The Efficiency-Equity Tradeoff of the Corporate Income Tax: Evidence from the Tax Cuts and Jobs Act

Layer 1: Overview

This paper estimates the firm- and worker-level effects of the corporate income tax cuts in the 2017 Tax Cuts and Jobs Act (TCJA) — the largest corporate tax cut in U.S. history — to inform the long-running efficiency-versus-equity debate over corporate taxation. The question matters because federal corporate tax reforms are rare, prior credible evidence comes mostly from subnational or small-economy variation (where factors are more mobile and the tax base smaller), and theory predicts alternate instruments behave differently, so existing estimates may not extrapolate to a major reform in a large advanced economy.

Identification exploits that TCJA cut the top C-corporation rate from 35% to 21% (a 40% reduction) while cutting the implied top rate for S corporations far less — from 39.6% to 37%, and to 29.6% for many via the new 20% Qualified Business Income deduction (a cumulative ~25% reduction). The authors use employer-employee matched federal tax records (corporate SOI files merged with W-2 and individual returns), tax years 2013-2019, on a balanced panel of large firms (>=50 employees and >=$1M sales each pre-period year): 15,490 firms and 108,430 firm-year observations. The main design is an event study / 2SLS comparing similarly sized C and S corps in the same industry-size bin, with firm and industry-size-year fixed effects and standard errors clustered by firm; entity-switchers are dropped. The identifying assumption is parallel trends absent the tax change (as in Yagan 2015), not random C/S assignment.

First stage: C corps’ marginal tax rate fell ~5.0 pp (s.e.=0.2) relative to S corps, raising the log net-of-tax rate ~6.6% (s.e.=0.2); C corps paid ~$2,100 (s.e.=341) less tax per worker. Real effects: C-corp sales rose 3.9 pp (s.e.=1.2) relative to S corps; pre-tax profits +3.0 pp (s.e.=0.7); after-tax profits +4.0 pp (s.e.=0.7); total payouts +21.9% intensive (s.e.=2.9) and +3.0 pp extensive (s.e.=0.5); employment +2.3% (s.e.=0.8); payrolls +3.4% (s.e.=0.8); net investment +2.9% (s.e.=0.4). The benchmark corporate elasticity of taxable income (pre-tax profits) is 0.46 (s.e.=0.11); after-tax-profit elasticity 0.61 (s.e.=0.11); investment elasticity 0.45 (s.e.=0.07). Worker earnings are flat for the bottom 90% (median wp50 coefficient -0.001, s.e.=0.004) but rise for the top 10%: +1.3% at the 95th percentile (s.e.=0.4), +4.8% at the 99th, and +4.8% for executives (top-5 paid; s.e.=0.7, earnings elasticity 0.73). Executive-pay gains barely shrink when controlling for firm performance (4.8% to 4.5%) and are concentrated among incumbents, consistent with rent-sharing rather than productivity.

Responses concentrate in capital-intensive industries and are not larger for cash-constrained firms, pointing to a cost-of-capital channel rather than liquidity. Via a stylized model, a $1 marginal cut in corporate tax revenue generates $0.44 in additional output; revenue falls $0.85 per $1 mechanical loss (total -$86 billion, 0.40% of GDP). Factor incidence: 51% of gains to firm owners, 10% to executives, 38% to high-paid workers, 0% to low-paid workers. Across the income distribution, 80% of gains accrue to the top 10% and 20% to the bottom 90%, with gains concentrated in the Northeast/West and large high-income cities. The corporate tax is ~twice as inefficient as the personal income tax but similarly progressive, suggesting margin-of-efficiency gains from shifting toward personal income taxation. Results are short-run and abstract from public-goods provision and deficit financing.

Layer 2: Deep Dive

What is the identification strategy and what are the main threats to it?

