(In)congruence across policy dimensions

Most studies on congruence have looked at the match between voters and their parties (Dalton, Reference Dalton1985; Ezrow and Xezonakis, Reference Ezrow and Xezonakis2011; Belchior, Reference Belchior2013). This perspective is often linked to the ‘responsible party’ or ‘party government’ model of representation which assumes that parties propose coherent political platforms, from which voters chose according to their ideological proximity and that, once elected, parties will implement their political programmes. As a result, if parties are congruent with their voters, governments should also ultimately represent the preferences of the majority of voters well (Pierce, Reference Pierce and Miller1999).Footnote ³ Another strand of research studies left–right congruence in relation to representative bodies, by matching the preferences of the voters with the position of the corresponding government or parliament (Huber and Powell, Reference Huber and Powell1994; Powell, Reference Powell2000; Blais and Bodet, Reference Blais and Bodet2006; Ferland, Reference Ferland2016). Within that tradition, research has focused on different aspects: citizens’ and policymakers’ median (or mean) preferences (Huber and Powell, Reference Huber and Powell1994; Powell, Reference Powell2000), the distribution of preferences among both citizens and political elites (see e.g. Golder and Stramski, Reference Golder and Stramski2010) or individual determinants of congruence (Bernauer et al., Reference Bernauer, Giger and Rosset2015).

Existing studies of government congruence almost exclusively focus on the left–right scale referring to ideological orientations rather than policy goals. This empirical approach presupposes that policy preferences across a wide range of issues can be summarized in a single left–right dimension. Yet, this assumption seems to be increasingly violated. Although a very large proportion of survey respondents have no problem placing themselves on such a continuum, it is not exactly clear what this position entails in terms of policy stances (Bauer et al., Reference Bauer, Barberá, Ackermann and Venetz2017). Political competition in Europe has evolved over the last decades, with non-economic issues gaining in importance (Kriesi et al., Reference Kriesi, Grande, Lachat, Dolezal, Bornschier and Frey2006). A cultural dimension has supplemented the traditional economic divide between pro-state intervention in the economy and pro-market stances. There individuals with universalistic and progressive values oppose those who hold more nationalistic and conservative views (Bornschier, Reference Bornschier2010). As a result, it is not obvious where a moderately left person (e.g. 4 on a 0–10 left–right scale) stands on specific policy domains. It could be that this person holds indeed moderately left preferences on all issues, but this positioning could for instance reflect a resolutely ‘leftist’ position on economic issues combined with a moderately conservative position on socio-cultural topics such as immigration or morality issues. Indeed, many survey respondents fail to associate specific policies with overall left–right orientations (Freire and Belchior, Reference Freire and Belchior2011; Dolezal et al., Reference Dolezal, Eder, Kritzinger and Zeglovits2013) rendering left–right incapable of providing an unbiased summary of citizen preferences across issues (Munzert and Bauer, Reference Munzert and Bauer2013; Rosset et al., Reference Rosset, Lutz, Kissau, Bühlmann and Fivaz2016). Accordingly, various studies call for distinguishing between preferences on different political themes.

Our article heeds this call and looks at congruence on the left–right scale as well as on two specific issues which are salient in the current context: redistribution and European integration. To be frank, this choice of issues is closely linked to data availability. At the same time, their inclusion in the citizen and elite surveys we use attests to the fact that these are also important issues on the European continent. One of these issues – European integration – can be linked to the cultural divide in societies between nationalistic and conservative views on the one hand and universalistic and liberal views on the other. Redistribution refers clearly to the economic cleavage. But what can we expect to find in relation to levels of policy congruence between the citizens and governments across specific issues?

The comparative research on issue congruence has mostly focused on the link between parties and their voters. Drawing on this literature, we can formulate some expectations with regard to citizen government congruence. First, it is one of the most consistent findings of the literature that congruence on general left–right attitudes tends to be relatively high (Thomassen and Schmitt, Reference Thomassen and Schmitt1997; Dalton and McAllister, Reference Dalton and McAllister2015). In contrast, studies focusing on more specific issues tend to come to more negative conclusions regarding the match between preferences of citizens and elites (Thomassen, Reference Thomassen2012; Freire and Belchior, Reference Freire and Belchior2013). Thus, our first hypothesis:

