Systematically Biased Beliefs About Inequality
By Bryan Caplan
I am a behavioral political economist. I think most political economy models are worthless. Unless you start with empirically sound assumptions about voter cognition and motivation, you’re wasting your time – and the time of everyone who reads you. What makes my stance especially stressful is that behavioral economists strangely neglect political economy. Most of the scholars with the tools to fix political economy are otherwise occupied.
Fortunately, thanks to noble exceptions, the intellectual climate is slowly improving. Case in point: Gimpelson and Treisman’s new NBER working paper on “Misperceiving Inequality.” The paper’s great from beginning to end, but here are some highlights.
What if the masses have little notion of how much wealth the elites have accumulated and whether the gap is growing or shrinking? What if even the rich cannot gauge how strong is the motive for the poor to revolt? In such cases, the neat link between actual inequality levels and political outcomes evaporates. The goal of this paper is to show that such uncertainty and misperception are ubiquitous. We present evidence from a number of large-scale, cross-national surveys that in recent years ordinary people have known little about the extent of income inequality in their societies, its rate and direction of change, and where they personally fit into the distribution. What they think they know is often wrong. This finding is robust to data sources, definitions, and measurement instruments. For instance, perceptions are no more accurate if we reinterpret them as being about wealth rather than income.
Fun self-referential argument:
A strange inconsistency underlies much recent scholarship. On the one hand, theories assume that individuals correctly perceive the income distribution. On the other hand, scholars complain that the data available to test these same theories–in developed democracies and, even more so, in poorer, less free societies–are “dubious” (Ahlquist and Wibbels 2012) and “massively unreliable” (Cramer 2005). Yet, if experts throw up their hands at the quality of the data, it is strange to assume the general public is better informed. And if analysts fault the figures available today–despite the most sophisticated statistical agencies the world has ever seen–data quality must have been much worse during the nineteenth century heyday of revolution and democratization.
The implications of this point for theories of redistribution, revolution, and democratization, are far reaching. If these are to be retained at all, they need to be reformulated as theories not about actual inequality but about the consequences of beliefs about it, with no assumption that the two coincide. We show that, although actual levels of inequality–as captured by the best current estimates–are not related to preferences for redistribution, perceived levels of inequality are… The actual poverty rate correlates only weakly with the reported degree of tension between rich and poor; but the perceived poverty rate is a strong predictor of such inter-class conflict.
Gimpelson and Treisman heavily rely on the ISSP survey, which showed respondents around the world five different income distributions, then asked them which one best-described their own country. Results:
Respondents turn out to be wrong about their country’s income distribution most of the time. Worldwide, 29 percent of respondents chose the “correct” diagram if we refer to their country’s post-tax-and transfer Gini and 24 percent got it right if we use the pre-tax-and-transfer measure. For reference, a purely random choice among the five possible answers would get the answer right 22.5 percent of the time for post-tax-and-transfer incomes and 20 percent of the time for pre-tax-and-transfer incomes.12 In other words, respondents worldwide were able to pick the “right” diagram only slightly more often than they would have done if choosing randomly.
Were most people at least close? To check this, we examined what proportion of respondents were within one diagram of the correct one (for instance, if the correct diagram was B, we measured how often the respondents picked A, B, or C). With only five options to choose between, getting within one place of the correct option is not a very difficult task. Picking randomly among the five diagrams, respondents should be within one place of the correct diagram 68 percent of the time if focusing on post-tax-and-transfer income and 43 percent of the time if focusing on pre-tax-and-transfer income. In fact, for post-tax-and-transfer income they were right 69 percent of the time, just one percentage point better than if they picked randomly.
G&T have a neat table summarizing the flimsy correlations between real and perceived inequality. Since it’s hard to read, I’ve transcribed it:
G&T go on to show that people around the world imagine they’re near the median income. Few people think they’re poor, and almost no one thinks they’re rich. Only 1% of people who own a second home think they’re in the top decile of their country’s income distribution!
Owning two houses is usually a sign of wealth. In all 40 LiTS countries, at most one in four respondents said that his or her family owned a second residence, and in all but three countries the frequency was less than one in six. Yet most such property owners did not consider themselves especially rich. On average, 60 percent of the secondary residence owners placed themselves in the bottom half of the income distribution. In Uzbekistan, only three percent of respondents lived in households with a second residence, yet almost two thirds of these thought their incomes were below the national median. Such anomalies were somewhat rarer in the developed countries. Still, in France, Italy, and Great Britain, 40 percent or more of second residence owners placed themselves in their country’s bottom half.
What about changes in inequality? Here, the correlation between reality and belief is actually negative.
The coolest part of the paper, though, are the regressions of support for redistribution on objective and subjective measures of inequality. As every shrewd politician suspects, perception crushes reality:
[N]either the pre-tax nor the post-tax actual income Ginis are positively related to support for redistribution at either the country or the individual level. However, perceived inequality is highly significant in both cases. In countries where inequality was generally thought to be high, more people supported government redistribution. But demand for redistribution bore no relation to the actual level of inequality. In fact, given the average belief about inequality, higher actual inequality was associated with lower demand for redistribution. Breaking down perceptions into their general and idiosyncratic components, we found a stronger effect of the general perception in the country than of the individual’s idiosyncratic perceptions. Still, both seemed to matter in the way expected.
The same goes for perceived class conflict:
At the country level, post-tax-and-transfer inequality was significantly associated with greater reported tension between classes, although pre-tax-and-transfer inequality was not (Table 10, panel A). However, the effect of actual inequality was dwarfed by that of perceived inequality, which was about three times larger. And the effect of actual inequality disappears if one controls for the country’s income and population (column 5).