. , Caminada K. ( 2011 ) Disentangling Income Inequality and the Redistributive Effect of Social Transfers and Taxes in 36 LIS Countries , LIS Working Paper Series No. 567, Luxembourg , Luxembourg Income Study (LIS ). White I. R. , Royston P. , Wood A. M. ( 2011 ) ‘ Multiple Imputation Using Chained Equations: Issues and Guidance for Practice’, Statistics in Medicine , 30 , 377 – 399 . Google Scholar CrossRef Search ADS PubMed World Bank ( 2008 ) International Comparison Program Database . Washington, DC , World Bank . Appendix A Table A1. Descriptive statistics ISSP 2009 ESS 2008 EVS 2008 Variables Mean SD Min. Max. Mean SD Min. Max. Mean SD Min. Max. Inequality attitude (5-pt. scale) 2.12 1.09 1 5 2.19 1.05 1 5 2.76 1.30 1 5 Inequality attitude (dummy) – – – – – – – – 0.46 0.50 0 1 Ln(income) 7.47 0.85 5.39 9.93 7.14 0.75 4.26 9.18 7.07 0.87 3.39 9.90 Male 0.46 0.50 0 1 0.46 0.50 0 1 0.45 0.50 0 1 Female 0.54 0.50 0 1 0.54 0.50 0 1 0.55 0.50 0 1 Age/10 4.88 1.71 1.8 8.9 4.84 1.78 1.8 8.9 4.84 1.76 1.8 8.9 Lower secondary or less 0.29 0.45 0 1 0.32 0.47 0 1 0.34 0.47 0 1 Upper secondary 0.46 0.50 0 1 0.44 0.50 0 1 0.44 0.50 0 1 Tertiary 0.25 0.43 0 1 0.24 0.43 0 1 0.23 0.42 0 1 Full-time employed 0.46 0.50 0 1 0.45 0.50 0 1 0.47 0.50 0 1 Unemployed 0.06 0.23 0 1 0.05 0.22 0 1 0.06 0.24 0 1 Student 0.05 0.23 0 1 0.06 0.23 0 1 0.05 0.22 0 1 Retired 0.25 0.43 0 1 0.24 0.43 0 1 0.25 0.43 0 1 Housekeeping 0.06 0.24 0 1 0.10 0.30 0 1 0.08 0.27 0 1 Other 0.12 0.32 0 1 0.10 0.30 0 1 0.09 0.29 0 1 ISSP 2009 ESS 2008 EVS 2008 Variables Mean SD Min. Max. Mean SD Min. Max. Mean SD Min. Max. Inequality attitude (5-pt. scale) 2.12 1.09 1 5 2.19 1.05 1 5 2.76 1.30 1 5 Inequality attitude (dummy) – – – – – – – – 0.46 0.50 0 1 Ln(income) 7.47 0.85 5.39 9.93 7.14 0.75 4.26 9.18 7.07 0.87 3.39 9.90 Male 0.46 0.50 0 1 0.46 0.50 0 1 0.45 0.50 0 1 Female 0.54 0.50 0 1 0.54 0.50 0 1 0.55 0.50 0 1 Age/10 4.88 1.71 1.8 8.9 4.84 1.78 1.8 8.9 4.84 1.76 1.8 8.9 Lower secondary or less 0.29 0.45 0 1 0.32 0.47 0 1 0.34 0.47 0 1 Upper secondary 0.46 0.50 0 1 0.44 0.50 0 1 0.44 0.50 0 1 Tertiary 0.25 0.43 0 1 0.24 0.43 0 1 0.23 0.42 0 1 Full-time employed 0.46 0.50 0 1 0.45 0.50 0 1 0.47 0.50 0 1 Unemployed 0.06 0.23 0 1 0.05 0.22 0 1 0.06 0.24 0 1 Student 0.05 0.23 0 1 0.06 0.23 0 1 0.05 0.22 0 1 Retired 0.25 0.43 0 1 0.24 0.43 0 1 0.25 0.43 0 1 Housekeeping 0.06 0.24 0 1 0.10 0.30 0 1 0.08 0.27 0 1 Other 0.12 0.32 0 1 0.10 0.30 0 1 0.09 0.29 0 1 Sources: ISSP (2009); ESS (2008); EVS (2008). View Large Table A1. Descriptive statistics ISSP 2009 ESS 2008 EVS 2008 Variables Mean SD Min. Max. Mean SD Min. Max. Mean SD Min. Max. Inequality attitude (5-pt. scale) 2.12 1.09 1 5 2.19 1.05 1 5 2.76 1.30 1 5 Inequality attitude (dummy) – – – – – – – – 0.46 0.50 0 1 Ln(income) 7.47 0.85 5.39 9.93 7.14 0.75 4.26 9.18 7.07 0.87 3.39 9.90 Male 0.46 0.50 0 1 0.46 0.50 0 1 0.45 0.50 0 1 Female 0.54 0.50 0 1 0.54 0.50 0 1 0.55 0.50 0 1 Age/10 4.88 1.71 1.8 8.9 4.84 1.78 