Does Inequality Matter?

How People Perceive Economic Disparities and Social Mobility

Report

18 November 2021

Available in:

English
français

Download PDF

2. How do people perceive economic inequalities?

Abstract

This chapter discusses how perceptions of and concern over income and earnings disparities vary across countries and change over time. It shows that such concern has increased strongly since the early 1990s and correlates with changes in conventional indicators of income inequality. The chapter then disentangles people’s perceptions of the current extent of disparities from their preferred level of disparities. It shows that perceived income and earnings disparities are wider in countries where inequality, measured by conventional indicators, is greater; the inference is that people incorporate information about disparities in their perceptions thereof. Perceived earnings disparities have grown considerably over time. However, people have partly adapted their preferences for equality and become more tolerant of inequality. The chapter also discusses how concern over income disparities is influenced by perceptions of the intergenerational persistence of advantages and disadvantages and by belief in the importance of hard work.

2.1. How people’s concern over income disparities has evolved

Concern about income disparities is great and growing

Concern over income disparities (Table 1.1 and Annex 1.A) is widespread in OECD countries. According to the latest available data from the International Social Survey Programme (ISSP) and Eurobarometer (Annex 2.A), the vast majority of people agree with the statement that income disparities in their country are too wide. Indeed, in 2017, an average of some 80% of respondents agreed, and almost half agreed strongly (Figure 2.1).

There are sizeable differences between countries, however. Shares of people who strongly believe income disparities are too wide range from 17% in Denmark to 63% in Hungary. Considerable differences are also observed between socio-demographic groups. The elderly, women and people who regard themselves as belonging to lower social strata all show higher levels of concern over income disparities (Ciani et al. (2021[1]) for more details).

Figure 2.1. Most people are concerned about wide income disparities in their countries

People’s concern over income disparities has long been growing (Figure 2.2 and Bussolo et al. (2019[2])).1 Across OECD countries, the share of respondents who strongly agree that income differences are too wide had been on the rise since the early 1990s before reaching its peak at the onset of the global financial crisis. An alternative source, the European and World Values Survey, reveals a similar pattern (Ciani et al., 2021[1]). Data from the latest ISSP waves (2017 and 2019) suggest that concern has slightly decreased, on average, in the decade since the onset of the global financial crisis.2

The trend in people’s concern mirrors the evolution of income inequality in OECD countries as described by conventional statistical measures. Indeed, the data from the Income Distribution Database increased between the mid-1980s and late 2000s and point to a somewhat flatter trend since.

Figure 2.2. Concern about income disparities increased over time before reaching a peak during the global financial crisis

The swell of concern in the two decades before the global financial crisis (for which data are most fully and widely comparable) spared only Norway and New Zealand (Figure 2.3).3 And the increase was steep in Australia, Switzerland and the United States, although it started from very low levels in those countries. In Italy, too, however, the share or people who felt strongly that income differences were too wide rose from a much higher 40% in 1988 to 70% in 2011. The increase was sharpest in Hungary, which was going through economic and political transition. Concern also rose in Poland with the political transition, albeit to a lesser extent.

Concern has abated a little in the last decade on average. Although in half of the countries with long-term observations it has actually grown or remained stable, it has dropped significantly in others. In Italy, concern dropped to its 1987 levels, after having reached a climax during the global financial crisis, which coincided with the sovereign debt crisis that hit the country in 2011 (the year in which ISSP 2009 was fielded in Italy). Similarly, in Austria and Poland, the latest available data suggest lower levels of concern than in the late 1980s. However, the trend is subject to caution as the data are from the ISSP 2017 wave and Eurobarometer 471/2017, respectively, which are not fully comparable with other waves.

Figure 2.3. In the two decades up to the global financial crisis, concern over income disparities increased in most countries

Changes in concern are related to changes in observed inequality

At any moment in time, cross-country differences in concern over income disparities do not match differences in the magnitude of inequality estimated with conventional statistical measures (Figure 2.4 and Gimpelson and Treisman (2018[3])). In some countries, nevertheless, levels of concern over income disparities are similar to the extent of inequality measured by conventional indicators. Nordic countries, for instance, exhibit both lower Gini indices and lower concern, while Turkey display high levels of inequality and concern, as do other Eastern and Southern European countries and Israel. The “low-Gini-high-concern” group includes some European countries that transitioned to a market economy, as well as France. By contrast, most English-speaking countries belong to the “high-Gini-low-concern” group, although their average levels of concern are quite close to the median and to levels observed in Japan and some Central and Southern European countries, such as Germany and Greece.

Figure 2.4. Concern over income disparities only weakly correlates with conventional measures of income inequality

By contrast, changes in concern over income disparities correlate positively with changes in conventional inequality indicators (Figure 2.5).4 In those countries where the Gini coefficient rose the most, concern over income disparities also increased more steeply, from which it may be inferred that people’s concern reflects the changes in income disparities that have occurred in their country over the years. The inference is in line with a body of literature that has highlighted how within-country differences in perceptions and concerns, either over time or across regions, tend to correlate with statistical estimates of inequality (Kerr, 2014[4]; McCall et al., 2017[5]; Bussolo et al., 2019[2]; Kuhn, 2019[6]; Colagrossi, Karagiannis and Raab, 2019[7]; Giger and Lascombes, 2019[8]; Xu and Garand, 2010[9]; Newman, Shah and Lauterbach, 2018[10]; Franko, 2017[11]).

Figure 2.5. People's concern reflects changes in inequality over time

At country level, changes in levels of concern over income disparities are more closely related to changes in inequality measured in disposable rather than market income (Table 2.1, Columns 1‑3). The inference is that people’s perceptions take into account the redistribution operated through income taxes and cash transfers. Thus, if market income inequality increases, but is offset by effective redistribution, then concerns will not change significantly. Similarly, if market inequality does not change, but redistribution weakens, then concern over inequality tends to rise.5

Changes in concern over income disparities do not seem closely related to macro-economic conditions. For example, higher employment rates and GDP per capita reduce concern, while lower unemployment seems actually to increase it, though not by much in either case (Table 2.1, Column 4).

The average respondent also seems to be concerned chiefly about income differences between the top and the middle of the distribution (Table 2.1, Column 5), while the gap between the median and the bottom of the distribution has no significant effect on the average level of concern. The inference is that the income growth dynamics of the middle class relative to the top is particularly important in explaining concern over income disparities in the country as a whole. The finding is also in line with the conclusions of Lupu and Pontusson (2011[12]), who argue that the structure of inequality – as captured by the distance between the 90th percentile and the median income – is a key driver of actual redistributive policies. Fisman et al. (2020[13]) also provide evidence that the relative standing of high-income individuals is a particularly salient determinant of individuals’ concern about income distribution. Their results show, moreover, that people keep an eye on the incomes of individuals just above them in the distribution.

Column 6 in Table 2.1 estimates whether alternative estimates of fiscal income inequality from the World Inequality Database yield results consistent with those of conventional indicators. The World Inequality Database relies on tax data and better captures the top of the distribution. Results suggest that the most powerful driver of concern is the share of income owned by the richest 10%, while the share owned by the richest 1% does not, per se, exert a significant effect.

Results are similar at the individual level, after controlling for a broad set of characteristics that influence concern. At the individual level, it is also possible to look at the impact on the entire range of possible responses to the ISSP question. When inequality rises in their countries, respondents are more likely to agree strongly that income disparities are too great and less likely to answer anything else (Figure 2.6).

	(1)	(2)	(3)	(4)	(5)	(6)
Gini market income (before taxes and transfers)	0.66*		0.29	0.10
	(0.35)		(0.36)	(0.56)
Gini disposable income (post taxes and transfers)		1.71**	1.55**	1.69**
		(0.67)	(0.71)	(0.73)
Unemployment rate				-0.01*		0.00
				(0.00)		(0.01)
Employment rate				-0.15		0.29
				(0.57)		(0.32)
GDP per head (logarithm)				-0.17		-0.15
				(0.25)		(0.10)
90th percentile vs median income ratio					0.29**
					(0.12)
median income vs 10th percentile ratio					-0.03
					(0.08)
Top 10% share (WID)						2.54**
						(1.20)
Top 1% share (WID)						-1.68
						(1.24)

Observations	78	78	78	78	78	84
Countries	29	29	29	29	29	28
Country fixed effects	Included	Included	Included	Included	Included	Included
Period fixed effects	Included	Included	Included	Included	Included	Included

Note: * denotes statistically significant at the 10% level, ** at 5%, *** at 1%. All coefficients can be read as percentage point changes, e.g. in column (1) a 1 percentage point increase in the Gini coefficient of market income is associated, on average, with a 0.78 percentage point increase in the share of respondents who strongly agree that income differences are too large. Standard errors clustered by country in parentheses. Results are from fixed (country) effects regressions, including period fixed effects. GDP per head is in logarithms, but the original values are expressed in constant prices and PPP (2015 USD PPP).

Source: OECD calculations from ISSP 1987, 1992, 1999, 2009, 2017 and Eurobarometer 2017 for concern over income disparities (see Figure 2.1 for the list of countries for which Eurobarometer is used); OECD Income Distribution Database for the Gini coefficient; World Inequality Database (WID) for the income share of the richest 10% and 1% (pre-tax national income, adults, including elderly (20+), household income of couples attributed to each individual assuming equal-split).

