Examining the Internal Revenue Code for Disparate Racial Impacts

Aside from a few provisions involving Native Americans,¹ the Internal Revenue Code contains no explicitly race-based provisions. Of course, it may nevertheless be the case — in fact, it clearly is the case — that many IRC provisions have racially disparate impacts. Nearly all of the myriad factors that go into determining federal income tax liability — including gross income, types of income eligible for favorable tax treatment (for example, capital gains and dividends), retirement savings, health insurance coverage, itemized deductions (including the deductions for home mortgage interest, state and local taxes, and charitable contributions), marital status, children in a household, and age — have different distributions for different racial and ethnic groups in the U.S. population. When tax-relevant attributes are differently distributed by race and ethnicity, disparate impacts of code provisions based on those attributes necessarily follow.

Although the pioneering studies of disparate racial impacts of code provisions were published a few decades ago,² interest in the topic has become more widespread in the past few years. Signs of increased interest appeared even before the protests over the killing of George Floyd and have intensified since. In January the Tax Policy Center and the Urban Institute produced an interactive internet feature, “Racial Disparities and the Income Tax System.”³ Following an introduction explaining that “some tax policies can . . . exacerbate income and wealth inequalities stemming from long-standing discrimination in areas such as housing, education, and employment,” the feature offers information on the racial distribution (the classifications are Asian, Black, Hispanic, and white) of taxpayer characteristics relevant to a wide range of income tax provisions. The featured topics include marital status, age, overall income, salary and wages, capital gains and dividends, retirement savings, dependent children, the earned income tax credit (IRC section 32), the passthrough deduction (IRC section 199A), health insurance, the standard deduction and itemized deductions, homeownership, charitable giving, and overall federal income tax burdens. “Racial Disparities” is far from alone; in the past few years there have been several essays, policy papers, and studies calling attention to the disparate racial impacts of various aspects of the federal income tax.⁴

Because the IRS does not ask taxpayers about their race on Form 1040, and because no government agency produces race-based data on the impact of federal income tax provisions (more on that below), the information provided in “Racial Disparities” is suggestive rather than conclusive. Nevertheless, on some topics the suggestion is quite strong. For example, data showing that “homeownership rates for White households were significantly higher than for people of color in 2018” does not quite prove that the home mortgage interest deduction (MID) disproportionately benefits white taxpayers, but it establishes a prima facie case for that proposition. Although the data suggest disproportionate benefits to white taxpayers in several areas, “Racial Disparities” is careful to note that in some areas the disparities appear to run in the opposite direction. Perhaps most significantly, immediately below a chart showing that white and Asian taxpayers are overrepresented among households with incomes above $150,000, and Black and Hispanic taxpayer are underrepresented, the authors observe that “overall, federal income taxes are progressive,” and that “therefore, federal income taxes help reduce income inequality and racial income gaps.”

The absence of government-produced race-based tax data is the focus of a forthcoming article by Jeremy Bearer-Friend.⁵ The article reviews the history of exclusively colorblind government-produced tax statistics and contrasts it with the long-standing practice of the federal government of producing analyses of the racial impacts of myriad nontax government programs. Although Bearer-Friend does not advocate including a race question on Form 1040, he explains that there are less intrusive ways of determining the distributions by race of the burdens and benefits of income tax provisions, and he strongly urges the IRS and Treasury to do so. He sees an important instrumental purpose to this data production: “I show that colorblind data conceal racial inequalities embedded in our tax code and prevent the remedy of that inequality.”⁶ Although he does not argue that every provision shown to disproportionately benefit whites or disproportionately burden people of color necessarily should be revised or repealed, he contemplates that that would be the appropriate legislative response in at least some cases.

One of the strongest indications of the increasing interest in this topic is a July 9 Zoom panel discussion, well attended and enthusiastically received, sponsored by the Section on Taxation of the Association of American Law Schools, on “how to identify and address policies with racist implications in the Federal Income Tax Course.”⁷

This article considers how a legislature committed to racial justice should respond to a convincing statistical demonstration that a code provision has disparate impacts by race. The question is important because, as noted above, there are few if any code provisions that do not have differential racial impacts in today’s United States. If every code provision with a significantly disparate racial impact is to be revised or repealed, a great deal of revising or repealing will be required. That would be true whether the targeted provisions are all provisions with demonstrated disparate racial impacts, or only those provisions the disparate impact of which is to the disadvantage of people of color.