The strategy is a difference-in-differences/event study (and 2SLS) comparing C corporations to S corporations in the same industry-size bin before and after TCJA, instrumenting the change in the log net-of-tax rate with pre-existing C/S entity status, with firm and industry-size-year fixed effects and firm-clustered standard errors. The identifying assumption is parallel trends in outcomes absent the tax change (not random C/S assignment), supported by (a) flat pre-trends in the event studies, (b) Yagan (2015) showing C and S trends were statistically indistinguishable 1996-2008, (c) the unexpected nature of TCJA before the 2016 elections limiting anticipation, and (d) industry-size-year fixed effects matching firms in similar product markets. Main threats: anticipatory/intertemporal tax shifting (some rate decline already in 2017; executive pay also trends up in 2017); other concurrent TCJA provisions (bonus depreciation, DPAD repeal, NOL/interest limitation, international); endogenous entity switching; differential industry-size composition; and general-equilibrium/SUTVA violations where C-corp gains could be S-corp mirror-image losses or where common wage effects are absorbed by time fixed effects.

What are the main mechanisms and how are they distinguished empirically?

The authors argue the dominant mechanism is a reduction in the cost of capital from the permanent rate cut, not liquidity relief and not primarily bonus depreciation. Evidence: (1) responses are larger in capital-intensive industries (profits and investment), consistent with the cost-of-capital first-order condition; (2) high-cash firms are if anything more responsive than low-cash firms, ruling out liquidity constraints (and thus income effects); (3) bonus depreciation is downweighted because many eligible firms do not claim it, much capital (intangibles, structures) is never fully expensed, C and S corps had near-identical expensing exposure (so the design differences them out), and the investment response is driven almost entirely by short-lived assets rather than the long-lived assets where accelerated depreciation is most valuable. A complementary dynamic-adjustment-cost model (Auerbach-Hassett 1992 with Foertsch 2018 cost-of-capital inputs) yields elasticities very similar to the benchmark.

What heterogeneity is documented?

By capital intensity: C corps in capital-intensive industries show significantly larger profit and investment responses (supporting the cost-of-capital channel). By liquidity: high-cash firms are no less (if anything more) responsive than low-cash firms, contrasting with Zwick and Mahon (2017). By firm size: no clear pattern in profits, median earnings, or investment, with only suggestive evidence that high-income-worker gains are larger in smaller firms. By worker position: earnings gains are concentrated entirely in the top 10% of the within-firm distribution and especially in executives, with zero gains below the 90th percentile. By worker tenure: gains are driven by incumbents, not new hires (consistent with rent-sharing). Geographically: gains concentrate in the Northeast and West and in large high-income commuting zones (e.g., ~3x the median CZ gain in New York City, ~5x in the San Francisco Bay Area).

What robustness checks are run?

Alternate specifications (Table 7): cohort(age)-by-year FE, state-by-year FE, firm-specific pretrend controls, 6-digit NAICS industries, reweighting S to match the C industry-size distribution, inverse-propensity-weighting, log-transformed outcomes, winsorizing at 5th/95th percentiles, and 2016-sales/payroll weighting — elasticities are stable. Alternate samples (Table 8): excluding firms with >$1B sales or >10,000 employees, excluding mismatched industries (C share >80% or <20%), excluding manufacturing (trade-war exposure), unbalanced panel, excluding public firms, excluding industries most exposed to DPAD/NOL/interest-limitation/bonus-depreciation provisions, excluding multinationals, dropping tax years 2017-2018 (anticipation/shifting), and dropping single-owner S corps (wage/profit reclassification). Entity switching rose only from ~0.1% to ~0.3% (profit-weighted) and is negligible. Most estimates stay within the benchmark confidence intervals.

How does this paper relate to and differ from closely related prior work?

It builds on the C-vs-S comparison design of Yagan (2015) but studies marginal corporate rate cuts rather than the 2003 dividend tax cut. It obtains an investment elasticity (0.45) very close to Chodorow-Reich et al. (2023)’s 0.52 despite a different identification strategy and sample. Its corporate ETI (0.46) is below state/local estimates (Giroud-Rauh ~0.50; Suarez Serrato-Zidar ~0.9; Bachas-Soto 3.0-5.0 in Costa Rica) but above typical personal-income ETIs (Saez et al. central 0.25), consistent with distortions scaling with factor mobility. Its incidence finding — that the corporate tax falls on capital and high-income workers — differs from Fuest et al. (2018), who find German municipal corporate tax hikes fall on low-skilled/marginally-attached workers (the authors note possible asymmetry between hikes and cuts and small-firm effects), and aligns with Risch (2024). It uses directly observed owner returns and the full earnings distribution, requiring weaker assumptions than Suarez Serrato-Zidar (2016, who infer owner returns structurally) and Fuest et al. (who assume negligible rental-rate changes).

