Job Market Paper

Who Earns? Universal Pre-K and the Division of Labor Within Households
Abstract
Using a stacked triple-difference design across five U.S. jurisdictions, I examine how universal pre-kindergarten affects the division of market work and earnings within married couples. UPK increases wives’ share of couple earnings by 3.4 percentage points, alongside increases in wives’ labor-force participation and their share of couple market work. These findings show that the effects of childcare policy extend beyond maternal employment to the division of work and earnings within couples.

Working Papers

Career Choice Under Relaxed Constraints: STEM OPT and the Career Paths of Foreign-Born Doctoral Graduates
Abstract
Using restricted-use Survey of Earned Doctorates data and a difference-in-differences design, I study how the 2008 STEM OPT extension affected employment commitments and initial sectoral placement. The extension increased the probability of a definite employment commitment at graduation by about 4–5 percentage points, with no detectable increase in industry employment. The evidence points instead to increased employment in academia.

Work in Progress

Risk Aversion and the Timing of First Births

Using the German Socio-Economic Panel and the 2007 Elterngeld reform, I study whether risk preferences shape first-birth timing and whether responses to parental benefits differ by women’s pre-reform risk aversion.

Leveling Up or Down? Pay Transparency and the Gender Wage Gap

I study whether mandatory salary-range disclosure in job postings narrows the gender wage gap through higher women’s wages, lower men’s wages, or both.

Asymmetric Observability in Common Agency: Theory and Evidence from Artificial Intelligence (with Afam Madu)

We develop a common-agency model to study how asymmetric observability shapes commercially motivated steering in generative AI. The model examines how user monitoring affects the influence of developer-side commercial incentives.

Other Projects

Do Models of Attitudes Toward Intimate Partner Violence Travel Across Countries and Time?

Using Demographic and Health Surveys data, I test whether machine learning models of IPV justification travel across countries and time. I find that models perform better when trained on local data than when transferred across countries or survey years, with little improvement from pooling countries.

Project and code available on request.