Polarization Between the Rich and the Poor? The Dynamics and Structure of Redistributive Preferences in a Comparative Perspective

Abstract

Citizens’ support for redistribution varies largely between and within countries. An important empirical challenge in this field is the scarcity of comparative data, which this study overcomes by designing a novel time-series cross-sectional dataset that spans more than three decades in seven European countries. Using nearly 300 surveys and a dyadic ratios algorithm, we estimate aggregate redistributive preferences for each country, as well as for population strata within countries based on household income. We then ask to what extent support for redistribution varies across the rich and the poor. We find that citizens are not systematically becoming more reluctant toward or more supportive of redistribution. While redistributive preferences of the rich and the poor do not strictly move in parallel, there is no polarization between the two. Moreover, both the demand for redistribution and the preference gap between the rich and the poor evolve in a cyclical way.

In his foundational work Politics: Who gets what, when, [and] how, Lasswell (1936) addresses a series of issues that remain recurring questions in politics today. The notion of politics as the resolution of how resources will be distributed has sparked much political economy and public policy research related to inequality and redistribution. At the same time, this extensive scholarship has overlooked an important and equally necessary question, namely “who wants what and when.” Classic political economy models have underestimated the role citizens play in the inequality–redistribution puzzle, in spite of the extensive macro polity literature highlighting its importance (Erikson et al., 2002). As a direct consequence, the question of what people want in this domain needs further clarification.

Systematic insights into redistributive preferences are often hampered by the scarcity of available measures. Our study sets out to fill that gap by providing unique measures of redistributive preferences and make them available to the public. By means of a comprehensive time-series cross-sectional design, we model the extent to which these preferences evolve across European democracies. Furthermore, we explore variation within individual countries, i.e., whether the preferences of different segments of the population move in different ways. In particular, we analyze the way income influences the structure and evolution of redistributive preferences. By distinguishing between the preferences of the rich and the poor within each country, we assess whether they diverge or converge over time and which income stratum primarily drives the observed dynamics.

Drawing from a large number of survey data sources and using a dyadic ratios algorithm, we construct country-year measures of redistributive preferences for country populations and three different income quintiles (low, middle, and high) in seven European democracies: France (1995–2017), Germany (1984–2016), Great Britain (1981–2017), Netherlands (1978–2017), Norway (1989–2017), Sweden (1985–2017), and Switzerland (1993–2017). The underlying dataset is comprised of 880 country-year observations and relies on 299 surveys, 1,384 item series, and 8,244 administrations. Our findings provide key insights into the comparative macro polity. Although there is some shared variance, we find little evidence of a cross-national trend in redistributive preferences. Within each country, public opinion trends do not have a systematic trajectory, but are rather cyclical. In line with extant scholarship, we find that different income groups, and most notably the rich and the poor, demand different levels of redistribution. Yet, our results also indicate there is more independent movement (less parallelism) between the rich and the poor in Europe compared to what scholars find in the USA. The preference gap between the rich and the poor is rather stable over time, meaning there is no clear polarization or depolarization between the two.

As a whole, our in-depth comparative analysis of the dynamics and structure of redistributive preferences broadens our understanding of the macro polity, public opinion, and issue publics, most notably beyond the typical Anglo-Saxon scope in this field. Additionally, the preference measures and overarching dataset provide an important resource for scholars with an interest in mass representation and comparative political behavior, both at the individual and aggregate level. It can serve as a valuable tool for the discipline to gain additional insights into the relationship between inequality and redistribution, as well as the empirical study of democracy more generally.