Statlink

Figure 2.6. When income inequality rises respondents become more likely to strongly agree that income disparities are too large

People’s concern over income disparities is shaped by their perceptions and preferences

In order to understand how and why concern over income disparities has changed with time in response to income inequality, it is crucial to recall that it combines two elements:

individuals’ perceptions of the extent of income inequality, i.e. what they think it is;
individuals’ preferences for income equality, i.e. what they think it should be.

An increase in inequality might therefore influence concern over income disparities in two ways:

People may incorporate information about rising inequality in their perceptions (perceptions adjust to reality), which could heighten their concern about inequality.
People’s preferences adapt to high inequality as they grow gradually more tolerant of inequality, so that they eventually prefer higher levels of inequality (Trump, 2018[14]).

Perceptions and preferences may combine differently in response to increases in inequality. Thus, when inequality grows, so might concern if people’s perceptions of inequality diverge from their preferences. Concern may not change if people overlook the signs, or if the changes in perception and preference balance each other out. And it may even fall, if people adapt to greater inequality.

Evidence as to how people’s concern over inequality evolves in the long run (Figure 2.3) and how it relates to inequality indicators (Table 2.1) suggests that adjusting perceptions to reality generally prevails over people adapting their preferences.

A crucial determinant of people’s concern about inequality is their own income – indeed, those at the bottom of the income distribution are more concerned about inequality than those at the top (Rueda and Stegmueller, 2020[15]). When inequality increases, the average level of concern may rise, too, but not only because people become more concerned about the overall level of inequality (Alesina and Giuliano, 2011[16]). Concern also grows because more people’s incomes fall below average and they perceive that their relative position in the distribution has worsened. This is the mechanism behind the standard Meltzer-Richard model (Meltzer and Richard, 1981[17]), whereby demand for redistribution rises as inequality grows, because the median voter becomes poorer than the average. The inference is that:

perceptions of both the overall level of inequality and the individual’s own position in the distribution are crucial;
the impact of inequality on concern about inequality and preferences for redistribution likely depends on relative income.

Chapter 3 discusses in detail the role of people’s own income – both real and perceived – in shaping perceptions of and concerns over inequality, which it relates to preferences for redistribution.

Concern over income disparities also depends on beliefs about the sources of such disparities. People who believe that hard work is a more important determinant of economic success than other factors are more inclined to accept that some individuals earn more than others as a consequence of their efforts (Fong, 2001[18]; Alesina and Giuliano, 2011[16]; Karayel, 2015[19]; Clark and D’Ambrosio, 2015[20]; Daniels and Wang, 2019[21]; Mijs, 2019[22]; Almås, Cappelen and Tungodden, 2020[23]). Conversely, those who believe that luck and sheer circumstance drive economic success are more concerned about inequality (Figure 2.7). And in countries where more people believe that parental wealth matters little for getting ahead in life and that hard work matters there is less concern over income disparities. Countries where there is a strong belief in equal opportunities include some Nordic countries – Iceland, Sweden and Norway, but not Denmark and Finland – and most English-speaking countries (see also the related evidence by Benson (2021[24]) on the importance given by respondents from the United Kingdom to meritocracy). While the United States is usually described as a country of social mobility, it is in fact at the median when it comes to the perceived importance of parental wealth. However, it is also the country that believes most strongly in the importance of hard work. Patterns in other countries are less cut and dried. Post-transition and Southern European countries tend to harbour the perception that parental wealth matters, although in some of them the average respondent believes that hard work pays. Although its level of concern over equality is almost the same as the average English-speaking country, Japan accords less importance to both parental wealth and to the virtue of hard work. Korea, by contrast, is at the opposite end of the spectrum, deeming parental wealth very important and having great faith in hard work.

The next two sections seek to further disentangle people’s perceptions of and preferences for inequality and how they evolve over time. Section 2.2 considers perceptions and 2.3 preferences. The discussion necessarily entails addressing people’s views of equality of opportunities and the virtue of hard work.

Figure 2.7. There is less concern over income disparity in countries where people believe that hard work, not parental wealth, is what matters for getting ahead in life

2.2. Behind concern over income disparities lie people’s perceptions of income and earnings disparities

Perceived income and earnings disparities are wide

One way to unbundle the different drivers of people’s concerns about income inequality is to ask them what they think the current level of economic inequality is (their perceptions), and what they would like it to be (their preferences).

Surveys which collect information on people’s perceived economic disparities focus on different outcomes. The OECD Risks that Matter survey asks about household income, while ISSP asks respondents about earnings disparities.6 Despite their differences, looking at perceptions of disparities in either economic outcomes is useful for two reasons. First, a single survey covering perceptions of both earnings and income inequality is not available. Moreover, data on both perceptions are not always available for all OECD countries and for all the relevant periods. In fact, the analysis of changes is possible only on ISSP for perceived earnings disparities. Secondly, one needs to take into account that perceptions of and preferences for earnings disparities might differ from those for income disparities. For example, people might be more tolerant of earnings disparities, because these can be attenuated by welfare transfers to households with low-earners. In fact, earnings are only one component of income and, therefore, concern over earnings disparities can be considered as one of the determinants of the overall concern over income disparities.

Most people perceive high levels of both income and earnings inequality. According to the results of the 2020 OECD Risks that Matter survey, average respondents believe that the share of their country’s total income that goes to the richest 10% of households is extremely large (Figure 2.8). In all 25 countries surveyed, the average perception is that the richest 10%’s share of national income is 42% – ranging from 38% in Denmark to 67% in Turkey. To put perceptions in perspective, the latest average estimate from the OECD Income Distribution Database is that the richest 10%’s share of disposable income is actually 25% in the countries which Risks that Matter surveyed.

ISSP 2009, which covers a wide set of countries, considered perceptions of earnings disparities. It found, on average, that median respondents believed that highly skilled earners (doctors and CEOs) earned around 9 times more than an unskilled factory worker. Yet there is substantial variation in countries’ perceived top-bottom earnings ratio, which ranges from 3 in Sweden to 26 in Korea.

Figure 2.8. Perceptions are that there is great income and earnings inequality

In most countries, the prevailing perception is that income inequality has increased in the last decade (Figure 2.9).7 A retrospective question is usually more likely to prompt the answer that inequality has been on the rise. Evidence from other surveys, like the French Baromètre d’opinion, the recent Ipsos MORI survey conducted in the United Kingdom for the Deaton review (Garret and Day, 2021[25]) and the American Election Studies (Macdonald, 2019[26]), report similar findings, with most respondents asserting that inequality has been on the rise in recent years.

Figure 2.9. Most people perceive that income inequality has increased in the recent decade

Most people believe that intergenerational income persistence is high and related to inequality of outcomes

As documented by the extensive literature (Alesina and Giuliano, 2011[16]), people’s opinions of intergenerational mobility play a crucial role in shaping their concern over current inequality of outcomes (either earnings or income). This crucial role is consistent with the interpretation of Benabou and Ok’s Prospect of Upward Mobility hypothesis (POUM) in an intergenerational perspective (Benabou and Ok, 2001[27]). POUM advances that people might be less concerned by their current situation if they believe that their offspring have good chances of climbing the income ladder. Furthermore, research has shown that income inequality and social mobility are negatively related, both across countries (OECD, 2018[28]) and within them, as the chances of scaling the income ladder are lower in areas with wider income disparities (Chetty et al., 2014[29]). Whether people’s perceptions are aligned with this finding – so that perceived intergenerational mobility is lower where perceived income or earnings inequality is greater – is less well known (see Alesina, Stantcheva and Teso (2018[30]) for evidence relative to different areas in the United States).

The latest Risks that Matter survey finds that, in OECD countries, people believe that a child from a household in the bottom 10% of the income distribution is highly likely still to be in there when s/he grows up (Figure 2.10, Panel A). The average share of respondents who hold that belief is 55%, ranging from 47% in Norway, Poland and Denmark to 64% in Austria. Women perceive intergenerational persistence as more prevalent than men, but do not believe the richest 10%’s share of income is as high as men do (Ciani et al., 2021[1]).

Figure 2.10. The poor are seen to have little chance of climbing the social ladder

It is possible to paint a more qualitative picture from ISSP respondents’ beliefs about the importance of family background (Brunori, 2017[31]). It considers two types of parental characteristics: family wealth and the parents’ education. There are conceptual differences between the two, even though they correlate. Wealthy parents, for example, might finance their offspring’s education or entrepreneurial activities. Similarly, highly educated parents might influence offspring’s success independently of family wealth – by transmitting different knowledge, for instance. Indeed, in France, Belgium, Spain and Chile, a much higher share of respondents agree that having well educated parents is more important for getting ahead in life than having wealthy ones (Figure 2.10, Panel B).

A perceived intergenerational persistence index, built by averaging the perceived importance of parents’ education and wealth, points to wide differences between countries (Figure 2.10, Panel B). Perceived intergenerational persistence is strong in Turkey and Poland, but slight in Finland, Norway and Denmark. The United States, often cited as the country where most people are confident of social betterment, ranks in the middle of the distribution.8

On average, people’s perceptions of intergenerational mobility are in line with the so-called “Great Gatsby Curve”, whereby greater inequality spells less next-generation upward mobility. In the Risks that Matter survey, perceptions of intergenerational persistence in the shares of income of the richest and poorest 10% are closely related (Figure 2.11). Similarly, ISSP 2009 finds the perception of very wide top-bottom earnings ratios is associated with the belief that family wealth and parental education are particularly important for success in life.9 These findings from Risks that Matter and ISSP are in line with experimental evidence suggesting that, when people are provided with pessimistic information about the level of inequality, they also weaken confidence in intergenerational mobility (McCall et al., 2017[5]; Davidai, 2018[32]; Browman, Destin and Miele, 2020[33]) and, similarly, when provided with pessimistic information on low mobility, perceptions of high inequality increase (Shariff, Wiwad D and Aknin, 2016[34]).

