National Income, Transfers, and the JCT’s Tax Rates

In a recent pamphlet framing a congressional hearing on high-income and high-wealth taxpayers, the staff of the Joint Committee on Taxation offered what I believe is its first-ever official judgment about trends in historical income inequality.¹ In focusing primarily on the federal level, the JCT staff concluded that after-tax, after-government-outlay, after-government-deficit (three mouthfuls collectively referred to here as post-government) income inequality has changed little in 50 years.

The JCT further employed this historical approach to confirm prior results found with its standard distributional model (which doesn’t extensively attempt to distribute federal outlays or deficits and hasn’t specifically been applied in a historical context) that the federal tax system is progressive throughout the income spectrum (although the analysis didn’t reach the granular level of the 40 or 400 highest income or wealth units).² The JCT reached these tax rate and post-government income conclusions by adopting much of the method and data from research by two Treasury Department and JCT staffers Gerald Auten and David Splinter, respectively.³

There’s a lot going into historical tax rate and income distribution questions, and on top of that related wealth analysis. While all of this is easily a few weeks of reading and interpretation, this article focuses mostly on tax rate issues that also have implications for post-government inequality. The JCT’s historical tax rate’s inconsistency with the principles of national accounting, which is more than just a matter of semantics or taste, spills over to one of the most cited statistics in tax policy considerations, federal taxes as a percentage of GDP, with ramifications for the JCT’s standard as well as other government agency income distributions.

Taxation Without NIPA Calibration

U.S. federal taxes as a percentage of GDP in 2018 generally are cited as about 16 or 17 percent, depending on details such as calendar-versus-fiscal year reporting and treatment of government social insurance and retirement plans. Like it or not,⁴ that summary statistic about taxes and broad economic activity often drives tax policy considerations and is commonly cited for international and historical comparisons — there have even been U.S. attempts to codify this ratio as a constraint on revenue collection. According to the Commerce Department’s National Income and Product Accounts (NIPA), in 2018 federal taxes were 17.3 percent of GDP on a calendar year basis, which also is how income distributions generally are presented.⁵

As a first principle in thinking about appropriate income distributions, the average tax rate in a distributional model (which usually focuses on a concept of income, which is narrower than GDP) should be consistent with whatever broad economic summary statistic is used, whether the 17.3 percent statistic or otherwise. Consistency doesn’t require that the average tax rate in a distributional model equal 17.3 percent (tax rates in distributional models tend to be higher than rates applicable to broader economic activity like GDP),⁶ but the derivation of the tax rate in the distributional model should adhere to the same rules by which the broader economic measure is calculated.

Table 1. JCT and National-Account-Based Average Tax Rates, 2018^a

	Not NIPA-Specified or Constrained				NIPA-Specified or Constrained
	Model	Federal Taxes	Income	Tax Rate	Model	Federal Taxes	Incomeb	Tax Rate
1. Income distribution models	JCT historical	$3,097	$20,504	15.1%	Piketty, Saez, and Zucman	$3,361	$17,690	19%
2. Actual or implied broad economic measure	Implied by JCT historical	$3,420	$23,426	14.6%	NIPA, federal taxes divided by GDP	$3,568	$20,612	17.3%
Sources: Auten and Splinter, supra note 3; JCX-24-21, supra note 1; NIPA; and PSZ, supra note 8. aAmounts in billions, everything on calendar year basis. The author is responsible for calculations. See Table 2 in Appendix for details on how broad economic measure implied by JCT historical income model was derived. bThe $17.7 trillion is NIPA national income; the $20.6 trillion is NIPA GDP.

Table 1 shows that for 2018 the JCT historical method of calculating tax rates is inconsistent with the NIPA statistic that federal taxes are 17.3 percent of GDP. Instead, the JCT’s overall average tax rate in its income distribution model of 15.1 percent⁷ implies a broader meta economic tax rate of 14.6 percent. The construction of these tax rates is described in the Appendix, with the main factors in this divergence being the JCT’s choices to deviate from NIPA by (1) treating tax refundables (that is, payments through the tax system that exceed tax liability) as negative taxes rather than as government outlay transfers, and (2) adding transfers to denominators of both the JCT distributional and the implied JCT broader economic measures (the latter exercise performed by me and not the JCT).