Part I of this article examines the question of how to identify the provisions that should be analyzed for their racial impacts; this might be called the “compared to what” question. The usual approach is to identify some supposedly neutral version of an ideal income tax, and to examine the race-based impacts of provisions of actual law that would not exist in an ideal income tax. Part I.A explains that this approach is deeply problematic, both because it is unclear why normative status should be conferred on any version of an income tax, and because the contours of an ideal income tax are themselves up for debate. Part I.B considers a different possible answer to the “compared to what” question: analyzing the racial impacts of changes in the law by assuming (usually implicitly rather than explicitly) the normativity of the pre-change version of the law as normative. Part I.B discusses the problems with this approach.

Part II considers how indirect effects of tax provisions (and of their repeal) may complicate race-based analysis. Part II.A examines how the analysis changes if one considers taxpayers claiming a deduction as primarily conduits for the delivery of federal dollars to others (a plausible view of both the charitable deduction and the deduction for SALT), or if one concludes that market forces deprive taxpayers of most of the economic benefit of tax benefits they nominally enjoy (a plausible view of the MID).

Part II.B turns to a recent article by Daniel Hemel and Kyle Rozema that describes and explains the counterintuitive fact that repeal of a tax provision disproportionately benefiting upper-income taxpayers may further enrich those taxpayers who had been able to take advantage of the provision, depending on what Congress does with the revenue gained from the repeal.⁸ In particular, they demonstrate that repealing the MID and using the revenue from the repeal to reduce the tax liability of every taxpayer by the same percentage would leave the highest-income taxpayers better off than under current law. Hemel and Rozema do not include racial effects in their analysis, but Part II.B explains how their analysis might extend to race in those situations in which disparate racial impacts result from racial differences in income or wealth distributions (or in areas strongly correlated with income or wealth, such as homeownership).

Part III considers the level of granularity at which statisticians and policymakers should analyze disparate racial impacts. The most common approach — perhaps more reflexive than considered in some cases — is to focus more or less by code section. The MID,⁹ for example, is usually considered an appropriate object of race-based analysis. But one could go more granular — for example, by focusing on the racial impact of the $750,000 ceiling on mortgage principal, or (more granular still) on the racial impact of the marriage penalty created by the $750,000 ceiling.

Conversely, one might decide even the entire MID is too granular for analysis, and that the question should be the overall racial impact of the federal income tax, netting all of the particular provisions. Or one might decide that even that is too narrow, and that the scope of analysis should be extended to include other federal taxes, or all federal, state, and local taxes. And even that is arguably too narrow; perhaps the scope of analysis should not be limited to taxes but should extend to government transfer payments and the benefits of other government spending (either federal only or considering all levels of government). There are no easy answers to this granularity question, but Part III examines the relevant considerations.

Part IV offers some concluding thoughts. In brief, although Bearer-Friend makes a persuasive argument for more and better information on the racial impacts of the various IRC provisions, in many cases it will be far from apparent how, if at all, Congress should respond to that information. One thing, at least is clear: If one is persuaded that a particular code provision is bad policy, even apart from its racial impacts, the fact that the provision also disproportionately benefits whites, or disproportionately burdens people of color, is a compelling – and perhaps in some cases decisive — additional reason for its reform or repeal.

I. Compared to What Ideal?

Tax provisions do not have racial impacts (disparate or otherwise) in a vacuum; rather they have impacts relative to some alternative. Thus, selecting the alternative for comparison is the crucial first step in the analysis. This section considers two plausible approaches to identifying alternatives — first, comparing provisions of current law with some version of a normative or ideal income tax, and second — useful only for evaluating changes in the law — comparing the tax treatment under prior law with the tax treatment under the new law (in effect, treating prior law as normative).

A. Comparisons With a Normative Baseline

On questions relating to the income tax base — in other words, inclusion and deduction provisions — there are two leading candidates for a normative baseline. One is so-called Haig-Simons income, sometimes referred to as the comprehensive income tax base. The classic formulation is by economist Henry Simons:

The other is the “normal structure of the individual income tax” concept used to identify tax expenditures for purposes of the tax expenditure budget.¹¹ Most code provisions deviating from the Haig-Simons definition also deviate from the normal structure, and so are considered tax expenditures. The two most significant differences are that the Haig-Simons definition encompasses imputed income from owner-occupied housing and consumer durables, and unrealized appreciation in assets, whereas nontaxation of both imputed income and unrealized appreciation is considered part of the normal structure. ¹² To take either Haig-Simons income or the normal structure as a normative starting point for analyzing the racial effects of deviations from the starting point is to accept, at least implicitly, the starting point as race neutral.