Figure 2.11. The perceived Great Gatsby Curve: perceptions of intergenerational persistence and economic disparities are closely related

Perceptions are correlated with conventional measures, but do not necessarily align with them

Perceptions across countries significantly correlate with conventional estimates of inequality, with reference to both income and earnings inequality (Figure 2.12). Similarly, where perceived intergenerational persistence is higher, also estimates of father-son elasticity in either earnings or education are higher (higher elasticity means that the son’s earnings/education are more strongly related with the father’s, indicating higher intergenerational persistence) (OECD, 2018[28]).10 The inference is that people form their perceptions of income inequalities and social mobility by incorporating at least some information on the real economic outcomes. The inference is also consistent with previous observational evidence from Kuhn (2019[6]), Bussolo et al. (2019[2]), Roth and Wohlfart (2018[35]) and Domènech-Arumí (2021[36]).

Figure 2.12. Perceived inequality and intergenerational persistence are consistently related to conventional indicators of the same phenomena

A series of in-survey experiments has tested how far people’s perceptions take in information about the extent of inequality. To that end, researchers fed to a randomized subset of respondents information on the current magnitude of inequality reported in studies or the media. They then compared the subjects’ perceptions with those of a subset of participants who had not been given the information. They found that the individuals that received information about high inequality perceived more inequality in outcomes and opportunities and were more concerned about it (Box 2.1), so corroborating observational evidence (e.g. Figure 2.12).

Box 2.1. Evidence from in-survey experiments

The evidence presented in this chapter bears out the hypothesis that, on average, people generally incorporate information about the extent of inequality in their perceptions. However, this evidence is based on observational data, which makes it hard to single out whether perceptions are indeed shaped by inequality. This because differences in perceptions between countries and changes over time might be influenced by other variables. To isolate the impact of new information about inequality on concerns and perceptions thereof, a growing body of literature has built upon survey experiments.

In these experiments, a randomly selected proportion of respondents is provided with information on the distribution of outcomes and opportunities in the population (or about their own position in the income distribution). Usually this information points to a large degree of inequality. Therefore, the hypothesis is that, if people incorporate such information, they should increase their perceptions of and concern over inequality. As the provision of the information takes place at random, comparing the answers of the group with information and the group without makes it possible to test this hypothesis. Some experiments also examine the role of other factors, such as trust, in explaining results.

Ciani, Fréget and Manfredi (forthcoming[37]) conduct a meta-analysis of the experiments that measure the impact of information on perceptions and concerns about economic inequality. Because the experiments use a heterogeneous range of measures, results are standardized with regard to the standard deviation of each measure in the control group (i.e. in the groups receiving no information). Most estimates of the effect of information on perceptions of and concern about economic disparities are positive, albeit to different extents. The average standardized effect across all studies is 0.17 in the United States, 0.15 in EU countries and 0.16 in other countries. Thus the provision of additional information produces an average increase in perceptions and concerns that is above 0.15 standard deviation, in line with the hypothesis outlined above. Meta-regression analysis shows that that the effect on perceptions is stronger than on concerns, but the latter is still sizeable (between 0.09 and 0.13 standard deviation) and statistically different from zero. This evidence therefore suggests that people interpret information correctly and incorporate it into their outlook.

Some experiments also back up the hypothesis that people partially adapt their preferences upon receipt of signals of increased or high inequality. Trump (2018[14]) finds that informing US and Swedish respondents about the true extent of inequality increases both perceived and preferred levels of earnings disparities. As a result, the effect on concern over income disparities is not statistically different from zero. Hoy and Mager (2020[38]) find that supplying respondents in the United States with the facts about the actual levels of inequality and social mobility lessens how strongly they agree with the statement that income differences are too wide. The change in perception, say the authors, is driven chiefly by respondents who, prior to the experiment, stated that high levels of inequality do not exist. Hoy and Mager (2020[38]) also interpret this as evidence that people increase their preferred level of inequality when they find out its true extent.

Perceptions and conventional indicators of inequality do not fully match. Perceptions of the richest 10%’s share of income and intergenerational income persistence among the bottom 10% exceed conventional measures. For instance, the top 10%’s average perceived share of income across the 25 countries in Risks that Matter is 52%, while the average estimate from the OECD Income Distribution Database is 25%. The Compare Your Income tool, which uses a different approach, displays similar findings for the richest 10%’s share of income.

As for statistical measures of how likely the poorest children are to be poor as adults, they are not available for all countries. Those that are reveal once more a divergence between perception and statistics. Respondents to the Risks that Matter survey in Italy and the United States believe that intergenerational persistence will affect respectively 53% and 52% of children in the poorest 10% of households. Statistical estimates find much lower shares –16% in a 1980 cohort for Italy (Acciari, Polo and Violante, 2019[39]) and 20% in 1980-82 cohorts for the United States (Chetty et al., 2014[29]).

These differences between perception and conventional indicators should not necessarily be interpreted as a measure of bias for three main reasons (detailed in Box 2.2):

1. People may think in terms of wealth, rather than income, even though the Risks that Matter questions refer explicitly to income.

2. Conventional estimates reflect methodological choices, while people probably use other, different definitions.

3. Questions are complex for respondents and estimated differences between perceived values and conventional estimates are highly sensitive to how the question is defined and framed.

The answers to the quantitative questions about perceived inequality and social mobility provide valuable and interesting results that go beyond the calculation of a bias. Despite the complexity of definitions and questions, people’s average perceptions consistently correlate with conventional estimates across countries, showing that they reflect real disparities. Looking at perceptions – particularly of income inequality, earnings disparities and intergenerational persistence – affords researchers insight into how people process information (Phillips et al., 2020[40]). Furthermore, as Chapter 3 shows, answers to these quantitative questions on perceived inequality and social mobility are powerful predictors of preferences for redistribution, both at the individual and country level. Lastly, they provide descriptions of the distribution (and polarization) of perceptions in the same country (see Chapter 4). Such descriptions are both richer and different from those derived from qualitative questions, where most people tend to respond with the same value, i.e., “agree”. Nevertheless, for methodological reasons, it is important to:

employ a wide array of perception measures, which should include qualitative questions;
analyse preferred disparities and how they diverge from perception.

Box 2.2. Understanding the differences between average perceptions and conventional estimates

Average perceptions of the magnitude of the richest 10%’s share of income and intergenerational persistence among the bottom 10% (measured by responses in the OECD Risks that Matter survey) tend to be significantly greater than conventional estimates. Such divergence should not necessarily be interpreted as bias for three main reasons.

First, it is likely that people think of a different or broader concept of economic outcomes, even though both questions specify income. The average perception of the level of the richest 10%’s share of income suggests that people might be thinking more of wealth than income (Balestra and Cohen, 2021[41]). Indeed, the average perception is closer to the top 10%’s share of household wealth, which is 53% for the 19 countries for which data are available in the OECD Wealth Distribution Database. It also seems to be closer to the top 10%’s share of fiscal income in the World Income Distribution Database, which paints a more accurate picture of top incomes, usually underrepresented in household income surveys. Across the 25 countries covered in Risks that Matter, the latest estimate of the top 10%’s share of fiscal income is 38%.

Second, conventional statistical indicators to measure economic inequality depend on a number of methodological choices that include:

the definition of income, e.g. which sources to include and exclude;
the adjustment of income for household size and needs, e.g. equivalence scales;
the population of reference, e.g. which cohorts to use for measuring intergenerational persistence) and others.

While these choices reflect statistical conventions and consensus between experts, there is no single method. And the sensitivity of numerical estimates to methodological variations makes it hard to find the “perfect” counterpart to each measure of perceptions, as people might unconsciously use other and different definitions.

Third, results are highly dependent on how questions are framed (Jachimowicz et al., 2020[42]). The importance of the question seems particularly relevant when it comes to determining bias with respect to “actual” values. The methodological debate between Eriksson and Simpson (2012[43]) and Ariely and Norton (2013[44]) shows that measuring income disparities as group shares of total income (like the top 10%’s income share in Risks that Matter) or income levels (as in the ISSP questions) might lead to different conclusions about whether respondents under- or overestimate income inequality.

As for intergenerational mobility, Swan et al. (2017[45]) find questions asked with reference to income quintiles or tertiles yield different results as to the “bias” of perceived intergenerational mobility in the United States. In fact, Swan et al. (2017[45]) suggest that measures of perception are better suited to questions that go beyond merely calculating bias and consider how perceptions shape attitudes towards redistribution and vary from group to group. In addition, as the Risks that Matter survey does, dividing the population into 10 income deciles is a complex exercise for respondents, and people may simply refer to “the rich” for the top 10% and “the poor” for the bottom 10%.