Introducing redistributive preferences

Since the 1930s, public opinion scholarship has included an important conceptual debate about the nature of public opinion itself: What constitutes public opinion (and what does not)? Contrary to most “essentially contested concepts” in political science (Gallie, 1955), there appears to be quite some agreement in the overall conceptualization of public opinion (see Converse, 1987). In line with much of the literature, we therefore think of public opinion as the collective or aggregate of true or long-term individual policy preferences, attitudes, and beliefs regarding a harmonized or specific set of issues and questions (Key, 1961; Page & Shapiro, 1992; Stimson, 1991).¹

Such a macro-societal conception of public opinion has a number of important implications. Not only does it provide an aggregate reflection of the so-called “breeding ground” of political phenomena, but it also allows political actors to respond to societal evolutions and develop appropriate political strategies (e.g., Druckman & Jacobs, 2006; Stevenson, 2001). In turn, the matching of political supply with political demand can be seen as a core characteristic of representative democracy (Dahl, 1971; Key, 1961; Manin, 1995). With that in mind, an accurate and comprehensive understanding of public opinion also has relevant normative implications (Erikson et al., 2002; Stimson et al., 1995; Stimson & Wager, 2020).

While there have been important debates about the nature and movement of public opinion as a whole, recent academic work also examines public opinion in more disaggregated forms, tracking the preferences of different population strata within each country. Most notably, public opinion scholarship explores preferences related to inequality, wealth, and income distribution, and particularly how these may differ between (income) groups (Enns & Kellstedt, 2008; Enns & Wlezien, 2011; Gilens, 2005, 2009; Kelly & Enns, 2010; Page & Jacobs, 2009; Ura & Ellis, 2008). Most such studies agree that different societal groups—whether defined by income, education, sophistication, race, or gender—have distinct demands.

While social and cultural issues such as immigration increasingly dominate the agendas across advanced democracies, the economy—as a political issue—continues to play a prominent role in party competition (Williams et al., 2016), electoral outcomes (Singer, 2011), and policy priorities (Bevan & Jennings, 2014). This is unsurprising considering how governments have faced recurrently periods of economic and financial hardship throughout the last decades, most notably the energy crisis in the late 1970s, the global economic recession in the early 1980s, the economic downturn in the early 1990s, the financial crisis in the late 2000s, the European debt crisis in the 2010s, and the on-going health crisis related to COVID-19. In light of the ensuing inequalities and polarization, citizens—and the political arena more generally—remain attentive to economic questions and concerns regarding government interventions (Bellucci et al., 2012; Jennings & Wlezien, 2011).

The role of government remains largely a directional or spatial issue for which “reasonable people can take either side” (Stimson, 1991). Citizens position themselves alongside a continuum of dis/agreement with respect to government involvement and redistribution by demanding more or less of it. This is in line with earlier theorizations, for example, by Page and Shapiro (1992), who argue that increasingly available survey data allow us to capture aggregated public demands or sentiments that concern relatively specific and highly salient societal points of contention over time. With that in mind, we understand peoples’ opinions for redistribution, which we refer to as redistributive preferences, as a comprehensive construct that describes the evolution of a set of collective preferences or aggregated opinions regarding the levels and impact of redistribution in society.

Such domain-specific interpretations of public opinion, as opposed to more universal ones like the “policy mood” (Bartle et al., 2019, 2020; Erikson et al., 2002; McGann et al., 2019), are not uncommon. We refer, for example, to studies utilizing government spending preferences (Jacoby, 2000; Shapiro & Young, 1989; Wlezien, 2004), mass preferences in policy-specific domains (Soroka & Lim, 2003; Wlezien, 1995), or policy-specific moods (Coggins & Stimson, 2019; Romero-Vidal, 2020; Van Hauwaert & English, 2019; Cinalli & Van Hauwaert 2021). Extant research studying preferences regarding redistribution often remains on the individual level and singles out partial components of redistributive preferences, ranging from targeted indicators, such as the support for the introduction of a two-tier welfare system (Bay et al., 2013) to more general measures, such as attitudes toward welfare policies (Svallfors, 2012), support for redistribution (Gonthier, 2017), attitudes toward redistributive policies (Breznau & Hommerich, 2019; Fernández & Jaime-Castillo, 2018; Finseraas, 2009; Peters & Ensink, 2015) or, more holistically, economic conservatism (Caughey et al., 2019). Alternatively, some studies rely on a group of survey items to estimate latent measures of preferences regarding redistribution. Cavaillé and Trump (2015), for example, distinguish between support for redistribution from the rich and support for redistribution to the poor, while Roosma et al. (2013) identify seven different dimensions of redistribution.

