Reference points, loss aversion, and redistribution
Economists, policymakers, and the general public are all well aware that the United States has a much higher level of income inequality than other developed countries. But what sets it apart from other nations isn’t the distribution of income but rather the level of redistribution. With high levels of inequality, economists would expect the median U.S. voter to push for more redistribution as the top earners pull further away from those on the middle and bottom of the income ladder.
So why doesn’t the U.S. government redistribute more?
A paper presented at the just-concluded annual Allied Social Sciences Associations meeting in Boston can shed light on this puzzle. The paper, by Columbia University economists Jimmy Charité and Raymond Fisman and Princeton University economist Ilyana Kuziemko argues that preferences for redistribution are affected by what the authors call “reference points.”
One of the insights of behavioral economics is loss aversion. In neoclassical economics, the utility gained from an extra dollar should be the same (but in the opposite direction) as the utility lost by losing a dollar. But research by behavioral economists and psychologists finds that losing a dollar causes a lot more disutility than the utility gained by gaining a dollar. This effect is why it feels much worse to lose $5 than to find a five-dollar bill on the ground.
Charité, Fisman, and Kuziemko present data that indicates loss aversion may explain why U.S. voters don’t redistribute more. To get at this problem, they run an experiment where they present respondents with a hypothetical situation in which two individuals receive money based on a coin flip. One person is given $15 and the other $5. The respondents are then asked to redistribute money between those two hypothetical people.
But the respondents are split into two groups and read slightly different prompts. The first group is told that the two hypothetical people receiving the money will only know the final amount of money they receive, not the money based on the coin flip. If the respondents redistribute $5 then they know the person who initially got $15 will only know in the end that the total amount received is $10.
In contrast, the second group is told that the two hypothetical people are aware of the money they received from the coin flip before redistribution. So again, if the respondents redistribute $5 then one of the people would know they had $15 but ended up with $10.
The results of the experiment show that these reference points actually change the level of redistribution quite a bit. In the first group, the respondents eliminate 94 percent of the income gap. But the group that has reference points and knows the recipients are aware of the redistribution only eliminates 77 percent of the gap. The level of redistribution is decreased by about 20 percent once reference points are included.
Of course, consideration of reference points isn’t the only factor explaining preferences for redistribution. Individuals might think that people deserve to keep a large portion of the money if it is based on skill or merit. Indeed, when the authors changed the reason for the distribution of income from a coin flip to SAT scores, the amount of redistribution declines to 56 percent. The decline in redistribution due to reference points is then about half of the decline caused by the distinction between luck and merit.
What’s so interesting about Charité, Fisman, and Kuziemko’s paper is that it indicates that the general public’s preferences for redistribution might be different from what is assumed in the classic optimal taxation research. Their research isn’t the first to point this out, but it provides more proof that the average person might not be a strict utilitarian. This in turn means that the classical economic model that undermines a fair bit of the conversation about the proper level of progressive taxation might be based on some flawed assumptions. The importance of that fact shouldn’t be lost on economists and policymakers.