Clearly, it is important to use different measures of perceptions and to elicit preferences. To that end, researchers should use different methods to elicit quantitative responses (e.g. asking them as shares of total income – as for the perceived share of income of the top 10% in Risks that Matter – or asking levels for different group – as for example in the perceived earnings of different occupations in ISSP) and support them with qualitative estimates (e.g. the perceived intergenerational persistence in the bottom 10% from Risks that Matter and the “get ahead in life” questions from ISSP).

Numerical questions about perceptions have two main advantages:

First, it is easier to frame the question in a way that pre-empts any judgement about whether a disparity is “too” large.
Secondly, they help better reflect the heterogeneity of perceptions among citizens of the same country. Answers to qualitative questions tend to bunch at certain values (e.g. “agree”), so masking a marked underlying heterogeneity across the population. When it comes to concern over income disparities, most people might agree that they are too wide in their country. Yet, in reality, the differences between what people believe the top 10%’s share of income is and what they think it should be vary widely (Chapter 4).

One way to compare countries’ perceptions and conventional estimates without looking at the precise definitional difference is to use both as yardsticks to rank the countries according to whether they score high, medium, or low. Table 2.2 ranks countries using the OECD Income Distribution Database estimate of the richest 10%’s share of income with the average perceived share from the Risks that Matter survey. As for intergenerational persistence, it compares the country ranking according to estimated earnings elasticity between fathers and sons (available for a wide set of OECD countries) with the perceived intergenerational persistence among the poorest 10%.

The results for income inequality are broadly consistent in the countries ranked top and bottom, with the Nordic countries exhibiting relatively low levels of measured and perceived inequality, and Chile, Mexico and Turkey showing high levels. As for intergenerational persistence, the Nordic countries again score low on both counts, and Austria, Germany and Chile relatively high.

	Income inequality		Intergenerational persistence
	Estimated income share that goes to the 10% richest (IDD, latest available year)	Perceived income share that goes to the top 10% richest (RtM 2020)	Estimated intergenerational earnings elasticity (sons observed late 2000s; OECD (2018[28]))	Perceived intergenerational income persistence in the bottom 10% poorest (RtM 2020)
SVN	Low	Low	-	-
BEL	Low	Medium	Medium	Medium
NOR	Low	Low	Low	Low
EST	Low	Low	-	-
DNK	Low	Low	Low	Low
POL	Low	Low	-	-
AUT	Low	High	High	High
FIN	Low	Medium	Low	Low
NLD	Medium	Low	Medium	Medium
CAN	Medium	High	Medium	Low
DEU	Medium	Medium	High	High
GRC	Medium	High	Low	High
IRL	Medium	Medium	Medium	High
CHE	Medium	Medium	High	Medium
ESP	Medium	Medium	Low	Medium
FRA	Medium	Low	High	Medium
KOR	Medium	Medium	Medium	Medium
PRT	High	High	Medium	Medium
ITA	High	Medium	High	Low
ISR	High	Low	-	-
LTU	High	Medium	-	-
USA	High	High	Medium	Low
TUR	High	High	Low	High
MEX	High	High	-	-
CHL	High	High	High	High

Note: The countries observed are ranked low, medium or high depending on the distribution of the indicator among the countries observed; for instance, “High” for estimated top 10% income share refers to the 8 countries with the highest values. The ranking for intergenerational persistence is calculated only for those countries for which the intergenerational earnings elasticity is available in OECD (2018[28]).

Source: OECD calculations from Risks that Matter 2020; OECD IDD for top income shares, OECD (2018[28]) for intergenerational earnings elasticity.

Statlink

Nevertheless, a country’s perceived and measured levels of inequality and intergenerational persistence may be very different. Some countries rank lower in perceived than conventionally measured inequality – e.g. Italy, Israel, Lithuania, France and the Netherlands – and some the other way round, such as Austria, Belgium, Canada, Finland and Greece. As for intergenerational persistence among the poorest 10%, perceptions thereof in France and Italy are lower than conventional indicators, as in Canada, Switzerland and the United States. Greece, Ireland and Turkey, however, rank higher in perceived than conventionally estimated intergenerational persistence.

Perceived top-bottom earnings ratios have grown over time

Perceived disparities as captured by the top-to-bottom earnings ratio have long increased significantly. They generally reached a peak during the global financial crisis, then fell the following decade. From the 1980s to the global financial crisis, the median perceived top-bottom earnings ratio grew in all 13 countries for which data are available (Figure 2.13 and Giger and Lascombes (2019[8])). On average, it doubled from 5 to 10 between the first ISSP wave and 2009. In the ensuing decade, though still higher than 30 years before, it fell from 10 to 8 (as ISSP 2019 shows).

Among countries observed in ISSP since 1987, the increase was especially steep in Australia (Leigh, 2013[46]) and the United States, as well as in countries transitioning to a market economy, such as Poland and Hungary. As for countries observed since 1992, the increase was marked in Germany, Italy and Slovenia. The fall since the global financial crisis has been particularly robust in Australia, where the perceived earnings disparities had reached a very high level in 2009. They have remained stable, however, in Germany and New Zealand – a possible explanation being that, in both countries, the latest ISSP survey was carried out in 2020, during the pandemic crisis.

Figure 2.13. The perceived top-bottom earnings ratio has increased over time

During the last 30 years the average perception of intergenerational persistence have only mildly increased, according to the index built from qualitative answers to ISSP about parental characteristics that are important to get ahead in life (Figure 2.14, Panel A). At country level, though, the picture is varied. Between the late 1980s or early 1990s and the global financial crisis, the increase was significant in Australia, Germany, Hungary, Slovenia and the United States on both counts of parental wealth and education (Figure 2.14, Panel B). The change was also sizeable in Poland, where, however, the perceived importance of coming from a wealthy family declined. In New Zealand, Italy, Austria and Sweden, by contrast, respondents to ISSP 2009 reported that persistence was lower than in the late 1980s or early 1990s. In the decade since the global financial crisis, perceived persistence lessened in Australia, Switzerland and, to a lesser extent, in the United Kingdom. It rose in the other countries, however, particularly in New Zealand and Italy, where it more than offset the fall observed in the previous two decades.

Figure 2.14. Changes in perceived intergenerational persistence of inequality were quite heterogeneous across countries

The pandemic has raised awareness of economic disparities

There is evidence that the ongoing pandemic and resulting recession have brought to light pre-existing inequalities (Blundell et al., 2020[47]). People’s awareness of income disparities and lack of intergenerational mobility might therefore have risen, too. Indeed, results from the Risks that Matter survey show that people who report having experienced any health or economic hardship during the COVID-19 crisis, either themselves or in their household, perceive greater inequality and intergenerational persistence than others (Table 2.3).11 (See OECD (2021[48]) for further discussion of household insecurity during the COVID-19 crisis.) The perception is not attributable to differences in respondents’ socio-economic status or demographic characteristics. Nor can it be put down to the reported changes in their households’ financial situation or their country’s macro-economic performance in the previous 12 months. Although people affected by COVID-19 might anyway have perceived higher levels of inequality, the impact of the pandemic and economic inequality may well have compounded those perceptions (Table 2.3).

	(1)	(2)	(3)	(4)	(5)	(6)
	Perceived richest 10%’s share of income			Perceived bottom 10% intergenerational persistence
Experienced health or economic hardship	2.6***	2.2***		3.1***	2.7***
during the pandemic	(0.4)	(0.4)		(0.4)	(0.4)
Experienced physical or mental health problems			2.5***			2.2***
because of the pandemic			(0.4)			(0.4)
Experienced job-related disruption during			-0.1			0.5
the pandemic			(0.4)			(0.4)
Had difficulties in making end meets			2.6***			1.1**
during the pandemic			(0.5)			(0.5)
Report that household financial situation	2.2***	1.3***	0.7	2.1***	1.0**	1.0*
worsened during the pandemic	(0.4)	(0.5)	(0.5)	(0.4)	(0.5)	(0.5)
Country fixed effects	included	included	Included	included	included	included
Household and individual characteristics		included	Included		included	included
Observations	25181	24526	24526	25181	24526	24526

Note: * denotes statistically significant at the 10% level, ** at 5%, *** at 1%. Robust standard errors in brackets. The results are based on OLS regressions, including country fixed effects and weighting by sample weights (rescaled so that weights add up to 1 in each country). Household and individual characteristics include age, age2, household size, number of children and dummies for: household disposable income decile, gender, educational level, employment status, marital status, size of town (including an indicator for missing value), housing tenure, perceived changes in national economy and household finance situation with respect to the previous 12 months (on a 5 point Likert scale from much worse to much better; the much better category has been combined with better because of its small size). The regression also includes a dummy for those who opted for “I prefer not to answer” in the question about having experienced physical or mental health problems. Risks that Matter was fielded in September-October 2020.

Source: OECD calculations from Risks that Matter 2020.

Statlink

2.3. To what extent do people tolerate inequality

Preferred economic disparities are lower and more homogeneous across countries

In all countries, what people think economic disparities should be (i.e. “prefer”) is considerably lower than what they perceive. In all the OECD and EU countries covered by ISSP 2009, the median preferred top-bottom earnings ratio – what people think earnings should be – is less than half of the ratio they perceive: 4 rather than 9 (Figure 2.15). Similarly, the Compare Your Income web tool shows that, in the OECD countries for which data are available, the preferred income share of the richest 10% is around 20 percentage points lower on average than the perceived level (Balestra and Cohen, 2021[41]).12

Figure 2.15. Perceived inequalities in economic outcomes are much greater than preferred ones

Preferred levels of inequality, in both earnings and income, are also more homogeneous across countries than perceived levels. The preferred top-bottom earnings ratio ranges from 2 in Sweden to 9 in Chile, compared to 3 and 20 respectively in perceived levels. The preferred income share of the richest 10% is lowest in Norway at 24%, and highest in Poland with 36%, while perceived shares vary from 40% to 60%. Most people actually accept some degree of inequality. Indeed, the median “should-be” earnings ratio is always far from 1, and the average preferred income share of the richest 10% is consistently larger than its equality value (i.e. 10%).

