Research

World War II spurred an unprecedented exodus of teachers from American schools, primarily men leaving for military service. Using a difference-in-differences framework that exploits geographic variation in the pre-war gender composition of the teacher workforce and students' ages, I estimate the effects of childhood exposure to this teacher supply shock on educational attainment and labor market outcomes. Increased exposure reduces high school and college completion, as well as adult income. States more affected by this shock responded by hiring more teachers on emergency licenses, resulting in a decrease in teachers' educational attainment. These changes in the composition of the teacher workforce persisted long after the war, suggesting that even temporary shocks to teacher supply can have lasting impacts on student outcomes through changes in teacher quality.

This paper revisits World War II's effect on female labor force participation (FLFP) using newly available complete count census data from 1940 and 1950. Compared to seminal papers in the literature—such as Acemoglu, Autor, and Lyle (2004) and Goldin and Olivetti (2013)—which rely on 1% census samples and use the state-level military mobilization rate of men as a source of variation in women's labor supply, I find substantially smaller effects of the war on FLFP in the complete count data. The mobilization rate is therefore a weak instrument for growth in FLFP, calling into question estimates that use mobilization to instrument for FLFP's effects on wages. To understand why earlier studies using 1% samples reached different conclusions, I repeat these analyses using 10,000 random 1% subsamples of the complete count data, comparable to those used in earlier work. In all cases, the distribution of estimates varies widely, suggesting that different census samples can generate very different conclusions when estimating the effects of state-level treatment. These findings challenge previous conclusions about WWII's impact on women's labor force participation and highlight broader methodological concerns for research using small census samples.

Between 1940 and 1960, the share of Americans with health insurance increased from less than one in ten to more than two-thirds. Large employers in manufacturing were most likely to offer these benefits. I use variation in states' pre-war industry mix and firm size to instrument for the growth in health insurance premiums per capita across states. Preliminary results suggest that greater insurance growth is associated with increased physician specialization and group practice but a decrease in physicians per capita and lower entry of new physicians. I also estimate the effects of health insurance growth on hospital admissions and mortality. This project provides new evidence on how physicians responded to the first widespread adoption of health insurance in the U.S.

We study the labor market and health consequences of wartime injuries during World War II. We link a novel dataset of individual-level WWII Army hospital records to enlistment records, the 1940 and 1950 Censuses, and Social Security mortality data to estimate the effects of combat injuries and illnesses on educational attainment, labor market outcomes, and longevity. The hospital records contain detailed information on diagnoses and treatments for wounded and ill soldiers, allowing us to examine how different types of injuries and illnesses have different long-run effects. We exploit variation in combat exposure across units and over time, as well as variation in injury type among wounded soldiers. By distinguishing among specific injuries and illnesses rather than using a single binary measure of combat exposure, this project will provide new evidence on the heterogeneous costs of combat from the largest military mobilization in American history.

We evaluate the long-run effects of school infrastructure spending financed by Title III of the National Defense Education Act of 1958. The program funded new construction and renovation for science, mathematics, and foreign language facilities in K-12 schools. We exploit variation across states in the share of funds used for each of these subjects and the overall allotment for each state. Using a difference-in-differences design that compares cohorts exposed to NDEA projects during school age to slightly older, unexposed cohorts, we study how targeted infrastructure shaped college major choice, occupational sorting, and earnings.