Government Redistribution and Public Opinion: A Matter of Contention or Consensus?

Abstract

Previous comparative research has been guided by the idea that the level of government redistribution accords with the degree of consensus on redistribution among citizens. By extending the scope of analysis to non-Western rich democracies, I offer an alternative account that associates public opinion with actual redistribution. I argue that it is not a broad consensus but a clearly formed contention among citizens that concurs with more redistributive governments. Using the International Social Survey Program (ISSP) 2016 data, this study compares social cleavages in redistributive preferences in 23 Organisation for Economic Co-operation and Development (OECD) countries. Countries with the least egalitarian governments, such as South Korea, Taiwan, Chile, and Israel, have broadly consented high-levels of support for redistribution. What distinguishes them from more redistributive countries is that those common redistributive cleavages such as income, education, and gender are either nonexistent or weak, indicating that the economically disadvantaged do not prefer redistribution significantly more than the advantaged. The statistical results support an explanation of the association between redistributive preferences and the size of redistribution based on “cleavage” rather than “consensus.”

Keywords:

• Government redistribution
• redistributive preferences
• social cleavages
• non-Western OECD countries
• ISSP

One of the main motivations that attracts researchers in redistributive studies to public opinion analysis comes from the expectation that citizens’ redistributive preferences can explain the size of government redistribution, significantly independent of party politics. That is, public opinion has an explanatory power in its own right for cross-national variation in redistributive efforts by governments (Brooks and Menza 2007). The underlying logic is that in democracies if people demand redistribution broadly, governments are to respond to these preferences regardless of their partisanship. Rising tides lift all boats. A broad consensus in support of redistribution can restrain right-wing governments’ pursuit of market-oriented policies to reduce taxes and social transfer benefits, while facilitating leftist governments’ egalitarian policies. Thus, the broader the support for redistribution, the larger the government redistribution between the rich and the poor (Brady and Bostic 2015; Huber and Stephens 2012; Rehm 2011). Once established, this positive association can be sustained by institutional feedback effects, as redistributive policies generate the policy constituents who have a stake in the welfare states (Pierson 1993, 2000).

Previous research motivated by this positive association has focused on the institutional conditions that can reduce the difference in preferences between the poor and the affluent, particularly the opposition by the latter who contribute more than they may benefit from redistribution. According to the existing literature, less disagreement in preferences renders redistribution a less conflictual policy issue between the affluent and the poor (and also between right-wing and leftist parties) and thus makes redistribution politically feasible. Research based on this “consensus logic” has examined either the aggregate support level for redistribution or the weakness of the key social cleavages in redistributive preferences (e.g., income) as the explanatory variable for the size of redistribution: the higher the aggregate support, the larger the size of redistribution. Without critical reflection on the different sociopolitical contexts in non-Western democracies, this positive association has been presupposed for all affluent democracies in numerous quantitative studies (Brady and Bostic 2015; Brooks and Menza 2007; Dallinger 2010; Huber and Stephens 2012).¹

Based on this studied association, previous research has sought to identify institutional arrangements that can maintain (or cultivate) a broad consensus on redistribution between the advantaged and the disadvantaged in the market distribution of income and risks (Brady and Bostic 2015; Cusack, Iversen, and Rehm 2006; Rueda 2018). According to these studies a broad consensus on redistribution is possible if the affluent consent to redistribution, while poor people’s support for redistribution is assumed given by their economic needs. For instance, a universalistic distribution of taxes and benefits will increase popular support for redistribution by making the affluent consent to redistribution more likely (Rothstein 1998) or rendering the critical cleavage for redistribution (such as income) substantially weaker (Beramendi and Rehm 2016). Given that the aggregate support for redistribution ranges between 50–90 percent of the population among rich democracies, the conditions that affect the preferences of the affluent have been the analytic focus.

By extending the scope of analysis to non-Western rich democracies, this study reexamined the consensus-based view on the association between the aggregate support for redistribution and the size of redistribution. Figure 1, a bivariate plot graph, presents the association for 23 affluent democracies, including non-Western countries. The widely assumed positive association no longer holds. Instead, the association reveals puzzling cases where high levels of support for redistribution correspond to the least egalitarian governments, such as Chile, Taiwan, Israel, and South Korea (hereafter Korea). According to the consensus-based view, the aggregate support for redistribution should be relatively lower in these countries, or income cleavage in preferences should be significantly strong (thus less consensus on redistribution). However, what recent research (Beramendi and Rehm 2016) found is exactly the opposite;² an exceptionally weak and insignificant income cleavage in Korea. In fact, the aggregate support in Korea and Taiwan is considerably higher than in most Western welfare states. In these two countries, almost everyone (78.5 and 82.5 percent of the population, respectively) agrees that government should reduce income differentials between the poor and the rich.³

FIGURE 1 Aggregate support for redistribution and the size of redistribution among 23 OCED countries in 2016.

Notes: Based on the author’s calculation of ISSP 2016 data for the aggregate preferences and OECD and LIS data for the inequality reduction by government (the difference of Gini coefficients between pre- and post-redistribution household incomes multiplied by 100). Country abbreviations: Chile (CL), Czech Republic (CZ), Denmark (DK), Finland (FI), France (FR), Germany (DE), Hungary (HU), Iceland (IS), Israel (IL), Japan (JP), Latvia (LV), Lithuania (LT), New Zealand (NZ), Norway (NO), Slovakia (SK), Slovenia (SI), South Korea (KR), Spain (ES), Sweden (SE), Switzerland (CH), Taiwan (TW), the United Kingdom (GB), and the United States (US).