Our operationalization of support for redistribution is more comprehensive. In this regard, we build on Lupu and Pontusson (2011), who study “aggregate support for redistributive policies.” Yet, while they measure preferences using a single survey item, we estimate latent preferences relying on a large number of survey questions. This allows us to account for the potentially multifaceted nature of our latent measure. Additionally, Lupu and Pontusson limit their analysis to the percentage of middle-income citizens who think that government is responsible for reducing income differences, whereas we also want to take the preferences of the first and fifth quintiles (rich and poor, respectively) into consideration.

Due to a lack of consistent data, even in countries with long-standing survey traditions, previous studies often remain limited in their temporal and geographic scopes. This leaves unexplored the nature and movement of the “median voter” and the citizenry as a whole. We add to this scholarship by designing comprehensive country-year measures of redistributive preferences for seven European democracies using a dyadic ratios algorithm. In what follows, we examine their structure and the patterns of variance between and within countries, comparing the aggregate opinions of different income groups. This will allow us to substantively assess whether redistributive preferences of the rich and poor become more or less polarized in the time period under analysis.

Data specifications: estimating redistributive preference measures

We rely on a unique dataset of redistributive opinion measures from seven countries: France (1995–2017), Germany (1984–2016), Great Britain (1981–2017), the Netherlands (1978–2017), Norway (1989–2017), Sweden (1985–2017), and Switzerland (1993–2017). With the exception of some longer series in the Netherlands and shorter series in Switzerland and France, they tend to go from the 1980s to 2017. The dataset includes an extensive selection of thematically grouped items, which have been asked on at least two separate occasions, from a wide range of high-quality national and international opinion surveys. We include all questions concerning positions toward the welfare state, social benefits, and public spending in fields that directly enhance redistributive policies (and therefore excluding defense, culture, and environmental protection spending). We also include items dealing directly with redistribution, taxation, and the role of the state in reducing inequalities, as well as items related to the desired degree of government involvement throughout economic and business functioning.²Table 1 includes an overview of the total number of questions (often across different surveys) and administrations in every country.

For each survey item, we compute the percentage of citizens wanting more redistribution and divide it by the cumulative percentage of those wanting more and those wanting less redistribution, while using each survey’s weight to account for design and sample biases. Higher values of the resulting “survey marginals” indicate relative preferences for more redistribution, while lower scores indicate preferences for less redistribution. It is important to note that absolute levels of redistributive preferences are less relevant here, since we seek to record and examine changes over time, which renders these values in and of themselves change variables. Marginals from questions repeated over time form independent item series.

The diverse frequencies of item series within a country complicate the identification of comparable trends. Some survey items we include have been asked twice, some more frequently. Yet, consistent item series that allow for comparisons through time and across space remain scarce given variations in question wordings, response categories, ordering of the survey, sampling methods, or the absence of data points following survey-specific or rotating modules. Yet, some examples of longer series in our countries inquire about the size of the public sector (Sweden, 31 administrations), the balance between tax reductions and social benefits (Great Britain, 31 administrations), the desirability of income difference (Switzerland, 24 administrations; Netherlands, 21 administrations), and general state intervention (France, 15 administrations).

Even if some individual series provide reasonable coverage, conclusions drawn from a single series are likely to be subject to several types of bias and measurement error (Diamantopoulos et al., 2012). Different item series might have different biases depending on the issues that it includes, the precise wording, and the response options it provided (Zaller & Feldman, 1992). While relying on a single item can provide important insights, it is paramount to note that it also restricts the analysis of public opinion in its classical sense, namely as a latent construct with a long-term equilibrium (Key, 1961; Page & Shapiro, 1992; Stimson, 1991).

