Update: The Washington Center for Equitable Growth has issued a statement regarding the Tax Foundation’s response to this critique.
The Tax Foundation last week released an analysis estimating that the recently released Tax Cuts and Jobs Act would increase U.S. Gross Domestic Product by 3.9 percent.1 The Trump administration and congressional Republicans have made it clear that they will justify their new tax proposals in large part on the basis of claims about economic growth and an associated increase in incomes, and proponents of the legislation immediately seized upon the Tax Foundation’s estimates as support for their claims. This note identifies two issues in the Tax Foundation’s analysis that suggest its estimated increase in GDP is probably substantially overstated.
First, the Tax Foundation appears to incorrectly model the interaction between federal and state corporate income taxes, thus overstating the effect of statutory rate cuts. Second, the Tax Foundation appears to treat the estate tax as a nondeductible annual property tax paid by businesses, which results in inflated estimates of the effect of repealing the tax. Appropriately addressing the issues raised in this note could reduce the Tax Foundation’s estimate of the increase in GDP that would result from the legislation to 1.9 percent—a reduction of roughly half—even if there are no other issues with the Tax Foundation’s estimates.
There is substantial uncertainty in this estimate of the GDP impact that would be generated by a revised version of the Tax Foundation model since it depends on an attempt to approximate the results of a model for which only partial documentation is publicly available. The estimate that would be generated by a revised model could be larger or smaller than this value.
Nonetheless, the significance of this difference raises important questions about the reliability of the Tax Foundation’s estimates. Critical assessment of the Tax Foundation’s analysis is particularly warranted, as some legislators have suggested that they might consider dynamic scores from organizations other than the nonpartisan Joint Committee on Taxation—the traditional source of nonpartisan estimates of congressional tax proposals—in determining the budgetary effects of the legislation.
This note does not attempt a complete assessment of the Tax Foundation model, which would be impossible without greater knowledge of the equations that make up the model. The criticisms raised in this analysis are based on inspection of publicly available estimates and documentation, as well as communication with Tax Foundation staff.2
The Tax Foundation appears to incorrectly model the interaction between federal and state taxes, inflating the effects of changes in the statutory corporate rate
The Tax Foundation model assumes that there is a fixed return required by investors. The long-run capital stock is set equal to the level at which a $1 increase in investment yields a pretax return sufficient to generate exactly that fixed required return after taxes. In computing the taxes on this $1 increase in investment, the Tax Foundation uses an estimate of the effective marginal tax rate on new investment.
It is well understood that the effective marginal tax rate on new investment can differ substantially from the statutory tax rate on business income. For example, in the case of a business tax system that allows full expensing—a policy under which businesses may deduct the full cost of any investment in the year the expense is incurred—the business-level effective marginal tax rate is zero, regardless of the statutory rate.3
Given this general finding, it is surprising that estimates generated by the Tax Foundation model suggest large effects of reductions in the statutory corporate rate, even after adoption of full expensing. The Tax Foundation’s estimates of House Republicans’ “Better Way” tax plan, for example, suggest that even after adopting full expensing and repealing the estate tax, reducing the corporate income tax rate from 35 percent to 20 percent would increase the size of the economy by 1.7 percent.4
A more recent Tax Foundation analysis concluded that full expensing for C corporations would increase GDP by 3 percent, a reduction in the statutory corporate rate from 35 percent to 20 percent would increase GDP by 3.1 percent, and both policies together would increase GDP by 4.5 percent.5 This analysis thus suggests not only a large effect of statutory rate cuts on top of expensing, but also finds that a 15 percentage point reduction in the corporate rate would have a larger effect than full expensing. This result is surprising, as expensing would be expected to have an effect roughly equivalent to reducing the statutory corporate rate to zero in a model like that used by the Tax Foundation.