According to previous research (Dallinger 2010), the level of economic development should explain much of the cross-country differences in the level of support for redistribution. Thus, if the level of economic development is controlled, the positive association between the support level and redistribution should hold. However, the four cases remain puzzling if we take into account the level of development. In 2016, Korea had a per capita income comparable to New Zealand and the Czech Republic (hereafter Czech), but Korea’s much higher level of support compared to the aforementioned two countries accords with an even smaller redistribution. Taiwan, Israel, and Chile pose a similar puzzle, showing much higher support but less redistribution than countries with comparable income levels (Germany, Czech, and Latvia).

By excluding these four deviant countries a weakly positive association between the support level and the size of redistribution seems to exist, but it only reveals a more general question about the explanatory power of the consensus-based approach. Dallinger (2010) claimed that once the level of economic development is controlled, institutional differences explain cross-country differences in the support for redistribution, which, in turn, accords with the size of redistribution. However, countries with a comparable income level and a similar welfare-regime type show remarkably different levels of support. Japan is clearly distinguished from the other East Asian countries. Likewise, Eastern European countries differ substantially: from Czech, with the lowest support level, to Slovenia, with the highest. Among social democratic welfare regimes, the support level is widely dispersed, from Denmark on the lowest side to Iceland on the highest. In fact, researchers have long been puzzled by the very low level of support in Denmark, one of the most redistributive countries.

Observing unexpected patterns in cross-national research, one way to respond is to question the quality of the data for those deviant cases from theoretical expectations and remove them from the analysis. The other strategy is to examine whether the lack of fit between theory and observations is rooted in the theory or its untested assumptions (Bechert and Edlund 2015). I chose the latter strategy since there is little ground to suspect the data quality for the deviant cases. Those four puzzling countries are regular participants in the International Social Survey Program (ISSP), which has a high scientific reputation among researchers in the field. In addition, 4 cases (out of 23) are not negligible to drop as outliers.

I use these deviant cases to reexamine the commonly assumed association in the study of redistributive politics. One theoretically plausible reason why broader support does not necessarily correspond to larger redistribution can be the problem of presupposing low-income earners’ preferences as given. The other reason can be the problem of assuming the same structure of redistributive cleavage in all countries. While scrutinizing these two assumptions of the consensus-based approach in the literature, I propose a cleavage-based approach to the association between redistributive preferences and the size of redistribution. I expect that the size of redistribution corresponds with the strength of cleavages rather than the degree of homogeneity in public opinion. To put it differently, it may not be a broad consensus on redistribution but rather a broad contention that accords with greater redistribution.

A CLEAVAGE-BASED ARGUMENT

Drawing on an alternative view on democratic process, I propose a cleavage-based argument regarding the conditions under which public opinion can affect the size of redistribution. The consensus-based argument sees a division in preferences (cleavage) as counterproductive to redistribution, and thus it seeks to identify the conditions to reduce the strength of the disagreement in preferences. In contrast, the contention-based argument sees a broad disagreement in public opinion as conducive to government’s pursuit of egalitarian policies. This is because it is dissent rather than consensus among citizens that reveals the problem as a social one and provides fertile ground for political mobilization to solve the problem (Korpi 1983; Piven and Cloward 1979; Schattschneider 1960; Solt 2008, 2010). Little or no contention on redistributive issues could mean that these issues have not been acknowledged (or discussed) as a social problem in need of government action. Thus, a clearly formed difference in public opinion on redistribution is a prerequisite rather than an obstacle to sustain (or increase) redistribution. Furthermore, whereas the consensus-based approach assumes the low-income individuals’ preferences as given, the poor (or the disadvantaged) may not necessarily demand government redistribution. They often acquiesce, attributing their economic plight to their own fault or misfortune (Gaventa 1980). If expressed, their policy preferences may not necessarily align with their economic interests (e.g., poor people’s opposition to the welfare state based on religious belief).

Through learning- and information-sharing processes on complex policy issues, unions and leftist parties can play a critical role in helping working class (and low-income earners’) preferences be closely aligned with their socioeconomic status (Iversen and Soskice 2015). Unless those who are disadvantaged demand redistribution consistently (and are politically mobilized as successfully as the opposition), governments may not respond to their demand. To study redistributive preferences beyond the scope of Western democracies, it is particularly important to acknowledge the potential variation of low-income earners’ preferences. In many non-Western democracies, leftist parties and trade unions have been under political repression and have remained weak, lacking an organizational mass base. In this context, whether low-income earners demand redistribution should be an empirical question rather than assumed.

Figure 2 illustrates two distinct ideal conditions under which public opinion induces greater redistribution: the graph on the left-hand side for the consensus-based approach and the graph on the right-hand side for the cleavage-based approach. For the former, the ideal condition for redistribution is the highest level of aggregate support, which is possible if the support among the affluent is as closely high as the support among the worse off. In contrast, the ideal condition for the latter is to have significantly higher support among the worse-off group than among the affluent group (or a significantly strong income cleavage in redistributive preferences). The aggregate support level does not need to be higher than a majority. For instance, if the aggregate support level is as low as 50 percent, government may sustain (or increase) redistributive policies if a majority of the people prefers redistribution strongly and coherently. It would help if the affluent oppose redistribution to a lesser degree, but strong demand by the worse off is the prerequisite (or a necessary condition) prior to getting consent from the affluent.

FIGURE 2 Two distinct approaches to redistributive preferences.

Notes: Three income groups are assumed to have a similar-size population. Redistributive preferences are scaled from 5 to 1, for which 5 indicates strong support for redistribution, 3 neutral, and 1 strong opposition to redistribution.

