Hemel and Rozema ask a question that we thought was well settled: What are the distributional consequences of repealing the mortgage interest deduction (“MID”)? Their conclusion (spoiler alert): It depends.
The article is important in two respects. First, it introduces a more sophisticated analysis of the distributional consequences from repealing (or limiting) the MID and other tax preferences. Through this exercise, the article also adds empirical perspective to the choices for measuring progressivity in tax reforms.
Building on David Kamin’s work on progressivity in tax reforms, the article begins with a result we’ll call the “Tax Burden Paradox”: Repealing a tax preference that provides a larger benefit to high-income taxpayers in dollar terms may in fact decrease those taxpayers’ share of society’s total tax burden. Consider the article’s example of a two-household society, one rich with pretax income of $100, and one poor with pretax income of $50, and progressive rates that tax the first $50 of income at 20%, and additional income at 40%. The rich and poor households pay $12 and $9, respectively, in mortgage interest. With the MID, the rich household pays $25.20 in taxes, or 75.4% of the society’s total $33.40 tax burden. If the MID is repealed, the rich household pays $30 in taxes, for a reduced 75% share of the society’s total $40 tax burden.
From the Tax Burden Paradox, Hemel and Rozema advance to their central argument, that any distributional analysis of MID repeal must consider the use of the revenues generated by the repeal (an estimated $79.7 billion in their static model). Using data from the IRS Statistics of Income and the NBER TAXSIM simulator, they calculate the distributional effects of MID repeal (as measured by changes in after-tax income at different income levels), assuming counterfactual uses of the increased revenues: (1) an equal-size rebate or public expenditure for every household, (2) a proportional reduction in every household’s tax liability, and (3) an across the board percentage reduction in all tax brackets.
Hemel and Rozema find that MID repeal has varying distributional effects depending on the counterfactual, leaving high-income taxpayers worse off in the case of the equal size rebate but better off in the case of the proportional reduction in taxes. Interestingly, the across the board percentage reduction leaves the top decile worse off, but the top percentile better off. The distributional effect becomes less ambiguous if, instead of repeal, the MID is capped at 28%, as in Hillary Clinton’s recent proposal. This reform would reduce inequality under all counterfactuals, but would raise only $2.7 billion.
The article is provocative, rich with implications, and certain to launch a thousand tax policy discussions. To start, what are we to make of the Tax Burden Paradox? While it is true that MID repeal in the article’s example reduces the rich household’s share of the overall tax burden, the total size of that burden increases, from $33.40 to $40. Conversely, with the MID, the rich household pays a larger share of a smaller overall burden. Consider a more extreme example: a reform that eliminates all taxes, except for a tax of one dime on every billionaire. The total social tax burden would be a little more than fifty bucks, even though the tax is borne 100% by billionaires.
One might conclude that the size of the tax burden—and not just relative shares—is a critical aspect of progressivity. Reconsider the article’s example as two separate taxes: Tax 1, which is the income tax with the MID, and Tax 2, the MID repeal, structured as an additional tax equal to the MID benefit for each taxpayer. The rich household bears 75.4% of Tax 1’s total $33.40 social burden. The rich household also bears 72.7% ($4.80) of Tax 2’s total $6.60 social burden.
From this perspective, MID repeal layers an additional tax that is also borne disproportionately by the rich household, albeit in a smaller proportion. That is, with MID repeal, we preserve all the progressivity of Tax 1, in terms of share of social burden, and gain the (slightly less) progressive Tax 2.
Hemel and Rozema convincingly argue that analyzing MID repeal’s distributional effects must also account for the use of use of the revenue generated, thereby enriching a developing literature on the shortcomings of analyzing the distributional effects of taxes in isolation. (A theme in Edward Kleinbard’s recent book). Their findings under the various counterfactuals confirm this view—if the government rescinds a tax preference for the wealthy with one hand, and gives the money back with the other, we shouldn’t expect any reduction in inequality. If, however, removing a tax preference of greater value to high-income taxpayers funds a rebate of equal value to all taxpayers, or of greater value to lower-income households, inequality should decrease.
The question then becomes: What counterfactual should we prefer? The article remains agnostic on this point, instead emphasizing the importance of context in analyzing distributional effects of MID repeal. Other plausible counterfactuals may be ripe for analysis. It might be interesting, for example, to test a counterfactual where the revenues are spent in same proportion as current government transfers or—to the extent estimates are available—total spending.
One of the many fascinating results in the article is the divergent effects of MID repeal among the top deciles and percentiles, presumably because of the MID’s receding significance at the highest income levels. These results suggest that tax policy tools for addressing inequality between the poor and the “merely wealthy” may fail in the context of high-end inequality. Similarly, the analysis of the minimal revenue raised from from Hillary Clinton’s proposed MID cap demonstrates the limited economic impact of tax reforms targeting the top of the ladder alone.