It is not self-evident, to put it mildly, why either Haig-Simons or the normal structure has such normative force. One might, instead, examine Haig-Simons and the normal structure themselves for racial effects compared to some other starting point; candidates might include a consumption tax, a head tax, and no tax at all. Although the choice of starting point should itself be defended, the relevant literature generally gives the question little or no attention.¹³

Even if one accepts that some comprehensive tax base can be identified as a neutral baseline from which to identify and analyze suspect provisions, it is unclear that it should be an income tax base. There is an enormous and inconclusive literature arguing the relative merits of taxing income and taxing consumption, and a fair summary might be that the proponents of the two tax bases have fought to a draw, both intellectually and politically. Consumption tax proponents offer plausible arguments that a consumption tax would be both fairer and more efficient than an income tax.¹⁴ And despite its label, the income tax has major consumption tax features, including the deferral of tax on unrealized appreciation and on most retirement savings. Thus, the current income tax is really an income-consumption hybrid.

In addition to the overall hybrid character of the tax, some provisions are themselves hybrids. For example, special low rates on capital gains and dividends are a hybrid between an income tax (which would tax at ordinary rates) and a consumption tax (which would impose no tax on unconsumed gains and dividends). Moreover, a good case can be made for the normativity of an income-consumption hybrid. It might, for example, be appropriate as a matter of principle to tax retirement savings on a consumption tax model, while taxing other savings on an income tax model.¹⁵

All this matters because several of the major income tax provisions likely to be scrutinized for disproportionately benefiting white taxpayers are not deviations from a consumption tax norm, or from a hybrid income-consumption tax norm. Deferral of taxation of retirement savings would then be consistent with the normative baseline, and thus would not be subject to scrutiny. The same would be true of the nontaxation of unrealized appreciation. And from a consumption tax perspective (although not necessarily from a hybrid tax perspective), any taxation of capital gains and dividends (not converted to consumption) would be over-taxation and would probably disproportionately burden white taxpayers.

It is also possible to construct normative baseline justifications for existing provisions that are usually considered deviations from both income tax and consumption tax ideals. For example, William D. Andrews has argued that “consumption” in the Haig-Simons definition of income “should be elaborated in a way that excludes medical services and whatever satisfactions one gets from making charitable contributions.”¹⁶ He explains in detail why he would categorize the medical expense deduction and the charitable contribution deduction not as tax expenditures, but as neither consumption nor accumulation, and thus not part of the ideal tax base.

Although Andrews limits his detailed discussion to the medical expense and charitable deductions, his analysis would clearly also justify the exclusion of health insurance (employer-provided or otherwise) from the tax base. Finally, he tentatively suggests that the MID “may be justifiable . . . if interest paid can sensibly be excluded in measuring personal consumption,” and that there is a similar “case to be made for deducting state and local taxes.”¹⁷ Andrews’s position on these matters is far from universally accepted,¹⁸ but if accepted it would exempt from scrutiny at least the medical expense and charitable deductions, and the exclusion for employer-provided health insurance, and arguably the deductions for mortgage interest and for SALT.

Similar baseline uncertainty exists regarding tax rate issues. The Haig-Simons definition says nothing about tax rates, and in its tax expenditure budget analyses the Joint Committee on Taxation considers the “existing rate structure” (including the standard deduction, understood as a de facto zero-rate bracket) as part of the normal structure of the tax.¹⁹ Nevertheless, there is a tendency to treat a flat rate tax as normative, and thus to look for race-based effects of progressive rates, on both the income side and the deduction side. On the income side, this is exemplified by the statement in “Racial Disparities” that:

In other words, progressive average rates mean that the income tax disproportionately burdens whites (evidently a commendable result in the view of “Racial Disparities”). “Racial Disparities” does not adequately explain how it arrives at this conclusion, but by concluding that the income tax reduces (after tax) racial income gaps immediately after explaining that the income tax features progressive average rates, the implication seems to be that progressive average rates reduce racial income gaps compared with the results under a flat tax. Thus, progressive rates are viewed as a deviation from the neutrality of flat rates, albeit a meritorious deviation.

When the focus is not on what is taxed, but on what is not taxed because of deductions or exclusions, progressive rates are again viewed as producing disparate racial effects, but this time to the detriment of Black and Hispanic taxpayers. Lipman, Mirkay, and Strand, for example, contend that “racialized tax deductions undermine progressivity”; they explain that “the benefit of tax deductions increases as a taxpayer’s household income rises, thereby providing the most significant benefits to taxpayers who . . . own their own homes” and who “are disproportionately white.”²¹ But if the deduction is understood to have an income-defining purpose — as Andrews claims for the medical expense and charitable deductions, and suggests for the mortgage interest and SALT deductions — then the effects on the deduction side are appropriately a mirror image of the effects on the income side.²²

Although they reach opposite conclusions about which racial groups are harmed and which are benefited, the analyses on both the income side and the deduction-and-exclusion side implicitly assume a flat rate norm. But despite the intuitive appeal of the notion that flat rate taxation is somehow neutral, the intuition cannot survive scrutiny. Perhaps the most thorough demolition of the normative status of a flat rate tax was accomplished more than three decades ago by Joseph Bankman and Thomas D. Griffith. Noting that some may perceive a flat tax as a principled compromise “between the perceived efficiency costs of a progressive tax and the perceived inequities of a lump sum tax,” they counter that this:

In short, because there is no justification for treating a flat rate tax as a neutral baseline from which to determine the race-based effects of progressive taxation, neither the income-side argument that progressivity disproportionately burdens white taxpayers nor its deduction-and-exclusion-side mirror image is persuasive.