Regime- (Country-) Specific Redistributive Cleavages

In public opinion studies the contentious view on democratic process has guided a stream of research that focused on regime-specific redistributive cleavages in preferences. These studies have viewed the existence of social cleavages as an essential condition to sustain a redistributive government (Andreβ and Heien 2001; Svallfors 1997). They have found multiple social cleavages in redistributive preferences, such as social class, income, employment sector, age, and gender. The existence of multiple social cleavages suggests potential ground for crosscutting cleavages and redistributive coalitions among different social groups in support of redistribution (e.g., high-income women who may support redistribution). More important is the observation that these social cleavages have regime-specific or country-specific characteristics. The type and strength of these cleavages may differ across countries. For instance, support for redistribution among women is stronger in social democratic welfare states. As women participate more in paid labor, their preferences become in favor of generous social insurance programs and publicly provided care services (Svallfors 1997).

Later research has refuted these findings on regime-specific redistributive cleavages. In particular, Dallinger (2010: 337) boldly concluded that “it is difficult to find evidence for the existence of regime-specific cleavages. Instead, support for state redistribution is structured by cleavages according to gender, class and benefit dependency in equal measure in all countries.” Once refuted, recent literature on redistributive preferences has rarely focused on regime- (or country-) specific redistributive cleavages.

Extending the scope of analysis beyond a small set of Western welfare states makes it necessary to revisit the debates on the cleavage structure in redistributive preferences. In Dallinger’s (2010) study, “all” countries actually meant all Western democracies. Research on particular countries has continuously produced evidence of the cross-country variation in the effects of individual characteristics (social cleavages) on redistributive preferences. On the one hand, low-income earners may not necessarily demand government redistribution more than the affluent. For instance, Shayo (2009) showed that in a country like Israel, national identity has a stronger influence on low-income earners’ redistributive preferences than does their economic positions, due to security concerns and nationalism. Song (1994) claimed that in countries like Korea and Taiwan, the anticommunist authoritarian past has suppressed the formation of pro-redistributive preference among the poor. In addition, since female labor force participation is substantially lower in East Asian countries than in other affluent democracies, we can expect that women in these countries may not have the same redistributive preferences as those in Nordic countries.

On the other hand, those groups known as economically disadvantaged may not have the same economic position in non-Western countries. Public-sector employees may have different redistributive preferences, depending on whether they constitute a large share of the labor force or a small number of government bureaucrats. In East Asian countries, public-sector employment has been limited to privileged segments of employees who have relatively high wages and benefits as well as legally guaranteed lifetime tenure (Kwon 1997). It is difficult to expect that public-sector employees in East Asia have more critical interests in redistribution than those working in the private sector. Similarly, union membership may have different effects on redistributive preferences, depending on the position of organized labor in the society; that is, whether union membership mainly covers labor market insiders or extends to outsiders as well (Rueda 2005; Rueda and Pontusson 2000). In East Asia and Latin America (Bellin 2000; Jung 2002, 2006), union membership is limited to high-income earners in large-size firms and state-owned enterprises (i.e., labor market insiders who constitute less than 10 percent of the labor force). Under such conditions, union members tend to align with big businesses rather than with labor market outsiders (Yang 2013). These findings suggest that redistributive cleavages by income, gender, union membership, and employment sector may differ substantially across countries rather than being the same in equal measure in all countries.

In sum, I expect that the structure of redistributive cleavages rather than the aggregate support level can account for cross-country variation in redistribution. I expect that countries with the least redistributive governments have weakly formed redistributive cleavages, whereas countries with the largest redistribution have strongly formed redistributive cleavages. I also expect to make sense of the puzzling cases in previous research. Denmark’s high-level redistribution has to do with its strongly formed redistributive cleavages, whereas low-level redistribution in Korea, Taiwan, Israel, and Chile has to do with nonexistent or weakly formed cleavages in these countries, independent of their varying levels of aggregate support for redistribution.

In what follows I introduce the data and methods. In turn, I document redistributive cleavage structures for a broader set of affluent democracies. The analytic focus is to compare those non-Western democracies previously considered as outliers with Western democracies, drawing a common logic. Subsequently I report and discuss the statistical results on the validity of the cleavage-based argument vis-à-vis the consensus-based argument regarding the association between the structure of public opinion and the size of redistribution.

DATA, MEASUREMENT, AND METHOD

The main data source for this study is the International Social Survey Program (ISSP) 2016 Role of Government module (ISSP Research Group 2018). The ISSP 2016 provides information on individuals’ redistributive preferences along with socioeconomic and demographic characteristics. It includes 23 affluent democracies, a much broader and more diverse set of countries than covered by previous studies. To select affluent democracies I used the membership of the Organisation for Economic Co-operation and Development (OECD) because its membership criteria require a functioning democracy and a high-level per capita income from a global perspective. Among 34 OECD countries, previous research has mainly focused on 14 to 16 countries with long-standing membership. The cases in this study cover 22 OECD countries and Taiwan in the ISSP 2016: Chile, Czech, Denmark, Finland, France, Germany, Hungary, Iceland, Israel, Japan, Korea, Latvia, Lithuania, New Zealand, Norway, Slovakia, Spain, Sweden, Switzerland, the United Kingdom, and the United States. Among them, 11 countries in Western Europe and North America have been the usual subjects in previous studies, whereas the other 12 countries have been neglected due to various reasons. Except for Japan, non-Western countries such as Chile, Israel, and Korea have been rarely included in previous research. Some smaller countries have garnered relatively less attention, including New Zealand, Iceland, and the two Baltic countries. Although Taiwan certainly falls under the category of affluent democracy, the country has rarely been studied in comparative research, mainly due to difficulties in accessing comparable data (Taiwan data is not available from the OECD database).

