Where do the others stand? Concern about Inequality and Perceptions of the wage gap

Samuel Bagshaw

Introduction

What do you think a doctor earns? And how much do you think they should earn? Using these two questions from the ISSP survey, as well as those concerning other occupations — chairman, shop assistant, unskilled worker and cabinet minister — I show that people who want to reduce inequalities are more concerned about levelling up low occupational groups’ wages than levelling down high wages. This article adds to the small literature on perceived inequalities, and the very large one on demand for redistribution. 

Explaining differences in demand for redistribution between people is an extensively researched field of economic literature. There are broadly two ways for reducing income redistribution: (i) an increase in the revenues of those at the lower end of the income distribution, (ii) a reduction in income of those at the high end of the income distribution. This article shows that overall people prefer levelling up low wages. 

One of the main proprieties of redistribution is that it adjusts a society’s social structure: wealth and income spreads between positions and mobility opportunities of individuals. Hence, one’s demand for redistribution depends mostly on the difference between what this person considers being the ideal social structure and what the social structure actually is. The closest they are, the less demand for redistribution arises.Two interpretations of this concept are to be considered (Clark and D’Ambrosio 2015)[2]. Either the individual is or can be part of the group about which he assesses the fair level of inequality. Either the individual is not and cannot be part of the group about which he assesses inequality — the normative view — in conditions closest to what John Rawls (1971)[7] describes as the “Veil of Ignorance”.

One implicit hypothesis usually made in the literature is that voters have an accurate perception of the real social structure they live in. Yet, as will be shown below, this is seldom the case.

For example, social mobility is despite common belief slightly higher in Europe. Indeed, the probability for a child from the bottom revenue quintile  to remain in the bottom quintile is of 33.1 percent in the US and below 30 percent in Europe (Alesina, Stantcheva and Teso 2018)[1]. Because of cultural stickiness, beliefs may differ from reality.

Duru-Bellat and Kieffer (2008)[4] observe that when asked about their own mobility, only half of the individuals have an accurate perception of it. This accuracy increases with one’s educational attainment. The survey administered in France gives that 45% consider that they have an upward trajectory, whereas only 32% really do. On the other hand, 27% consider that they occupy a lower position than their father, though only 10% do. The authors find out that optimistic subjective social mobility is positively correlated with: the higher the actual position, the lower the father’s occupation, being a male, higher education and more advanced age. 

As people have a tendency to live in homophilious social circles and as they have limited perception of the overall social structure, people tend to consider themselves closer to the mean than they actually are (Evans and Kelly 2004)[5]. Respondents that are in contact with people from a large variety of economic backgrounds have a more accurate estimation of their own relative wealth. In a tailor-made survey led in Argentina, Cruces (2013)[3] show that those living in areas with higher density and therefore encountering regularly people with more diverse backgrounds tend to have a weaker bias towards the mean, further confirming Evans and Kelly (2004)’s mechanisms. Minkoff and Lyons (2019)[6] add a new layer to this hypothesis by showing that it is neighbourhood income diversity, rather than local inequalities, that is the most relevant factor in explaining differences in perception of inequality. The relevancy of diversity over inequality may seem counter-intuitive, but as the authors state “a poor neighbourhood with a unique millionaire in it would appear to the observer or resident as just a poor neighbourhood”. While high inequality may remain hidden to the untrained eye, diversity is very often exacerbated.

Many factors bias the perception of inequality, to an extent to which untangling the effects of a difference in preferred inequality and perceived actual inequality may be difficult. In the first part of this article, I present a model putting forth the mechanisms at play in perception biases. Then, using these results to control for all variables that may have an effect on perception rather than ideal social structure, I show that people are more interested in levelling up low wages than levelling down high wages to reduce inequalities, regardless of their macroeconomic effects.

Mechanisms:

Before delving in the empirical analysis, it is important to have in mind the mechanisms at play to understand what is being measured. To do so, I have generated a two-dimensional canvas on which there are a hundred points, where one point represents one individual. To each point is assigned a value that represents its wealth (or income, it does not really matter). 