In contrast, the right-hand columns of Table 1 signify that the historical distributional model of Thomas Piketty, Emmanuel Saez, and Gabriel Zucman (the PSZ model),⁸ which the JCT seems to shadow-box in the pamphlet, generates a 19 percent tax rate that’s consistent with the NIPA method that finds that 2018 federal taxes were 17.3 percent of GDP.⁹ Much of the difference between the PSZ and JCT tax rates is at the lower end of the income spectrum because of the JCT’s treatment of refundables and inclusion of transfers in the tax rate denominator.¹⁰

One odd element of the JCT pamphlet is its relaxed usage of the term “national income.” When first used in the pamphlet, national income is defined as “the total amount of money earned within a country,”¹¹ which seems intended to match the NIPA definition of national income (assuming that by money earned within a country, the JCT means income wherever generated by U.S.-owned factors of production). But the denominator the JCT staff uses for tax rates, “pretax/after-transfer national income,” is well in excess of NIPA national income ($20.6 trillion versus $17.7 trillion in 2018, as noted in Table 1), yet nevertheless interchanged in the pamphlet with the plain term “national income.”¹² The JCT doesn’t flag that its tax rate denominator is well above NIPA national income, and income totals are only offered for subgroups (for example, the bottom 50 percent of observations) and only for the pretax/after-transfer concept, so this divergence is not quantitatively obvious.

A second oddity concerns the JCT’s assertions that its tax rate calculation is conventional while the PSZ method is prone to “exaggerated tax rates.”¹³ It’s true that the Congressional Budget Office and Treasury staffs, like the JCT, treat tax refundables as negative taxes. But PSZ, the Commerce Department’s Bureau of Economic Analysis which is responsible for NIPA, and the Office of Management and Budget in its historical tables treat refundables as transfers instead of as negative taxes. Perhaps most surprisingly, given that the JCT pamphlet refers readers to Auten and Splinter’s paper for data and method,¹⁴ Auten and Splinter¹⁵ themselves treat refundables not as negative taxes but as government transfer payments.

Moreover, while the CBO and Treasury staffs as well as the JCT in its production of its standard distribution include transfers in tax rate denominators, only the Treasury staff matches the JCT in the magnitude and breadth of transfers included in the tax rate denominator.¹⁶ It also shouldn’t be forgotten that anybody who refers to federal taxes as a percentage of GDP is, by default and like PSZ, not adding the amount of transfers to that tax rate denominator. The JCT and others are engaging in a form of double counting (double counting that in effect understates capital income) by computing tax rates with, in NIPA parlance, both sources and uses (that is, income and consumption) in the denominator.

The exaggerated tax rates that the JCT refers to concern how in treating refundables as transfers and not including transfers in the tax rate denominator, some high positive tax rates can appear at low-income percentiles under the PSZ method. However, the opposite situation, very high negative tax rates, occurs under the JCT method as refundables increase, which will be relevant for pandemic payments in 2020 and 2021 and as refundables, particularly the child tax credit refundables, rise in the near term.

Origins of Tax Rate Inconsistency

The roots of this tax rate inconsistency between the government agency distribution models and the 17.3 percent NIPA meta rate stem from a mix of the following: tax-centric hegemony, data limitations (at least in the past), issue topicality, how the federal individual income tax often defers taxation of factor income, other federal taxes often tax consumption, selective fidelity to comprehensive income taxation, and the limitations of cross-section analysis. As a result, the JCT and other agency distributional models can seem ad-lib with the only common overriding constraint being the reflection of federal taxes on a cash-flow basis (and sometimes not even that).

As a tax person who likes to frame questions in tax terms, I get it. Most distributional models have their origins in, or rely heavily on, administrative tax data. Congressional tax committees and the Treasury Department have turfs to defend and/or expand. However, what’s happened in the inequality area is that too much has been jammed into tax rates. For example, transfers ought to be addressed in the post-government inequality analysis that doesn’t suffer from form-over-substance transfer-versus-negative-tax issues or open-ended tax rate denominators.