Our starting point is, therefore, a set of item series for each country. In order to group them and get a single measurement of the latent support for redistribution in each of the seven countries, we employ a dyadic ratios algorithm (Stimson, 1991, 2018). This method is conventional in macro polity research (Bartle et al., 2011, 2019; Enns & Kellstedt, 2008; Kelly & Enns, 2010; Stimson et al., 1995), but much more limited across issue-specific applications of public opinion largely due to data restrictions. Notable exceptions use the algorithm to construct measures of presidential approval (Carlin et al., 2015a, 2015b, 2018), immigration opinions (Jennings, 2009; Van Hauwaert & English, 2019), Muslim-related issues (Cinalli & Van Hauwaert 2021, support for independence (Romero-Vidal, 2020), environmental concerns (Brulle et al., 2012), and gender equality (Koch & Thomsen, 2017), among others.

The dyadic ratios algorithm presupposes that each single-item time series can be considered a valid indicator of redistributive preferences, and therefore the change between any two values within that time series (a dyad ratio) is a relative indicator of change in redistributive preferences over time. Through an iterative process, the algorithm estimates the covariance between the dyadic ratios of each item. From this covariance, it then calculates validity measures for the different dyad ratio series and uses these to estimate the best possible latent measure. The dyad ratio series combined and adjusted according to their covariance compose redistributive preference values at each available user-defined interval (in our case, every year). In order to maintain accurate patterns of variance over time, we use these data-driven estimates without applying any further smoothing. In other words, we combine the independent single-item series and estimate a combined redistributive preference measure for each country.³

We highlighted earlier that different survey items tap into different components of the latent redistributive preference measure and, therefore, gauge it to various extents. In turn, that means—almost like in a factor analysis—each item series has a separate loading on the underlying measure, depending on how closely it relates to the latent construct. On average, the first dimension of the estimation captures nearly 60 percent of the overall variance (see Table 2).⁴ This supports our idea that redistributive preferences are to a large extent captured as a one-dimensional construct.⁵ We standardize the measures using country means because the absolute values are dependent on each country’s individual series and we are mostly interested in change over time. Figure 1 displays the corresponding measures of redistributive preferences in each country over time.

While scholars tend to argue there is a neo-liberal trend across advanced democracies (e.g., Schram, 2015), this is not something we see translated in the evolution of European publics, at least not to the same unambiguous extent. Rather, the defining feature of redistributive preferences is cross-country variability, without a clear-cut cross-national pattern. Redistributive preferences appear to have a cyclical character in all countries, alternating periods of higher and lower demand for redistribution. The standard deviations of the redistributive preference measures in Table 2 further illustrate the cyclical nature of these preference measures. Most notably, redistributive preferences remain relatively stable in France and Switzerland, while they move much more in the Netherlands and Germany.

Redistributive preferences covary—although not perfectly—over time and a trend in one country appears to be reflected in other countries. The average inter-item covariance of 0.10 between the seven measures indicates that, while most of their movement remains country-specific, they share some cross-country movement.⁶ We include the results from a principal factor analysis in the final two columns of Table 2. It further illustrates there is a shared consideration across most countries that accounts for more than a third of the cross-country variance (EV = 2.51).⁷ This suggests a certain degree of parallelism in redistributive preferences between countries.

A descriptive examination of redistributive preferences indicates some common movement across European democracies, which could be referred to as a “European redistribution mood.” Yet, there is important country-specific movement of redistributive preferences as well.⁸ Based on the uniqueness estimates in Table 2, about half of the variance remains country-specific on average. This does not only indicate national particularities—and, thus, independent movement, but it also suggests that the movement and variance of citizens’ redistributive preferences vary within countries.

Redistributive preferences across income groups

Within-country patterns of variance

Considering the extensive country-specific variance, it is difficult to speak of a clear and homogeneous cross-national trend. The question remains, however, whether the public preference structure can be explained by patterns of variance within national publics. In and of itself, this would not be surprising, as objective group interests, such as income, are likely to shape the redistributive preference structure (Enns & Wlezien, 2011). There is extensive evidence that lower-income groups typically favor more redistribution and welfare spending, while more affluent citizens prefer more market freedom (Erikson & Tedin, 2019; Kuhn, 2013; Luttig, 2013; Peters & Ensink, 2015). However, little is known about how the observed variance between income groups evolves through time.