When looking at other issues than left–right, congruence appears to be particularly low on cultural issues. Dalton (Reference Dalton2017), for instance, finds that the match between parties and their voters is lowest on the authoritarian–libertarian divide as well as on gender issues. Regarding the direction of this incongruence, the majority of current research documents that party voters are more conservative than party elites. For instance, European voters have been consistently found to be more Eurosceptical than political elites (Thomassen and Schmitt, Reference Thomassen and Schmitt1997; Mattila and Raunio, Reference Mattila and Raunio2006). Country studies also reveal more conservative stances of the public as compared to the parties they vote for. Portuguese voters display more authoritarian stances than their parties (Freire and Belchior, Reference Freire and Belchior2013). The same conclusion can be drawn from an analysis on The Netherlands showing that voters are more in favour of law and order than the parties, especially among left party supporters (Thomassen, Reference Thomassen2012). Research is, however, not completely unanimous on this matter as on issues of immigration there is little gap between parties and voters and, when such a gap exists, voters tend to be more liberal than elites (Lefkofridi and Horvath, Reference Lefkofridi and Horvath2012). Taken together, however, these results show that, when there are differences between parties and voters, parties tend to hold more liberal views on cultural issues than their voters. This is consistent with the assumption that politicians are more tolerant than the general population, notably due to their socialization in party structures (Sullivan et al., Reference Sullivan, Walsh and Shamir1993). Therefore, we provide our second hypothesis:

In relation to economic issues, research has generally found smaller gaps between voters and parties than for cultural issues (Freire and Belchior, Reference Freire and Belchior2011; Dalton, Reference Dalton2017). There is, however, also evidence for a systematic bias with citizens being generally more left or pro-state intervention than elites. This has been documented in the Netherlands where party voters favour more income equality than their representatives, especially among the voters of right parties (Thomassen, Reference Thomassen2012). In Switzerland a similar trend has been identified with the parliament as a whole holding views that are significantly less favourable to social security and redistribution than the median citizen (Rosset, Reference Rosset2013). There is also an emerging literature on responsiveness showing that policy preferences of the relatively poor (and left-leaning on economic issues) citizens are less well translated into economic policy and that government decisions tend to follow the preferences of richer (and right-leaning) citizens (Donnelly and Lefkofridi, Reference Donnelly and Lefkofridi2014; Peters and Ensink, Reference Peters and Ensink2015). In sum, based on the results of previous research, we formulate the following hypothesis:

One of the potential explanations for the differences that we expect in relation to the direction of congruence for cultural and economic issues is related to representational inequalities. The current literature points to a systematic bias in the extent to which citizens’ preferences are reflected in political institutions or public policy (Gilens, Reference Gilens2012). Social inequalities, so the argument goes, might translate into political inequalities with lower levels of issue congruence among socially disadvantaged citizens. At least two mechanisms could account for such a systematic bias. First, disadvantaged social groups tend to turn out less to vote and thus have proportionally a more limited influence on election results. Studies show, for example, that on average highly educated citizens tend to participate more in elections. The same is true for lower income groups that display lower levels of participation in most European countries (Armingeon and Schädel, Reference Armingeon and Schädel2015; Schäfer, Reference Schäfer2015). Second, there is also an under-representation of individuals sharing those groups’ characteristics among political elites. This descriptive under-representation is very high in relation to features such as education and wealth (Pontusson, Reference Pontusson2015). Regarding these two categories, representative bodies tend not to include many (or oftentimes any) members of very large groups of citizens defined in these terms, and this despite of the fact that these features have been consistently found to be associated with individuals’ policy preferences.

Insofar as individuals’ characteristics in terms of income or education are related to specific policy preferences, these two mechanisms may tilt governments’ positions in the direction of advantaged social groups. These preferences, however, are not easily summarized in a single dimension as the systematically under-represented groups tend to hold policy preferences that are to the ‘right’ in some policy domains, but to the ‘left’ in other policy areas.