Refundables and traditional transfers (for example, public assistance and Medicaid) are essentially the same thing — means-tested payments generally unrelated to how taxes are applied to production or consumption — with the only difference that the refundables are conveyed through the tax system. In both cases, the government writes checks to, or on behalf of, beneficiaries, which is the classic definition of government outlays. Identifying refundables as negative taxes elevates form over substance and contradicts the principle of tax expenditures (and refundables are the tax expenditures most like government outlays) that seeks to break down artificial barriers between tax benefits and government outlays such as traditional transfers.

Tax rates across countries and over time shouldn’t be affected by whether stimulus or pandemic payments are channeled through the tax system, or whether in lieu of traditional public assistance a basic grant is transmitted through the tax system.¹⁷ There also can be some weird sorting within the refundables. Why would stimulus or pandemic payments through the tax system not be treated as negative taxes but refundable child tax credits and the like would be treated as negative taxes? Isn’t all this just means-tested public assistance that should be looked at together and away from how we generally think of tax rates?

Moreover, in ascertaining income inequality, average tax rates wouldn’t seem to matter as much for equity purposes as post-government income results.¹⁸ For targeting both equity and efficiency, marginal tax and outlay rates that take account of tax, transfer, and other phaseouts would seem to matter a lot more than average tax rates. It might be fruitful for the JCT and other government staffs to spend less time making average tax rates into some kind of stand-alone equity measure that includes transfers and more time looking at marginal tax and outlay rates.¹⁹

Decades ago, the JCT and the CBO began expanding their income classifiers to cover topical issues in healthcare and retirement. As a result, more double counting became evident in the tax rate denominators as both financing and spending (that is, sources and uses) of healthcare and retirement were added. Interestingly, even though it has been a periodic policy topic through the years, the government agencies generally have not added accrued capital gains to the income classifier,²⁰ nor did the JCT staff see fit to discuss accrued capital gains even in this pamphlet that ostensibly was supposed to focus on high income and wealth.²¹ As something like a carbon tax gets more attention, one wonders if carbon consumption will make it into the government agency tax rate denominators ahead of general accrued capital gains.

One rationale (and the most appealing justification, though ultimately not winning, to me) for including transfers and things like private retirement benefits as well as payroll taxes and other financing means in tax rate denominators stems from the Schanz-Haig-Simons or comprehensive income tax angle. But ideally, comprehensive income taxation should be modeled and examined on a longitudinal basis, yet all the models discussed here are cross-sectional. These cross-section models target the government’s cash flows, with demographics and time patterns putting both program contributors for something like Social Security right next to program beneficiaries, often within the same income grouping. At least for government transfers, it seems reasonable to omit transfers from tax rate denominators because the government is endogenous to the level and form those payments take.

Put differently, there’s no need to avail comprehensive income taxation of government transfers when the government itself can tailor those outlays internally on a pretax basis to meet equity needs. For Social Security, its trust fund and benefit setup already reflect government decisions about progressivity. If it’s felt that the overall progressivity of Social Security, taking account of contributions and benefits, should change, that doesn’t necessarily require taxing more benefits — there could be an adjustment made directly in the benefit formula — and is best looked at in a post-government context and not crammed into an ad hoc tax rate story.

Solving Tax Rate Inconsistency

Most of the policy-based references to the ratio of federal taxes to GDP adhere to the national accounting standard of treating tax refundables as transfers rather than negative taxes, and not double counting uses and sources in the tax rate denominator. There are three ways to deal with the tax rate discrepancy between federal taxes as a share of GDP and the average tax rates characterizing the JCT and other income distribution models that don’t adhere to NIPA calibration: Ignore the inconsistency, change the NIPA result, or change the income distribution models.

I don’t believe that the current practice of not acknowledging the discrepancy is satisfactory. Historical, international, and other comparisons policymakers are interested in can get twisted by this discrepancy, as indicated by the difference between the 2018 NIPA result that federal taxes are 17.3 percent of GDP and the comparable implied result consistent with how the JCT looked at historical income distributions of 14.6 percent. The discrepancy invites policy whipsawing: The 17.3 percent is high enough for someone to claim that taxes shouldn’t be increased overall to raise revenue, while the JCT says that taxes are already progressive and post-government income inequality hasn’t changed much so there’s no equity reason to change taxes (even though the JCT’s historical distribution is consistent with a meta tax rate result of 14.6 percent and not 17.3 percent).