Regarding the evolution of group opinion, the existing literature offers two contrasting views. A first perspective holds that group opinions, whether defined by income (Kelly & Enns, 2010), gender (Kellstedt et al., 2010), education (Enns & Kellstedt, 2008), or party identification (Ura & Ellis, 2012), track each other closely enough through time to speak of a uniform or parallel movement. While this is particularly prominent in the USA, to the best of our knowledge comparative evidence of parallelism remains scarce. An alternative perspective argues that, in addition to being different, group opinions also change differently through time, i.e., they do not move in parallel and are rather asymmetric. From a comparative perspective, Peters and Ensink (2015) find such different evolutions in support for redistribution between income groups (see also Brooks & Manza, 2006).⁹ If group opinions are truly evolving asymmetrically, the question remains whether they are diverging or converging over time. A conclusive answer eludes us at the moment, as the scarcity of cross-sectional time-series dampens the study of the potential polarization of redistributive preferences between the rich and the poor.

In order to estimate and examine the evolution of income group opinions, we divide country populations into income quintiles and focus on the top, middle, and bottom groups. For simplicity reasons, we refer to the former and latter groups as the rich and the poor, respectively. Since survey data usually offer income ranges instead of absolute values, the number of citizens in each group does not always represent an exact quintile. Usually, the group with lower income represents a higher share of the sample than the top quintile. We correct these errors by using the weight measure for each survey. For each income group, we repeat the estimation process detailed in the previous section: we calculate the marginals of our selected items and then estimate the central tendency of redistributive mass preferences for that income group through a dyadic ratios algorithm. The result is a single redistributive measure for each income group in every country.

Figure 2 displays these preference series for the low-, middle-, and high-income categories in each country. Since the absolute values are dependent on each country’s individual series, we standardize the measures using cross-income group country means. It is immediately evident there is a persistent preference gap between income groups. Higher-income groups generally demand less redistribution than their lower-income counterparts, which is in line with previous findings (Peters & Ensink, 2015).

Table 3 further validates that lower-income groups—on average—desire more redistribution (mean = 70.1) than middle-income groups (mean = 69.1) and their more affluent counterparts (mean = 62.4). This observation holds for most countries, with the notable exception of Scandinavian countries, where the middle groups demand more redistribution than lower-income groups. A set of paired t-tests in the last two columns of Table 3 further suggests that—in nearly all instances—the levels of support for redistribution between income groups differ significantly (at least, on average and through time). The t-tests also suggest that the level of redistributive support by the middle-income groups is closer to the level of the low-income groups than that of the high-income groups, with average cross-national differences of 3.6% and 19%, respectively. This is a similar observation to what Peters and Ensink (2015) found for a more limited time frame (2002–2010) in 25 European countries.

Researchers have shown that the redistributive preferences of the American middle class typically approach those of the rich, diverging from what some scholars tend to refer to as a pronounced “underclass” (Enns & Wlezien, 2011; Soroka & Wlezien, 2008). This illustrates the welfare gap in the USA, as well as the growing levels of inequality between the poor and other socio-economic classes. At the same time, our findings illustrate that patterns from the USA are not necessarily generalizable to other contexts. After all, we do not observe the same kind of evolution across Europe. Without exception, we find that middle-income groups across Europe tend to situate their redistributive preferences closer to those of the poor than those of the rich. Often times, it is not even close.

Table 3 further indicates that low-income groups hold relatively stable redistributive preferences, whereas demands for redistribution are more volatile among higher-income groups. This pattern, which holds across most countries, might appear counter-intuitive at first sight, seeing how Converse (1964) and Zaller (1992) imply greater stability in opinions for the more “sophisticated” strata. Yet, because of the common overlap between income and education, we can argue that the top quintile pays most attention to political and economic changes. In turn, this is why wealthier citizens are able to update their preferences more frequently, in reaction to both political and economic events (Enns & Kellstedt, 2008; Erikson et al., 2002; Ura & Ellis, 2008). Moreover, demand for redistribution among the rich might be conditioned to external factors such as a crisis or an election, whereas lower-income groups might demand more redistribution by default.