What are the policy implications and their scope conditions?

On efficiency: a $1 cut in corporate tax revenue yields $0.44 of additional output, and current U.S. top corporate rates appear below the revenue-maximizing rate (revenue falls only $0.85 per $1 mechanical loss). The corporate tax is ~twice as inefficient as the personal income tax but similarly progressive, and 3-4x more progressive than the payroll tax while being 2-3x as inefficient — implying that shifting the federal revenue mix toward personal income taxes could raise efficiency without much loss of progressivity. On equity: the cuts are regressive in the short run, with 80% of gains to the top 10% (24% to the top 1%, 56% to the 90-99th percentiles), 0% to low-paid workers, and 17% flowing to foreign equity holders. Scope conditions: estimates are short-run (through 2019, pre-COVID); they hold welfare equal to output (ignoring utility curvature); they assume a representative consumer (no consumer-price channel) and equal redistribution of revenue; they abstract from deficit financing, public-goods provision, and long-run productivity/wage effects; and the very largest C corps have no S-corp analogue, so their responses are not well identified.

What other significant findings, extensions, or caveats appear?

Employment increases reflect predominantly reallocation of workers across sectors rather than net new hiring, which the authors account for in the aggregate analysis (and is why incidence focuses on wages, not employment). New investment gains are in short-life assets (e.g., computers), with no change in long-life machinery or structures. Firms returned excess profits via dividends and buybacks but did not increase equity or debt issuance, and shareholder-payout results are robust to excluding multinationals (so the repatriation holiday is not the driver). Executive pay shifted forward into 2017 (bonuses) to be deducted at the higher pre-cut rate. Caveats flagged by the authors: rent-sharing tests are suggestive not dispositive (conditioning on post-treatment outcomes; unobserved hours/effort; short two-year horizon); private-income components are precisely estimated but the welfare confidence interval includes zero (up to ~0.4% of GDP); and long-run channels (productivity, lower prices, real wages) and offsetting cuts to public services/transfers are outside the analysis.

Key Concepts

C corporation vs. S corporation: The two legal entity types whose divergent TCJA tax treatment provides identification. C corps pay corporate income tax directly (rate cut 35% to 21%) and their dividends are taxed at the shareholder level; S corps pass income through to up to 100 individual U.S. shareholders who pay ordinary income tax (top rate cut 39.6% to 37%, or 29.6% with QBI), with no corporate-level or dividend tax.

Implied marginal tax rate (for S corps): Because S corps pay no entity-level tax, their firm marginal rate is constructed as the ownership-share-weighted average of the individual marginal income tax rates of the firm’s owners, computed from linked personal returns (e.g., two equal owners at 25% and 35% imply 30%).

Corporate elasticity of taxable income (ETI): The percent change in the corporate tax base (pre-tax profits) per percent change in the net-of-tax rate; the paper’s benchmark is 0.46. Following Feldstein (1999), it summarizes the deadweight loss / efficiency cost of the tax under negligible income shifting and income effects.

Net-of-tax rate: One minus the marginal tax rate, ln(1-tau); the object firms optimize against, used to scale reduced-form effects into elasticities. TCJA raised C corps’ log net-of-tax rate by ~6.6% relative to S corps.

Cost-of-capital channel: The mechanism by which a lower tax rate (or higher expensing parameter theta) reduces the user cost of capital phi = r(1-theta*tau)/(1-tau), raising capital demand, labor demand, and firm scale — the paper’s preferred interpretation, distinguished from liquidity effects.

Marginal excess burden: dW/dT, the change in welfare (output, defined as private income plus tax revenue) per dollar of corporate tax revenue; estimated so that $1 of foregone corporate revenue generates $0.44 of additional output.

Incidence across the income distribution: An extension of factor incidence that assigns owners’ capital gains back to workers using the Distributional Financial Accounts (since many workers hold equity and many owners work), yielding the result that 80% of tax-cut gains accrue to the top 10% of earners.

Rent-sharing: The channel whereby earnings gains accrue to incumbent high-paid workers and executives rather than to new hires (the marginal unit of labor), with executive pay only weakly tied to firm performance — interpreted as workers/executives capturing a share of excess after-tax profits.