Monday, October 5, 2015

How much would increasing top income tax rates reduce inequality? Good question.

Last week, William Gale, Melissa Kearney, and Peter Orszag (henceforward “GKO”), all of the Brookings Institution, published the result of a simulation ostensibly showing that increasing top-bracket ordinary income tax rates would have little impact on inequality. Unfortunately, their negative conclusion arises more or less automatically from the measures of inequality they report, which do not capture variation in inequality in the tails of the income distribution very well. This morning, John Quiggin made that point, which I expand on below.

The specific scenarios GKO model are increasing the top marginal tax bracket from its current 39.6% to either 45% or 50% and rebating the revenues to low-income households. Their primary analysis simply quantifies the mechanical effect of this policy on income inequality, assuming no behavioral response to tax changes. In this discussion, I confine my attention to the effect of the tax increases on tail inequality and do not discuss the issue of redistribution.

There are two key reasons why these tax scenarios do not affect inequality very much. First of all, the rich earn a great deal of their income in categories other than “ordinary income,” to which these tax rates apply. Those categories are grouped under the somewhat-misleading heading “capital gains,” though tax advisors are skilled at re-categorizing many things to fit the broad definition. This is the nature of the controversy over the “Carried Interest Loophole,” a tax shelter available to employees of Private Equity, Venture Capital, and Hedge Funds—meaning that the beneficiaries of that loophole, who increasingly comprise the top 1% and top 0.1%, would be significantly shielded from the GKO scenarios. News stories in recent years have focused on the low effective tax rates paid by Mitt Romney and Warren Buffett, precisely because those very high-income individuals avail themselves of those loopholes. By looking only at tax rates on ordinary income, the Brookings scenarios ignore them.

The other reason why this scenario doesn’t do much for inequality is that GKO measure inequality by the Gini Coefficient. But the scenarios only affect individuals comfortably within the top 1% of the income distribution. The top tax bracket kicks in at taxable income of $457,601 for married couples filing jointly, which is above the threshold for the top 1% of the taxable income distribution. The Gini Coefficient is insensitive to measuring inequality in that group. That is why Thomas Piketty and his many collaborators on the World Top Incomes Database focus on measured top income shares. Other measures of inequality that GKO consider, like the ratio of 99th to 50th percentile income, are also inadequate for the same reasons. Indeed, one of the authors, Orszag, conceded as much on Twitter, and Lucas Goodman, a graduate student who worked on the analysis, made a similar point on his personal blog as well.

Using data that Goodman kindly provided to my colleague Ben Zipperer, it’s possible to estimate what happens to inequality as measured by the top 1% income share under GKO’s scenario. Goodman reported the Gini Coefficients in GKO’s after-tax income data for both the bottom 99% of the distribution and for the distribution as a whole, either under current tax law or under the GKO scenario of increasing the top bracket rate to 50%. Using the formula for converting partial distribution Gini Coefficients to top income shares derived by Alvaredo (2011), it appears that the policy would reduce the top income share in GKO’s data (after-tax income distribution of tax units) by about one percentage point. That drop is actually quite substantial. The context is not precisely the same, but for comparison, the Congressional Budget Office estimates that the after-tax income share for the top 1% of households increased by a total of 5.2 percentage points between 1979 and 2011, so the reduction in the top 1% income share as a result of the GKO scenario is just under 20% of the total increase in inequality over the whole period the CBO analyzes—a much more substantial impact than the one highlighted by GKO.

GKO are continuing a debate that played out this past year concerning policies that would combat income inequality and wage stagnation. At a public appearance sponsored by the Hamilton Institute in February, Lawrence Summers said that focusing on education is a distraction from the challenge facing workers, which is that “there aren’t enough jobs.” The following month, he, Kearney, and Bradley Hershbein published an analysis that argued that a 10% increase in the share of men with college degrees would not substantially impact tail inequality. John Schmitt and I commented on that analysis here, and the authors responded here. All of us agreed with the conclusion that dramatically expanding higher education attainment would not affect inequality very much because it wouldn’t impact the tail of the income distribution. Now, in essence, GKO are saying “well, sure, education wouldn’t affect tail inequality, but neither would higher taxes on the rich,” and they’ve constructed a scenario and employed ill-fitting metrics to support that rhetorical move.

A final word about behavioral responses: the most important research on the effect of top marginal tax rates on the behavior of those liable to pay them is the recent paper by Piketty and his coauthors Emmanuel Saez and Stefanie Stantcheva (henceforward “PSS”), which my colleague Nick Bunker summarized here, and which I discussed before here. The paper describes three elasticities by which top tax rates affect economic behavior. Aside from the classic “supply-side” elasticity, which conjectures that people work less when they’re taxed more, there are two additional responses. PSS’s second elasticity (not original to their paper) is the re-categorization of labor income as capital gains or some other tax-preferred category, and their third elasticity (which is their original contribution) is that in response to higher taxes, rich individuals face a diminished incentive to bargain for a share of the corporate pie.

If the scenarios GKO model were in fact implemented with no other changes, top-bracket taxpayers would enjoy an even larger incentive to re-categorize their income as capital gains than they already do, limiting the ability of such rate changes alone to affect inequality without also tightening up the loopholes available to the rich. On the other hand, they would also have less reason to bargain hard in wage negotiations, which would increase the effect of such a policy on inequality. PSS provide evidence that the third elasticity is more important in explaining behavior than the second. Furthermore, they show that declines in top marginal tax rates across countries are the primary driver of rising tail inequality, implying that a reversal of that policy would have the opposite effect. These questions are far from settled in the literature, so future discussion of the effect of top tax rates and the tax system more broadly on inequality should aim at confirming or disproving the arguments in PSS, not on scoring rhetorical points.