It’s unlikely that policymakers will stop using the NIPA ratio of federal taxes to GDP because the concept is just too embedded, so the second option is out. That leaves the third option: Discipline in framing income distributions is needed for historical and international comparisons, as the Bureau of Economic Analysis recognized in its 2015 decision to treat refundables as transfers,²² and to prevent form-over-substance classification issues. Yet full adherence in distributional models to the NIPA approach, including all NIPA components, seems extreme in isolating the observations in cross-section models that for demographic reasons tend to be sorted exclusively into programmatic financers/contributors or beneficiaries.

Instead, a riff on the third option could be to impose the NIPA quantitative constraints on income distributions, so the full amount of national income is distributed in the various pre- and post- concepts and for tax rate purposes as well as income share results, and double counting is dispensed with, but apply a half-and-half contributor/beneficiary split at the pre-level. Thus, reflecting Social Security and other retirement benefits as well as health outcomes in income distributions doesn’t require JCT-like double counting of both finance and usage in a tax rate denominator — instead, a half-and-half approach could be applied that doesn’t violate the NIPA quantity constraint (and doesn’t make the income measure lopsided in favor of noncapital income). This half-and-half method would add longitudinal flavor to enrich cross-section distribution models so as not to make older people look like such tax scofflaws.²³

Conclusion

It’s a little surprising that the JCT staff as an institution chose to take an official position on historical income inequality, a topic that hasn’t been a routine or even occasional JCT task. Of course, it’s possible that the JCT staff was asked to look into historical income inequality by the House Ways and Means Committee for the hearing. And it’s possible that the JCT staff firmly believe the research of Auten and Splinter and saw this as the time and method to publicly affirm that, and also perhaps believed that the JCT’s standard distributional approach is buttressed by the historical modeling in the pamphlet.

In any case, the JCT could have neutrally surveyed the issues and the literature (including Auten and Splinter) without essentially adopting the methods and results of Auten and Splinter’s historical income research, but chose not to. Now that the JCT has taken a view on historical income inequality that has been linked by the JCT itself to its standard distribution for policy proposals, both of these methods can be looked at on their own and in relation to each other. For example, relative to NIPA national income, both the JCT’s standard and historical income distributions are overweighted with noncapital income, causing tax rates at the upper end to be overstated.

Details one might expect would come with the JCT’s taking a stance on historical income inequality, such as how to reconcile distributional model tax results with NIPA principles as discussed above, are slim in the pamphlet or effectively referred to other papers (including a featured referral to a paper by Treasury and JCT staffers that, as noted above, isn’t consistent with the JCT pamphlet’s treatment of refundables for tax rate calculations). It’s also possible that this may be the first in a series of JCT pamphlets on income and wealth distribution that will flesh the topic out over time.

It’s unclear why the JCT staff didn’t explore accrued capital gains much or look at the very top of the income and wealth spectrum even though the hearing topic was high income and high wealth — perhaps that, too, had to do with what was requested, or time and resource limits. Because the JCT staff didn’t present much on those topics, what was obscured is that with a model that takes account of accrued capital gains or, under a strict NIPA approach that excludes capital gains, a model with a broader capital income distribution, a desire to address tax rate regressivity or post-government income regressivity at the very top likely would have to employ tax changes.

This article has focused on tax rate construction in the JCT pamphlet and thus didn’t get very far into the JCT’s historical post-government income distribution results. Procedurally and substantively the tax rates and post-government income results can be two separate things, but the JCT’s deviation from NIPA calibration in (1) the calculation of tax rates and (2) general noncapital income lopsidedness in the tax rate denominator would seem to foreshadow the JCT historical income distribution.²⁴

A preview of my take on historical income distribution is that PSZ’s view on tax evasion is compelling, and it’s a good thing when there’s a wealth inequality story that complements an income inequality story. PSZ’s pre-government income matches up well with their pre-government wealth. Because I prefer the CBO’s size-discounted family unit over PSZ’s adults as the unit of observation (though it’s understood why PSZ chose adults only), my read is that historical income inequality has trended somewhere between the PSZ and CBO findings.

