The Washington PostDemocracy Dies in Darkness

Should the top tax rate be 73 percent?

Top tax rates used to be much, much higher. Peter Diamond and Emmanuel Saez want to go back to that. (Visualizing Economics)

Most arguments about tax brackets in the United States are over a percentage point or two. Obama wants the top rate to rise 4.6 points to 39.6 percent; Republicans want it to stay at 35 percent. Despite the revenue costs of low rates, some self-styled deficit hawks want the top rate to go even lower. Simpson-Bowles and Domenici-Rivlin, for instance, both put it at 28 percent.

But some tax experts think this is all much too timid. Two economists, Nobel laureate Peter Diamond at MIT and Emmanuel Saez at Berkeley, have argued that the optimal top bracket is 73 percent, higher than at any point in the United States since the 1960s. Many analysts get sticker shock at that number, but the Diamond/Saez paper lacked a detailed response until AEI released a paper by Aparna Mathur, Sita Slavov and Michael Strain attempting to debunk it.

The disagreement is all about what the word "optimal" means. Economic models often aim at maximizing "welfare," an amorphous term usually used to refer to the degree to which preferences are satisfied, to which people get what they want. Welfare, it is important to stress, is not just money.

That's because economists assume that money becomes less valuable as people get richer. If Bill Gates were to lose $1 billion tomorrow, his life wouldn't get appreciably worse. If a poor American were to lose only $10,000, that'd be a catastrophe. This result is backed up by the empirical literature. Researchers have found that "emotional well-being" maxes out at an income of about $75,000, bolstering the idea that an extra $1 doesn't do the upper-middle and upper classes any good.

How much less valuable money gets, however, is a source of considerable disagreement. Diamond and Saez, in keeping with most economists, assume that it decreases until it isn't very different from zero. $1, in this view, is almost literally meaningless to top earners, just as the empirical studies have found. Mathur, Slavov and Strain dispute this, for three reasons.

First, they argue that this understates the value of that $1 -- not to the top earner, but to lower earners. The reason is that high marginal tax rates discourage people from going into high-earning professions that may be socially valuable. If a bright biology major is scared away from med school by high taxes, that's bad not just for him but for all his potential patients.

Diamond and Saez argue that there's no empirical evidence for effects of this type. And indeed, it's hard to see those effects when one looks at the U.S. in the 1950s, '60s and '70s, when tax rates varied from 70 to 90 percent. The two sides also disagree about how people respond to taxes in the short run. Diamond and Saez argue that high earners' income varies only mildly in response to changes in taxes, whereas Mathur, Slavov and Strain think the effect is larger.

But the biggest point of divergence between the two sides is not actually economic. In a rarity for an economics paper, Mathur, Slavov and Strain concede that the biggest problem with Diamond and Saez is that they're wrong about political philosophy, about what we should be caring about when we set tax policy.

Diamond and Saez are utilitarians. They want a tax policy that maximizes welfare, full stop. That implies taking money from people who don't get much welfare out of it (the rich) and giving it to those who get a lot of welfare out of it (the poor). How people got that money is immaterial. That bothers Mathur, Slavov and Strain. "Social welfare in those models does not depend on how the rich got rich," they note. "Did the rich invent products that most of us can’t imagine living without? Or did they get rich by lobbying for favorable regulations for their industries?" The former, they argue, is more legitimate than the latter, and more deserving of protection from the government.

It's a good reminder that when they get down to brass tacks, you can't always resolve policy arguments like this merely by determining what the data says or what a good model implies. There are inevitably value judgments involved.