Why are high-income and low-income earners not substantially polarized in their support for progressive income taxation? This article posits that the affluent fail to recognize that they belong to the high-income income group and this misperception affects their preferences over progressive taxation.
To explain this mechanism theoretically, I introduce a formal model of subjective income-group identification through self-comparison to an endogenous reference group. In making decisions about optimal tax rates, individuals then use these subjective evaluations of their own income group and earnings of other groups.
Relying on ISSP data, I find strong evidence for the model’s empirical implications: most high-income earners support progressive taxation when they identify themselves with a lower group. Additionally, individuals who overestimate the earnings of the rich are more likely to support progressive taxation.
[Keywords: taxation, preferences, inequality, public opinion, subjective income class, social comparison]
…More specifically, I demonstrate that most people, even the affluent, support progressive tax rates when they believe it would be someone richer than them who would disproportionately bear the extra tax burden. This belief is mostly driven by the difficulty in precisely identifying high-income individuals and their income. For most citizens being affluent is a fuzzy concept that is hard to define. Everyone—high-income and low-income individuals alike—is confident that Bill Gates or Mark Zuckerberg are among those with high income. Nobody would oppose the notion that an individual who lives on the Upper East Side in Manhattan, drives a Ferrari, and takes vacations on an exotic island would be considered rich. However, how do people classify the owners of the most beautiful house on their block or the person in their neighborhood who has a nice car? Who do they think are the high-income earners? More importantly, how do people assess their affluence?
…The following analysis relies on the 2009 Social Inequality International Social Survey Programme (ISSP 2009), which asks a variety of questions about perceptions of economic inequality, self-placement, and preferences on redistributive policies. The analysis was restricted to countries where information on income allowed the generation of 10 deciles and where the respondents were asked to report gross household income before taxes and other deductions. The sample covers 22 countries and around 8,000 respondents.2 It thus provides rich individual-level data on perceptions and preferences over welfare policies, as well as all the important control variables.
First, I examine the determinants of subjective self-placement. Then I proceed to explore how subjective self-placement and assessments of high-income group’s affluence levels affect preferences over progressive taxation.
…Before presenting the empirical results, it is interesting to look at ‘Self-Placement’ and ‘Income Distance’ to a CEO descriptively. The aim is to establish whether most high-income individuals place themselves in the middle, as well as to investigate the nature of perceptions pertaining to the income distance to a CEO.
Figure 2 shows the distribution of Self-Placement and Income Distance to a CEO by objective income deciles of the respondents. Although the analytical scope of Figure 2 is limited, it is immediately clear that when asked to place themselves on a 10-point scale, most respondents place themselves between the 4th and the 6th groups. Although self-placement increases with the objective income decile, the magnitude of this increase is not very substantial. The median value of Self-Placement of the respondents below the median household earnings averages around 5, whereas it is 6 for those above the median.
Figure 2 also reveals valuable insights about the subjective perceptions of respondents on the income distance from a CEO. The range of perceived distance ranges from −5 to +20. The horizontal line shows the logarithmic transformation of the highest CEO to average worker pay ratio of a company in the United States, the country with the highest overall proportion in the sample. The logarithmic transformation of the highest CEO-employer compensation ratio in the United States is 4.06 ( et al 2019), whereas the logarithmic transformation of the average CEO-employer compensation ratio is only 2.42 (2019).
Looking at the distribution of the perceived distances and the actual numbers, it is clear that many people overestimate the distance by a considerable margin. This figure thus shows that some respondents’ best guess about the yearly earnings of a CEO is substantially larger than the highest earner’s salary in their country. These numbers reveal that some people think about prototypes that do not exist when they are prompted to think about the income levels of the rich…Looking at the right side of the figure, respondents who belong to the top decile place themselves in the higher groups, around 7, only when they think they earn more than a typical CEO in their country. As their impression of the income of an average CEO increases, they start underestimating their position substantially.
[‘“Bloomberg spent $500 million on ads. The US population is 327 million”, Rivas wrote. “He could have given each American $1 million and still have money left over. I feel like a $1 million check would be life-changing for most people. Yet he wasted it all on ads and still lost.”’]
…The most striking result perhaps relates to the individuals who place themselves in high groups but still believe they earn substantially less than a CEO. In line with the prediction of Proposition 3.3 which posits that an individual who identifies with the higher-income group still prefers a progressive tax rate if she believes that the other members of the high-income group are substantially richer than her, this figure shows that the predicted probability of supporting progressive taxation of an individual who places herself in the top income group is substantially high, 0.95, when that individual unrealistically overestimates a typical CEO’s earnings.
…Perhaps one of the most interesting findings of this paper is that the well-known “middle-income bias” found in public opinion surveys can be systematically explained. When individuals compare themselves either to the superrich or the superpoor, they tend to infer that they are situated around the middle of the income ladder. This, of course, has severe effects on their political preferences.