Preferred earnings disparities are larger in more unequal countries

A possible explanation for the weak correlation between concern over income disparities and conventional measures thereof (Figure 2.4) is that people in countries with greater inequality tend to be more tolerant of it.13

Evidence from preferred top-bottom earnings ratios lends support to this hypothesis (Figure 2.16). In countries where gross earnings disparities are greater, so are preferred disparities (as measured in ISSP surveys). Australia and the United States are cases in point. The median respondents in both countries not only perceive high top-bottom earnings ratios – 23 in Australia and 20 in the United States in ISSP 2009, compared to the OECD average of 9. They also prefer them – with ratios of nearly 7 versus the OECD average of 4.

The hypothesis, by contrast, is not supported by evidence from the Compare Your Income webtool. It finds that the preferred income shares of the richest 10% are no higher in high-inequality countries. One possible explanation of the contrasting evidence for and against the hypothesis is that people are more willing to accept higher earnings disparities (as the ISSP surveys show) rather than household income difference (as in the Compare Your Income data) so they adapt their preferences more easily to actual levels. Indeed, preferences might be more homogeneous with respect to disparities in household income, which takes account of taxes and transfers.

Figure 2.16. Preferred income and earnings disparities are higher in countries that are more unequal when it comes to earnings but not income

Another interpretation is that people build their notion of a preferred top-bottom earnings ratio in relation to what they believe is the current level (Osberg and Smeeding, 2006[49]; Pedersen and Mutz, 2018[50]), as confirmed by the close correlation between the logarithms of perceived and preferred ratios at the individual level (0.69 within the 2009 ISSP wave).14 The same does not apply to perceived and preferred income shares of the richest 10%, which are almost uncorrelated at the individual level in the Compare Your Income data. Correlation may be so weak because respondents’ view the richest 10%’s income share as a more distant concept, since they seldom think of themselves as belonging to the richest 10% (Balestra and Cohen, 2021[41]). Respondents are, therefore, more likely to think about the preferred richest 10%’s share of income from a purely altruistic point of view, even if they are part (or may be part in the future) of that group. As a result, answers are more homogeneous and closer to an “idealistic” setting.

Preferred earnings disparities have increased over time

Between the late 1980s and the global financial crisis preferred disparities increased, but by less than perceived disparities (Figure 2.17), so only partly counteracting them. The gaps between preferred and perceived top-bottom earnings ratios is a measure of concern over earnings disparities, because it captures the tension between what people perceive and what they would be willing to accept. In line with Schneider (2011[51]), this gap is calculated as the logarithmic difference between the two ratios, or as the ratio of ratios. Changes in the gap may be attributed to rises in perceived ratios or to increases in preferred ratios (Figure 2.18, Panel A).

Figure 2.17. The preferred top-bottom earnings ratio has increased over time, though less than the perceived ratio

Figure 2.18. Over time, the growing preference for higher earnings disparities has only partially offset perceptions of greater disparities

Although perceived ratios grew in most countries, so did acceptable levels of disparities. As a result, people’s concern over earnings disparities has been weakened by the change in preferences. In fact, if preferred earnings disparities had not increased, the average increase in the gap between perceived and preferred disparities in OECD countries would have been almost twice as large. This offsetting effect has been particularly strong in Australia, but also in Poland and Hungary, countries which started from very low acceptance of wide earnings disparities – in the late 1980s, median ratios were 2/1 and 2/2, respectively (similar to Norway and Sweden in the early 1990s).

In the decade following the global financial crisis there was a slight dip in the preferred top-bottom earnings ratio in most of the countries observed by ISSP up to 2019 (Figure 2.18, Panel B). In Norway, Switzerland and the United Kingdom, the fall compensated for the fall in the perceived ratio and, to a lesser degree in Australia, Italy and Slovenia. However, only in Slovenia did the gap between perceived and preferred earnings disparities fall to the levels of the early 1990s. Finally, in New Zealand and Germany, where the latest ISSP wave was conducted during the COVID-19 pandemic, preferred earnings disparities fell while perceived ones rose, so spelling growing concern.

The increase in the preferred magnitude of earnings disparities between the late 1980s and the global financial crisis was steeper in countries where the perceived extent of disparities grew the most. This trend may reflect “adaptive preferences” – as people become accustomed to living in a less equal society they show increasing tolerance of it (Benabou and Tirole, 2006[52]). However, the change over time in the preferred top-bottom earnings ratio might also be explained by the tendency of respondents to build their notion of “preferred” disparities according to their perceptions of income differences, as discussed above. Although the evidence is not sufficient to choose one explanation over the other, it is important to stress that perceptions of greater earnings disparities tend to be offset by preferences for them, which yields a less pronounced rise in concern over earnings disparities. There is also some evidence from in-survey experiments in Sweden and the United States that people adapt their preferences to information that current levels of inequality are high (Trump, 2018[14]). The evidence to that effect, however, is still limited.15

Preferred levels of earnings and income disparities may also have increased because of the spread of the belief that hard work, rather than luck or personal circumstances, is what matters for getting ahead in life (Mijs, 2019[22]). And such beliefs might actually self-reinforce over time. Indeed, Alesina and Angeletos (2005[53]) propose a model in which the widespread belief that hard work matters more than luck might give rise to a society in which both redistribution and taxes are low. Such a society would enshrine the conviction that individual effort determines individual success, and the initial meritocratic belief would end up being proved correct by reality (Piketty, 1995[54]). Initial international differences in meritocratic beliefs, attributable, for instance, to history, could ultimately lead to two societies with different welfare regimes.

Beliefs in the importance of hard work for getting ahead in life grew in most countries between the late 1980s and the global financial crisis (Figure 2.19 and Mijs (2019[22])). However, unlike the prediction of the model by Alesina and Angeletos (2005[53]) that differences between countries would widen over time, there were in fact signs of convergence, as the countries which changed the most were those that initially assigned less importance to hard work.

Between 2010 and 2019, however, the perceived importance of hard work seems to have fallen back, according to country data available in ISSP 2019. Its average level is once again what is was in the late 1980s and early 1990s. This can partially explain the limited extent of the fall in concern over income and earnings disparities since the global financial crisis, even though the perceived top-bottom earnings ratio has shrunk.

Figure 2.19. The belief in hard work for getting ahead in life grew in the two decades up to the global financial crisis

Round-up: Perceptions of wider disparities explain to a large extent the increase in concern

There are important cross-country differences in both levels of and changes in concern over income disparities. To explain them, it is crucial to disentangle the influence of perceived and preferred disparities in outcomes (such as earnings), perceived intergenerational persistence, and beliefs in the importance of hard work for getting ahead in life.

Columns 1‑3 in Table 2.4 show that greater perceived earnings disparities increase concern over income disparities, while greater preferred disparities reduce them. The belief that parental characteristics matter for getting ahead in life increases concern over income disparities, while the importance of hard work has the opposite effect. Importantly, the literature on experimental surveys confirms the role of all those factors. Information-related experiments endorse the importance of perceptions of economic disparities and intergenerational persistence (Box 2.1), while laboratory experiments confirm the importance of belief in hard work (Durante, Putterman and van der Weele, 2014[55]; Almås, Cappelen and Tungodden, 2020[23]).

	(1)	(2)	(3)	(4)	(5)	(6)
	Cross-country regression, 2009 wave			Country fixed effect regression, 1987-2019
Perceived top-bottom earnings ratio	0.432**		0.296*	0.382***		0.448***
	(0.173)		(0.164)	(0.0642)		(0.0778)
Preferred top-bottom earnings ratio	-0.594**		-0.428*	-0.465***		-0.534***
	(0.214)		(0.213)	(0.103)		(0.124)
Perceived intergenerational		0.711***	0.613**
persistence index		(0.202)	(0.236)
Fraction that believe that hard		-0.441	-0.441*
work matters to get ahead in life		(0.266)	(0.248)
Fraction that believe that hard					0.258	-0.410
work matters to get ahead in life					(0.233)	(0.238)
Observations	28	30	28	62	80	62
Countries				21	25	21

Note: *** denotes statistically significant at the 1% level; ** 5%; * 10%. Country level regressions.

Source: OECD calculation from ISSP 1987, 1992, 1999, 2009, 2019.

Statlink

The most relevant factor in the rise in concern over income disparities between the late 1980s and the global financial crisis was the growing gap between perceived and preferred earnings disparities. Panel A in Figure 2.20 proposes a simple descriptive assessment of the relative importance of the different factors in explaining changes over time in concern over income disparities:16

the gap between perceived and preferred earnings disparities,
perceived intergenerational persistence,
belief in hard work.

The growing gap between perceived and preferred earnings disparities plays a significant role in most countries.17 By contrast, changing perceptions of intergenerational persistence has little impact in most countries (save for Australia, Germany, Slovenia and the United States). Finally, the rise in the perceived importance of hard work reduced concern in all countries, although only to a limited extent in some.

