The revenue maximizing rate for federal capital gains taxes is a matter of considerable discussion and disagreement in the economics literature. Agersnap and Zidar (2021) use state level variation in capital gains taxes to measure the elasticity of revenue with respect to the tax rate. The authors use state-level panel data from 1980 to 2016 and implement a direct-projections approach to examine the dynamic impact of capital gains taxes on realizations over a 10-year time horizon. The authors then adjust for state-level migration decisions to estimate a policy-relevant elasticity at the federal level. This elasticity is used to derive the revenue maximizing federal tax rate and is also used to calculate to calculate the MVPF of a reduction in the capital gains rate.
Applied to tax policy reforms, the MVPF measures the benefit-cost ratio associated with a cut to the tax rate, in which the total benefits are the mechanical tax savings for those who benefit from the tax cut, while the costs are the effects of the tax reform on government tax revenues, i.e. the sum of the mechanical and behavioral revenue effects. In Appendix F of the paper, Agersnap and Zidar (2021) use this approach to calculate the MVPF of a reduction in the capital gains tax rate.
Using the same style of notation as in Saez, Slemrod and Giertz (2012), they write the formula for the MVPF as:
In this case, the authors a tax reform that applies the same marginal change d \tau in the capital gains tax rate across all tax brackets, and so the mechanical effect can simply be written as:
This is the total amount of capital gains realizations multiplied by the marginal tax change.
The behavioral effect associated with the reform can be written as the average tax rate, multiplied by the change in capital gains realizations:
where the second line follows from the definition of the realization elasticity, \varepsilon. Combining the above equations, the authors find the formula for MVPF in their setting to be:
This formula resembles the one for MVPF calculations in the case of a linear tax. Unlike equivalent formulas from the top income tax literature, a Pareto parameter does not appear here, as the authors are assuming a reform that changes the tax rate on the entire tax base of all realized capital gains. However, to account for different tax rates across brackets, the formula contains both the average tax rate, which determines the effect of behavioral distortions on government revenue, and the marginal tax rate, which is relevant for individual decision-making. The differences between these two concepts are discussed in further detail in Section 4.4 of Agersnap and Zidar (2021).
The authors approximate the marginal tax rate τ by the average combined maximum state and federal tax rate, which was 27.82% in 2017, the last year for which they have data. For the average tax rate, they take the average federal tax rate of 17.79% and add the 2016 average population-weighted state tax rate of 6.27%, for a total average tax rate of 24.06%.
In the body of the paper, Agersnap and Zidar (2021) derive estimates of the realization elasticity ε across multiple time horizons and models. Their primary specification uses a 10 year time horizon as it incorporates behavioral effect in the short, medium and longer run, and thus gives the broadest possible picture of the average behavioral effect. In that primary specification, the authors calculate the elasticity to be \varepsilon = 1.87. Using that elasticity in the equation above produces an MVPF of 2.65.
The authors also consider several alternate elasticity calculations. They consider a case where only large tax changes are used in calculating the realization elasticity. That approach produces an elasticity of 1.48 and an MVPF of 1.97. They also consider a case where the elasticity calculation includes controls for other taxes. That approach produces an elasticity of 1.01 and an MVPF of 1.51.
The authors note that these estimates are based on an assumption that there are no spillovers between capital gains and other tax bases. If such spillovers do occur, the MVPF associated with capital gains tax cuts would be lower than estimated here. Also, while the denominator of the MVPF captures net costs to government revenue and thus society at large, the benefits that appear in the numerator of the MVPF accrue only to a specific set of people – the beneficiaries of the policy in question. In this setting, the beneficiaries of a capital gains tax cut are an average of taxpayers weighted by their total amount of capital gains realizations. The estimates from the baseline model thus suggest that for every $1 that a marginal capital gains tax cut would cost the government, it would provide $2.65 worth of benefits to the weighted average payer of capital gains taxes, while the other models imply a somewhat lower benefit amount.
MVPF = 2.7
Agersnap, Ole, and Owen M. Zidar. The Tax Elasticity of Capital Gains and Revenue-Maximizing Rates. American Economic Review: Insights, Forthcoming.
https://www.aeaweb.org/articles?id=10.1257/aeri.20200535&&from=f
Saez, Emmanuel, Joel Slemrod and Seth H. Giertz (2012). “The Elasticity of Taxable Income with Respect to Marginal Tax Rates: A Critical Review.” Journal of Economic Literature, 50(1), 3-50. DOI: https://doi.org/10.1257/jel.50.1.3