1. Introduction
Income inequality has risen in many democracies over the last several decades, but these states have rarely responded with tax and transfer policies that significantly counter this increase. The lack of a policy response to rising inequality has two broad interpretations. The first is that voters prefer more redistribution but that democracy is failing to deliver the policies that voters want, either because institutions fail to represent them (e.g., money counts for more than votes) or because governments are constrained in meeting voter demands by considerations of policy efficiency or global competitiveness. The second is that voters simply do not demand more redistributive tax and transfer policies.
This second account, however, immediately raises the question of why preferences for redistribution are not increasing in the presence of rising inequality. The existing literature has emphasized a range of possible answers to this question. Voters may be concerned about the efficiency costs of redistribution (Durante et al., Reference Durante, Putterman and van der Weele2014; Stantcheva, Reference Stantcheva2020). They may hold beliefs about (a) the fairness of economic outcomes (McCall and Kenworthy, Reference McCall and Kenworthy2009; Page and Jacobs, Reference Page and Jacobs2009; Alesina and Giuliano, Reference Alesina, Giuliano, Bisin and Benhabib2011; Cavaille and Trump, Reference Cavaille and Trump2015; Dimick et al., Reference Dimick, Rueda and Stegmueller2016; Bechtel et al., Reference Bechtel, Liesch and Scheve2018; Roth and Wohlfart, Reference Roth and Wohlfart2018; Kerschbamer and Müller, Reference Kerschbamer and Müller2020; Stantcheva, Reference Stantcheva2020), (b) the deservingness of the rich and the poor (Piketty, Reference Piketty1995; Gilens, Reference Gilens1999; Fong, Reference Fong2001; Alesina and Glaeser, Reference Alesina and Glaeser2004; Krawczyk, Reference Krawczyk2010; Durante et al., Reference Durante, Putterman and van der Weele2014; Ballard-Rosa et al., Reference Ballard-Rosa, Martin and Scheve2017; McCall et al., Reference McCall, Burk, Laperrière and Richeson2017; Stantcheva, Reference Stantcheva2020), or (c) the fairness of state policies (Scheve and Stasavage, Reference Scheve and Stasavage2016, Reference Scheve and Stasavage2022; Limberg, Reference Limberg2020). Any of these factors may constrain support for redistribution despite high and increasing inequality.
Another line of research, however, suggests a much simpler answer to the question: voters lack information about inequality, either in the aggregate or with respect to their own standing in the distribution (Bartels, Reference Bartels2008; Cruces et al., Reference Cruces, Perez-Truglia and Tetaz2013; Kuziemko et al., Reference Kuziemko, Norton, Saez and Stantcheva2015; McCall et al., Reference McCall, Burk, Laperrière and Richeson2017; Boudreau and MacKenzie, Reference Boudreau and MacKenzie2018; Fernández-Albertos and Kuo, Reference Fernández-Albertos and Kuo2018; Gimpelson and Triesman, Reference Gimpelson and Triesman2018). As a result, their preferences are formed based on incomplete information and do not reflect the increase in inequality. Our article aims to address the question of whether redistributive preferences differ from status quo policies and, if not, whether incomplete information about inequality explains why.
Our study focuses on redistribution through the income tax system and examines preferences in Austria and Germany. We measure preferences for marginal tax rates on different income brackets, combine these stated preferences into a single measure of progressivity, and experimentally investigate the sensitivity of these preferences to different information about inequality. We fielded original surveys of 2,100 adults representative of the population in Germany and Austria. Our main strategy for measuring multidimensional tax preferences employs menu estimates in which respondents chose their preferred marginal tax rate for each existing income tax bracket. We then combined these preferred rates, across the income distribution, into a single measure of progressivity preferences by calculating the implied Kakwani index (Kakwani, Reference Nanak1977).
Our estimates suggest that the mass public in both countries prefers tax policies that deviate to some extent from current ones. They would prefer more progressive taxes in Austria, though mostly through lower taxes on low and middle incomes, and—perhaps surprisingly—somewhat less progressive taxes in Germany. Taken together, neither set of estimates is consistent with the idea that there is significant unmet demand for dramatically more progressive tax policies on the whole.