In the decade from 2010 to 2019 (Panel B), the decline of the perceived importance of hard work in all countries save Italy led to rising concern. The contribution of perceived intergenerational persistence was again heterogeneous and, on average, quite small. In Germany and New Zealand, where the last ISSP wave was fielded during the COVID‑19 pandemic, all three factors contributed positively to concern over income disparities.

Figure 2.20. Long run changes in concerns about economic disparities have been mostly influenced by the larger gap between perceived and preferred earnings disparities

References

[39] Acciari, P., A. Polo and G. Violante (2019), “And Yet It Moves”: Intergenerational Mobility in Italy, National Bureau of Economic Research, Cambridge, MA, http://dx.doi.org/10.3386/w25732.

[53] Alesina, A. and G. Angeletos (2005), “Fairness and Redistribution”, American Economic Review, Vol. 95/4, pp. 960-980, http://dx.doi.org/10.1257/0002828054825655.

[16] Alesina, A. and P. Giuliano (2011), “Preferences for Redistribution”, in Handbook of Social Economics, Elsevier, http://dx.doi.org/10.1016/b978-0-444-53187-2.00004-8.

[30] Alesina, A., S. Stantcheva and E. Teso (2018), “Intergenerational Mobility and Preferences for Redistribution”, American Economic Review, Vol. 108/2, pp. 521-554, http://dx.doi.org/10.1257/aer.20162015.

[23] Almås, I., A. Cappelen and B. Tungodden (2020), “Cutthroat Capitalism versus Cuddly Socialism: Are Americans More Meritocratic and Efficiency-Seeking than Scandinavians?”, Journal of Political Economy, Vol. 128/5, pp. 1753-1788, http://dx.doi.org/10.1086/705551.

[44] Ariely, D. and M. Norton (2013), “American’s desire for less wealth inequality does not depend on how you ask them”, Judgment and Decision Making, Vol. 8/3, pp. 393-394.

[41] Balestra, C. and G. Cohen (2021), “Income Inequality through People’s Lenses: Evidence from the OECD Compare your Income Web-tool, *, Paris, forthcoming.”, OECD Papers on Well-being and Inequalities, forthcoming.

[27] Benabou, R. and E. Ok (2001), “Social Mobility and the Demand for Redistribution: The Poum Hypothesis”, The Quarterly Journal of Economics, Vol. 116/2, pp. 447-487, http://dx.doi.org/10.1162/00335530151144078.

[52] Benabou, R. and J. Tirole (2006), “Belief in a Just World and Redistributive Politics”, The Quarterly Journal of Economics, Vol. 121/2, pp. 699-746, http://dx.doi.org/10.1162/qjec.2006.121.2.699.

[24] Benson, R. et al. (2021), Attituted to inequalities, Institute for Fiscal Studies, https://ifs.org.uk/inequality/attitudes-to-inequalities/.

[47] Blundell, R. et al. (2020), “COVID‐19 and Inequalities”, Fiscal Studies, Vol. 41/2, pp. 291-319, http://dx.doi.org/10.1111/1475-5890.12232.

[33] Browman, A., M. Destin and D. Miele (2020), “The Perception of Economic Inequality Weakens Americans’ Beliefs in Both Upward and Downward Socioeconomic Mobility”, Mimeo.

[31] Brunori, P. (2017), “The Perception of Inequality of Opportunity in Europe”, Review of Income and Wealth, Vol. 63/3, pp. 464-491, http://dx.doi.org/10.1111/roiw.12259.

[2] Bussolo, M. et al. (2019), “I Perceive therefore I Demand: The Formation of Inequality Perceptions and Demand for Redistribution”, Policy Research working paper, No. WPS 8929, World Bank Group, Washington, D.C.

[62] Campos-Vazquez, R. et al. (2020), “Perception of inequality and social mobility in Mexico”, AFD Research papers.

[60] Cansunar, A. (2018), “Redistribution by the Rich: Information, Perceptions, and Preferences”, Duke University, PhD Dissertation.

[59] Chambers, J., L. Swan and M. Heesacker (2015), “Perceptions of U.S. Social Mobility Are Divided (and Distorted) Along Ideological Lines”, Psychological Science, Vol. 26/4, pp. 413-423, http://dx.doi.org/10.1177/0956797614566657.

[63] Cheng, S. and F. Wen (2019), “Americans overestimate the intergenerational persistence in income ranks”, Proceedings of the National Academy of Sciences, Vol. 116/28, pp. 13909-13914, http://dx.doi.org/10.1073/pnas.1814688116.

[29] Chetty, R. et al. (2014), “Where is the land of Opportunity? The Geography of Intergenerational Mobility in the United States *”, The Quarterly Journal of Economics, Vol. 129/4, pp. 1553-1623, http://dx.doi.org/10.1093/qje/qju022.

[37] Ciani, E., L. Fréget and T. Manfredi (forthcoming), “Learning about inequality and preferences for redistribution: A meta-analysis of in-survey informational experiments”.

[1] Ciani, E. et al. (2021), “Perceptions of inequality across OECD and EU countries: Long term trends and recent evidence”, OECD WISE Working Paper series No. 2.

[20] Clark, A. and C. D’Ambrosio (2015), “Attitudes to Income Inequality”, in Handbook of Income Distribution, Elsevier, http://dx.doi.org/10.1016/b978-0-444-59428-0.00014-x.

[7] Colagrossi, M., S. Karagiannis and R. Raab (2019), “The Median Voter Takes it All: Preferences for Redistribution and Income Inequality in the EU-28”, Publications Office of the European Union, No. 2019/6, http://dx.doi.org/doi:10.2760/797251.

[21] Daniels, J. and M. Wang (2019), “What do you think? Success: is it luck or is it hard work?”, Applied Economics Letters, Vol. 26/21, pp. 1734-1738, http://dx.doi.org/10.1080/13504851.2019.1593930.

[32] Davidai, S. (2018), “Why do Americans believe in economic mobility? Economic inequality, external attributions of wealth and poverty, and the belief in economic mobility”, Journal of Experimental Social Psychology, Vol. 79, pp. 138-148, http://dx.doi.org/10.1016/j.jesp.2018.07.012.

[58] Davidai, S. and T. Gilovich (2015), “Building a More Mobile America—One Income Quintile at a Time”, Perspectives on Psychological Science, Vol. 10/1, pp. 60-71, http://dx.doi.org/10.1177/1745691614562005.

[36] Domènech-Arumí, G. (2021), “Neighborhoods, Perceived Inequality, and Preferences for Redistribution: Evidence from Barcelona”, mimeo, https://sites.google.com/site/domenechweb/jmp.

[55] Durante, R., L. Putterman and J. van der Weele (2014), “Preferences for Redistribution and Perception of Fairness: An Experimental Study”, Journal of the European Economic Association, Vol. 12/4, pp. 1059-1086, http://dx.doi.org/10.1111/jeea.12082.

[43] Eriksson, K. and B. Simpson (2012), “What do Americans know about inequality? It depends on how you ask them”, Judgement and Decision Making, Vol. 7/6, pp. 741-745.

[13] Fisman, R., I. Kuziemko and S. Vannutelli (2020), “Distributional Preferences in Larger Groups: Keeping up with the Joneses and Keeping Track of the Tails”, Journal of the European Economic Association, http://dx.doi.org/10.1093/jeea/jvaa033.

[18] Fong, C. (2001), Social preferences, self-interest, and the demand for redistribution, http://www.elsevier.nl/locate/econbase.

[11] Franko, W. (2017), “Understanding Public Perceptions of Growing Economic Inequality”, State Politics & Policy Quarterly, Vol. 17/3, pp. 319-348, http://dx.doi.org/10.1177/1532440017707799.

[25] Garret, C. and H. Day (2021), Perceptions of Inequality in the UK, Institute of Fiscal Studies, https://ifs.org.uk/inequality/perceptions-of-inequality-in-the-uk-quantitative-survey-for-the-ifs-deaton-review/.

[8] Giger, N. and D. Lascombes (2019), “Growing income inequality, growing legitimacy: A longitudinal approach to perceptions of inequality”, Unequal Democracies Working Paper, No. 11, University of Geneva.

[3] Gimpelson, V. and D. Treisman (2018), “Misperceiving inequality”, Economics and Politics, Vol. 30/1, pp. 27-54, http://dx.doi.org/10.1111/ecpo.12103.

[38] Hoy, C. and F. Mager (2020), “American Exceptionalism? Differences in the Elasticity of Preferences for Redistribution between the United States and Western Europe”, ECINEQ Working Paper.

[42] Jachimowicz, J. et al. (2020), Inequality in Researchers’ Minds: Four Guiding Questions for Studying Subjective Perceptions of Economic Inequality, Center for Open Science, http://dx.doi.org/10.31234/osf.io/gn2z5.

[19] Karayel, A. (2015), “Income Inequality Tolerance and Preferences for Redistribution in Turkey”, European Journal of Economics and Business Studies, Vol. 1/3, pp. 98-105, http://journals.euser.org/index.php/ejes/article/view/492 (accessed on 7 February 2021).

[61] Kenworthy, L. and L. Mccall (2008), “Inequality, public opinion and redistribution”, Socio-Economic Review, Vol. 6/1, pp. 35-68, http://dx.doi.org/10.1093/ser/mwm006.