On economic issues, research has shown that individuals tend to hold less pro-state intervention preferences when their income increases, possibly reflecting self-interested attitudes (Cusack et al., Reference Cusack, Iversen and Rehm2006). Education also gives individuals more confidence in their own capacity of faring well on the job market or coping with difficulty. Holding income constant, education is consequently found to be negatively associated with demand for redistribution (Rehm, Reference Rehm2009). It, thus, seems that if politicians are affected by their social background, they should be more to the right of citizens on economic issues, which has been documented in a number of studies (Carnes and Lupu, Reference Carnes and Lupu2015; Rosset, Reference Rosset2016). In relation to cultural issues of internationalization and universalistic values, however, the descriptive representation gap is likely to have the opposite effect with politicians being on average more libertarian or more ‘left’ than the citizens. Indeed, the new social cleavage revolving around cultural issues linked in particular to globalization is one that particularly divides along educational lines with better educated people displaying on average much more liberal (Stubager, Reference Stubager2006) and pro-European preferences (Hakhverdian et al., Reference Hakhverdian, Van Elsas and Van der Brug2013). To some extent, the more pro-EU stances of political elites (Mattila and Raunio, Reference Mattila and Raunio2006) could be read in the light of the fact that the social background between citizens and their representatives differs considerably. If politicians’ social background plays a role in the way parties perceive political problems, we would expect the party system to be tilted to the right on economic issues and tilted to the left on sociocultural issues and that there would be a similar tendency for governments. Interestingly, the existence of such a pattern would likely be associated with a close congruence between governments and citizens if solely measured on a left–right scale, but low levels of congruence on specific policy domains.

If these patterns are indeed associated with inequalities in how governments reflect preferences of citizens from different societal groups, preferences of individuals with high income and educational status should also be better reflected than the preferences of citizens belonging to complementary groups. Given that income is associated with specific preferences on economic issues and education on cultural issues, we can expect poor citizens to be particularly incongruent with the preferences of their government on economic issues and those with relatively low education not to have their preferences taken into account on cultural issues. As both education and income tend to be associated with the better educated usually earning higher salaries, the poor are also likely to be under-represented on cultural issues and the relatively lower educated on economic issues, too. However, the magnitude of these gaps should be smaller than for the policies most closely associated with the individual characteristics considered. Thus, our two hypotheses regarding representational inequalities are the following:

Data and methods

Congruence has different conceptual dimensions. One dimension refers to the choice of representative agents as reference points. It has been studied with reference to parties or party systems (Ezrow and Xezonakis, Reference Ezrow and Xezonakis2011; Lefkofridi and Horvath, Reference Lefkofridi and Horvath2012), legislatures, governments (Stecker and Tausendpfund, Reference Stecker and Tausendpfund2016), or combinations of these (Golder and Stramski, Reference Golder and Stramski2010). Another dimension refers to the question of whether agents are modelled as unitary or collective reference points; or, to put it in the words of Golder and Stramski (Reference Golder and Stramski2010), whether we study one or many citizens and representatives respectively. This one-to-many dichotomy can be further refined, when the focus lies on specific groups of voters (e.g. Belchior and Freire, Reference Belchior and Freire2013).

These different options to conceptualize congruence lead to quite different results and conclusions. It is therefore important to make the choices that are appropriate for the research question at hand. With regard to the representative agent, we take the government as our reference point. This brings us as close as possible to an answer to the question of whether different groups (e.g. rich vs. poor) are equally well served by the outputs of a political system. With regard to the citizenry, we take on different perspectives, which we explain in detail in the following. Most importantly, we move beyond absolute values of distances as we are also interested in the direction of (in)congruence.

There are also different approaches when it comes to choosing the actual data with which congruence is measured. In most studies, congruence between citizens and parties or governments is measured based on how citizens place themselves on a scale and how they perceive parties to be positioned on that same scale. This can be easily done, for example, with data from the Comparative Study of Electoral Systems (e.g. Blais and Bodet, Reference Blais and Bodet2006). Unfortunately, such data are only available for the left–right axis, while information on issue specific preferences for citizens and parties are spread across different sources. In analysing congruence between citizens and governments on multiple issues across time we, hence, face an unfortunate data limitation. We dealt with this problem by combining different sources: measures of citizen preferences are taken from the EES (2014) and ESS (all six waves, 2006–2016), for party positions we refer to CHES (2006, 2010, and 2014).