This result raises the question of why, in Austria and Germany, preferences do not favor greater progressivity. We investigate whether a lack of information about inequality provides a compelling answer. We randomly assigned respondents to four treatment groups and a control group. The first two treatment groups provided respondents with information about aggregate inequality in the form of the 99/50 and 99/10 income ratios. The other two treatment groups elicited the respondent’s income and then reported the percentage of income earners who earned more, or who earned less, than they did, which we term positional inequality information. When considering the effect of these informational treatments on respondent preferences for income tax progressivity, in the German case, we find that exposure to information about the fraction of the German population that earned less than the respondent served to increase progressive preferences and additionally find that information about earnings of individuals at the 99th relative to the 50th percentiles also increased progressivity of income tax preferences. However, while these findings suggest some role for inequality information in affecting tax preferences, we find no similar effect in Austria.
This note makes two main contributions. The first is to measure the multiple-dimensional tax preferences of representative samples of citizens in Austria and Germany. We find that respondents have progressive tax preferences but ones that are at most modestly more progressive than current policies. This finding is important for the literature on the political economy of redistribution, as it suggests that weak preferences for redistribution through the tax system are likely one reason that tax policy has not responded to higher inequality in countries like Germany. This is an important corrective to the literature which has often emphasized special interests or some other failure of representation to explain the lack of a policy response to rising inequality (e.g., Bartels, Reference Bartels2008).
The second contribution of the article is to provide new evidence on whether lack of information about inequality helps to explain why tax preferences are not more progressive. Existing studies examining the effect of information about inequality on policy preferences report contrasting results. A number of correlational studies emphasize that misperceptions about inequality explain the disconnect between inequality and support for redistributive policies (see, e.g., Bartels, Reference Bartels2008; Gimpelson and Triesman, Reference Gimpelson and Triesman2018). In experimental work, Stantcheva Reference Stantcheva(2020) reports that providing U.S. respondents with information of the income distribution increases preferences for tax progressivity.
Several additional studies provide experimental evidence that inequality information influences policy preferences over taxes, but only for various subpopulations of respondent types or certain country samples (Cruces et al., Reference Cruces, Perez-Truglia and Tetaz2013; Karadja et al., Reference Karadja, Mollerstrom and Seim2017; Boudreau and MacKenzie, Reference Boudreau and MacKenzie2018; Fernández-Albertos and Kuo, Reference Fernández-Albertos and Kuo2018; Bublitz, Reference Bublitz2020). In contrast, Kuziemko et al. Reference Kuziemko, Norton, Saez and Stantcheva(2015) experimentally manipulate information about inequality in the United States and find that the treatment increases concern about inequality but only slightly influences policy preferences over taxes and transfers. In sum, the literature appears mixed on whether, and under what conditions, information about inequality is likely to matter politically, a point emphasized in a recent meta-analysis of studies on the redistributive effects of providing information about economic inequality (Ciani et al., Reference Ciani, Fréget and Manfredi2021).
Our article is closest to Kuziemko et al. Reference Kuziemko, Norton, Saez and Stantcheva(2015), but we study Germany and Austria, experimentally vary information about both aggregate inequality and the respondent’s relative position in the income distribution, and adopt a measurement strategy that allows us to study the multidimensional characteristics of the tax system as a concrete policy instrument for redistribution.Footnote 1 Our results strike somewhat of a middle-ground between the two existing camps. On the one hand, we do recover evidence that information about inequality—the 99/50 income ratio, and information about how many individuals earn less than a respondent—increases preferences for tax progressivity in Germany. However, we also note that not all forms of information about inequality have such an effect, nor do we find treatment effects in the Austrian sample.
The findings from these studies are starting to accumulate. There is little evidence of a large average treatment effect of information on redistributive policy opinions (Ciani et al., Reference Ciani, Fréget and Manfredi2021). The studies that do find effects of any magnitude typically find them only for a subset of voters. These results are important for understanding differences between objective and perceived inequality (Hvidberg et al., Reference Hvidberg, Kreiner and Stantcheva2023). But for the larger question of whether societies would choose substantially greater amounts of redistribution if only everyone knew the actual extent of inequality, the answer that is emerging is a negative one.