The difficulty or impossibility of finding a neutral baseline to serve as the starting point for disparate impact analysis arises in still more areas, beyond questions of tax base and rate structure. For example, analyses of income tax marriage penalties — whether generally or in terms of disparate racial impacts — compare tax liabilities produced by the joint return rules with the tax liabilities that would be produced if marital status were not a tax determinant.²⁴ It is possible to do marriage penalty analysis without taking a normative position. For example, a study of effects of the income tax treatment of marriage on the choice between marriage and unmarried cohabitation need not make any claims about how the income tax should treat marriage.

However, a normative position is necessarily involved in any claim of unfairness in marriage penalties, including unfairness in the racial distribution of marriage penalties. For claims of that sort, the normative baseline, whether implicit or explicit, is a system in which marital status is not a tax determinant. Although there are certainly powerful normative arguments in favor of separate returns,²⁵ a normative case can also be made for joint returns.²⁶ Any analysis of disparate racial impacts of marriage penalties is only as strong as the normative case for separate returns, and that case — although strong — is less than overwhelming.

The marriage penalty issue can be used to illustrate an important general point about analyses of possible racial disparities in the effects of the income tax. On the one hand, anyone who is convinced of the normative case for joint returns will not be persuaded by an argument that one or more racial groups are disadvantaged by the difference between tax liabilities under joint returns and liabilities under a hypothetical normatively wrong system in which marriage is not a tax determinant. On the other hand, someone who is convinced of the normative case for abandoning the joint return system will find the case for abandonment all the more compelling if it is demonstrated that joint return marriage penalties fall disproportionately on Black and Hispanic taxpayers. To put the point more generally: If a persuasive case can be made that a particular provision — whether it be the joint return, the mortgage interest deduction, or anything else — is bad tax policy quite apart from any disparate racial impacts,²⁷ the case for repeal or reform (and the prospects for political success) may be greatly strengthened if it can be shown that the provision also disparately impacts taxpayers of color.²⁸

B. Evaluating Changes in the Tax, as Contrasted With Evaluating ‘Snapshots’ of the Law

Most analyses of the racial impacts of tax provisions take what might be called a snapshot approach, comparing the current state of the law with some supposedly race-neutral normative alternative. A different approach — not really a competitor, because it addresses a different question — is to examine the effect of changes in the law by comparing racial impacts under the prior version with the impacts under the new version. Given the difference in the questions asked under the two approaches, it is no surprise that the answers may also be quite different. For example, under conventional analysis (setting aside for the moment all the problems with conventional analysis discussed elsewhere in this paper) there is good reason to think the SALT deduction disproportionately benefits white taxpayers, even after the recent imposition of the $10,000 deduction cap by the Tax Cuts and Jobs Act of 2017.²⁹ On the other hand, an analysis of the racial impacts of the 2017 change would in all likelihood conclude that the disallowance of SALT deductions over $10,000 disproportionately burdened white taxpayers relative to prior law. Again, the two approaches are not in conflict: They merely produce different answers to different questions.

Of course, a study of the race-based distributional effects of the TCJA could focus on the overall effect of the legislation without separately analyzing the distributional impact of each aspect of the legislation, or it could proceed change-by-change. In fact, the most prominent study of the TCJA’s racial impact took the overall approach. ³⁰ (This essay’s discussion of the appropriate level of analytical granularity is deferred to Part III.)

If distributional analysis of changes is used only descriptively, to identify race-based winners and losers from legislative changes, the approach is not problematic (although the results may be). But if the approach is used normatively — say, as the basis for an argument against a proposed change in a provision disproportionately benefiting people of color, which, if enacted, would reduce but not eliminate the disproportionality — there may be a problem. That perspective attaches normative significance to whatever current law happens to be, and the normative status of current law must be established rather than simply assumed. And, of course, treating current law as normative in this context is difficult or impossible to square with the snapshot version of race-based analysis of code provisions, which operates from the premise that current law is entitled to no presumption of normativity.