The movement of group opinions is not fully independent, as we observe in Figure 2 and Table 3. When support for redistribution increases for one income group, it tends to increase for the other income groups within a country as well. The average within-country inter-item correlation of 0.40 indicates that, while redistributive preferences covary, they do not move in perfect symmetry.¹⁰ This independent movement is nicely captured by the factor analysis of redistributive preferences across income groups presented in Table 4. Within each country, all series load on a single factor that accounts for an average of 58% of the variance across income groups. The uniqueness estimates indicate redistributive preferences exhibit a certain degree of parallel movement, but to a much lesser extent than earlier research typically shows.¹¹

Our analyses so far indicate common movement across Europe, shared movement between income groups and income group-specific movement. The question we, then, ask is whether certain income groups show more common movement than others. In that regard, Table 5 presents the results of a cross-national principal factor analysis of the preference measures for each income group. We notice that high-income groups share more variance than the other income groups, thereby indicating more common movement across countries. We also see this reflected in the more sizeable uniqueness measures for low- and middle-income groups. One might speculate that low-and middle-income groups are more influenced by national dynamics, whereas high-income groups are more permeable to international trends (and each other).

While the ensemble of analyses so far only presents indicative evidence, we do notice that the variance of our preference measures depends on the objective group interest (income), the individual country setting, and the temporal evolution. A more detailed analysis of variance (ANOVA) suggests that, on average, the preference measures are indeed characterized by differences between income groups (27.7%), across countries (10.2%), and through time (14.5%).¹²

Figure 3 plots the explained variance by income and time in each country. Following Enns and Wlezien’s (2011) suggestion that the variance explained by time provides a measure of parallel movements across income groups in the same unit, we find more parallel movement in France, Great Britain, and the Netherlands than in other countries. This is in line with earlier observations from Figure 2 and Table 3. Conversely, income groups in the Scandinavian countries explain the preference structure to a much greater extent than in other countries.

Polarization of redistributive preferences?

In addition to structural differences in levels, Figure 2 highlights that periods of convergence and divergence between income group opinions alternate. The combination of heterogeneous movement between income groups and the cyclical nature of the preference dynamics result in a fluctuating and contingent preference gap. To examine this gap and its potential persistence or even enlargement in more detail, we calculate the through-time difference in standardized redistributive preferences of the rich and the poor. Figure 4 plots the evolution of this distance, as well as the corresponding fitted line, providing original insights into the potential polarization dynamics between income groups.

The fitted lines indicate the extent to which the gap between the redistributive preferences of the rich and the poor follows a consistent trend. A negative slope would suggest that group preferences converge, whereas a positive slope would indicate increasing levels of polarization. Yet, as the overall slope coefficient approaches zero in all countries, it is safe to say we find little to no cross-national evidence of systematic polarization in redistributive preferences between the rich and the poor. The distance between their preferences is not constant or symmetric, but shows discontinuous periods of polarization with a long-term equilibrium. For example, in Germany we find a distinct period of polarization between the mid-1990s and early 2000s. This is then followed, as in every other instance, by a period of preference convergence.

Such heterogeneity further highlights that differences between income groups are irregular across countries. We have already alluded to this on various occasions by positing that not all movement is shared (see also Table 3). Yet, the preference gap itself does not reveal which population stratum is actually driving the observed short-term polarization periods. To tackle this, we analyze which income group moves further away from the previous equilibrium when the preference gap grows. In particular, we are interested in the relationship between both low- and high-income group preferences with those of the middle category.

Table 6 presents the results of different OLS regressions, with all variables standardized. The first three models report pooled estimates of the middle-income group preferences predicted by its own lag and different combinations of the low- and high-income group preferences. It is important to note these models are not meant to imply or assess causal relationships among the redistributive preference series. They are merely a tool to identify potential asymmetry in the movement. Models four to six estimate the first difference of middle-income group preferences expressed as a combined function of its own lag and the first differences of the lower- and higher-income group series. This ensures that inferences derived from the models are not the result of spurious relationships among non-differenced income group preference series.