I performed two statistical analyses. The first analysis was ordinary least square (OLS) regressions for each country to figure out country-specific cleavage structures in redistributive preferences. The subsequent regression analysis was undertaken to assess the validity of the contentious argument against the consensus argument regarding the association between redistributive preferences and the size of redistribution.

In the first analysis I regressed social group characteristics on individuals’ redistributive preferences for each country. The analytic focus of this country-by-country regression was to map out which social cleavages are significantly present in each country and how the structure of redistributive cleavages differs among countries.

The dependent variable was the responses to the question of whether government should or should not reduce the income differentials between the poor and the rich. I used this survey item following the lead of previous studies (Brady and Bostic 2015; Beramendi and Rehm 2016; Cusack, Iversen, and Rehm 2006; Dallinger 2010; Rehm 2011). Instead of recoding it as binary responses, I used the original 1–5 scale (5 = definitely should be, 4 = probably should be, 3 = cannot choose, 2 = probably should not be, 1 = definitely should not be). The aggregate support for redistribution is calculated by averaging the score of this measure.

I included a set of individual-level characteristics that have been studied as explanatory variables for redistributive preferences. These variables are relative income position (low, middle, and high), education level, gender, age, employment status (unemployed or not), employment sector (public or private), and union membership. I recoded these variables to assign the higher values to the studied disadvantaged positions (e.g., education level was recoded to span the range of 6, no formal education to 0, postgraduate degrees). By recoding them in reverse order the positive effects of these variables will indicate the positive preferences of the worse-off groups and the negative preferences of the well-off groups. I also included a set of individual characteristics as control variables, such as religiosity (frequency of religious activity), residence (urban to rural), and marital status. Although these variables are not directly associated with the beneficiary status of redistribution, previous research has found them to be influential factors for various reasons.

Income was measured as a categorical variable with three relative income groups: low, middle, and high income.⁴ Low income is income lower than half the mean income in the country, while middle income is between half and twice the mean income. High income is above twice the mean income. The reference group is those with high income. Thus, the coefficients of the low-income position indicate to what extent these low-income people prefer redistribution compared to the high-income people.

Education level was measured as a continuous variable from 6 (no formal education) to 0 (postgraduate degrees). Education level can be considered as a good proxy for social class in comparative research that includes non-European countries.⁵ Given that education level shapes occupation skills and lifetime income, it may affect redistributive preferences. The expectation is that the higher the education level, the less supportive the individual will be of redistribution.

According to previous studies, women, public-sector employees, union members, the elderly, and the unemployed are more supportive of redistribution than men, private-sector employees, nonunion members, the younger, and the employed (Andreß and Heien 2001; Dallinger 2010; Svallfors 1997). The gender variable was binary, in which female was coded as 1. Public-sector employment was coded as a binary variable as well: 1 for public-sector employees, 0 for all the others (private-sector employees, never in paid work, and do not know). For union membership current member was coded as 2, once a member as 1, and all the others as 0, which includes those who never had membership, refused to answer, or did not know about their membership status. Employment status was coded as binary, 1 for those unemployed who seek a job and 0 for all the others including the employed, those in housework, the retired, those in training, and the disabled. I measured age as a continuous variable: the higher the value, the older the age. Those who refused to answer and those aged over 100 were dropped.

Urban and rural residence was coded on a scale from 5, for big city, to 1, for rural farm. There exist mixed expectations for this variable. It is claimed that in developing countries, urban residents support redistribution more than rural residents because the former are working in industrial occupations and are exposed to conditions that form class consciousness, whereas the latter are often smallholder peasants (Haggard, Kaufman, and Long 2013). Yet the opposite expectation is plausible as well. In East Asian countries, industrialization processes driven by central planning have marginalized rural residents in terms of earnings and level of living. In Western democracies, deindustrialized areas (small- and medium-size cities) appear to have declining economic prospects and need redistribution, whereas large cities have reaped economic gains from globalization and the transition to a service economy.

For marital status, married or in partnership was coded as 3, separated as 2, and never in a civil partnership as 1. The expectation is that stable marital status has a negative association with redistributive preferences. For religiosity, the frequency of religious activity attendance was measured on a scale from 0 (never attend) to 7 (several times a week). The expectation is that the higher value has negative effects on redistributive preferences (Stegmueller et al. 2012).

In the second analysis I performed country-level OLS regressions to assess the association between the structure of public opinion and the size of the redistribution for 23 affluent democracies. The analytic focus was to compare the effects of redistributive cleavages on the size of redistribution with that of the aggregate support level.

The dependent variable was the size of redistribution, defined as the inequality reduction by government through taxes and social transfers. Following the convention in the literature, it was measured by the difference of Gini coefficients between disposable income and market income distribution. Except for Taiwan, the data for all the other countries were drawn from the OECD (OECD 2018). Taiwan’s data were calculated using the LIS database according to the OECD convention (LIS 2018), for which the household income is equivalized by household size. After the equivalization, Gini coefficients were calculated for both market and disposable incomes.

The main explanatory variable was the strength of country-specific redistributive cleavages, which is the coefficient of each social cleavage effect on redistributive preferences in the first regression analyses. The higher coefficient for a particular cleavage effect indicates stronger contention between the advantaged and the disadvantaged groups on redistribution. For instance, the higher coefficient for income effects represents stronger demand among low-income earners for redistribution compared to high-income earners. I measured the coefficients for seven categories: income, gender, union membership, employment sector, education level, age, and unemployment status.