2. Public opinion about tax progressivity
Tax progressivity is fundamentally multidimensional in that it is a function of tax rates applied on different levels of income and on consumption. Citizens in modern states are subject to an array of different types of taxes that together determine the overall progressivity of the tax system. The objective of this section is to measure public preferences over tax progressivity in Austria and Germany. Our main approach employs menu estimates and asks survey respondents to state their preferred tax rate on each dimension. We then combine those preferences about rates into a single Kakwani index measure.Footnote 2
2.1. Data and methods
We fielded nationally representative surveys in Austria and Germany.Footnote 3 Both studies were conducted by YouGov over the internet.Footnote 4 The Austrian survey was conducted in October 2018 with a sample size of 2,100 adults. The German survey was fielded in March 2019 and also had a sample size of 2,100. We implemented the menu method by presenting respondents with sliders with which they could indicate preferred tax rates for each of the eight (Austria) or five (Germany) existing income tax brackets. The sliders were set at the status quo tax rates for each bracket.Footnote 5
2.2. Menu estimates
Figure 1 reports our menu estimates for Austria and Germany. For each tax bracket, we report box plots that indicate the current actual rate, the median preferred rate, and the interquartile range. Generally, the figure suggests that, as estimated by the median respondent, Austrians prefer substantially lower taxes on below-median incomes, and Germans prefer slightly higher taxes on very high incomes.Footnote 6 At the same time, preferred tax rates within each bracket vary noticeably among respondents. As a result, the status quo rates almost always fall within the interquartile range of tax preferences.
Figure 1. Menu estimates of preferred tax rates in Austria and Germany.
However, a critique of this analysis of the menu estimates is that we focus our attention on the estimates for each income bracket individually but do not have a formal strategy for assessing whether overall tax preferences are more or less progressive. To address this issue, we used the menu estimates for income taxes to construct the Kakwani progressivity index (Kakwani, Reference Nanak1977), P, which measures overall tax progressivity (Kakwani, Reference Nanak1977). The Kakwani index can be defined as:
\begin{equation*} P = (G-G^*) \frac{1-t}{t} \end{equation*} where G and 
$G^*$ are the pre- and post-tax Gini indices, respectively, and t is the average tax rate. The Kakwani index can equivalently be computed as twice the area between the Lorenz curve for incomes and the concentration curve for taxes.Footnote 7 Figure 2 illustrates the Kakwani index further. In it, twice the yellow part equals the Kakwani index, and twice the area between the Lorenz curve and the green 45∘ line equals the Gini index.
Figure 2. Illustration of Kakwani index.
The index measures the extent to which tax burdens—or, in our case, tax burdens under respondents’ preferred rates elicited by the menu method—increase with rising incomes in Austria and Germany; it assigns a higher score to more progressive tax systems and a lower score to those that are less progressive. The Kakwani indices based on current and median preferred rates are 0.207 (Austria, current), 0.221 (Austria, median preferred), 0.180 (Germany, current), and 0.168 (Germany, median preferred). Strikingly, in Austria, respondents prefer somewhat more progressive tax systems, and in Germany, respondents prefer somewhat less progressive tax systems—a fact that can be inferred from the analysis of individual rates but is crystallized by the Kakwani index.
3. Inequality information and tax opinions
One frequently advanced reason for why voters have not increased their demand for a more progressive tax schedule is that they underestimate the extent of rising inequality in society. They might, for example, be rationally inattentive and form their tax preferences based on the information to which they are exposed, which does not accurately reflect society-wide inequality.
In this section, we evaluate arguments about the effects of differential forms of information about inequality by randomizing survey participants in Germany and Austria into four different informational treatments that provide current statistics about the pre-tax distribution of individual incomes in their respective countries. We manipulated two types of information. Treatment arms A.1 and A.2 show aggregate data about the income distribution, whereas treatment arms B.1 and B.2 present positional, or individualized, information about economic disparities relative to the respondent’s annual income. The participants were randomly assigned to receive one of these treatments (or were assigned to the control group that received no information about inequality) at the beginning of the survey. Tax preference questions followed the informational treatment, which allows us to estimate the causal effect of the informational treatments on the choice behavior in the menu questions.