1.8 8.9 4.84 1.76 1.8 8.9 Lower secondary or less 0.29 0.45 0 1 0.32 0.47 0 1 0.34 0.47 0 1 Upper secondary 0.46 0.50 0 1 0.44 0.50 0 1 0.44 0.50 0 1 Tertiary 0.25 0.43 0 1 0.24 0.43 0 1 0.23 0.42 0 1 Full-time employed 0.46 0.50 0 1 0.45 0.50 0 1 0.47 0.50 0 1 Unemployed 0.06 0.23 0 1 0.05 0.22 0 1 0.06 0.24 0 1 Student 0.05 0.23 0 1 0.06 0.23 0 1 0.05 0.22 0 1 Retired 0.25 0.43 0 1 0.24 0.43 0 1 0.25 0.43 0 1 Housekeeping 0.06 0.24 0 1 0.10 0.30 0 1 0.08 0.27 0 1 Other 0.12 0.32 0 1 0.10 0.30 0 1 0.09 0.29 0 1 ISSP 2009 ESS 2008 EVS 2008 Variables Mean SD Min. Max. Mean SD Min. Max. Mean SD Min. Max. Inequality attitude (5-pt. scale) 2.12 1.09 1 5 2.19 1.05 1 5 2.76 1.30 1 5 Inequality attitude (dummy) – – – – – – – – 0.46 0.50 0 1 Ln(income) 7.47 0.85 5.39 9.93 7.14 0.75 4.26 9.18 7.07 0.87 3.39 9.90 Male 0.46 0.50 0 1 0.46 0.50 0 1 0.45 0.50 0 1 Female 0.54 0.50 0 1 0.54 0.50 0 1 0.55 0.50 0 1 Age/10 4.88 1.71 1.8 8.9 4.84 1.78 1.8 8.9 4.84 1.76 1.8 8.9 Lower secondary or less 0.29 0.45 0 1 0.32 0.47 0 1 0.34 0.47 0 1 Upper secondary 0.46 0.50 0 1 0.44 0.50 0 1 0.44 0.50 0 1 Tertiary 0.25 0.43 0 1 0.24 0.43 0 1 0.23 0.42 0 1 Full-time employed 0.46 0.50 0 1 0.45 0.50 0 1 0.47 0.50 0 1 Unemployed 0.06 0.23 0 1 0.05 0.22 0 1 0.06 0.24 0 1 Student 0.05 0.23 0 1 0.06 0.23 0 1 0.05 0.22 0 1 Retired 0.25 0.43 0 1 0.24 0.43 0 1 0.25 0.43 0 1 Housekeeping 0.06 0.24 0 1 0.10 0.30 0 1 0.08 0.27 0 1 Other 0.12 0.32 0 1 0.10 0.30 0 1 0.09 0.29 0 1 Sources: ISSP (2009); ESS (2008); EVS (2008). View Large Table A2. Pearson correlations between macro-level variables Variables 1a 1b 2a 2b 3 4 1a. Target efficiency (OECD) 1.00 1b. Target efficiency (WC) 0.87*** 1.00 2a. Size of public transfers (OECD) 0.37 0.47* 1.00 2b. Size of public transfers (WC) 0.57* 0.70** 0.52* 1.00 3. Gini (market income) 0.47* 0.21 0.39 0.32 1.00 4. GDP(PPP)/capita −0.63** −0.36 −0.14 −0.50* −0.47* 1.00 Variables 1a 1b 2a 2b 3 4 1a. Target efficiency (OECD) 1.00 1b. Target efficiency (WC) 0.87*** 1.00 2a. Size of public transfers (OECD) 0.37 0.47* 1.00 2b. Size of public transfers (WC) 0.57* 0.70** 0.52* 1.00 3. Gini (market income) 0.47* 0.21 0.39 0.32 1.00 4. GDP(PPP)/capita −0.63** −0.36 −0.14 −0.50* −0.47* 1.00 Note: ‘WC’ = Wang and Caminada (2011). * P < 0.05; **P < 0.01; ***P < 0.001. View Large Table A2. Pearson correlations between macro-level variables Variables 1a 1b 2a 2b 3 4 1a. Target efficiency (OECD) 1.00 1b. Target efficiency (WC) 0.87*** 1.00 2a. Size of public transfers (OECD) 0.37 0.47* 1.00 2b. Size of public transfers (WC) 0.57* 0.70** 0.52* 1.00 3. Gini (market income) 0.47* 0.21 0.39 0.32 1.00 4. GDP(PPP)/capita −0.63** −0.36 −0.14 −0.50* −0.47* 1.00 Variables 1a 1b 2a 2b 3 4 1a. Target efficiency (OECD) 1.00 1b. Target efficiency (WC) 0.87*** 1.00 2a. Size of public transfers (OECD) 0.37 0.47* 1.00 2b. Size of public transfers (WC) 0.57* 0.70** 0.52* 1.00 3. Gini (market income) 0.47* 0.21 0.39 0.32 1.00 4. GDP(PPP)/capita −0.63** −0.36 −0.14 −0.50* −0.47* 1.00 Note: ‘WC’ = Wang and Caminada (2011). * P < 0.05; **P < 0.01; ***P < 0.001. View Large Table A3. List of countries included in the analysis (Tables 3–4) OECD (2008) Wang and Caminada (2011) ISSP 09 ESS 08 EVS 08 ISSP 09 ESS 08 EVS 08 Countries M1 M3 M5/7 M2 M4 M6/8 Austria x x x x x x Belgium x x x x x x Czech Rep. x x x x x x Denmark x x x x x x Estonia n/a n/a n/a x x x Finland x x x x x x France x x x x x x Germany x x x x x x Great Britain x x x x x x Hungary x x x x x x Netherlands x x x x x x Norway x x x x x x Poland x x x x x x Portugal x x x n/a n/a n/a Slovak Rep. x x x x x x Slovenia n/a n/a n/a x x x Spain x x x x x x Sweden x x x x x x Switzerland x x x x x x Turkey x x x n/a n/a n/a OECD (2008) Wang and Caminada (2011) ISSP 09 ESS 08 EVS 08 ISSP 09 ESS 08 EVS 08 Countries M1 M3 M5/7 M2 M4 M6/8 Austria x x x x x x Belgium x x x x x x Czech Rep. x x x x x x Denmark x x x x x x Estonia n/a n/a n/a x x x Finland x x x x x x France x x x x x x Germany x x x x x x Great Britain x x x x x x Hungary x x x x x x Netherlands x x x x x x Norway x x x x x x Poland x x x x x x Portugal x x x n/a n/a n/a Slovak Rep. x x x x x x Slovenia n/a n/a n/a x x x Spain x x x x x x Sweden x x x x x x Switzerland x x x x x x Turkey x x x n/a n/a n/a Sources: ISSP (2009); ESS (2008); EVS (2008); OECD (2008); Wang and Caminada (2011). View Large Table A3. List of countries included in the analysis (Tables 3–4) OECD (2008) Wang and Caminada (2011) ISSP 09 ESS 08 EVS 08 ISSP 09 ESS 08 EVS 08 Countries M1 M3 M5/7 M2 M4 M6/8 Austria x x x x x x Belgium x x x x x x Czech Rep. x x x x x x Denmark x x x x x x Estonia n/a n/a n/a x x x Finland x x x x x x France x x x x x x Germany x x x x x x Great Britain x x x x x x Hungary x x x x x x Netherlands x x x x x x Norway x x x x x x Poland x x x x x x Portugal x x x n/a n/a n/a Slovak Rep. x x x x x x Slovenia n/a n/a n/a x x x Spain x x x x x x Sweden x x x x x x Switzerland x x x x x x Turkey x x x n/a n/a n/a OECD (2008) Wang and Caminada (2011) ISSP 09 ESS 08 EVS 08 ISSP 09 ESS 08 EVS 08 Countries M1 M3 M5/7 M2 M4 M6/8 Austria x x x x x x Belgium x x x x x x Czech Rep. x x x x x x Denmark x x x x x x Estonia n/a n/a n/a x x x Finland x x x x x x France x x x x x x Germany x x x x x x Great Britain x x x x x x Hungary x x x x x x Netherlands x x x x x x Norway x x x x x x Poland x x x x x x Portugal x x x n/a n/a n/a Slovak Rep. x x x x x x Slovenia n/a n/a n/a x x x Spain x x x x x x Sweden x x x x x x Switzerland x x x x x x Turkey x x x n/a n/a n/a Sources: ISSP (2009); ESS (2008); EVS (2008); OECD (2008); Wang and Caminada (2011). View Large © The Author 2017. Published by Oxford University Press and the Society for the Advancement of Socio-Economics. All rights reserved. For Permissions, please email: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

Socio-Economic ReviewOxford University Press

Published: Jul 1, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off