The level of economic development was included as a control variable and measured by the per capita GDP, denominated by purchasing power parity–adjusted international dollars.⁶

The expectation is that the stronger the cleavage, the larger the redistribution. In addition, the analysis will reveal which cleavage corresponds to the size of redistribution more significantly than the other cleavages.

I chose to use this approach with two separate regressions instead of frequently used models in the literature, such as multilevel regressions or two-stage regression models. The main reason for this choice is that the primary research interest is not the (country-level) determinants of individual-level redistributive preferences but rather the association between the two aggregate-level variables, the size of the redistribution and the aggregate-level structure in public opinion.

COUNTRY-SPECIFIC REDISTRIBUTIVE CLEAVAGE STRUCTURE

In Table 1, I report the redistributive cleavage structure for eight countries, including the four outliers (Korea, Taiwan, Israel, and Chile) from the consensus-based approach (broad support but small redistribution) and the four countries that represent each type of Western welfare regime (the United States, Germany, Denmark, and Czech). The results for the remaining 15 countries are in Appendix A.

TABLE 1 Country-Specific Redistributive Cleavages in OECD Countries

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(9)
	TW	KR	IL	CL	US	DE	DK	CZ
Low income	0.100	−0.003	0.029	0.085	0.396***	0.305***	0.749***	−0.065
	(0.07)	(0.12)	(0.10)	(0.07)	(0.12)	(0.11)	(0.29)	(0.17)
Middle income	0.169***	0.053	−0.027	−0.093	−0.011	0.137*	0.485*	−0.099
	(0.06)	(0.11)	(0.07)	(0.06)	(0.11)	(0.08)	(0.26)	(0.08)
Public sector	0.003	0.151	0.106	0.110	0.081	0.064	0.385***	−0.059
	(0.08)	(0.13)	(0.07)	(0.09)	(0.10)	(0.06)	(0.10)	(0.09)
Gender	−0.043	0.084	0.045	−0.020	0.272***	0.070	0.208**	0.269***
	(0.05)	(0.08)	(0.06)	(0.06)	(0.08)	(0.06)	(0.10)	(0.08)
Union member	0.004	0.006	−0.151***	0.037	0.047	0.173***	0.147**	0.139**
	(0.03)	(0.07)	(0.04)	(0.06)	(0.12)	(0.04)	(0.07)	(0.07)
Age	−0.001	−0.005	−0.000	−0.004**	−0.009***	0.005***	0.016***	0.010***
	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)
Edu. level	−0.031	−0.022	0.029	0.010	0.045	0.117***	0.055	0.189***
	(0.02)	(0.03)	(0.02)	(0.02)	(0.03)	(0.03)	(0.04)	(0.03)
Unemployed	−0.046	0.101	0.249*	0.149	0.469**	0.263**	0.145	0.085
	(0.10)	(0.23)	(0.15)	(0.10)	(0.20)	(0.13)	(0.32)	(0.19)
Urban/rural	0.036	0.042	−0.006	0.025	0.011	0.021	0.008	0.086***
	(0.02)	(0.04)	(0.02)	(0.02)	(0.04)	(0.02)	(0.04)	(0.03)
Marital status	0.020	−0.174	0.026	−0.001	−0.120	−0.111	−0.374***	−0.175
	(0.07)	(0.12)	(0.10)	(0.07)	(0.10)	(0.07)	(0.13)	(0.12)
Religiosity	0.005	−0.019	−0.014	0.021*	−0.040**	−0.048***	−0.005	0.019
	(0.01)	(0.01)	(0.01)	(0.01)	(0.02)	(0.01)	(0.03)	(0.02)
_cons	3.833***	4.180***	4.354***	4.565***	3.782***	4.059***	2.295***	3.369***
	(0.17)	(0.26)	(0.20)	(0.15)	(0.28)	(0.18)	(0.41)	(0.24)
R^2	0.013	0.031	0.023	0.016	0.059	0.054	0.060	0.070
N	1951	1027	1134	1232	1367	1635	1019	1168

Note: Standard errors are in parentheses.

*p < 0.1, **p < 0.05, ***p < 0.01.

Clearly the effects of individual characteristics on redistributive preferences differ substantially among the 23 affluent democracies. The low-income individuals do not necessarily demand redistribution significantly more than do the high-income earners. Gender effects also vary from insignificantly negative to significantly positive. The variability of the effects across the countries is found in almost all independent variables. The size of the coefficient together with the significance level represents the strength of redistributive cleavages by individual characteristics (or social groups). The redistributive cleavage can be said to be stronger if the coefficient is significantly larger.

What distinguished the four outlier countries foremost from the others is the lack of (or the weakness of) significant redistributive cleavages between economically advantaged and disadvantaged groups. In Taiwan, Korea, Israel, and Chile redistributive cleavages are not significantly present in terms of income, educational level, gender, and public-sector employment. This indicates that in these countries low-income people do not prefer redistribution significantly more than do high-income people. Likewise, the less educated, women, and public-sector employees do not have more pro-egalitarian preferences than do the highly educated, private-sector employees as well as those who are not in paid work. Even the unemployed do not prefer redistribution significantly more than those employed or not in the labor market. Among those significantly formed cleavages, the effects are the opposite of those in Western democracies. Union members support redistribution less than do nonmembers in Israel, and the elderly are less pro-redistribution than are the young in Chile.