Appendix

Table 2 details the 2018 tax rate results shown in Table 1, starting with the NIPA tax rate of federal taxes as a percentage of GDP. The adjustments involve estate and gift tax receipts, U.S. federal taxes paid by foreigners, a net rest-of-the-world income adjustment (foreign income of U.S. persons less U.S. domestic income of foreigners), the subtraction of consumption of fixed capital, and addition of a statistical discrepancy to get from federal taxes as a percentage of GDP to Table 1’s refined definition of federal taxes divided by NIPA’s national income. This latter result, 19 percent, is also the overall federal tax rate in PSZ’s distribution model.

To get to the overall tax rate in JCT’s historical model that’s deduced from JCX-24-21 (the JCT didn’t provide total taxes for its historical model), refundables are treated as negative taxes, certain taxes are omitted, and other adjustments are made. Government transfers (also referred to as government social benefits paid to domestic persons) are added to NIPA’s national income along with an income adjustment to match the JCT’s historical tax rate of 15.1 percent.

Finally, some of these adjustments are reversed to arrive at the JCT meta tax rate implied by the JCT’s historical distribution model. The JCT meta rate is here the equivalent of NIPA federal taxes as a percentage of GDP, with the JCT meta rate adjusted to be consistent with how the JCT historical distribution tax rate is derived. The primary adjustments are that the JCT meta tax rate includes transfers in the tax rate denominator (consistent with what JCT does for its historical distribution tax rate) and treats refundables as negative taxes, unlike the NIPA-based (federal taxes/GDP) or the PSZ distributional tax rate which is consistent with how the NIPA-based (federal taxes/GDP) rate is constructed.

Table 2. Mapping NIPA to Distributional Model Tax Rates, 2018^a

	Source for Numerator and Denominator			Tax Rate	Notes
	Source	Numerator	Denominator
1. (Federal taxes)/GDP	NIPA	$3,568	$20,612	17.3%	Table 1 entry
2. Tax adjustments for distributions:
a. Estate and gift taxes	Tax data	$20
b. Tax paid by foreigners	PSZb	-$227
3. Income adjustments in distributions:
a. Net rest-of-world income	NIPA		$285
b. Consumption of fixed capital	NIPA		-$3,265
c. Statistical discrepancy	NIPA		$58
4. Intermediate result (also PSZ; denominator is NIPA national income)	[1] + [2] + [3]	$3,361	$17,690	19%	Table 1 entry
5. JCT historical tax adjustments:
a. Refundables as negative taxes	NIPA	-$148
b. Tax omissions, other adjustmentsc	Calculation	-$116
6. JCT historical income adjustments:
a. Government transfersd	NIPA		$2,923
b. Othere	Calculation		-$109
7. JCT historical tax rate (deduced from JCX-24-21)	[4] + [5] + [6]	$3,097	$20,504	15.1%	Table 1 entry
8. Tax adjustments to get JCT meta rate	-[2] - [5b]	$323
9. Income adjustments to get JCT meta rate	-[3]		$2,922
10. Implied JCT meta tax rate, comparable to (federal taxes)/GDP	[7] + [8] + [9]	$3,420	$23,426	14.6%	Table 1 entry
Sources: Auten and Splinter, supra note 3; JCX-24-21, supra note 1; NIPA; and PSZ, supra note 8. aAmounts in billions, everything on calendar year basis. The author is responsible for calculations. bPSZ, Appendix I, Aggregate tables S. A12, 12b, 12c. cInclude Federal Reserve Board receipts and current transfer receipts, and (possibly) different allocation of U.S. federal taxes to foreigners than PSZ made. dThe source is Auten and Splinter, supra note 3. Not sure whether refundables that are treated as negative taxes have also been included in the transfers (which may not be intended). eAdjustment required to reach JCT historical tax total.