Firms adjust their capital stock through both investment in new capital and maintenance of existing capital. Omitting the maintenance decision creates two problems for standard models of tax reform. First, tax reforms reduce the cost of new investment but not maintenance, so the user cost of capital falls less than standard models imply. Second, elastic maintenance means depreciation rises when investment becomes cheaper, breaking the one-for-one mapping from investment to capital. I study the importance of these channels using freight railroad regulatory filings and corporate tax returns. The estimates imply the tax elasticities of capital, output and wages are about 70\% as large as implied by standard investment models, while investment elasticities only map into 2/3 as much capital. Applied to the 2017 Tax Cuts and Jobs Act, I find that it generated \$240B less corporate capital than conventional estimates.

How far is the tax system from optimal, and how much does it matter? I develop a sufficient statistic that answers both questions: the sine of the angle between the welfare and revenue gradients of the tax system. When the two are aligned, the system is optimal. I apply the framework to the postwar tax system, which shifted from drawing nearly half its revenue from corporate taxes to less than 15 percent. A small optimal reform in the 1950s would have increased welfare by 0.5% of GDP, but the system became Ramsey-aligned by the 1980s, leaving few gains from reform since then.

Using a randomized information experiment on Italian firms, we show that the real investment response to inflation expectations depends systematically on the horizon of those expectations and on the pledgeability of capital. The tangible investment semi-elasticity with respect to expected inflation nearly doubles as the forecast horizon extends from six months to five years, and is roughly four times larger than the corresponding intangible response. These joint facts are difficult to reconcile with standard valuation or money illusion mechanisms but are consistent with a Mundell–Tobin financing channel in which firms lock in long-term nominal borrowing to fund tangible assets. 

Measured tax incidence is biased when agents differentially select out of the observed tax base. When low-passthrough agents are more likely to exit, measured incidence overstates the true domestic burden. I show that correcting this bias requires only the exit share. Applying this framework to tariffs, where rerouting generates detectable exit, I find that a ten percentage point tariff rate increase raises exit probability by 10-30 percentage points. Correcting for selection implies that the incidence of tariffs on the U.S. during the 2018-19 trade war was 85\% as large as commonly supposed, and 75\% as large for intermediate business inputs. Existing tax provisions can discourage exit---corporate tax deductibility reduces the return to avoidance---but a major reform in 2017 weakened this channel, reducing tariff revenue by \$2-9 billion.

This note extends smooth local projections to panel data and uses Monte Carlo methods to explore the bias and variance properties of smooth panel local projections (SPLP). SPLP allows researchers to penalize the impulse respond toward a polynomial, while standard local panel projections (PLP) are nonparametric but result in theoretically unappealing IRFs because they are too lumpy. Relative to PLP, SPLP has appealing properties in smaller samples.

Investment subsidies depress the market for used capital through two channels. On the supply side, claiming the subsidy on an asset accelerates tax deductions, increasing the capital gains tax on its eventual sale. On the demand side, subsidies typically apply only to purchases of new capital, making used assets relatively more expensive for buyers. Using cross-sectional variation in exposure to bonus depreciation, I show that higher subsidies cause firms to stop selling used capital entirely. This reduction in used capital supply depresses entry of young firms, which disproportionately rely on used capital. The 2017 Tax Cuts and Jobs Act reduced both wedges by cutting the applicable capital gains tax rate and extending bonus eligibility to used property. Consistent with the mechanism, capital sales rose sharply, reversing the pre-TCJA pattern. I develop a general equilibrium model calibrated to the reduced-form elasticities to quantify the value of  eliminating policy distortions in the used capital market.

Abstract: Modern Money Theory (MMT) has risen to prominence in popular policy debates within macroeconomics. MMT economists argue for creating a job guarantee program, which they argue would generate price stability. Using a benchmark model of time consistency supplemented with a job guarantee, we conclude that once policymakers' incentives are considered, the job guarantee does nothing to help stabilize prices. We compare this program to a competing proposal to maintain price stability and full employment, NGDP targeting.