Recently in this space, I criticized an op-ed that claimed to resolve a paradox related to inequality and public policy. Ilyana Kuziemko and Stefanie Stantcheva argued that while Americans are “deeply troubled about the current level of income inequality,” support for government policy to reduce it is low. Based on a series of randomized experiments they conducted with Emmanuel Saez and Michael Norton, Kuziemko and Stantcheva speculated that rising inequality has eroded trust in government, resolving the paradox.
In my previous essay, I argued that there is little evidence to indicate that Americans are particularly concerned about inequality, so their lack of interest in having the government intervene should be unsurprising. Here I want to draw attention to a problem with the conclusion of Kuziemko and her colleagues that providing people with information about inequality reduced trust in government.
In their experiment, some survey respondents were provided information about their ranking in the income distribution and about inequality levels. Receiving this information produced a decline in expressed levels of trust in government. Kuziemko and her colleagues conclude that,”emphasizing the severity of a social or economic problem appears to undercut respondents’ willingness to trust the government to fix it-the existence of the problem could act as evidence of the government’s limited capacity to improve outcomes more generally.” But the information in their survey did not simply emphasize the severity of inequality, it exaggerated economic hardship.
Respondents randomly selected to receive information about inequality first input their “annual household income” and were told the share of “US households” that earn less than their own “household.” But the information the survey gave respondents made them feel richer than they were. I typed into the survey form the 2011 median household income according to the Census Bureau-$50,054. The survey should have told me that “my” household was richer than 50 percent of American households-that’s what the median is. Instead, I was told I was richer than 66 percent of households.
In fact, what the information provided by the survey told the subject was the percentage of tax returns that have less gross income than the household income she reported. Tax returns are not households. Two roommates living together, a cohabiting couple, a married couple filing separate returns-all of these constitute one household but two tax returns. More to the point, a sixteen-year-old burger-flipper or a fulltime college student with a work-study job are also distinct tax returns even if they live at home. Furthermore, gross income on tax returns (AGI with adjustments put back in) is not “household income” as most people think of it. For example, non-taxable public transfers-including most Social Security benefits and all welfare benefits-are excluded. So are the tax-favored employee benefits commonly deducted from paychecks, such as health insurance premiums, retirement plan contributions, and flexible spending accounts.
The result of these differences between the income of households and the gross income of tax returns is that the median for the former is quite a bit bigger than the median for the latter (and the same is true for other parts of the income distribution, such as the “richest ten percent” or the”poorest third”). The survey tool reports that $33,800 is the median “household income”-one-third less than the actual median.
The respondent, then, “learned” that she was richer than she was, and if she correctly thought that her standard of living was average before responding, she learned that it was better than average. More people were doing worse than her than she thought, and fewer people were doing better than her. The next step in the survey drove that home by inviting her to move a slider to see how “households” with different income levels rank compared with other households. This step reinforced that Americans were poorer than they actually were.
Different subjects were shown additional screens subsequently. However, everyone randomized to receive the information about inequality proceeded through the rest of the survey-with its questions about policy preferences and trust in government-having been given this overly-negative data about how Americans are doing economically. The subjects randomized to bypass the informational screens were not primed in this way. The design of this experiment does not allow us to assess whether getting accurate information about the distribution of household income reduces trust in government. Instead, trust in government may be eroded by getting anxiety-provoking (and inaccurate) information.
Interestingly, Kuziemko and her colleagues report results from a separate experiment they conducted indicating that among below-median households, being primed with negative information about the state of the economy reduces opposition to inequality and support for redistribution and for progressive approaches to deficit reduction. It may be that attempting to convince middle class Americans that economic insecurity is more pervasive than it is will prove counterproductive to those who wish to help the truly insecure.
I have argued elsewhere that conventional accounts on the left do, in fact, systematically overstate both the extent of economic insecurity and the strength of the evidence that income inequality is harmful. It would be a regrettable irony if an excessive and distorted focus on inequality turns out to be more harmful to struggling families than income inequality itself.