How Much Do Taxes Matter?

Christina and David Romer’s new paper, “The Incentive Effects of Marginal Tax Rates: Evidence from the Interwar Era,” is available as an NBER working paper (if you are so lucky). Given the current debates about taxes, the paper is likely to garner some attention.

In the central section of the paper, Romer and Romer regress reported taxable income against the policy-induced change in marginal after-tax income share. The after-tax income share is the percentage of your gross income that is left after taxes; policy-induced changes are those caused by tax changes rather than be macroeconomic changes. They do this for the top 0.05% of the income distribution, broken down into ten sub-groups by income, because the income tax only affected the very rich during the interwar years.

Their headline finding is that “The estimated impact of a rise in the after-tax share is consistently positive, small, and precisely estimated” pp. 15–16). They find an elasticity of taxable income with respect to changes in the after-tax income share of 0.19.

Advocates of lower tax rates are sure to seize on this as evidence that higher tax rates depress incentives to work. But that’s hardly what the paper says. First of all, the Romers’ elasticity estimate is lower than earlier empirical estimates that are largely based on the postwar period. To put this in perspective, an elasticity of 0.19 implies that tax revenues would be maximized with a tax rate of 84 percent; that is, you could raise taxes up to 84 percent before people’s reduced incentives to make money would compensate for the higher tax rates.

Second, remember that this is a study of the super-rich: not the top 1%, but the top 0.05%. These are the people whom one would expect to have the highest income elasticity, precisely because they don’t need the marginal dollar. Elasticities tend to be lower for ordinary people because they need to cover their expenses.

Finally, the left-hand-side variable for the main regression is reported taxable income. Taxable income can change both because people are earning less income and because they are engaging in tax strategies to reduce their taxable income. As Emmanuel Saez, Joel Slemrod, and Seth H. Giertz conclude in “The Elasticity of Taxable Income with Respect to Marginal Tax Rates: A Critical Review” (pp. 49–50):

while there is compelling U.S. evidence of strong behavioral responses to taxation at the upper end of the distribution around the main tax reform episodes since 1980, in all cases those responses fall in the first two tiers of the Slemrod (1990, 1995) hierarchy—timing and avoidance. In contrast, there is no compelling evidence to date of real economic responses to tax rates (the bottom tier in Slemrod’s hierarchy) at the top of the income distribution.

In other words, recent U.S. history shows that when you raise taxes on the rich, they don’t stop trying to make money: they just pay their lawyers and accountants more to avoid paying taxes. The solution to that is a simpler tax code with fewer exclusions and deductions.

This post originally appeared at The Baseline Scenario and is posted with permission.

One Response to "How Much Do Taxes Matter?"

  1. Carlos Webb   March 26, 2012 at 7:02 am

    Why is it a bad thing to de-incentivize work among the well-to-do?

    Heavy taxation only counts as unfair when it burdens those in the lower ranges of income who don’t have enough to survive. In pre-Revolutionary France, the nobility was exempt from most forms of taxation. It didn’t notably incentivize them to do much of anything…and it cost many of them their heads.

    Debates over incentives and marginal utility in economics are silly, and reflect the pre-suppositions of the disputants. The real issue is not efficiency, but entitlement. And it needs to be solved, like any moral argument, through collective debate as expressed through voting preferences — not by appeal to false economic relationships.

    I’m in political sympathy with the Romers, but I find the notion of taking one gross macroeconomic variable (“cause”) and another gmv (“effect”), and running a regression line thorough them, then declaring this to be proof of a state of mind (incentive!) to be misguided.