Research

Does the Earned Income Tax Credit cushion local labor markets from wage shocks? I show that the answer depends on where a shock lands on the credit’s schedule. In the phase-out, a smaller credit offsets part of any wage gain, so greater generosity should reduce the employment response; in the phase-in, the credit adds to each dollar earned, so it should amplify the response. I test this using the uneven 1993–1996 EITC expansion across commuting zones, measured with predetermined family-size shares, and two wage shocks that hit different parts of the earnings distribution: local economic booms and Chinese import competition. A $100 higher maximum credit lowers single women’s employment response to a 1% local wage gain by 0.075 percentage points and raises less-educated women’s employment loss from $1,000 of Chinese imports per worker by 0.8 percentage points. The effects are concentrated in the education groups whose position on the schedule the theory points to, and no such pattern appears in the 1980s, before the credit varied with family size. The results imply that nonlinear transfers can amplify local shocks, not only cushion them.

Published: 13 August 2026, Social Indicators Research.

This paper examines how increases in wives' market value, measured through wages, shape intra-household tradeoffs in Japan. Using a Difference-in-Differences strategy, I exploit minimum wage hikes to study their effects on wives' relative income and time allocation. Evidence from the Japanese Household Panel Survey shows that higher minimum wages reduce the gender wage gap by encouraging wives—particularly those with young children—to enter the labor force and extend their working hours. Wives without young children respond by increasing employment while reducing time devoted to home production, whereas those with young children primarily adjust by sacrificing leisure. By contrast, husbands exhibit limited responses to wage changes. These findings highlight that narrowing the gender wage gap, while promoting women's labor market participation, can also intensify their household burden. The results underscore the need for policies that address not only income disparities but also inequalities in unpaid work and time use within households.

with Hao Li. Reject & Resubmit, Journal of Health Economics.

We study how income support after widowhood affects health in the short and long run, using Social Security Survivors Benefits as our setting. Exploiting the age-60 eligibility threshold, we use a fuzzy regression discontinuity design and data from the Health and Retirement Study to show that benefit eligibility lowers self-reported health and increases diagnosed heart disease and diabetes. We find little evidence of changes in health behaviors or food spending. Instead, eligibility increases out-of-pocket medical spending and prescription drug use, consistent with benefits relaxing liquidity constraints on medical care and bringing previously undiagnosed conditions to clinical attention. We then exploit the 1965 reform that lowered the earliest eligibility age for reduced widow benefits from 62 to 60. Survival to age 70 increased more rapidly among birth cohorts more exposed to the reform, with no comparable pattern in Canada, while all-cause mortality declined more steeply in states with larger widow populations. Taken together, the short-run rise in diagnosed illness is consistent with greater detection through expanded access to care, and greater exposure to benefits is associated with improved survival in the long run.

with Victoria Prowse.

The labor market costs of import competition need not fall on the workers it displaces. To shed light on who bears these costs, we use the Product Complexity Index to split the China trade shock by the type of production it displaces: a complex shock to industries whose products, such as semiconductors, require many capabilities, and a non-complex shock to industries whose products, such as apparel, require few. Using 722 U.S. commuting zones and shift-share instruments built from Chinese exports to other high-income countries, we find that both shocks reduce manufacturing employment, with similar estimated effects. The complex shock causes a persistent decline in total employment through 2019, while the non-complex shock has no statistically significant effect: non-manufacturing gains offset its manufacturing losses. Less precise education-stratified estimates suggest that high school dropouts have the largest long-run total employment losses under both shocks and that, under the complex shock, college graduates lose manufacturing but gain non-manufacturing employment, while dropouts' losses arise mainly in non-manufacturing. If the costs of the complex shock fall partly on workers outside manufacturing, as crowding out would imply, programs conditioning eligibility on job loss at trade-affected firms, such as Trade Adjustment Assistance, may miss them.

with Kevin Mumford.

Reducing application frictions reshapes student-college sorting, with effects that differ sharply across the quality distribution. Exploiting staggered adoption of the Common Application across more than 2,000 four-year institutions from 2002–2019, I find that reducing application costs shifts students toward highly selective institutions while disadvantaging medium-selective schools. Most-selective institutions experience substantial enrollment and entering-class quality gains, whereas medium-selective schools experience persistent enrollment losses and widening quality gaps. Least-selective schools initially attract higher-ability students as safeties. A Bartik shift-share design based on predetermined origin-state enrollment shares confirms the enrollment effects. College Scorecard earnings data are consistent with these sorting patterns. The results suggest that reducing application frictions generates a sorting cascade that concentrates talent at elite institutions and contributes to the decline in college enrollment at less-selective schools.

with Anthony Yim.

with Kevin Mumford and Ben Zou.