Table 1: Distribution of the points on the micro-simulation

Value	9	10	11	12	13	14	19	25	35	50	100
Frequency	10	13	15	20	15	7	6	5	4	3	2
c.f.*	10	23	38	58	73	80	86	91	95	98	100

*increasing cumulative frequency

The values have been arbitrarily assigned so that the overall Gini coefficient is equal to 0.3, as it is in average the case for the countries in my study. As for the spatial distribution of the points, the only rule used is that each of them is normally distributed around a mean corresponding to its rank, from the smallest value to the highest. For example, the third smallest point is normally distributed around µ = 3 and its value is 9, while the nineteenth point has a value of 10 and is distributed around µ = 19 (as the 10 smallest points are equal to 9 and the 13 next ones to 10). The usage of this criteria to distribute points allows those with similar values to be in similar places. As such, the homophilic circle axiom holds; i.e. the axiom stating that similar people live in similar places. For an illustration, see figures 1 and 3 which give two different iterations of the micro-simulation.

To represent limited perception, I have generated a circle around a random point, different at each iteration. The only inequalities perceived by the point in question are those within the circle. The difference between perceived and overall inequalities are represented in figures 2 and 4, respectively corresponding to the simulations 1 and 3. As can be seen, and no matter how many iterations are made, perceived inequality is always lower than real inequality (the in-circle Lorenz curve always being above the overall one). Moreover, because local inequalities sharpen following the wealth distribution, as wealth increases so does the accuracy in estimating real inequality: see the differences between figures 2 and 4, the latter centered around a low position of wealth equal to 10 and of perceived inequality equal to 0.04, while the former of position 13 perceives a Gini coefficient of 0.14 (the real Gini being equal to 0.3). To sum up, there is a general bias when estimating inequality level leading to underestimation of overall inequality. The closer the in-circle values are to the overall values, the more accurate the estimation will be. Thus, this bias can be dampened by multiple factors among which: an increase of the diameter of the circle (i.e. increasing perception), increasing density or increasing σ² (i.e. an increase in social mixing), levelling down the upper values or increasing inequality among the lower ones.

This model may be criticised for its oversimplification of a complex phenomenon. Indeed, it is static so the effect of social mobility is totally eclipsed; perception would gain in realism if it were a gradient fade in obscurity rather than a simple in-circle out-circle differentiation; perfect procession of information within the circle is unrealistic; there may be polarizing forces which push people to overestimate inequalities; perception is not only local, especially nowadays, it is also stimulated by medias, themselves often produced in big cities and therefore highly unequal environments. Yet, it is in the simplicity of this model that lies its strength, as homophilia and limited access to information are sufficient in representing various stylised facts about inequality misperception, facts that will be confirmed in the following sections of this article.

Urban area in which in respondent lives, past upward mobility and educational attainment which all affect the social diversity the respondent has encountered and thus their perception of inequality has been controlled for in the next part of this article.

Data:

The database used for this article is the module on social inequality of the International Social Survey Program (ISSP 2019) regrouping 44 975 respondents from 29 different countries^[1].

Explained variable:

To capture concern about inequality, I have used Q10:

“Some people feel angry about differences in wealth between the rich and the poor, while others do not. How do you feel when you think about differences in wealth between the rich and the poor in [COUNTRY]? Please place yourself on a scale of 0 to 10, where 0 means not angry at all and 10 means extremely angry”.

See Figure 6 below for distribution of concern about inequality.

Explanatory variable:

To construct my explanatory variables, I have used two questions from the survey:

Q2 “we would like to know what you think people in these jobs actually earn. Please write in how much you think they usually earn”;

Q3 “Next, what do you think people in these jobs ought to be paid? How much do you think they should earn.”

The jobs in question being doctor in general practice, chairman of a large national corporation, shop assistant, unskilled worker in a factory and cabinet minister in the government. Using these responses, I first built five different variables, one for each job, for which I express the difference between what respondents think others earn in a given job and what they think they should earn:

Results :  A Higher Concern for Fairness in the Lower End of the Distribution

The effect of each occupation taken individually on concern about inequality is displayed in tables 2 and 3. We can distinguish two groups within the occupations: on one hand shop assistant and unskilled worker — constituting the low occupations — and on the other hand chairman, cabinet minister and to a lesser extent doctor — constituting the high occupations —. Any increase in the perceived wage gap for the latter, i.e. when the respondents consider that they earn one percent relatively more than they should, is correlated with an increase in concern about inequality. On the opposite, an increase in perceived wage gap for the low occupations is negatively correlated with concern about inequality; put more intuitively, those who consider that the lower occupations earn less than they should tend to be more concerned about inequality. The coefficients associated with the low occupations are not just strongly significant, they also are of high magnitude. For example, in table 2 the coefficient for chairman is equal to 0.511 while the one for unskilled worker is equal to -1.268, about two and a half times stronger. This is suggestive that people are in average more concerned about low occupations earning relatively less than they should rather than high occupations earning relatively more. Interestingly, it is when we control for occupation and country fixed effects that the gap between coefficients for high and low occupations is the largest (table 2). I have a hypothesis for explaining this phenomenon using the mechanisms presented in section 3. People in similar occupational groups have similar perceptions about how much others earn, as they often work and have studied in similar areas, so the variation of perceived earnings is quite low within the same country and the same occupational group. As respondents answers about what people do earn are more and more similar, variations in the perceived wage gap are more and more explained by preferred income. Also, most people’s jobs correspond — by definition — to low occupations, so the data reflected in the table is more representative of low occupations than of high ones. These people logically more concerned about the well-being of those close to them, thus explaining the simultaneous increase in magnitude of coefficients for low occupations and decrease for high occupations in explaining concern about inequality.

As for the other variables, past and future prospects of mobility, as well as right wing political inclination and increase in self social positioning all are negatively correlated with concern about inequality (table 3), as they all are linked to the belief about which individuals are in control of their social trajectory and inequalities are fair as reflective of merit. Symmetrically, union members tend to be more concerned about inequality. Deductions from table 3 must be drawn with caution, as the number of respondents drop when asked about their political inclination and may not be as robust.

Conclusion:

This article adds to existing literature on demand for redistribution both on a theoretical and an empirical perspective. On one hand, by providing a simple theoretical framework that illustrates the mechanisms at play in explaining biases when estimating the inequality level within a large group.  On the other hand, it shows empirically that when people want to reduce inequalities, they are more concerned about low occupations earning less than they should than high occupations earning more than they should.

Samuel Bagshaw : Normalien au sein du département d’enseignement et de recherche Sciences Humaines et Sociales, il obtient en 2024 un Master 2 de recherche en économie publique à Paris 1. En parallèle, il est assistant de recherche en sociologie quantitative au Centre Emile Durkheim à l’Université de Bordeaux où il s’intéresse aux effets de l’égalité de genre entre parents sur la performance scolaire des enfants. Alliant méthodologie économique et question de recherche sociologique, il rédige un mémoire sur la perception des écarts de salaires et la préoccupation par les inégalités dont il livre ici les principaux résultats.

References

1. Alberto Alesina, Stefanie Stantcheva, and Edoardo Teso. Intergenerational mobility and preferences for redistribution. American Economic Review, 108(2):521–54, 2018.
2. Andrew E Clark and Conchita d’Ambrosio. Attitudes to income inequality: Experimental and survey evidence. In Handbook of income distribution, volume 2, pages 1147–1208. Elsevier, 2015.
3. Guillermo Cruces, Ricardo Perez-Truglia, and Martin Tetaz. Biased perceptions of income distribution and preferences for redistribution: Evidence from a survey experiment. Journal of Public Economics, 98:100–112, 2013.
4. Marie Duru-Bellat and Annick Kieffer. Objective/subjective: The two facets of social mobility. Sociologie du travail, 50:e1–e18, 2008.
5. Mariah DR Evans and Jonathan Kelley. Subjective social location: Data from 21 nations. International Journal of Public Opinion Research, 16(1):3– 38, 2004.
6. Scott L Minkoff and Jeffrey Lyons. Living with inequality: Neighborhood income diversity and perceptions of the income gap. American Politics Research, 47(2):329–361, 2019.
7. John Rawls. A theory of justice. The Belknap, 1971.

---

^[1] Countries included being Australia, Austria, Bulgaria, Chile, Taiwan, Croatia, Czech Republic, Denmark, Finland, France, Germany, Iceland, Israel, Italy, Japan, Lithuania, New Zealand, Norway, Philippines, Russia, Slovenia, South Africa, Suriname, Sweden, Switzerland, Thailand, Great Britain, United States, Venezuela.

---

---