The different models consistently illustrate that the redistributive preferences of middle-income groups are more strongly related to the preferences of low-income groups than high-income groups. Comparing the first two models, we notice the coefficient of low-income groups is about 20% larger than the coefficient of higher-income groups. This remains in the combined third model, but the difference between them is not statistically significant (p > .05). The fourth model draws a similar substantive conclusion: Conform to earlier indications, the association between low- and middle-income group preferences is considerably stronger than the association between middle- and high-income group preferences. Here, the difference between them is statistically significant (p ≤ .05). With the same interpretation from the differenced model, we can be sure that our cross-national findings are not sheer artifacts of temporal dynamics in the raw data.

It is clear from Figure 4 that the difference in redistributive preferences of high- and low-income groups is most dynamic in Germany (sd = 1.165). In other words, the movement of the preference gap indicates that polarization in Germany fluctuates quite a bit through time. With that in mind, we address two additional questions. First, we wonder whether the initial findings also hold if we examine Germany independently. Model 5 indicates this is effectively the case. Even more, as the coefficient for high-income groups is not significant in Germany, we find clear evidence that redistributive preferences of middle-income groups move together with the preferences of low-income groups. Second, we also examine if Germany might have been driving the earlier pooled results from models 1–4. Model 6 endorses and strengthens the earlier findings, suggesting there is no significant association between changes in preferences among the high- and middle-income group preferences when we exclude Germany from the sample. Overall, this provides additional credence to our substantive takeaways.

Altogether, we find clear evidence of asymmetry in the underlying dynamics of redistributive preferences. That is, the cyclical patterns of more or less polarization are not the result of symmetric changes between the redistributive preferences of low- and high-income groups. Rather, our findings suggest that middle-income group preferences are more strongly associated with low-income group preferences, while high-income group preferences exhibit more idiosyncratic movement. As a result, the movement of income group preferences may be thought of as an asymmetrical process with separate explanatory variables. Even though we do not find any evidence of long-term polarization in redistributive preferences, our analysis reveal the short-term asymmetric movement between income groups.

Conclusion

The interdependence between the socio-economic environment and political outputs such as public policies is part of the empirical and normative framework of contemporary democracies. According to Easton (1965), citizens and their preferences (public opinion) provide the necessary link between the two. Applied to the inequality–redistribution puzzle, this places citizens’ redistributive preferences at the center of the analysis. Yet, the role of citizens and their preferences for redistribution in this equation remains underdeveloped. We believe that the primary reason is a lack of data and corresponding measures enabling time-series cross-sectional analyses. This study aims to provide a solution to this problem. Drawing from an unprecedented number of public opinion surveys, we design a set of redistributive preference measures that cover an average of about three decades across seven European democracies. Through a dyadic ratios algorithm, we estimate measures for the population as a whole, as well as population strata based on income. The resulting measures enable new insights into when certain segments of the population update their demands for more or less redistribution. Moreover, its time-series cross-sectional character allows us to assess the extent to which we can speak of trends within, between, and beyond countries. Our measures offer a relevant and publicly available resource for scholars interested in the evolution of redistributive preferences and, more generally, the dynamics of public opinion across Europe.

Our analysis takes a three-folded approach. First, we explore the dynamics of national redistributive preference measures. On average, we find no evidence that European citizens are systematically becoming either more reluctant toward or more supportive of redistribution. Put differently, there is no single cross-national and longitudinal trend toward more (or less) demand for redistribution. While we observe some common movement among European publics, much of the observed movement remains unique to the country. This leaves considerable variance to be explained by within-country dynamics. Nonetheless, our analysis suggests that around half of the variation of redistributive preferences is shared across countries.

Second, we delve into the disparities in both levels and movement of redistributive preferences of low-, middle-, and high-income groups. Regarding preference levels, the data reveal a persistent gap between income group opinions. We find that, as expected, low-income citizens consistently demand more redistribution than their middle- and high-income counterparts. The distance between the groups is not constant but fluctuates through time in a non-systematic but cyclical way. The preferences of the middle group are closer to those of the low-income group across Europe. Regarding the movement of group preferences over time, we observe more variation in the preferences of high-income groups. At the same time, while income groups within and across countries share some movement, we find that considerable movement remains unique to each income group.