It should be noted, however, that arguments for and against proposed changes in the tax laws are routinely made on the basis of the fairness or unfairness of the distributional effects of proposed reforms relative to existing law, without anyone bothering to explain why existing law should be treated as a normative baseline. Anyone taking the same approach in the context of race-based analysis would at least have plenty of company.³¹

Finally, a version of the point made earlier, using the marriage penalty as an example, also applies here. The appeal (both theoretical and political) of an argument for or against proposed changes based on their racial impact will be greatest when the race-based argument is paired with a non-race-based argument — for example, expansion of the EITC would be good policy without regard to racial impacts, and it would be especially beneficial to Black and Hispanic working parents.

II. Taking Into Account Indirect Effects of Tax Provisions, and of Their Reform or Repeal

A. Of Conduits and Implicit Taxes

The usual approach in examining a code provision for disparate impacts is to focus on the nominal incidence of the provision, without considering whether its ultimate economic effects might be different — perhaps very different — from the nominal incidence. For example, “Racial Disparities” notes that “92 percent of the total benefits from the tax deduction for charitable contributions go to the top 20 percent of taxpayers by income, who are significantly less likely to be people of color.”³² This strongly suggests that a detailed study would find that the tax reductions from the claiming of the charitable deduction go disproportionately to white taxpayers.

However, the conventional policy justification of the charitable deduction is that the deduction’s purpose is not to benefit the taxpayers claiming the deduction, but to benefit the charitable donees and, ultimately, the individuals who benefit from the activities of the donees.³³ Suppose a donor in the 37 percent bracket is willing to give up $6,300 of after-tax income to benefit a charity. In the absence of the charitable deduction, the donor would simply give the charity $6,300. Given the deduction, however, the donor might give the charity $10,000 because the after-tax cost of a $10,000 donation is only $6,300. In this situation, the donor captures none of the benefit of the deduction; rather, the donor serves as a conduit for transferring $3,700 from the federal government to the charity and its beneficiaries.³⁴ From this perspective, the key question for the racial impact analysis is not the racial distribution of taxpayers claiming the deduction, but rather the racial distribution of those who benefit from the activities of the donee charities. This would require a truly daunting empirical research project, given that taxpayers give to many types of charities, and that different types of charities have different racial distributions of their beneficiaries.

In the end, such a project might still find disparate racial impacts, but the existence of disparate impacts seems much less self-evident when the focus is on ultimate benefits and burdens, rather than simply on the taxpayers claiming the deduction.

There is another complication. The above example imagined (plausibly enough) a taxpayer who increases the amount of her donation by the amount of her tax savings. But some taxpayers may be super-responsive to the existence of the deduction, increasing donations by more than the tax savings, so that they are poorer with the deduction than they would be in its absence. Other taxpayers may have the opposite response, increasing their donation by less than the amount of their tax savings (or even not at all), so that they end up richer because of the deduction.

A study of the ultimate racial impacts of the charitable deduction would also have to take into account these different elasticities of charitable giving by considering not only the racial impacts of the deduction on the beneficiaries of charities (as under the pure conduit analysis), but also the racial distribution of the burdens borne by more elastic donors and the benefits received by less elastic donors. Again, a comprehensive study — taking into account racial distributions of both charitable beneficiaries and of donors with different elasticities of giving — might find disparate racial impacts, but the question is enormously more complicated than is suggested by “Racial Disparities,” which focuses simply on the distribution of the nominal tax benefits of the deduction.

Similar issues arise in connection with the SALT deduction. “Racial Disparities” notes that in 2018 (with the $10,000 SALT deduction cap in effect) taxpayers with incomes over $100,000 enjoyed 90 percent of the federal income tax savings from the SALT deduction, and that Black and Hispanic households are 27 percent of all households, but only 16 percent of households with incomes over $100,000.³⁵ The implication, which is likely correct, is that a thorough study would find that a disproportionate percentage of the nominal tax savings from the SALT deduction go to white and Asian households.

Like the charitable deduction, however, the SALT deduction is commonly understood as using the taxpayers claiming the deduction as a conduit for transferring federal funds to state and local governments.³⁶ The idea is that the deduction makes it politically possible for the states to impose higher taxes than they otherwise could on affluent taxpayers able to claim the deduction, so that the taxpayers themselves end up no better off than in the absence of the deduction. Instead, the taxpayers serve as conduits for the transfer of funds from the federal government to the states. If the ultimate beneficiaries of the deduction are the beneficiaries of whatever the state and local governments do with their increased tax revenue attributable to the federal deduction, the racial impacts are much more complex and uncertain than under the nominal incidence approach of “Racial Disparities.” Nor is this the only daunting empirical question. The racial impacts depend not only on the distribution by race of the benefits of state and local government spending, but also on the extent to which the level of state and local taxation is sensitive to the SALT deduction. As with the charitable deduction, an analysis limited to the racial distribution of nominal federal tax benefits tells us little about the racial distributions of the ultimate benefits and burdens of a federal tax provision.

