Rawls Part III: Progressive Taxation and Tradeoffs Between the Minimum, Mean, and Maximum Levels of Income(long)
Part I here, Part II here; a brief post on Rawls here.
Both Bailey and Alterman referenced Rawls to justify redistribution, and Rawls himself was strongly in favor of redistributing income (“Social and economic inequalities…[must be] to the greatest benefit of the least advantaged members of society”). But this is a fairly extreme position; it rejects any policy that makes most people better off while making some at the bottom a little bit worse off. Earlier, I said this entails “an extremely high degree of risk-aversion”. Why? Suppose you are in the Original Position, designing the structure of society and its rules, but not knowing where you will be born into the world. Consider a simplified example, with two possible states of the world (holding all else equal; e.g., prices, choices of goods, freedoms,…):
System A: 10% of the population makes $10,000/year and 60% make $40,000/year, and 30% make $100,000/year.
System B: 10% of the population make $9000/year and 90% make $90,000/year.
Rawls’ take implies that you are only concerned about what happens to you if you land in the “least-advantaged” sector of society (no amount of increased benefit if you get lucky can outweigh the decreased benefit in the bad state), so System A is clearly preferable. But many people, liberals included, would argue that System B is better. For example, they might point out that in System A, average income is $56,000 while in System B it is $81,000.
If you prefer B, does that imply that you are really saying that the goal of economic policy should be to maximize average income? (Note: maximizing average income is equivalent to maximizing total income, meaning that any set of policies that does one necessarily does the other). This latter view was first formally decribed by John Stuart Mill’s mentor, Jeremy Bentham (e.g., Of the Principle of Utility) and was also advocated by one of Rawls’ contemporaries, John Harsanyi (who shared the 1994 Nobel Prize in Economics in 1994 with John Nash and Reinhard Selton). But to evaluate the idea of maximizing average (or total) income, consider System C:
System C: 95% make $10,000 and 5% make $5 million
In System C, average income is $259,500 but most people reject C in favor of B–the reward is not worth the risk even though the expected income is higher in System C (ex-post, the fortunate 5% might argue strongly that System C is very just).
So can anything definitive be said about the ideal, or just, tradeoff between the degree of risk and inequality in society and the average level of income? Some political scientists and economists have tried, using experiments and surveys, and the results are consistently in favor of redistributing income, but never to the extent that Rawls advocated.
In 1992, Political Scientists Norman Frohlich and Joe Oppenheimer published a book(see a review here), Choosing Justice: An Experimental Approach to Ethical Theory, that included the results of a series of experiments. They took groups of five undergraduates and explained Rawls’ maxi-min principle (maximize the well-being of the least well off) and Harsanyi’s utilitarian principle (maximize the average level of well-being), and also described two in-between rules that involved reducing the average in exchange for increasing the minimum possible income. They then asked the students to design rules for how to divvy up society’s income. The experiments were set-up so that the students’ could only increase the minimum payoffs by also decreasing the average. After they designed rules, each group member randomly drew their place in society according to the rules they designed while behind the veil, and actual cash was paid out accordingly. The experiments were akin to presenting systems like the ones I described earlier and saying pick the one you like best, we’ll randomly draw your place in that system, and pay you according to where you land.
83 groups were told that they had to unanimously agree on a system, and all were able to do so (this is important, as it shows that the Veil is a useful construct–people actually can agree on a set of rules). But of those 83 groups, only 10 chose Harsanyi’s rule of maximizing average payoffs, and just 1 chose Rawls’ to maximize the lowest payoff they could receive. The other 72 groups all were willing to accept a reduction of their lowest possible payoff in exchange for an increase in the average payoff, but not to the extent of maximizing the minimum possible income level.
A similar study (see Table 1, p. 7) using students from India found similar results-the students were consistently willing to accept reductions in both the maximum income and the average income in exchange for increases in the floor income, though not nearly to the degree Rawls argued. Here is a subset of their results:
So, comparing B1 to A, all but 4% of those surveyed were, holding the mean constant, willing to reduce the maximum payout by 14,000 in exchange for increasing the minimum payout by 7,000. That is, they favor a society that redistributes income downward. But only one third of the same students thought that B2 was preferable to A–for the remaining two thirds, increasing the floor by 1,250 was not worth reducing the mean and the maximum by more than 50%. Johansson-Stenman, Carlsson, and Daruvala (Economic Journal, April 2002, Vol. 112 Issue 479, p384) found similar results using European students.
What does this all mean? Appealing to Rawls and the Veil of Ignorance as a justification for progressive taxation (as opposed to either flat or regressive) is on balance valid, in the sense that real people when asked to make decisions from behind a pseudo-veil prefer increasing the minimum even when it entails some reduction in the average and maximum levels of income. But that only addresses the question of whether taxes should be progressive at all, not the question of how progressive they should be.
The data do exist to do a more directly relevant experiment. We could assemble statistics on the income distribution (minimum, percentiles, mean, and maximum) for various countries (the OECD has such data) and then replicate these studies, but to my knowledge this has not been done. Could economists use surveys like this to evaluate specific policy proposals, like the President’s latest tax package? Sure, just as soon as they find a way to agree on the distributional impacts of various proposals (meaning don’t hold your breath).
P.S. While he was in favor of redistribution in general, Rawls was not politically active and did not advocate using his logic to evaluate specific policy proposals. Matthew Yglesias, who went to Rawls’ memorial service, recounts the words of Rawls’ colleague, Tim Scanlon [Matt’s words]: “Scanlon then noted that perhaps times were changing and mentioned the Rawls reference on last night’s West Wing. Then he got all professorial and noted that the veil of ignorance is not supposed to be applied to political issues in isolation, but rather to the basic structure of society as a whole. The tax system is, of course, part of the basic structure, but he cautioned against looking at it in isolation from the rest.”
P.P.S. Thanks to Kevin Drum for helping me get rid of the big open space.