These results are unusual for a model that determines the capital stock based on the effective marginal tax rate on new investment. Changes in the statutory rate could have an effect on the economy independent of the effective marginal tax rate in a model that assumes shifting of real economic activity across borders in response to changes in statutory rates. The Tax Foundation model, however, does not appear to include this channel, so this cannot explain the results.6
Another potential reason for the large effects of the statutory rate reduction could be that the Tax Foundation might compute effective marginal tax rates on intellectual property using a different methodology than for other types of assets. Some analysts have suggested that traditional effective marginal tax rate computations are inappropriate for intellectual property because the nature of such investments differs. But there is no indication that the Tax Foundation has chosen this approach.7
One factor that may partially explain these unexpected results could be that the Tax Foundation’s estimates of proposals to provide full expensing do not include expensing of inventories. If this is the case then proposals to reduce the statutory corporate rate would reduce the effective marginal tax rate on inventories, yet proposals for full expensing would not. As inventories account for less than 10 percent of the corporate capital stock, this could have a modest but noticeable impact on the results.8
In response to an inquiry about the growth from cutting the statutory rate, even after enacting expensing and repealing the estate tax in the “Better Way” plan, Tax Foundation staff attributed this growth to interactions between the federal corporate income tax and property, excise, and state and local taxes. On its surface, this does little to resolve the mystery, as state and local taxes are generally deductible from federal corporate income taxes, and the deductibility of such taxes would seem likely to moderate or eliminate most interactions.9 A recent paper from the Tax Foundation, “Measuring [the] Marginal Tax Rate on Capital Assets,” offers a potential explanation for why the Tax Foundation’s model might generate this surprising interaction.10
Specifically, the formula in this paper for computing the service price of capital—the pretax return gross of depreciation required for an investment to yield the required after-tax return—appears to suggest that federal corporate taxes are deductible from state corporate taxes when determining the value of depreciation allowances, but also that federal corporate taxes are not deductible from state corporate taxes when determining the rate at which profits are taxed.11
This apparent inconsistency would inflate the estimated tax rate on investments by undervaluing the depreciation deductions to which a business is entitled. More importantly, it would create a potentially spurious interaction that would cause reductions in the statutory corporate tax rate to reduce the effective marginal tax rate in the model—even when no such effect exists in practice. It would also likely cause the model to understate the growth impact of expensing.
To get a rough sense of the quantitative significance of this interaction, assume that the Tax Foundation’s recent estimate of the growth impact of expensing for C corporations and a 20 percent corporate rate is an approximately valid estimate for the impact of a proposal to expense capital investment including inventories. (This assumption is likely not precisely correct. Expensing of inventories would tend to increase the estimate and the potentially spurious interaction with state and local taxes likely could increase or decrease it depending on other details of the model.)
In models like that used by the Tax Foundation in which the effective marginal tax rate drives economic behavior, reducing the statutory corporate rate to zero would tend to yield the same result as providing full expensing. Thus, a rough estimate of the impact of reducing the statutory rate would be the ratio of the proposed reduction in the statutory corporate rate to the current statutory rate multiplied by the economic impact of full expensing including inventories. This suggests an estimate of the GDP impact of reducing the statutory corporate tax rate from 35 percent to 20 percent of 1.9 percent, rather than 3.1 percent.
This approximation for a revised estimate of the impact of reducing the statutory rate from 35 percent to 20 percent in the Tax Foundation model is subject to several sources of uncertainty. First, it assumes that the diagnosis of a spurious interaction between federal corporate taxes and state corporate taxes in the Tax Foundation model is correct. Second, it uses the Tax Foundation’s recent estimate of the impact of full expensing for C corporations and a 20 percent corporate rate as an estimate of the impact of full expensing including inventories. This could be understated if the impact of expensing inventories on GDP is particularly large, due to the impact of the federal-state interaction on the growth impact of expensing, or due to other interactions with the corporate rate such as that discussed in the next section. Third, it uses a linear approximation for the relationship of the rate cut proposal to a full expensing proposal, which likely slightly overstates the adjustment.
Notwithstanding this uncertainty, using this estimate of the increase in GDP from a reduction in the statutory rate from 35 percent to 20 percent would reduce the growth effect from 3.9 percent to 2.7 percent, a reduction of about one third. Even if the appropriate adjustment is somewhat overstated, it would still reflect a substantial change in GDP. Notably, this potentially spurious interaction would appear to affect any estimate generated by the Tax Foundation model for proposals that change the corporate tax rate—not only its estimates of the Tax Cuts and Jobs Act.
The Tax Foundation appears to treat the estate tax as an annual property tax paid by businesses, which would overstate the effects of repeal
As noted above, the capital stock in the Tax Foundation model is set at the level at which a $1 increase in investment yields a pretax return sufficient to generate a fixed return required by investors. In computing the tax rate for this computation, the Tax Foundation appears to incorporate a cost associated with the estate tax that would be equivalent to treating the estate tax as an annual nondeductible property tax paid by businesses.12 The tax rate appears to be based on a measure of the average estate tax liability divided by the capital stock, which is then grossed up to reflect a marginal rate.13
Treating the estate tax as a property tax paid by businesses would not be an accurate description of the practical operation of the tax. The estate tax applies to the transfer of personal estates with a value of more than roughly $11 million at death for couples.14 If this is, in fact, the modeling assumption adopted in the Tax Foundation model, it would seem to suffer from several flaws.