Differences between nominal impacts and ultimate impacts are not limited to code provisions where taxpayer-as-conduit analysis is clearly relevant. The MID, for example, has no clear analogue to charities or state and local governments (and the individuals who benefit from their operations) as the ultimate beneficiaries of the deduction. Nevertheless, it is generally thought that much of the nominal tax benefit of the deduction is capitalized into the price of homes.³⁷

If the tax benefit were fully capitalized into housing prices, that would mean a taxpayer buying a home financed with a tax-favored mortgage would be no better off than if the deduction did not exist and she bought the same home at a lower price, financing the purchase with a nontax-favored mortgage. In that case, no taxpayer would enjoy any net benefit from the MID, with the result that there would be nothing left to analyze for disparate racial impacts. In the real world, the truth is probably somewhere in between: Price effects significantly reduce, but do not eliminate, the net benefit to homeowners of the MID.³⁸ In that case, the possibility of disparate racial impacts remains, but any impacts will be significantly smaller (because the net tax benefit itself is smaller) than if price effects are disregarded.

B. Analyzing Reform Packages in Their Entirety

In a recent article, Hemel and Rozema challenge the conventional wisdom that the MID disproportionately benefits upper-income taxpayers.³⁹ They acknowledge that the nominal tax savings from the deduction are skewed toward the high end of the income distribution, but they claim that an analysis that stops there is incomplete (in a way different from the price-effect aspect discussed above). They explain that repeal of the MID would necessarily involve not only the repeal itself, but also a legislative decision about what to do with the increased tax revenue resulting from the repeal. A complete distributional analysis of the repeal would have to consider not only its effect, but also the effect of whatever Congress decided to do with the resulting revenue. Although there is virtually no limit as to what Congress might do with the revenue, one plausible use would be to leave total tax revenue unchanged by decreasing everyone’s taxes by the same percentage, with the percentage determined by the revenue from repeal.

Consider a simple two-taxpayer model, similar to one used by Hemel and Rozema.⁴⁰ Richer taxpayer has gross income of $200,000 and pays $40,000 of deductible mortgage interest. Poorer taxpayer has gross income of $80,000 and pays $16,000 of deductible mortgage interest. There is a three-bracket income tax, with a zero rate on the first $40,000 of taxable income, a 20 percent rate on the next $40,000, and a 30 percent rate on all income above $80,000. With the MID in place, and each taxpayer qualifying for the deduction, Richer’s tax liability is $32,000, and Poorer’s is $4,800.⁴¹

Under a straightforward analysis, the MID reduces Richer’s tax by $12,000 ($40,000 * 30 percent) and reduces Poorer’s by $3,200 ($16,000 * 20 percent) — a seemingly classic upside-down subsidy. Repeal of the deduction, without any other change, would increase tax revenue by $15,200; Richer would pay $44,000, and Poorer $8,000. However, if Congress is content with the same $36,800 total revenue as before repeal, it can decrease effective tax rates across the board by approximately 29.23 percent. Richer would then pay about $31,139 ($44,000 * 70.77 percent), and Poorer would pay about $5,661 ($8,000 x 70.77 percent). Thus, repeal of the MID, made revenue-neutral by a proportionate decrease in average tax rates, would be a regressive change, transferring $861 from Poorer to Richer.⁴²

This approach — repeal of a seemingly regressive tax benefit to finance an across-the-board proportionate reduction in average rates —is not inherently regressive; the regressivity of the result depends on the distribution of the tax burden (after repeal of the tax benefit but before the rate reduction) being more unequal than the pre-repeal distribution of the tax benefit. In the example, Richer bore about 78.9 percent ($12,000/$15,200) of the burden of repeal but enjoyed about 84.6 percent ($44,000/$52,000) of the benefit of the rate reduction. And because the total burden of repeal (taking into account both taxpayers) equaled the total rate cut in dollars (taking into account both taxpayers), the difference between burden and benefit percentages was to the advantage of Richer.

As Hemel and Rozema acknowledge, Thomas Griffith pointed out the possibility of results like the above decades ago. ⁴³ What Hemel and Rozema add is an empirical analysis showing that this is not just an interesting theoretical possibility, but that based on data for tax year 2010, repeal-plus-proportionate-rate-cuts would have had significantly regressive effects. For example, the average household in the top 1 percent of the income distribution would have gained $13,208 annually from this approach.⁴⁴

Hemel and Rozema do not consider race-based impacts, but their analysis could be extended in that direction. In many cases, suggestions of disparate racial impacts disfavoring Black and Hispanic taxpayers are based on evidence that a provision disproportionately benefits upper-income taxpayers, and on the underrepresentation of Black and Hispanic taxpayers in the upper-income ranges.⁴⁵ In these situations, if the provision’s racial impacts are determined by comparing the current state of affairs with a counterfactual in which the provision does not exist, the choice of the counterfactual is crucial. Counterintuitively, the findings of Hemel and Rozema suggest that even some of the provisions most widely considered to disproportionately favor white taxpayers may not do so when compared with plausible counterfactuals.