3.1. Inequality information experiments
We randomized respondents with equal probability into one of the following four treatments or the control group. Each treatment consisted of a separate survey page that displayed information on income distributions; we present the text for the German survey below:Footnote 8
• Treatment A.1 (99/50): “In 2015, German individuals with incomes in the top 1% earned at least 124,997 euros annually before taxes, while German individuals with incomes in the middle of the income distribution (where half of individuals earned more, and half less) earned 21,000 euros annually. This means that in 2015 the ratio between the top 1% of earners and those at exactly the middle was 5.95.”
• Treatment A.2 (99/10): “In 2015, German individuals with incomes in the top 1% earned at least 124,997 euros annually before taxes, while German individuals with incomes near the bottom of the income distribution (bottom 10%) earned at most 4,536 euros annually. This means that in 2015 the ratio between the top 1% of earners and the bottom 10% was 27.56.”
• Treatment B.1 (earn more): Respondents are asked to enter their gross monthly individual income. Then, the survey displays the following sentence: “X% of German individuals who have incomes earn more than you.”
• Treatment B.2 (earn less): Respondents are asked to enter their gross monthly individual income. Then, the survey displays the following sentence: “X% of German individuals who have incomes earn less than you.”
3.2. Effects on knowledge about inequality
We conducted a series of manipulation checks to investigate whether our treatments improved knowledge about inequality.Footnote 9 In both countries, we find that individuals who received positional inequality information (Treatments B.1 and B.2) were significantly better able to identify their location in the income distribution later in the survey. In addition, we find that individuals assigned to treatments providing aggregate inequality information (Treatments A.1 and A.2) were generally more likely to identify correctly income levels of relevant groups subsequently; while these results are always significant in the German case, we find weaker evidence on our manipulation check for the 99/50 treatment in Austria. Generally speaking, then, we recover consistent evidence that both sets of treatments did increase respondent information about income inequality in society.
3.3. Effects on tax preferences
Table 1 reports results of our informational treatments on the individual-level Kakwani Index implied by a respondent’s tax preferences for Germany and Austria.Footnote 10 In Germany, we find evidence that exposure to information about aggregate inequality in society—particularly information about the relative earnings of the very rich relative to average earners (A.1)—increased support for overall progressivity of the income tax system, as captured in the Kakwani index. We also find that positional information telling respondents about the proportion of individuals who earn less than they do (B.2) significantly increased preferences for progressivity. These effect sizes are economically meaningful: they correspond to 32% and 37% of the increase in progressivity of the German income tax system from 1995 to 2015.Footnote 11 However, as we noted before, the tax system has not become substantially more progressive over the last few decades. As such, while the information effects are meaningful compared to historical changes, they do not represent substantially more progressive preferences in absolute terms. That said, it is also the case that we find no significant effect for the other two informational treatments in Germany. In Austria, we find no evidence that providing respondents with any form of information about inequality had systematic effects on the Kakwani index.Footnote 12
Table 1. Kakwani index (Germany and Austria)
Notes:
** Table reports estimates from OLS regression of respondent preferred Kakwani index on experimental treatment assignment. In Germany, the difference between the Earn more and Earn less coefficients is statistically significant. We cannot reject that the effects of the two aggregate inequality treatment effects are equal.Robust standard errors in parentheses. ***
$p \lt $ 0.01,
$p \lt $ 0.05, *
$p \lt $ 0.1.
In Germany, while essentially identical information was provided to respondents in the “earn more” as opposed to the “earn less” treatment, we do not recover any systematic evidence that learning what percentage earned more than they did affected respondents’ preferences over progressivity. In this case, our results suggest that framing respondents’ position relative to those below them is more likely to prove politically potent than efforts to emphasize one’s position below the rich. To interpret this result, it is useful to distinguish between the possibility that greater progressivity results from the rich learning that they are not “middle class” (and therefore exhibiting greater downward altruism), or instead the poor learning that they are poorer and thus increasing their demands for redistribution (Cavaille and Trump, Reference Cavaille and Trump2015). As displayed in Figure 3, our results in the German case appear more consistent with the latter interpretation: while we observe large and significant increases in the Kakwani score among respondents in the lowest income tercile, we do not detect systematic effects for higher income groups.
Figure 3. Positional inequality information and own income in Germany. The figure presents the average treatment effect from each positional inequality treatment by income tercile, on respondent preferences for progressivity, as captured by the Kakwani index. 95% confidence intervals reported.