First, the Tax Foundation model is a model of homogeneous capital. In other words, there is nothing unique about assets held by one business relative to those held by others or about the capital provided by one investor relative to another.15 If capital is homogeneous, then there is no reason the marginal supplier of capital would necessarily be subject to the estate tax at all. Indeed, the underlying assumption of the model appears to be not only that increases in investment financed by increases in saving are uniform (in percentage terms) across the population, but also that they translate into increases in assets held at death, which also need not be true.16
Second, as noted above, the Tax Foundation assumes a fixed required rate of return. The assumption that there is a fixed rate of return required by investors is often referred to as the small open economy assumption and is justified on the basis that there is a global pool of capital ready and willing to finance profitable investments in the United States. Tax Foundation staff have endorsed this label and justification in descriptions of their model.17 Yet assuming that foreign investors are the marginal source of finance can dramatically change the effects of the tax system on the economy. In the extreme case in which capital is perfectly mobile across countries—the assumption made by the Tax Foundation—investor-level taxes in the United States that do not apply to foreign investors become irrelevant to the determination of the capital stock.18
While there are cases in which foreign residents are subject to the U.S. estate tax, there are numerous ways for foreigners to avoid the tax through planning or careful selection of assets. There is no compelling reason to think that in a small open economy model, foreign residents would treat the U.S. estate tax as fully marginal in their decision-making.
Third, this modeling approach appears to treat the estate tax as a nondeductible tax for the purposes of business income taxes.19 If the tax is nondeductible, then businesses must not only pay the estate tax but also corporate income tax on the additional return they earn to cover the estate tax. This modeling assumption would appear to create spurious interactions between different taxes. Suppose, for example, the United States adopted full expensing at both the federal and state levels. Corporate income taxes would then not affect the rate of return on an investment. But because the Tax Foundation model appears to treat the estate tax as a nondeductible tax, the statutory corporate tax rate would still discourage investment because it would increase the effective estate tax rate. Thus, even if the U.S. estate tax did affect the effective marginal tax rate in a small open economy model, the modeling approach would still appear to be inappropriate.
In the context of the Tax Cuts and Jobs Act, the Tax Foundation estimates that repealing the estate tax would increase GDP by 0.9 percent.20 But if this estimate is based on a decision to model the estate tax as a property tax paid by businesses, then this result is likely inappropriate. In a small open economy model with homogeneous capital, zero would probably be a more appropriate estimate. Reducing the impact of estate tax repeal from 0.9 percent to zero would represent a roughly 25 percent reduction in the growth effects of the plan.
It is important to note that the estate tax could, at least in theory, have important effects on the level of output in models that reject other assumptions of the Tax Foundation’s model. But this would require more substantial changes to the Tax Foundation model that would have significant consequences for estimates of the effect of other policies. In a closed economy model, for example, it might be reasonable to assume that some portion of the additional savings necessary to finance an increase in investment would come from wealthy families subject to the estate tax. Or, as another possibility, recognizing the heterogeneous nature of capital and the frequency of company founders and their heirs among the wealthiest Americans, one could argue for impacts based on indirect taxation of new businesses. Yet pursuing this latter approach would require a more significant recognition of the estate tax as a burden on labor rather than capital than the Tax Foundation’s current marginal tax rate computations suggest.
This note is far from a complete assessment of the Tax Foundation model, and conducting such an assessment would be challenging given the limited information publicly available about the model. Nevertheless, this note concludes that the Tax Foundation’s model probably includes at least two important flaws. Addressing these two flaws could reduce the Tax Foundation’s estimates of the growth impact of the Tax Cuts and Jobs Act roughly in half. The significance of this change raises substantial questions about the reliability of the Tax Foundation’s estimates of economic growth and suggests that policymakers should be skeptical of the results, especially when considering substituting these estimates for those from traditional legislative scorekeepers at the nonpartisan Joint Committee on Taxation.