These surveys present an interesting selection of position items that can be meaningfully matched onto each other. Moreover, CHES and ESS in particular provide a perspective on the multi-dimensional congruence between citizens and governments over a period of several election cycles in each of the countries. Matching these different sources raises various methodological problems. The wording between CHES and ESS varies as well as the scales for the issue specific questions (see Table 1). Hence, to what extent a position on the citizen scale matches a position on the government scale is uncertain. Assessing the direction of incongruences is, hence, on shaky ground. Instead of postponing this research until more congruent party and citizen surveys are available, we tried to optimally address these limitations. Most importantly, we use the European Election Study of 2014 as a bridging measure between ESS and CHES. The advantage of the EES is that it includes a question about citizens’ position with the same wording and scale as the CHES data measuring party positions. Although the match is less exact in relation to European integration, the EES provides a more fine-grained response scale than the ESS, which arguably reduces potential limitations linked with rescaling. Inspecting citizen preferences from both EES and ESS allows us to assess the direction of incongruences with greater confidence. In fact, for most countries (see Appendix in Supplementary material) the direction (though not the exact distances) of incongruences turn out to be the same for ESS and EES.Footnote ⁴

Table 1. Question wording of policy congruence variables

ESS = European Social Survey; EES = European Election Study; CHES = Chapel Hill Expert Survey.

Matching citizens’ self-placement with expert placements of parties also raises methodological problems beyond the exact scales and question wordings. Adams et al. (Reference Adams, Ezrow and Somer-Topcu2011), for example, show that voters are sensitive to their perceived ideological distance to parties, while voters’ perceptions are not necessarily in line with ‘objective’ measures such as scores generated from manifestos or expert surveys. Yet, this problem affects our research only marginally. First, we are not interested in congruence between citizens and parties but between citizens and governments. On government policy, voters are likely to be better informed than about mere manifesto rhetoric (Adams et al., Reference Adams, Ezrow and Somer-Topcu2011: 380). Second, as shown by Powell (Reference Powell2009: 1481–1482), the results of congruence studies do not change much when measures of party positions are derived from manifestos, expert surveys, or citizen placements.

We generate congruence measures in the following way. Citizen preferences are taken from different items of the ESS, while measures of party preferences are taken from the CHES (Bakker et al., Reference Bakker, de Vries, Edwards, Hooghe, Jolly, Marks, Polk, Rovny, Steenbergen and Vachudova2015). From the CHES we calculated government policy positions. To single-party majority governments we attribute the positions of the governing party. When it comes to coalition governments we take the standard approach of using the seat-weighted average of the positions of all cabinet parties (Warwick, Reference Warwick2001; Martin and Vanberg, Reference Martin and Vanberg2011; Curini et al., Reference Curini, Jou and Memoli2012: 249–250). Coalitions do not exclusively compromise but also engage in logrolls (Laver and Shepsle, Reference Laver and Shepsle1996) or invoke intra-coalitional vetoes. Nevertheless, the seat-weighted average seems to be a reasonable approximation of the actual policy output of a coalition (Warwick, Reference Warwick2001; McGann, Reference McGann2006; Martin and Vanberg, Reference Martin and Vanberg2014).

We calculate the policy proximity between citizens and governments on three items: general left/right, redistribution, and European integration. Congruence scores on the issue of redistribution, for example, were generated by using the question of whether the ‘government should take measures to reduce differences in income levels’ in the ESS and from the expert rating of parties according to their ‘position on redistribution of wealth from the rich to the poor’ in the CHES. We recoded all items into the same polarity and standardized all response categories on a 0–1 scale. Accordingly, individual congruence scores vary between 0 (perfect congruence) and −1/1 (perfect incongruence). Table 1 gives a detailed overview of the question wording and the scale of the different items we combined from ESS and CHES.

Given the different timelines of CHES, ESS, EES, and cabinets, we took various steps to provide for an optimal combination of our measures of government policy positions and citizen preferences (see Figure A1 in the Appendix of Supplementary material). In a first step, we assigned each cabinet party the position from the closest CHES wave (2006, 2010, or 2014). From these positions, we calculated government policy positions in the aforementioned way. In merging government positions with citizen preferences, we determined for each respondent in the ESS which cabinet she or he faced at the exact day of the interview. This day-exact matching is important as the actual ESS field periods vary considerably and are sometimes carried out astonishingly distant from the official year of an ESS round. The ParlGov infrastructure (Döring and Manow, Reference Döring and Manow2016) has been very helpful for the task of merging the two surveys. We use the 2006, 2008, 2010, 2012, 2014, and 2016 ESS rounds for our analysis.Footnote ⁵ We kept all countries that provided for complete information on government positionsFootnote ⁶ and that were part of at least seven of these eight ESS waves. Overall, we analysed 165,316 citizens from 16 countries (see Appendix in Supplementary material).