In contrast, in the four Western welfare regimes multiple redistributive cleavages are present in a stronger and significant way. In the United States, Germany, and Denmark it is clear that low-income people have a significantly greater preference for redistribution than do high-income earners. In Germany, more than 75 percent of the adult population supports redistribution (if coded as binary responses), but this does not mean that it has a homogenously popular preference for redistribution. Instead, redistributive preferences appear clearly differentiated between the proponents and the opponents by multiple social cleavages. In Czech, although there is no significant cleavage by income group, other multiple social cleavages are significantly present, such as gender, union membership, age, and education level.

The gender effects are significant in the United States, Denmark, and Czech, but not in Germany. This finding reaffirms the findings from previous research on regime-specific social cleavages (Svallfors 1997). In the former three countries, gender identity constitutes a significant social cleavage for redistribution. The effects of unemployment also differ, with significantly positive effects in the United States and Germany but not in Denmark and Czech. The age effects are significantly positive except for the United States. This finding is at odds with a well-known example of policy feedback effects. The US Social Security and Medicare programs have been cited as typical examples of feedback effects that make the benefit recipients of these programs (the elderly) support redistribution. The American elderly might not consider these programs as redistributive but rather as social insurance programs because they pay out what these individuals contributed to the programs in earlier years.

As for the remaining variables, the urban/rural cleavage is significantly formed in Czech, where urban residents are more supportive of redistribution than are rural residents (and also in the United Kingdom and less significantly in Sweden and Hungary). The opposite effects are found in France, Iceland, and Lithuania, where rural residents are more pro-redistribution (see Appendix A). The religiosity effects are inconsistent as well, ranging from significantly negative in the United States, Germany, Hungary, and Spain to significantly positive in Chile, Latvia, and Slovakia. The effects of marital status are negatively significant in six countries: Denmark, the United Kingdom, New Zealand, France, Norway, and Sweden.

From this comparison of redistributive cleavage structures, the puzzles from the consensus-based approach can be resolved. In the four deviant countries (Korea, Taiwan, Israel, and Chile) almost everyone supports redistribution (about 8 out of 10 people in binary responses), yet these countries have the least redistributive governments. However, the common characteristics of these countries are that low-income individuals (or women, less educated, the elderly, public-sector employees, and the unemployed) do not prefer redistribution significantly more than do the affluent (or men, highly educated, the young, private-sector employees, and the employed). This can provide a less conducive condition for the political mobilization and representation of their interests. In contrast, Denmark is on the other side of the puzzle, as it has one of the most redistributive governments despite the lowest level of aggregate support for redistribution. In Denmark, however, multiple social groups with disadvantaged positions prefer redistribution significantly more than do those advantaged groups. The coefficient of income effects is the largest among the 23 countries. In addition, public-sector employees, women, union members, the elderly, and those with less stable marital status clearly support redistribution more than their counterpart groups.

The Strength of Redistributive Cleavages and Government Redistribution

In this section I report the results from the aggregate-level regression analysis and assess the validity of the cleavage-based argument for government redistribution vis-à-vis the consensus-based argument. Model 1 in Table 2 shows the effects of the aggregate support for redistribution on the size of redistribution with the control for the level of economic development. As opposed to the expectation by the consensus-based argument, the effects of the aggregate support for redistribution are not only insignificant but also negative. Subsequently, Models 2–8 present the effects of the strength of redistributive cleavages by seven social groups. Among them, the strength of redistributive cleavage by educational level and age have the most significant and positive effects on the size of redistribution. Redistributive cleavages by union membership also have positive but less significant effects on redistribution. Regarding the explanatory power of each cleavage, the strength of the educational level cleavage accounts for 47 percent of the cross-country variation in the size of redistribution (R² = 0.468). The strength of the age cleavage accounts for 36 percent of the cross-country variation.

TABLE 2 Strength of Redistributive Cleavages and Government Redistribution

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Aggregate	−0.014
support level	(0.04)
Per-capita GDP	−0.000
	(0.00)
Income cleavage		0.028
		(0.05)
Gender cleavage			0.122
			(0.1)
Cleavage by				0.198*
union membership				(0.11)
Cleavage by					0.057
employment sector					(0.11)
Cleavage by						0.660***
educ. level						(0.15)
Age cleavage							4.867***
							(1.43)
Cleavage by								0.090
employment status								(0.07)
_cons	0.217	0.140***	0.134***	0.132***	0.141***	0.101***	0.130***	0.127***
	(0.17)	(0.02)	(0.02)	(0.01)	(0.02)	(0.01)	(0.01)	(0.02)
R^2	0.008	0.014	0.068	0.135	0.012	0.468	0.356	0.080
N	23	23	23	23	23	23	23	23

Note: Standard errors are in parentheses.

*p < 0.1, **p < 0.05, ***p < 0.01.

The effects by income group position and gender cleavages are positive but not significant. The insignificance of income cleavage effects is at odds with the expectation from previous research. The reason may have to do with the sample selection. For instance, in Beramendi and Rehm (2016) the analysis is restricted to working-age employed individuals while excluding the retired, students, the unemployed, and those in housework. These excluded individuals comprise about 47 percent of the adult population (namely, the voting population) in the 23 countries of this study. Given that the elderly and the unemployed are typical pro-redistribution groups, the income cleavage among the working-age employed may have limited explanatory power on the size of redistribution. It may also have to do with the income measure. Instead of using a continuous income variable, I used relative income positions—low, middle, and high—to measure the low-income people’s preference compared to that of high-income people. In addition, since education level shapes lifetime income, the cleavage effect by education level can be more influential than that of temporal income position.