There are several important caveats. First, even assuming proportionate rate cuts are the appropriate counterfactual, repeal is not inevitably regressive. Hemel and Rozema analyzed the SALT deduction and the charitable contribution deduction using the same approach as for the MID, and found that repeal-plus-rate-cuts would also be regressive in the case of the SALT deduction but would not be regressive in the case of the charitable deduction.⁴⁶

Second, Congress need not use the revenue gained from a provision’s repeal to finance proportionate reductions in average tax rates. As Hemel and Rozema demonstrate, if Congress used the revenue from repeal of the MID (or of the SALT or charitable deduction) to finance equal-dollar universal cash grants, the change would be highly progressive.⁴⁷ Or if Congress simply used the revenue from repeal to decrease the size of the budget deficit, the challenge for the analyst would be to distribute the benefits of that decrease among future taxpayers (by income levels or by race, depending on the analysts’ interests).

Third, the Hemel-Rozema empirical results were for tax year 2010. Considering the TCJA’s new limitations on itemized deductions — the $10,000 ceiling on SALT deductions, the reduction in the mortgage principal ceiling from $1 million to $750,000, and the near-doubling of the standard deduction — Hemel and Rozema might reach different results applying their analysis to tax year 2020. Fourth, their analysis cannot be extended to disparate racial impacts — at least not in any straightforward way — when a provision’s disparate impacts are not closely tied to differences in income distribution by race. In the MID context, for example, Felipe Chacon has found that even controlling for income (and age) Blacks are 56.9 percent less likely to own a home with a mortgage than whites.⁴⁸ Chacon plausibly suggests this may be due, at least in part, to racial differences in intergenerational wealth transfers; in particular, “Minorities have less access to financial assistance on down payments from relatives.”

Apart from these four caveats, the Hemel-Rozema analysis has two major implications for race-based analyses of the effects of code provisions. One implication is theoretical. If an income tax without an MID, and featuring proportionate average rate reductions financed by MID repeal, would not be considered to have disparate racial impacts,⁴⁹ and if Black and Hispanic taxpayers fare better under current law (featuring the MID without the rate reductions) than they would under the reform, it is unclear that there is a valid race-based objection to current law. The second implication is practical. No matter how racially disproportionate the benefits or burdens of an existing provision may be, whether repeal of the provision would be an improvement depends on what accompanies repeal — namely, what Congress does with the revenue raised by repeal of a taxpayer-favorable provision, or on how Congress finances repeal of a taxpayer-unfavorable provision. To put the point in Hemel-Rozema terms, reform advocates should fully specify their counterfactual, and should make sure that their counterfactual, if enacted, would be an improvement over the status quo.

III. What Level of Granularity?

The usual approach to examining the IRC for racial effects is provision-by-provision, separately evaluating the MID, the charitable deduction, the EITC, the section 1 rate structure, capital gains rates, the treatment of retirement savings, and so on. In addition to the practical difficulties (largely due to the lack of government data on race-based effects) and theoretical conundrums described earlier, there is the question of whether this provision-by-provision strategy is the best approach.

Consider the EITC, for example. One representative recent study found that “21 percent of Black women receive the EITC, more than double the 9 percent share of white women who receive it. Women of color also tend to receive a larger average EITC than white women.”⁵⁰ If the overall effect of the EITC is the right level of analysis, the EITC grades out well. On the other hand, it is possible to drill deeper and analyze sub-rules within the EITC. Taking this approach, Dorothy Brown found that the marriage penalty embedded in the EITC’s phaseout rules fell disproportionately on Black households, as did the two-child limit on the number of qualifying children.⁵¹ One can imagine other provisions under which the effects might be reversed — for example, if the MID overall disproportionately benefits white taxpayers, it seems likely that two 2017 legislative innovations — the reduction of the principal ceiling from $1 million to $750,000 and the elimination of the deduction for home equity interest — disproportionately burden white taxpayers. In all such cases, the level of granularity of analysis is critical.

Of course, if racial impacts are studied just to expand our knowledge, there is no need to choose a single level. A study can examine both the overall effects of the EITC and separately examine the effects of phaseout rules, qualifying child limitations, and so on. But when the studies are conducted for the purpose of identifying provisions that should be revised or repealed to eliminate disparate racial impacts, the granularity question cannot be avoided. Are the effects of the phaseout and qualifying child rules of the EITC acceptable because of the good overall effects of the EITC, or must those rules be revised to eliminate their disparate impacts, regardless of the EITC’s overall effects? On the one hand, the goal of eliminating all racial effects that disadvantage people of color has theoretical appeal. On the other hand, there are two major problems with that approach.