Empirical results

In a first step, we present descriptive results on multi-dimensional representation in Europe. We use graphics as this sheds light on several important things: the distribution of citizen preferences, the gap between the policy preferred by the majority of citizens and the stances of the government (the absolute distance between the median citizen and the position of the government), and the direction of this gap (whether citizens are to the left or the right of the government). Figure 1 gives a first nuanced picture on representation in four selected countries (a less busy presentation is given by plots for each country in the Appendix in Supplementary material). The violins summarize the distribution of citizen preferences in all six waves of the ESS on the three selected issue dimensions. A vertical line represents the citizen median, a hollow circle the mean of these distributions. The asterisk gives the respective mean of citizen preferences in the EES 2014. Triangles represent the mean preference for different subgroups – grey triangles for people with high and low income, white triangles for people with high and low education. High income and high education categories correspond to respondents belonging to the two highest deciles regarding these characteristics in their countries, whereas income and low education correspond to the two lowest deciles. Finally, crosses represent the position of all governments that were in place during the period of investigation. Various insights can be derived.

Figure 1. Distribution of citizen preferences and governments in four countries: Denmark, Germany, Poland, and Sweden.

First the shape of distributions varies considerably, while there are some typical issue-related patterns (see also Figure 2). On the left–right dimension citizen preferences come closest to a normal distribution with a peak in the middle of the scale. With a view to redistribution, preferences are clearly right-skewed while preferences on European integration are quite diverse across countries. The mean positions of different sub-groups correspond to our expectations: People with below-average income and education prefer more redistribution than people who are better off. The ordering is reversed when the same groups are asked about European integration. On the general left–right-axis no clear group-sorting emerges.

Figure 2. Multi-dimensional citizen–government congruence across countries and issues.

The positions of the government relative to the citizenry are quite variable. Denmark, for example, witnessed governments that were both to the left and right of the mean citizen on the issues of redistribution and on the general left–right axis. We note that the mean preferences of the EES correspond rather closely to the EES results with regard to redistribution and the general left–right axis. There are, however, quite striking disparities with regard to European integration.

This nuanced perspective is further condensed in Figure 2. The violins again sketch the preference distribution and grasp the location of the median citizen (middle line). The line running through the zero of the x-axis now corresponds to the mean position of all governments. Accordingly, when the median takes a negative value, the majority of citizens are to the left of the government and when it takes a positive value, the majority of citizens are to the right of the government.

The first conclusion we can draw from this country-specific perspective is that there are some substantial variations in the level of congruence across policy dimensions. As expected, the highest level of congruence exists on the left–right scale. In a majority of countries, left–right is the dimension on which the match between citizen and government preferences is the closest. In contrast, the more specific policy domains show a considerably higher degree of incongruence.

Regarding the direction of incongruence, we also find interesting issue-specific patterns. On the left–right scale, citizens are more often to the left of the government. In fact, only in France the pattern is inversed. However, the magnitude of citizens–government incongruence is lower than on the two more specific issues. On European integration, there is also a clear directional pattern. Citizens are on average much more Eurosceptical than governments. In all countries the gap between the median citizen position and the government position is substantial and goes in the same direction. Regarding redistribution, it appears that citizens are, on average, much more in favour of redistribution than their governments, with the exception of Denmark where the position of the median citizen matches with the government position.

This first overview already highlights that measuring policy congruence only on the left–right scale over-estimates the level of agreement between citizens and elites, which is in line with our first hypothesis. Furthermore, it suggests that moving beyond absolute congruence values provides interesting insights that would otherwise be blurred. Most importantly, it reveals that on issues like redistribution and European integration, the congruence gap and its direction is large and systematic. Citizens are clearly more to the left on redistribution and much less pro-EU integration. These findings are consistent with our hypotheses 2 and 3. These hypotheses were based on a representation bias argument which implied that wealthy and educated individuals are systematically better represented by governments on these specific issues. We test this argument in the following section.