[4] Kerr, W. (2014), “Income inequality and social preferences for redistribution and compensation differentials”, Journal of Monetary Economics, Vol. 66, pp. 62-78, http://dx.doi.org/10.1016/j.jmoneco.2014.03.002.

[6] Kuhn, A. (2019), “The individual (mis-)perception of wage inequality: measurement, correlates and implications”, Empirical Economics, http://dx.doi.org/10.1007/s00181-019-01722-4.

[46] Leigh, A. (2013), Battlers and Billionaires: The Story of Inequality in Australia, Black Inc.

[12] Lupu, N. and J. Pontusson (2011), “The Structure of Inequality and the Politics of Redistribution”, American Political Science Review, Vol. 105/2, pp. 316-336, http://dx.doi.org/10.1017/s0003055411000128.

[26] Macdonald, D. (2019), “Trust in Government and the American Public’s Responsiveness to Rising Inequality”, Political Research Quarterly, Vol. 73/4, pp. 790-804, http://dx.doi.org/10.1177/1065912919856110.

[5] McCall, L. et al. (2017), “Exposure to rising inequality shapes Americans’ opportunity beliefs and policy support”, Proceedings of the National Academy of Sciences, Vol. 114/36, pp. 9593-9598, http://dx.doi.org/10.1073/pnas.1706253114.

[17] Meltzer, A. and S. Richard (1981), “A Rational Theory of the Size of Government”, Journal of Political Economy, Vol. 89/5, http://dx.doi.org/10.1086/261013.

[22] Mijs, J. (2019), “The paradox of inequality: income inequality and belief in meritocracy go hand in hand”, Socio-Economic Review, http://dx.doi.org/10.1093/ser/mwy051.

[10] Newman, B., S. Shah and E. Lauterbach (2018), “Who sees an hourglass? Assessing citizens’ perception of local economic inequality”, Research & Politics, Vol. 5/3, p. 205316801879397, http://dx.doi.org/10.1177/2053168018793974.

[57] Norton, M. and D. Ariely (2011), “Building a Better America—One Wealth Quintile at a Time”, Perspectives on Psychological Science, Vol. 6/1, pp. 9-12, http://dx.doi.org/10.1177/1745691610393524.

[48] OECD (2021), Main Findings from the 2020 Risks that Matter Survey, OECD publishing, https://doi.org/10.1787/b9e85cf5-en.

[56] OECD (2021), The impact of shifting demographic and family structures on inequality and its policy implications.

[28] OECD (2018), A Broken Social Elevator? How to Promote Social Mobility, OECD Publishing, Paris, https://dx.doi.org/10.1787/9789264301085-en.

[49] Osberg, L. and T. Smeeding (2006), ““Fair” Inequality? Attitudes toward Pay Differentials: The United States in Comparative Perspective”, American Sociological Review, Vol. 71/3, pp. 450-473, http://dx.doi.org/10.1177/000312240607100305.

[50] Pedersen, R. and D. Mutz (2018), “Attitudes Toward Economic Inequality: The Illusory Agreement”, Political Science Research and Methods, Vol. 7/04, pp. 835-851, http://dx.doi.org/10.1017/psrm.2018.18.

[40] Phillips, L. et al. (2020), Inequality in People’s Minds, Center for Open Science, http://dx.doi.org/10.31234/osf.io/vawh9.

[54] Piketty, T. (1995), “Social Mobility and Redistributive Politics”, The Quarterly Journal of Economics, Vol. 110/3, pp. 551-584, http://dx.doi.org/10.2307/2946692.

[35] Roth, C. and J. Wohlfart (2018), “Experienced inequality and preferences for redistribution”, Journal of Public Economics, Vol. 167, pp. 251-262, http://dx.doi.org/10.1016/j.jpubeco.2018.09.012.

[15] Rueda, D. and D. Stegmueller (2020), “Preferences that Matter: Inequality, Redistribution and Voting”, mimeo.

[51] Schneider, S. (2011), “Income Inequality and its Consequences for Life Satisfaction: What Role do Social Cognitions Play?”, Social Indicators Research, Vol. 106/3, pp. 419-438, http://dx.doi.org/10.1007/s11205-011-9816-7.

[34] Shariff, A., Wiwad D and L. Aknin (2016), “Income Mobility Breeds Tolerance for Income Inequality: Cross-National and Experimental Evidence”, Perspectives on Psychological Science, Vol. 11/3, pp. 373-380.

[45] Swan, L. et al. (2017), “How should we measure Americans’ perceptions of socio-economic mobility?”, Judgment and Decision Making, Vol. 12/5.

[14] Trump, K. (2018), “Income Inequality Influences Perceptions of Legitimate Income Differences”, British Journal of Political Science, Vol. 48/4, pp. 929-952, http://dx.doi.org/10.1017/S0007123416000326.

[9] Xu, P. and J. Garand (2010), “Economic Context and Americans’ Perceptions of Income Inequality*”, Social Science Quarterly, Vol. 91/5, pp. 1220-1241, http://dx.doi.org/10.1111/j.1540-6237.2010.00729.x.

Annex 2.A. Data sources

International Social Survey Programme

The ISSP is a long-standing survey that focuses on social topics. It collects the perceptions and opinions of a representative sample of respondents in a wide set of countries. Each year it addresses a specific subject. The Social Inequality module has been conducted in waves in 1987, 1992, 1999 and 2009. It is fielded by local ISSP committees on a representative sample of a country’s population. The year of the survey varies from country to country, but is usually within 2 years of the “module year” – e.g. 2008-11 for the 2009 module. The 2017 module, which addressed social networks, also included questions on income disparities and preferences for redistribution, but not all the other variables appear in this report. The main variables which do appear are consistent throughout the different years and across different countries. There are some exceptions, such as perceived and preferred earnings for different professions, discussed in the relevant sections of this report.

Eurobarometer 471/2017

The Eurobarometer is a survey carried out on an annual basis to monitor public opinion in European member and candidate countries. It comprises a standard part and a special-issue part. The special Eurobarometer 471/2017 focused on “Fairness, inequality and intergenerational mobility”. It surveyed the population aged 15 or older in the 28 member states, with a sample of around 1 000 respondents per country.

Risks that Matter

The OECD Risks that Matter (RtM) survey is a cross-national survey that examines people’s perceptions of social and economic risks and how well they think government addresses those risks. The survey was conducted for the first time in two waves in the spring and autumn of 2018. The 2020 survey, conducted in September-October 2020, draws on a representative sample of over 25 000 people aged 18‑64 years old in 25 OECD countries.

Consistent with other surveys, RtM is implemented online by Respondi Limited using samples recruited online and over the phone. Respondents are paid a nominal sum (around EUR 1 or EUR 2 per survey). Sampling is based on a modified form of quota sampling with sex, age group, education level, income level, and employment status (in the last quarter of 2019) used as the sampling criteria. Survey weights are used to correct for any under- or over-representation based on the five criteria. The target and weighted sample is 1 000 respondents per country.

Compare Your Income

Compare your income (CYI, www.compareyourincome.org) is a webtool developed by the OECD to give users in OECD countries the opportunity to compare their perceptions of income inequality with statistics on the subject from the Income Distribution Database. To start, people are asked to provide some basic socio-demographic information on their gender, country of residence, age, household size and household net income. Then, they are asked where they think they fit in their country’s income distribution and what minimum income they would need in order to not be considered “poor”. There is also a question on how they think their country’s population is distributed according to income level and how they would like it to look if it was up to them. In addition to these questions, following a modular approach, short ad hoc modules focusing on specific inequality-related issues were added to the survey over the course of the years.

To mitigate the non-representativeness of the CYI samples and achieve more accurate estimates, a weighting scheme was developed. This allowed to balance out and compensate for over- and under-representation of some population groups, in the sample and between countries (for more details, see Balestra and Cohen (2021[41])). After data cleaning, only those country samples with at least 1 500 observations were retained for analysis.

	1987		1992		1999		2009		2017
AUS	1 663	5.7	2 203	5.4	1 672	3.6	1 525	5.5	1 317	6.1
AUT	953	2.2	1 027	2.8	1 016	3.0	1 019	3.6	1 200	1.1
BEL							1 115	3.2	1 001	0.4
CAN			1 002	2.4	974	2.8
CHE	987	3.9					1 229	0.5	1 066	1.5
CHL					1 503	2.2	1 505	0.9
CZE					1 834	0.7	1 205	0.5	1 405	0.5
DEU			3 391	4.6	1 432	3.6	1 395	2.5	1 701	1.9
DNK					1 823	2.8	1 518	3.6	1 079	4.4
ESP					1 211	1.2	1 215	1.6	1 733	4.0
EST							1 005	0.1	1 005	2.2
FIN							880	3.6	1 074	6.1
FRA					1 889	1.0	2 817	1.6	1 489	3.0
GBR	1 212	2.9	1 066	2.5	804	3.6	958	2.7	1 595	5.2
GRC									1 010	0.1
HUN	2 606	4.1	1 250	1.8	1 208	0.7	1 010	0.2	1 007	0.1
IRL									1 004	3.4
ISL							947	0.3	1 450	2.5
ISR					1 208	0.6	1 193	1.2	1 267	2.4
ITA	1 027	1.0	996	0.3			1 084	2.8	1 029	0.7
JPN					1 325	7.8	1 296	5.4	1 609	7.3
KOR							1 599	0.6
LTU							1 023	2.2	1 052	0.3
LUX									504	4.2
LVA					1 100	0.7	1 069	0.7	1 000	1.1
MEX									1 002	1.8
NLD									1 040	1.1
NOR			1 538	1.7	1 268	1.4	1 246	4.5
NZL			1 239	3.8	1 108	3.5	935	3.1	1 357	2.7
POL	3 937	52.8	1 636	5.9	1 135	6.3	1 263	0.9	997	1.1
PRT					1 144	1.0	1 000	0.5	1 089	0.2
SVK					1 082	0.6	1 159	0.5	1 404	0.1
SVN			1 049	1.7	1 006	1.8	1 065	1.3	1 047	1.8
SWE			749	3.1	1 150	1.6	1 137	2.7	1 125	3.5
TUR							1 569	2.8	1 521	0.4
USA	1 564	4.7	1 273	2.3	1 272	6.6	1 581	4.4	1 173	2.6

Source: Yellow shaded cells refer to data from Eurobarometer; in Poland, 1987, half (50.1%) of the sample was not asked the question.