First, it would require empirical study of the impacts of literally thousands of code subsections and sub-subsections, and repeal or revision of all subsections and sub-subsections determined to have objectionable disparate impacts. It is unclear that the necessary effort is within the capacity of either researchers or Congress. Second, the simplest way to eliminate the disparate impacts of the EITC’s phaseout and qualifying child rules would be to repeal the EITC itself. The offending rules would be gone, but it would be hard to view that as an improvement in the racial equity of the code.⁵²

It is also possible to move up, rather than down, from the usual provision-by-provision analysis. Assuming the MID disproportionately benefits white taxpayers,⁵³ is it necessary to amend or repeal the deduction to eliminate the disproportionate impact, or is it sufficient that the provision is a part of a federal income tax that (in the words of “Racial Disparities”), “help[s] reduce income inequality, and racial income gaps?”⁵⁴ Just as, in the case of the EITC, there is the possibility that the overall good distributional effects of the provision make it unnecessary to eliminate every subprovision with disparate impacts in the opposite direction, there is the possibility that the overall distributional effects of the federal income tax make it unnecessary to revise or repeal every provision that disproportionately benefits white taxpayers.⁵⁵ Nor is the entire income tax an obvious stopping point for moving up the level of analysis. It might make more sense to focus on the overall distributional effects of all federal taxes — payroll, excise, estate and gift, corporate income, and individual income. Or, better yet, to focus on the overall distributional effects of all federal taxes and spending, taking into account both cash transfers and the distribution of the benefits of other forms of federal spending. And even that would not be the most comprehensive approach — a truly comprehensive approach would consider the overall distributional effects of all federal, state, and local taxing and spending.⁵⁶

The point here is not to argue for any particular level of granularity; rather it is to emphasize the unavoidability of the granularity question in race-based analyses of federal income tax provisions, and to suggest the question merits more consideration than it has received to date.

IV. Conclusion

The case for government collection of data on the racial impacts of federal income tax provisions, as set forth by Bearer-Friend, is compelling.⁵⁷ The government already collects race-based data on virtually every other important activity of the federal government, and in any event more knowledge is always a good thing. The harder question is what changes in the law, if any, should follow when the data reveal racial disparities.

Taking the charitable deduction as an example, there would be several difficult steps between the future IRS production of data showing that the nominal benefits of the charitable deduction are enjoyed disproportionately by white taxpayers, and the conclusion that the tax preference for charitable giving should be repealed or reformed to eliminate the disparate racial impact. First, there is the need to establish a normative baseline from which current law departs. In particular, this will involve arguing against Andrews’s claim that a charitable deduction belongs in an ideal income tax.⁵⁸ Second, it is necessary to consider the ultimate racial distribution of the benefits of current law rather than just the nominal distribution. Do donors ultimately capture the benefit of the deduction, or does most or all of the federal tax expenditure end up benefiting charitable organizations and their beneficiaries? The ultimate racial distribution of benefits may not be to the disadvantage of people of color, even if the nominal distribution is. Third, one must propose a reform package that improves racial equity when the package is considered in its entirety. As Hemel and Rozema demonstrate, this may require something other than using the revenue from repeal of an objectionable tax preference to finance an across-the-board rate cut.⁵⁹ Finally, the granularity question must be addressed. This requires consideration, for example, of whether it is necessary to eliminate any subsection or sub-subsection of the EITC that disproportionately burdens (or fails to benefit) Black or Hispanic families, even if the overall racial impacts of the EITC are commendable. Taken together, these are significant challenges.

It does not follow, however, that there should be no policy changes in response to new and better data on disparate racial impacts in the code. Bearer-Friend points out one example of what might be considered low-hanging fruit. Referencing findings that EITC-eligible Hispanic parents are less aware of the existence of the credit than other parents, and so claim it less often, Bearer-Friend accurately describes increasing awareness of the credit among eligible Hispanic parents as “a policy intervention that is neither complex nor particularly controversial.”⁶⁰ It is not particularly controversial because — far from requiring a change in the substantive law — it is designed to bring real-world results closer to the results called for by the substantive law. But the most important use of evidence of disparate racial impacts of tax provisions, as explained earlier, will almost certainly be as an argument — sometimes outcome-determinative — for repealing or reforming a provision that constitutes bad tax policy even apart from its racial effects. “This is bad tax policy, and the burden of the bad policy falls disproportionately on people of color,” may succeed with Congress where a simple “this is bad policy” would not.