Representational inequalities

One way to measure representational inequalities is to disentangle the citizenry into different categories and analyse whether the opinions of privileged social groups are more closely reflected in the stances of governments than the preferences of other groups. To assess how congruence varies by income and education, we move to an individual-level analysis in which we compute issue specific congruence measures for each respondent based on his or her proximity with the government position. We estimated regression models for each country and issue predicting the mean policy congruence of rich and poor as well as highly and low-educated citizens. Most importantly, we control for the distance of each citizen to the median voter on an issue. This reflects the notion that incongruences of certain societal groups are not problematic per se. If, for example, the views of the low educated are more distant to what the median voter thinks than the views of the highly educated, a better representation of the latter’s view might even be seen as normatively desirable. This becomes, however, a problem if we hold the distances to the median citizen constant and still representational biases prevail.

Furthermore, we also control for age, gender, and political interest. Political interest, in particular, may have an impact on how precisely people will situate themselves on the different issue scales (de Vries and Giger, Reference de Vries and Giger2014; Mayne and Hakhverdian, Reference Mayne and Hakhverdian2017). In the models for different income groups, we also control for education and vice versa (see Table A1). However, we do not include other explanatory variables such as government ideology or issue salience. To be sure, these variables are important, when we aim to explain the variance in congruence in an outcome-centred approach (Lax and Phillips, Reference Lax and Phillips2012). As our factor-centred approach wants to estimate the representation gaps between different societal groups, however, we should only include as statistical controls those variables that may influence this relationship (George and Bennett, Reference George and Bennett2005: 80). Of course, salience is an important variable when it comes to an overall assessment of incongruences. Normatively, incongruences are the more disconcerting the more salient a specific policy area is in a particular country (Reher, Reference Reher2015).

Figure 3 present the results (see the Appendix in Supplementary material for the country-specific regression tables). We contrast the regression coefficient for advantaged and disadvantaged groups in society, specifically looking at the two top deciles in terms of income and education vs. the two bottom deciles.Footnote ⁷ Circles represent the difference in the effects of top and bottom group identifiers in terms of income and diamonds present the same results in relation to education. When the differences between the two groups are significant at a p.<0,001 level, the symbols are black, when this difference is insignificant it is filled with grey. The centre of the x-axis, marked by a grey vertical line, corresponds to a situation when both effects are equal, i.e. advantaged and disadvantaged groups enjoy same levels of congruence with their government. When the symbols are to the left of the grey line, disadvantaged groups (i.e. low educated, low income) are more congruent with their government than their advantaged counterparts (i.e. high educated, high income). When the symbols are to the right, the opinions of advantaged groups are better reflected in the position of the governments than those of the corresponding disadvantaged groups.

Figure 3. Multi-dimensional citizen–government congruence separated by income and education.

To begin with, it appears that on all three congruence measures, a better representation of advantaged groups than disadvantaged groups is much more frequent than the opposite situation. The over-representation of high income and high educated is the least prevalent in the case of left–right ideology. It is still more frequent that the rich are significantly better represented than the poor in terms of left–right orientations, but the pattern in relation to education is much less clear cut. With regard to European integration, the better representation of advantaged groups is particularly visible. On that issue, it appears that the contrast between high and low education groups is particularly strong and the former enjoy higher levels of congruence. The pattern is similar in relation to income, but the magnitude of the differences between high- and low-income groups is on average smaller than for education. With regard to redistribution, advantaged groups also enjoy better representation than their counterparts much more frequently than the other way around. In contrast to European integration, the differences are generally greater between high- and low-income groups than between groups defined by education. There are, however, more exceptions to the generally described pattern with a better representation of low income in four countries (Bulgaria, France, Greece, and Poland).

Taken together, the results presented in Figure 3 support hypotheses 4 and 5. They show that inequalities in the representation of income and educational groups might contribute to explaining the direction of the overall gap between the stances of the median citizen and the position of governments in a given country. However, the magnitude of the difference between advantaged and disadvantaged groups is rather limited. Comparing the magnitude of these differences with overall levels of incongruence (see Figure 2), it is clear that systematic inequalities can only constitute a partial explanation for the high levels of incongruence found between governments and their citizens. Indeed, although the position of privileged groups is on average closer to the stances of governments, there is still a substantial gap: most governments exhibit less redistributive and more pro-European stances than even their privileged citizens. A potential explanation for this pattern could be found in the fact that governments face a trade-off between representing their citizens and acting responsibly (Mair, Reference Mair2009). Redistribution and European integration are precisely issues in which political elites are bound by economic factors or international agreements and might, for these reasons, take stances that are far away from the wishes of the public.