In a democracy, income inequality should in theory correct itself. The poor majority should vote to tax the rich and divide the proceeds among themselves. But that’s not happening in the United States. In fact, inequality has been rising for decades.

Scholars suggest various explanations. Here’s one possibility that few have discussed: Poor voters may oppose a policy of higher taxes and income transfers because they wrongly think they’ll be victims of it.

Defying logic

The idea that in democracies the poor soak the rich goes back to Aristotle. The economists Allan Meltzer and Scott Richard formalized the logic in a classic paper. If the majority gets to set a flat tax rate and share the revenue equally among all voters, then when incomes are especially unequal, the majority should set the rate higher.

But in the United States, things haven’t worked out that way. Rather than expropriating the wealthy, low-income voters in several swing states helped put a billionaire in the White House who then slashed taxes on the rich. As Nicholas Carnes and Noam Lupu showed here last year, polls suggest most Trump voters had above-average income. But the crucial states of Wisconsin, Michigan and Pennsylvania all saw big drops in the share of low-income voters supporting the Democratic candidate. Had large minorities of the poor not backed Trump in these states, Hillary Clinton would have won.

Why don’t poor Americans use their voting power to shrink the income gap? Some observers blame the Democratic Party for abandoning blue-collar workers to cater to educated elites. Others see the influence of racial divisions or the power of tycoons and lobbyists to distort the agenda.

These factors may well contribute. But there’s another piece to the puzzle.

What you don’t know can hurt you

In a recent article, Vladimir Gimpelson and I show that in many countries — the United States included — most people simply don’t know how high inequality is or where they fall in the distribution. Nor can they say reliably how it’s been changing.

On many different surveys, answering a range of questions, the average respondent guesses right about inequality only slightly more often than chance. Respondents do a little better in rich countries than in poor ones — but not by much.

For a citizen to judge how redistribution will affect her, she needs to know how her income compares to those of others. Is she among the rich who pay more in taxes than they get back from public programs — or the poor, who come out ahead?

Many don’t know. In 2015, Ipsos MORI asked respondents in 28 countries to guess the average annual wage. That year in the United States, the correct answer was around $59,000, according to the OECD. The average guess was less than half that — about $26,000. (Even if we suppose respondents had in mind post-tax not pretax income, the guesses were way too low.)

Another survey in 2007 asked Americans about average household income. The median respondent guessed $40,000. According to the Census Bureau, average household income that year was $69,193 — again, much higher than the typical answer.

Suppose that — as the Meltzer-Richard theory predicts — those with below average wages or favor income redistribution while those with higher wages or income oppose it. Clearly, a lot of Americans who fall below the true average think they are above it. That implies that many who would benefit from redistribution wrongly think that they would lose.

The Lake Wobegon effect

This fits with another common bias. People tend to think they are closer to the middle of the income distribution than they actually are. As in Garrison Keillor’s fictional town where all children are above average, residents of many countries like to think they are close to the mean. The poor underestimate how much the rich earn, while the rich underestimate the poverty of the poor.

Some misperceptions appear extreme. For instance, among the 5 percent of U.K. respondents who said they or their family owned two homes, a whopping 40 percent thought they were in the bottom half of the income distribution. Meanwhile, 19 percent of Americans who said they had “often” or “sometimes” gone without enough food in the previous year placed themselves — implausibly — in the highest-earning half.

Across countries, actual levels of inequality bear little relation to the percentage who think government should reduce the income gap. But the level of inequality citizens perceive correlates strongly with demand for more redistribution.

If people are misinformed, would education solve the problem? Not necessarily. Some research suggests people resist changing their minds even when confronted with contrary facts.

For instance, Eric Lawrence and John Sides found that respondents, even after learning average income was higher than they had thought, were no readier to support aid to the poor or job training programs. Ilyana Kuziemko, Michael Norton, Emmanuel Saez and Stefanie Stantcheva, using an online survey, randomized whether respondents received information about inequality, but found this had little or no effect on their attitudes toward income tax rates and transfer programs.

The source of information matters. People are readier to believe a message when it comes from a person — or political party — that usually argues the opposite. If Republicans said inequality was too high and advocated aid to the poor, that would get attention. Of course, that’s unlikely to happen. All parties portray reality in ways that serve their base.

Wealth may be different. Unlike with income, Americans overestimate the asset share of the top 1 percent. And attitudes toward the estate tax seem more responsive to facts. Telling respondents how few Americans are subject to it — less than 1 percent — increases the share that support the estate tax (by 11 percentage points) and that favor raising its rate (by 36 percentage points).

Misperceptions are certainly not the only reason poor Americans resist higher taxes to pay for programs that would benefit them. Culture, racial prejudice and political messaging also contribute. But the desire of millions to see themselves as more or less average may help explain why they don’t use their voting power to shape public policies in their interest.

Daniel Treisman is a professor of political science at the University of California at Los Angeles.