	Observed in period:					Total
	1987-88	1991-93	1998-2001	2008-11	2017-18
AUS	0	0	1	1	1	3
AUT	0	0	0	1	1	2
BEL	0	0	0	1	1	2
CAN	0	1	1	0	0	2
CHE	0	0	0	1	1	2
CHL	0	0	1	1	0	2
CZE	0	0	1	1	1	3
DEU	0	1	1	1	1	4
DNK	0	0	1	1	1	3
ESP	0	0	0	1	1	2
EST	0	0	0	1	1	2
FIN	0	0	0	1	1	2
FRA	0	0	1	1	1	3
GBR	1	1	1	1	1	5
HUN	0	0	0	1	1	2
ISL	0	0	0	1	1	2
ISR	0	0	1	1	1	3
ITA	1	1	0	1	1	4
JPN	0	0	1	1	1	3
LTU	0	0	0	1	1	2
LVA	0	0	0	1	1	2
NOR	0	1	1	1	0	3
NZL	0	1	1	1	0	3
POL	0	0	0	1	1	2
PRT	0	0	0	1	1	2
SVK	0	0	0	1	1	2
SVN	0	0	0	1	1	2
SWE	0	1	1	1	1	4
USA	1	1	1	1	1	5
Total	3	8	14	28	25	78

Notes

← 1. Unless explicitly stated, here and in the rest of the report the focus is only on the share of respondents who strongly agree with the statement “Differences in income in [country] are too large”. We do so for two main reasons: (i) a large majority of respondents agree with the statement; (ii) the strongest variation over time is observed in the share of people who strongly agree. Ciani et al. (2021[1]) provide a more extensive discussion and comparison with alternative measures.

← 2. It should be noted that the 2017 ISSP questionnaire shows some important differences from the 1987, 1992, 1999 and 2009 waves, as it specifically focuses on social inequalities and does not collect the full set of variables used below to investigate and explain the evolution of concerns. In detail, the ISSP 2017 module focuses on social networks. The question about income disparities is designed as the one in ISSP 1987, 1992, 1999 and 2009. However, 1987, 1992, 1999 and 2009 asked the questions about income disparities after asking the respondent to assess earnings in a wide range of occupations and to state what should be a “fair” level of earnings. This might influence answers to the subsequent, more general question about income disparities. The 2019 wave, though, is fully comparable with the previous waves, but has been released only for a limited set of countries and is still being carried out in others. Nevertheless, comparing the data from either wave to ISSP 2009 confirms a slightly downward trend.

← 3. Only a small share of these changes in concerns about income disparities is explained by socio-demographic compositional changes (see Ciani et al. (2021[1])).

← 4. Figure 2.5 and Table 2.1 use only data up to 2017 for two reasons: (i) some of the ISSP 2019 countries are observed in a year for which there are not yet any inequality indicators; (ii) it enable a wider coverage of countries within a single wave, rather than combining data from multiple waves.

← 5. These results are robust to a range of sensitivity checks such as removing period dummies, controlling for year of interview (rather than period dummies), using a first difference estimator, using only series with the old income definition from IDD, excluding data from Eurobarometer (or including a dummy for related data points). Adding data points from the Luxembourg Income Study (LIS) leads to similar estimates for both column (3) and (4), although the estimates for column (4) become statistically insignificant; this might be due to the additional measurement error induced in LIS by merging different sources of data. The results are not influenced by a single country. See Ciani et al. (2021[1]) for the full tables of results.

← 6. Inequality in earnings and income might differ substantially for different reasons. For example, household income also includes non-employment revenues and earnings refer only to individuals without accounting for her household.

← 7. Posing this question is not the same as looking at how individual concern over income inequality has changed over time. In fact, trends in concerns about income disparities (Figure 2.2) actually suggested a decrease in concerns during the last decade.

← 8. Osberg and Smeeding (2006[49]) also show that, according to these ISSP indicators, the United States is not an exceptional case in terms of perceived intergenerational mobility. Nevertheless, Alesina, Stantcheva and Teso (2018[30]) suggest that residents in the United States overestimate the probability that children from poor families could climb the social ladder. The literature is not unanimous on this (McCall et al., 2017[5]; Cheng and Wen, 2019[63]).

← 9. The association in ISSP is weaker than in Risks that Matter. One explanation is that ISSP measures perceived disparities in earnings and intergenerational persistence along with wealth and education, therefore combining different dimensions. Unlike ISSP, both measures in Risks that Matter refer to the income distribution.

← 10. While for perceived inequality of outcomes it is possible to compare perceptions and estimates which refer broadly to the same aspect (e.g. top income shares or earnings disparities), for intergenerational persistence it is more difficult to do so. There are two reasons. First, the index built on ISSP is qualitative and captures two dimensions, one of which (intergenerational persistence of wealth) is not covered by internationally measurable statistics. Second, although there is an interest in capturing intergenerational income persistence, estimates are available only for a few countries, as most available conventional estimates are based on earnings and education.

← 11. Experiencing any health or economic hardship includes physical or mental health problems linked to the pandemic, job-related disruption during the pandemic, or difficulties in making ends meet during the pandemic.

← 12. Results from Compare Your Income regarding the perceived richest 10%’s income share may differ from those of Risks that Matter for multiple reasons, including the period of observation (from May 2015 to May 2020 for CYI, from September to October 2020 for RtM) and the methodology (online opt-in survey users for CYI, online panel-based survey for the RtM). Nevertheless, for the 20 countries included in both tools, the average perceived income share of the richest 10% is quite similar. The correlation is 0.75 and the average difference is only 2.9 percentage points. Results from the 27 OECD countries analysed by Balestra and Cohen (2021[41]) also confirm that the perceived top 10% is correlated with the OECD Income Distribution Database (IDD) values (averaged across available years from 2015). The correlation is 0.5 and the linear fit has a slope of 0.7, statistically significant at the 1% level.

← 13. For example, Roth and Wohlfart (2018[35]) show that individuals who grew up in periods of high income disparities are more tolerant of current levels, even when their perceptions are in line with reality.

← 14. This is the anchoring effect discussed by Pedersen and Mutz (2018[50]).

← 15. The only other experiment available was carried out by Campos-Vazquez et al. (2020[62]), and it does not corroborate Trump’s evidence for Mexico. In fact, they find that providing respondents with information on the actual extent of income inequality or level of intergenerational mobility does not affect people’s preferred levels of inequality and intergenerational mobility.

← 16. The main limit of such exercise is that the different variables used in the decompositions are imperfect measures of the underlying concepts (e.g. perceived top-bottom earnings ratio refer to earnings and not income). However, it is still helpful to understand the relevance and direction of the different contributions.

← 17. The stronger importance of perceived and preferred disparities is highlighted by country fixed-effects regressions (Table 2.3, Columns 4 to 6), which exploit the change over time in the different dimensions. These fixed-effects regressions are, however, based on a small number of countries (21) observed at least twice in all relevant variables (and make it possible to include only the importance of wealthy parents, because the importance of educated parents and hard work are not observed in one of the intermediate waves). Hence the need to treat the results with caution.

Featured topics

Agriculture and fisheries

Climate change

Development

Digital

Economy

Education and skills

Employment

Environment

Finance and investment

Governance

Health

Industry, business and entrepreneurship

Regional, rural and urban development

Science, technology and innovation

Society

Taxation

Trade

Energy

Nuclear energy

Transport

Featured topics

Agriculture and fisheries

Climate change

Development

Digital

Economy

Education and skills

Employment

Environment

Finance and investment

Governance

Health

Industry, business and entrepreneurship

Regional, rural and urban development

Science, technology and innovation

Society

Taxation

Trade

Energy

Nuclear energy

Transport

Countries A - C

Countries D - I

Countries J - M

Countries N - R

Countries S - T

Countries U - Z

Regional and global engagement

Countries

Countries A - C

Countries D - I

Countries J - M

Countries N - R

Countries S - T

Countries U - Z

Regional and global engagement

Publications

Publications

Featured publications

Data

Data

Featured data

News & Events

News & Events

Featured Events

About OECD

About

Engage with us

Work with us

Featured topics

Agriculture and fisheries

Climate change

Development

Digital

Economy

Education and skills

Employment

Environment

Finance and investment