Matthew Yglesias has a post on how progressive taxation can improve the pre tax distribution of income. Now, for a long time, I have been meaning to write a semi scholarly paper on beneficial incentive effects of taxes, thus I am even more than usually peeved at being shown up by a philosophy concentrator half my age.

Yglesias' point is that progressive taxes give firms an incentive to distribute their payroll equally. He notes as evidence that in the 50's with high income taxes and no great society programs, the US income distribution was much more equal than it is now. I add that high tax countries have much more equal pre-tax income distributions than the US does.

I think the argument is clearly sound. The magnitude of the effect is hard to estimate, but it is hard not to believe that there is some such effect. The closest thing I know to this argument in the Economics literature is an article by Macdonald and Solow in the AER in 1984 which Yglesias probably didn't read in nursery school. To me a point is that, even with very strong assumptions, economic theory suggests that no tax market outcomes are Pareto efficient which is a very weak result compared to welfare maximizing.

There are many reasons why incentive effects of taxes might be welfare improving. Actually most of the other arguments imply that taxes can be part of a Pareto improving intervention.

1) also in the post if executive compensation is inefficiently high because there are imperfect mechanisms to enforce shareholder rights, then the effect of taxes on executive compensation can bring the economy towards the efficient outcome.

2) When people talk about economic incentives they often use phrases like "selling out" (at least they used to). This implies that people believe that there is a choice between doing something socially useful and making money. It contrasts wildly with the idea that the social value of work is equal to its market return.

Imagine for a second that the people who say "selling out" are not totally confused. Let's say that they know of cases in which they can chose between helping society and making money (examples below). Thus greed and altruism compete. Progressive taxation weakens the greed motive. This can be good.

It is clear why such effects don't appear in standard simple economics models. According to standard simple assumptions, economic agents are totally selfish so they balance greed and laziness not greed, laziness and altruism. No one really believes people are selfish, but economists tend to assume that it is usually a relatively harmless assumption when considering behavior in response to taxes.

So the promised examples. How can income maximizing work and social welfare maximizing work differ ? The simplest case (which I really honestly thought of on my own long ago) is that doing something for a poor person can be worth more utils and fewer dollars than doing something similar for a rich person. Consider two doctors, one works in Beverly Hills, the other treats homeless people. The first should certainly have a higher income. Most people would guess that the second does more for human health. If you think homeless people are un helpable think of some other underserved group.
This is close the Yglesias' argument. The basic problem with the market is that the happiness of a rich person counts for more in dollars than the happiness of a poor person. This means that how much you help people (in utils) and how much money you make are different. This means that there is a conflict between the desire to help people and the desire to make money. This means that the incentive effects of taxes can be good.

Now more extreme examples. One way in which greed can be bad and even Pareto inefficient is if it leads to corruption of public officials. Now is the corruption is achieved with bribes, taxes are no help as bribes are not reported. What about the revolving door ? In the model public policy is distorted by bribes paid in the form of high wage jobs (implicitly) promised to the public officials after they retire. I don't for a minute address the question of whether this model has anything to do with reality (perish the thought). If public officials had any shame about selling out and taxes were highly progressive, this form of corruption wouldn't work.

It is often hard to prosecute fraud. Also breach of implicit contracts is legal but may have bad economic effects similar those due to, well, breach of contract. If there are implicit contracts greed is not so good. That is, at least, it is not the efficiency maximizing incentive

For all of the greed vs laziness vs altruism arguments it is impossible to evaluate the sign of the effect on welfare of the incentive effects of taxes without knowing which of the two choices (greed vs altruism or greed vs laziness) is more elastic. I sure guess that the bad incentive effects are stronger than the good ones (I'm lazy as mud). However, the good effects are completely ignored in most welfare calculations.

There are other cases in which incentive effects of taxes can be desirable.

One is if there is dissipative signalling. This is a Harvard topic (do to A Michael Spence who went on to be dean of the FAS). In a simple model, students learn nothing useful at university but go there to prove they are smart enough to get through it without too much effort. It is possible two write a model in which it is rational to go to university even though nothing useful is taught (job market signalling) The level of pointless and exhausting signalling (showing off) required to convince employers that one is smart is endogenous. Progressive taxation implies that less effort is wasted on signalling. It is possible to write models in which progressive taxation is Pareto improving.

Now greed is not the only problem here. Many of Dean Spence's former students don't show off for money.

I am sure there are many other examples.