In sum, the statistical results support the cleavage-based argument on the association between the structure of public opinion and government efforts for redistribution. My findings suggest that it is not the level of consensus on redistribution (or the weakness of redistributive cleavages) but the strength of redistributive cleavages that corresponds to the size of redistribution. Among the cleavages, the strength of the redistributive cleavage by education level accounts for the largest share of the cross-country variation in the size of redistribution in affluent democracies.

In Figure 3, I plot the association between the strength of redistributive cleavage by education level and the degree of inequality reduction by government. It shows a strong positive association between these two country-level variables. In countries where less-educated individuals prefer redistribution more strongly than do highly educated individuals (the stronger the cleavage), the larger the redistribution. Those outliers from the consensus-based approach (see Figure 1) are no longer deviant cases from the cleavage-based approach to redistributive preferences. Governments redistribute more in countries where the redistributive cleavage by education level is stronger.

FIGURE 3 Size of redistribution and strength of redistributive cleavage by education level.

CONCLUSION

The findings of this study challenge the widely presupposed assumption that consensus between the beneficiaries and the contributors of redistribution is the necessary condition for greater redistribution within affluent democracies. This study found that it is the strength of redistributive cleavages (or a broadly formed contention) not the weakness of redistributive cleavages (or a broad consensus) that accords with greater redistribution. In particular, the strength of cleavage by educational level and age are closely associated with government efforts for redistribution. Drawing on the contentious logic of democratic process, I argue that clearly formed cleavages in preferences between the proponents and the opponents of redistribution are conducive to redistribution rather than being obstacles to it.

I acknowledge that this study neither specified nor tested the causal directions, but only found the association between the strength of redistributive cleavage and the size of redistribution. As the previous literature has agreed, existing redistribution can affect the preference formation through the policy feedback loop, which makes it difficult to disentangle which comes first between the two variables. Finding the causal direction was not the main purpose of this study. Rather, it was limited to assessing an alternative argument on the association between the structure of public opinion and the size of redistribution.

The findings of this study reaffirm that redistributive references are not necessarily given by their economic needs (Katznelson and Weingast 2005), while highlighting the variability in the effects of certain individual characteristics on preference formation across countries and welfare regimes. As opposed to the consensus-based approach to redistributive conflicts (e.g., Beramendi and Rehm 2016), my findings suggest that a strong cleavage in preferences may not be counterproductive for redistribution. This alternative argument on the association between public opinion and redistribution may help understand the puzzle as to why countries like Korea, Taiwan, Israel, and Chile have the least redistributive governments despite absolute majority support for redistribution. In these countries, redistributive cleavages are weak and insignificant, indicating that redistributive preferences remain undifferentiated between the proponents and the opponents. In Denmark, although the aggregate support is as low as in the United States, multiple social groups in disadvantaged positions support redistribution significantly more than their counterpart groups.

In contrast to the four least redistributive countries, multiple significant cleavages are present in all Western OECD countries, including the Baltic countries. Take the example of Czech, Korea, and Israel, which all have comparable income levels. Although Czech has a substantially lower level of aggregate support than that in Korea and Israel, its government redistributes much more. What distinguishes the former from the latter two countries is clearly formed redistributive cleavages. In Czech, women, union members, the elderly, and the less educated prefer redistribution significantly more than do men, nonunion members, the young, and the highly educated.

One of the contributions this study attempted to make is the extension of an analytic scope to non-Western affluent democracies, which previous research has tended to neglect or has treated as outliers. This study used such countries that do not fit the existing theoretical framework as the lever to seek an alternative explanation. The cleavage-based argument is an attempt to find a common logic that applies to both mature and emerging welfare states.

This study leaves several issues to future research for further examination. One obvious possibility to develop is to analyze multiple waves of ISSP data and examine the causal relationship between the structure of redistributive cleavages and more (or less) redistributive efforts by government. In addition, research on the determinants of individual-level redistributive preferences can take a refreshed view on regime-specific redistributive cleavages.

ACKNOWLEDGMENTS

For valuable comments the author wish to thank the journal editor and reviewers as well as colleagues at the Swedish Institute for Social Research, Stockholm University.

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the author.

Additional information

Notes on contributors

Young-hwan Byun

Young-hwan Byun is a researcher at the Swedish Institute for Social Research (SOFI), Stockholm University. His main research interest lies in institutional analysis of welfare states, politics of redistribution, income distribution with a focus on the middle class, and social policy development in East Asia.

Notes

1 The positive association is assumed, with the condition that the level of economic development is controlled for.

2 Beramendi and Rehm (2016) employed a cleavage-based approach but presupposed the consensus-based argument that less divided preferences between the poor and the rich are conducive to larger redistribution. They focused on the income-based cleavage and found that the effects of income cleavage on redistributive preferences are greater if the tax system is more progressive and the benefit distribution is low-income targeting. Based on this finding, they suggested that stronger income cleavage would lead to less redistributive government. They treated the Korea case as an outlier without further examination.

3 The author’s calculation based on the ISSP 2016 data. The original responses were recoded as binary variables (agree or disagree), as in Brady and Bostic (2015). For the 23 OECD countries included in the ISSP 2016 data, the mean of the aggregate support is 74.2.

4 The income is household income equivalized by household size, following the OECD convention (divide the income by the square root of the household size).

5 In studying non-European countries, one of the major problems in using social class measures comes from a substantially large discrepancy between class consciousness and the objective class positions. This discrepancy can be attributed to different historical developments of class politics and democracy, such as political repression against the working class or a privileged position of unionized workers in formal sectors. In research on a broad scope of countries a more common strategy to measure social status is to use income and education level (see Brooks and Svallfors 2010; Brady and Bostic 2015).

