How to Tax Capital Gains to Reduce Inequality

12/30/2012 12:00:00 PM
In my previous post I noted that Krugman's argument about capital-biased technological change doesn't really warrant his conclusion--that we should tax capital income to finance entitlements. Perhaps a more succinct summary of my point would be that Krugman's argument directly contradicts the best known result from the Ramsey optimal tax literature (named after Frank P. Ramsey, the economist who started the literature in the 1920s), which is that in the long run we should cut capital gains taxes to zero.

The problem with the Ramsey model, however, is that it does not allow for heterogeneity between households, nor does it allow for non-linear tax functions. This is highly problematic since that makes the Ramsey model basically useless both for representing the real world--the tax function in the US is actually highly non-linear--as well as for studying income inequality. There is, however, a completely alternative modeling paradigm for studying optimal tax policy that allows for both non-linearity and heterogeneity, which I will call the Mirrlees literature (after economist James Mirrlees, who started the literature in the 1970s).

I will attempt to summarize the Mirrlees results for capital taxation: interest income and dividends should be taxed at highly progressive rates, but capital gains remains a big question mark. There are some cute things you can do to produce a model in which the optimal capital gains tax is non-zero--usually somewhere in the 2% to 5% range--but they really don't take into account the reality that capital gains income gets super-concentrated in a few number of people. On the other hand, we don't want to significantly distort the accumulation of capital over time.

I have been toying with a policy proposal that aims to rectify the two concerns--the Ramsey concern about arbitrage and accumulation of capital, and Krugman's concern about wealth inequality. My proposal is to institute a moderately high capital gains tax of, say, 35%, but also include a high life-time deductible amount so that households don't pay any capital gains tax on the first, say, $1 million of capital gains income over their lifetimes. Essentially this creates a progressive capital gains tax that differs from typical progressive taxes in the fact that the tax brackets are determined by life-long capital gains, not annual gains. This eliminates the problem that arises with taxing annual capital gains at progressive rates, which results in distorting investment incentives in favor of short-run rather than long-run investments.

Strictly speaking, the Mirrlees literature clearly proves that my policy is sub-optimal: it would result in shifting some of the risk from high to low-income households when the optimal thing to do is just the opposite. However, the policy side-steps some of the more dubious (In my opinion) aspects of the Mirrlees models, and would achieve a reduction in wealth inequality without necessarily penalizing new capital formation. That is, so long as the life-time deductible amount is high enough, we can achieve both a substantial amount of redistribution and the same level of capital formation as in the zero-tax equilibrium (depending somewhat on how risk aversion and income are related in household preferences). Intuitively, my reasoning here is that any profitable investments foregone by the rich due to the high marginal tax can be taken up by the poorer households who have not yet reached the $1 million life-time limit and therefore face a 0% marginal tax rate. This tax policy would distort investments to favor smaller amounts by more households (that is, a more even distribution), but unlike the progressive annual tax is not skewed towards less overall investments, nor to shorter-run investments.