When considering the effectiveness of information on aggregate inequality, we also find that information about the earnings of the richFootnote 13 relative to “average” Germans (the 99/50 ratio) systematically increased progressivity preferences as captured by the Kakwani index. Our alternative aggregate inequality treatment—which provided instead information about earnings of the rich relative to “poor” Germans (the 99/10 ratio)—had no effect on respondents’ tax policy views. One possibility is that this difference in responsiveness is related to differences in beliefs about the “deservingness” of the poor and middle class for their economic status and the effects of these beliefs on redistributive policy preferences (Durante et al., Reference Durante, Putterman and van der Weele2014; Ballard-Rosa et al., Reference Ballard-Rosa, Martin and Scheve2017). If respondents view individuals at the bottom of the income distribution as less “deserving,” perhaps because that outcome is viewed as related to a lack of effort, they are less likely to respond to the inequality information in forming their tax preferences. Alternatively, it could be that the 99/50 treatment was more likely to be relevant given tendencies for individuals to identify with the “middle class” (Lupu and Pontusson, Reference Lupu and Pontusson2011); if most respondents believe they are close to the middle of the income distribution, they may have found information about this group’s position more important. In either case, this suggests that societal portrayals of information regarding “average” citizens may resonate more strongly than those that explicitly compare the earnings of the rich to those of the poor.
4. Discussion
While economic inequality has expanded substantially over the past several decades, few countries have sought to ameliorate disparities in income through the use of more progressive taxation. In this note, we show, first, that citizens in Austria prefer somewhat higher progressivity while Germans prefer somewhat lower progressivity than in the status quo. In Germany, the magnitude of the difference in progressivity between median elicited preferences and the status quo equals 23% of the increase in progressivity in the German tax system from 1995 to 2015. However, preferences are in favor of less, rather than more progressivity. Since the tax system has not become much more progressive over the last few decades, these preferences are not dramatic decreases from the status quo. This suggests that, absent heightened public demand for much higher taxes, the reluctance of politicians to increase taxes on the rich may be consistent with the overall view of the public.
Yet why citizens do not demand more redistribution—in the form of greater taxes on the wealthy—still presents a puzzle. One influential explanation that has been put forward suggests that if citizens merely knew how unequal the income distribution really is, they would respond by demanding heightened progressivity in income taxes. Our results suggest that certain kinds of information about inequality may induce heightened preferences for tax progressivity, at least in the German case; however, the lack of consistent effects for all types of information, as well as the lack of any systematic effects on the Austrian sample, suggest that greater attentiveness to the interaction between citizen beliefs and types of inequality information is warranted.
This note also suggests what kinds of information may have greater impact: in particular, the German results indicate that informing people about how many people are poorer than they are is likelier to elicit progressive responses than will telling them how many are richer. In addition, we also detect that information comparing “average” Germans to the rich are likely to be more politically consequential than accounts that contrast the rich and the poor.
However, while our study does offer some evidence that (certain kinds of) information about inequality may elicit greater demands for tax-based redistribution, we also emphasize that the magnitudes of our findings, while not inconsequential, are hardly so great as to suggest that—even when information changes preferences—informing people about inequality would lead to substantially higher demand for progressivity.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/psrm.2025.10018. To obtain replication material for this article, https://doi.org/10.7910/DVN/7ZWYDJ.
Acknowledgements
We thank the editor and three anonymous referees for helpful comments that have improved the paper. We also thank seminar participants at the 2020 Annual Meeting of the American Political Science Association.
This study was conducted in compliance with relevant laws on human subjects research and received approval from human subjects review boards at Stanford University (no. 39652), the University of North Carolina (no. 16-2946), and New York University Abu Dhabi (no. 046-2017). Data and supporting materials necessary to reproduce the numerical results in the article will be made available upon publication.
Funding statement
This work was supported by the Department of Political Science at Stanford University, the Global Populisms Project at the Freeman Spogli Institute, New York University Abu Dhabi, the Weatherhead Center for International Affairs at Harvard University, and the Austrian Chamber of Labour (the statutory representative body for employees in Austria).
Competing interests
None.
Open access