6 The World Bank Database, “World Development Indicators: Exchange Rates and Prices” at http://wdi.worldbank.org/table/4.16#. Data on Taiwan’s GDP and PPP are from the International Monetary Fund’s Database, “World Economic Outlook: Implied PPP Conversion Rate” at https://www.imf.org/external/datamapper/PPPEX@WEO/OEMDC/ADVEC/WEOWORLD/THA.

 

APPENDIX A Country-Specific Redistributive Cleavages in OECD Countries

	JP	UK	NZ	FR	CH	FI	IS	NO	SE	HU	LV	LT	SK	SI	ES
Low income	0.318**	0.530***	0.600***	0.027	0.411***	0.459***	0.188	0.400***	0.326**	0.066	0.424***	−0.171	0.186	0.178*	0.215**
	(0.13)	(0.12)	(0.14)	(0.09)	(0.12)	(0.12)	(0.15)	(0.15)	(0.17)	(0.14)	(0.12)	(0.14)	(0.14)	(0.10)	(0.09)
Middle income	0.133	0.170*	0.110	−0.231***	0.109	0.171*	−0.027	0.049	−0.102	−0.044	0.140	−0.200**	0.108	0.203***	0.084
	(0.10)	(0.09)	(0.11)	(0.08)	(0.09)	(0.10)	(0.08)	(0.11)	(0.13)	(0.08)	(0.09)	(0.08)	(0.07)	(0.06)	(0.06)
Public sector	0.147	0.097	0.096	0.343***	0.089	0.127	0.075	0.146*	0.124	0.072	0.073	0.268***	0.090	0.043	−0.037
	(0.13)	(0.09)	(0.13)	(0.08)	(0.08)	(0.08)	(0.08)	(0.08)	(0.09)	(0.08)	(0.08)	(0.09)	(0.07)	(0.06)	(0.08)
Gender	0.057	0.143*	0.146*	−0.024	0.098	0.261***	0.206***	0.225***	0.254***	0.109	−0.024	−0.121	0.149**	0.104*	0.042
	(0.07)	(0.08)	(0.08)	(0.07)	(0.07)	(0.08)	(0.07)	(0.07)	(0.09)	(0.07)	(0.08)	(0.08)	(0.07)	(0.06)	(0.05)
Union member	0.073	0.109*	0.279***	0.115**	0.019	0.117**	0.102	0.184***	0.216***	v0.108*	0.040	0.100	0.030	0.117***	0.003
	(0.05)	(0.06)	(0.07)	(0.05)	(0.05)	(0.05)	(0.06)	(0.05)	(0.06)	(0.06)	(0.07)	(0.08)	(0.06)	(0.05)	(0.04)
Age	0.002	0.005*	−0.001	0.006**	−0.006**	0.008***	0.012***	0.012***	0.013***	0.002	0.004	0.004	0.005**	0.003	−0.001
	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)
Edu. level	0.127***	0.032	0.012	0.107***	0.117***	0.138***	0.123***	0.055**	0.081***	0.030	0.064**	0.073**	0.099***	0.103***	0.047***
	(0.04)	(0.03)	(0.02)	(0.02)	(0.03)	(0.03)	(0.03)	(0.03)	(0.03)	(0.03)	(0.03)	(0.03)	(0.03)	(0.02)	(0.02)
Unemployed	0.195	0.424**	0.160	0.445***	−0.042	0.238	0.418	0.259	0.494	0.207	0.175	−0.035	0.463***	0.256**	0.105
	(0.24)	(0.21)	(0.22)	(0.17)	(0.20)	(0.15)	(0.28)	(0.21)	(0.36)	(0.17)	(0.13)	(0.15)	(0.15)	(0.11)	(0.08)
Urban/rural	0.016	0.155***	0.038	−0.070**	0.022	0.016	−0.062*	−0.006	0.059*	0.047*	0.016	−0.117***	0.005	−0.005	0.015
	(0.04)	(0.04)	(0.03)	(0.03)	(0.03)	(0.03)	(0.04)	(0.03)	(0.03)	(0.03)	(0.03)	(0.03)	(0.03)	(0.03)	(0.02)
Marital status	−0.134	−0.261**	−0.375***	−0.220**	−0.137	0.023	−0.129	−0.173*	−0.325***	0.100	−0.048	−0.095	0.001	−0.015	−0.042
	(0.10)	(0.11)	(0.12)	(0.09)	(0.10)	(0.09)	(0.09)	(0.10)	(0.11)	(0.10)	(0.10)	(0.11)	(0.09)	(0.08)	(0.07)
Religiosity	0.000	−0.001	0.025	−0.014	0.015	0.007	−0.010	0.017	0.043	−0.061**	0.072***	0.019	0.026*	0.003	−0.024**
	(0.03)	(0.02)	(0.02)	(0.02)	(0.02)	(0.02)	(0.03)	(0.02)	(0.03)	(0.03)	(0.02)	(0.02)	(0.01)	(0.01)	(0.01)
_cons	3.710***	3.012***	3.625***	4.608***	4.194***	3.443***	4.172***	3.391***	3.028***	3.958***	3.732***	4.694***	3.913***	4.365***	4.458***
	(0.24)	(0.28)	(0.29)	(0.22)	(0.21)	(0.21)	(0.23)	(0.21)	(0.29)	(0.23)	(0.21)	(0.23)	(0.21)	(0.18)	(0.14)
R^2	0.020	0.054	0.056	0.073	0.048	0.072	0.066	0.070	0.080	0.021	0.045	0.053	0.059	0.069	0.018
N	1474	1154	1209	1383	1052	1139	1100	1160	1031	867	979	959	1041	1011	1757