Income Tax Cuts

What the evidence says

Cross-country panel evidence puts income taxes below corporate taxes but above consumption taxes on a scale of harm to long-run growth: comparing tax types, Johansson (2008) finds corporate taxes are most harmful to GDP per capita, personal income taxes second, and consumption taxes least harmful, though the paper does not attach a number to the gap between them. At the level of individual firms, corporate income tax cuts show a repeated, time-lagged, and directionally consistent innovation effect. Atanassov (2019), using a firm-level difference-in-differences design around state tax changes, finds that a tax cut increases patent counts by 0.63 patents at 3 years and 0.79 at 4 years, and citations per patent by 0.75 at both 3 and 4 years, relative to firms that saw no cut; tax increases produce the mirror-image decrease (0.25 and 0.30 fewer patents; 0.35 and 0.37 fewer citations per patent). R&D expenditure increases starting 1 year after a cut, before the patent effect shows up — consistent with a cash-flow channel that shows up in spending before it shows up in output.

A separate literature on R&D-specific tax incentives complicates how much of that reported response is real. Zhao (2021), studying a Chinese R&D tax notch with a bunching design, finds reported R&D investment increases by 25% for large firms, 17% for medium firms, and 10% for small firms — but estimates that 24.2% of that reported increase is relabeling of existing expenses rather than new R&D. The real component still produces a productivity gain: a 100% increase in real R&D increases total factor productivity by 9% in the same model.

On personal income tax rates specifically, Johansson (2008) finds that high top marginal rates decrease productivity growth by reducing entrepreneurial activity, and correspondingly that reducing top marginal rates increases economy-wide productivity in OECD countries with a large share of industries with high enterprise-creation rates. The same paper finds that highly progressive personal income tax schedules decrease GDP per capita through two channels — lower labor utilization and lower productivity — though it does not separate how much of the effect comes from each channel.

At the U.S. federal reform level, Malkov (2021) models three historical income tax changes — the Tax Reform Act of 1986, the Economic Growth and Tax Relief Reconciliation Act of 2001, and the Tax Cuts and Jobs Act of 2017 — and finds each increased aggregate welfare for married couples, by 0.55%, 0.44%, and 0.62% of aggregate labor income respectively. The paper attributes most of that welfare gain to increased female labor-force participation rather than to existing workers increasing their hours, and reports welfare gains per dollar of tax revenue reduction ranging from 0.63 USD (for the 1993 Omnibus Budget Reconciliation Act) to 1.10 USD (for the 2017 Tax Cuts and Jobs Act).

The Earned Income Tax Credit — a income-tax-linked transfer rather than a rate cut — shows a split response by marital status. Shirley (2020), using birth-timing as a source of exogenous variation, finds EITC receipt increases earnings and employment for unmarried women in the 12 months after their first child's birth, including those with a high school degree or less. For married women the same design finds effects that are usually small and insignificant, except negative earnings effects for married women with a high school degree or less. Decomposing the mechanism, the direct income effect of EITC receipt is negative across all groups of women, while the information effect of first receiving the credit is positive for unmarried women and negative for married women.

Older evidence on state and local business taxes is directionally suggestive but not quantified: Bartik (1991) reports that variations in state and local business taxes have effects on area growth of a size most policymakers would treat as meaningful, and that recent studies at the time mostly agreed on both the direction and the approximate size of that effect — but no claim here supplies the actual magnitude. The same paper notes a business tax cut may fail to increase jobs in a state or metro area if the tax cut is paired with a deterioration in public services to business.

Where the evidence splits

No contradictions in the current sources.

Thresholds and dose-response

Zhao (2021) finds the reported R&D response to a corporate tax notch scales with firm size: large firms report a 25% increase in R&D investment, medium firms 17%, and small firms 10%, all under the same policy notch.

Atanassov (2019) finds a timing threshold in the innovation response to a corporate tax change: no effect appears in the years before the tax change, and most of the change in patenting and citation outcomes occurs 2 or more years after the tax change — the measured patent effect grows from 0.63 additional patents at year 3 to 0.79 at year 4, and tax increases show the same delayed, mirror-image pattern.

Open empirical questions

• None of the sources here quantify the growth or GDP magnitude of a personal income tax cut outside of Malkov's welfare-model estimates and Johansson's directional cross-country ranking — no dollar-for-dollar GDP or employment multiplier is given.
• Distributional effects of income tax cuts by income level are only addressed for the EITC and only by marital status and education; no source here breaks out effects by income decile more broadly.
• Long-run fiscal feedback (how much of the revenue loss from a cut is offset by growth or labor-supply response) is only measured here as a static welfare-per-dollar figure for U.S. federal reforms, not as a dynamic revenue estimate.
• No source here separates the labor-utilization channel from the productivity channel in Johansson's progressive-tax-schedule finding, so the relative size of each channel is unmeasured.