In a third and final step, we explore the fluctuating and contingent preference gap between low- and high-income groups. The evolution of this gap over time shows no evidence of any systematic trend toward polarization between the rich and the poor. At the same time, while examining the short-term cyclical nature of the preference structure, we show that patterns of convergence and divergence are not the result of symmetric change between the redistributive preferences of the rich and the poor. Whereas American scholarship typically shows that the redistributive preferences of the middle class resemble those of the rich, our analysis identifies a strong association between low- and middle-income preferences, both in levels and changes. Consequently, high-income citizens support redistribution considerably less, although their preferences are more volatile. Given the idiosyncratic movement of their preferences, our evidence suggests that the rich drive and maintain the preferences gap.

Our study presents an unprecedented analysis of redistributive preferences across European democracies, as well as how they differ between income groups. It starts to fill an important gap in various literatures, ranging from aggregate public opinion and comparative political behavior in Europe to empirical democracy and comparative political economy studies more generally. Yet, perhaps even more importantly, our study illustrates that research and findings from the USA should not be automatically exported to other (European) democracies. Particularly when it comes to public opinion, there is something to be said about “American exceptionalism.” The extensive parallelism of preferences that notable scholars in the field find in the USA, or Anglo-Saxon countries more generally, is not something we see translated to mainland Europe. While we provide a first analysis and observation, the question of why this is the case remains. We speculate this might be related to the differences in electoral systems, but it might also relate to Anglo-Saxon countries being liberal-market economies where large segments of the middle class simply “go private,” eroding the social contract and therefore aligning much more closely with the rich.

The novel dataset we present in this study allows for the comprehensive analysis of redistributive preferences as a dependent, independent, or control variable. Our dataset can facilitate further analysis of various components of a broader comparative macro polity and may be used to study a wide variety of research questions: What shapes redistributive preferences? Which group interests are accounted for in politics and redistributive policy more specifically? What does income group heterogeneity mean for the inequality–redistribution puzzle? Our data allow researchers to provide new insights into these questions and to obtain a broader and deeper understanding of the dynamics of public preferences and their interplay with the political and economic context.

Supplementary Data

Supplementary Data are available at https://doi.org/10.7910/DVN/1XRAYZ and IJPOR online. Replication data are available at https://doi.org/10.7910/DVN/TBPW4Z.

Footnotes

Acknowledgements

We would like to thank four anonymous reviewer and the journal editors for their extensive comments, as well as their patience with our requests for additional time. We appreciate both and take neither for granted. We would like to express our gratitude to John Bartle, Ryan Carlin, Sebastian Dellepiane-Avellaneda, Juan J. Fernández, Tim Hellweg, Greg Love and Christian Welzel for valuable input on various versions of this manuscript (and its spin-offs). This paper was part of a research project we presented in the “Macro Opinion in Comparative Perspective” workshop at the 2018 ECPR Joint Sessions. We thank John Bartle and Tim Hellwig for bringing such a great group of scholars together, as well as all participants for their valuable feedback, comments and suggestions.

Conflicts of interest: None declared.

References

Steven M. Van Hauwaert PhD is an Associate Professor (Senior Lecturer) in Comparative Politics at the University of Surrey and the principal investigator of the Global Public Opinions Project (GPOP). His primary research interests are in the fields of comparative political behavior and public opinion, as well as populism and other challenges to democracy. His most recent academic contributions have appeared in, among others, the European Journal of Political Research, the European Political Science Review, Politics, and West European Politics. He is also an associate editor of the ECPR journal Political Research Exchange (PRX) and the Methods and Measurement section of Frontiers in Political Science.

XavierRomero-Vidal PhD is a postdoctoral Research Associate at the University of Cambridge. He holds a PhD in political science from the Leuphana University of Lüneburg. His main research area is the study of the evolution of public opinion and political behaviour from a comparative perspective.