- Tax Foundation, “Details and Analysis of the 2017 Tax Cuts and Jobs Act” (2017), available at https://taxfoundation.org/2017-tax-cuts-jobs-act-analysis/. ↩
- For two publicly available pieces of documentation, see Tax Foundation, “Overview of the Tax Foundation’s Taxes and Growth Model” (2017), available at https://taxfoundation.org/overview-tax-foundation-s-taxes-and-growth-model/ and Huaqun Li, “Measuring Marginal Tax Rate on Capital Assets” (Washington: Tax Foundation, 2017), available at https://taxfoundation.org/measuring-marginal-tax-rate-capital-assets/. ↩
- More precisely, the effective marginal tax rate would be zero for the business tax system judged in isolation. ↩
- Kyle Pomerleau, “Details and Analysis of the 2016 House Republican Tax Reform Plan” (Washington: Tax Foundation, 2016), available at https://taxfoundation.org/details-and-analysis-2016-house-republican-tax-reform-plan/. This statement assumes that the estimates in that report are presented in the order they were stacked for purposes of estimation. ↩
- Scott A. Hodge, “The Jobs and Wage Effects of a Corporate Rate Cut” (Washington: Tax Foundation, 2017), available at https://taxfoundation.org/jobs-wage-effects-corporate-rate-cut/. ↩
- Including this channel in the Tax Foundation model would change its other estimates in significant ways. For example, in estimating the effects of the House “Better Way” plan, the Tax Foundation concluded that a border adjustment would have a negative impact on GDP. If shifting real economic activity in response to the statutory rate were an important effect, this would tend to suggest the border adjustment would have a positive impact on GDP. ↩
- Li, “Measuring Marginal Tax Rate on Capital Assets.” ↩
- Congressional Budget Office, “Taxing Capital Income: Effective Marginal Tax Rates Under 2014 Law and Selected Policy Options” (2014), available at https://www.cbo.gov/publication/49817. The CBO analysis excludes certain types of capital that appear to be included in the Tax Foundation’s model and thus likely overstates the share of the capital stock consisting of inventories in the Tax Foundation model. Note, however, that inventories would be a larger fraction of produced capital (equipment, structures, inventories, and intangibles), which is the subset of capital that can respond to changes in taxation. ↩
- The taxes would still have a direct effect on the economy even as deductibility moderates or eliminates interactions with federal corporate taxes. Sales taxes on investment goods are generally depreciated as part of the cost of the investment, but the effect is the same. ↩
- Li, “Measuring Marginal Tax Rate on Capital Assets.” ↩
- Some states allow corporations to deduct federal corporate income taxes from their income on their state corporate tax returns, but most do not. For a summary, see the Federation of Tax Administrators, “Range of State Corporate Income Taxes” (2016), available at https://www.taxadmin.org/assets/docs/Research/Rates/corp_inc.pdf. However, the potential inaccuracy of the underlying assumption is likely less consequential than the apparent inconsistency. ↩
- Li, “Measuring Marginal Tax Rate on Capital Assets”; Tax Foundation, “Overview of the Tax Foundation’s Taxes and Growth Model.” ↩
- Tax Foundation staff pointed to Aldona Robbins and Gary Robbins, “The Case for Burying the Estate Tax” (Irving, Texas: Institute for Policy Innovation, 1999), available at http://www.ipi.org/ipi_issues/detail/the-case-for-burying-the-estate-tax as a reference. ↩
- The exemption is roughly $5.5 million per person and thus a couple can shelter from tax roughly $11 million. However, for nonresidents, the estate tax exemption is only $60,000. ↩
- The Tax Foundation appears to rule out shifting between pass-through and corporate status and allows for a different composition of assets within sector. Thus, strictly speaking, this assumption applies within sector. ↩
- In addition, depending on the details of the marginal rate computation, the Tax Foundation may be assuming that the marginal charitable contribution resulting from an increase in wealth is zero. ↩
- Alan J. Auerbach and others, “Macroeconomic Modeling of Tax Policy: A Comparison of Current Methodologies” National Tax Journal, forthcoming, available at https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3060608; Kyle Pomerleau and Stephen J. Entin, “The Tax Foundation’s Model Results in Context” (Washington: Tax Foundation, 2016), available at https://taxfoundation.org/tax-foundation-s-model-results-context/; Gavin Ekins, “Time to Shoulder Aside `Crowding Out’ As an Excuse Not to Do Tax Reform” (Washington: Tax Foundation, 2017), available at https://taxfoundation.org/crowding-out-tax-reform/. ↩
- This observation applies to a larger set of taxes that appear to have large effects in the Tax Foundation model, including taxes on capital gains and interest, but these are not directly relevant to the Tax Foundation’s score of the Tax Cuts and Jobs Act. ↩
- Li, “Measuring Marginal Tax Rate on Capital Assets”; Tax Foundation, “Overview of the Tax Foundation’s Taxes and Growth Model.” The estate tax is implicitly nondeductible because it is divided by one, minus the marginal corporate tax rate. ↩
- Tax Foundation, “Details and Analysis of the 2017 Tax Cuts and Jobs Act.” ↩