Thomas Helgerman - Research

Published and Accepted Papers

How the 1963 Equal Pay Act and 1964 Civil Rights Act Shaped the Gender Gap in Pay

(with Martha Bailey and Bryan Stuart, Quarterly Journal of Economics, 139(3): 1827-1878, 2024)

In the 1960s, two landmark statutes—the Equal Pay and Civil Rights Acts—targeted the long-standing practice of employment discrimination against U.S. women. For the next 15 years, the gender gap in median earnings among full-time, full-year workers changed little, leading many scholars and advocates to conclude the legislation was ineffectual. This paper uses two different research designs to show that women’s relative wages grew rapidly in the aftermath of this legislation. The data show little short-term changes in women’s employment but some evidence that firms reduced their hiring and promotion of women in the medium to long term.

Working Papers

The Role of Federal Policy in Women's Entry into Medicine

During the 1970s, women’s representation in medical schools approximately tripled, mirroring similar increases in dental, law, and business programs. Using a newly constructed dataset on medical school admissions, this paper studies the role of anti-discrimination policy in increasing women’s access to medical training. My empirical strategy leverages variation across medical schools in the potential loss of federal funding by failing to comply with the non-discrimination mandate in Executive Order 11246. I find that the threat of sanction led to a sharp increase in the percentage of women enrolled in first-year medical school classes. The effects of this policy were likely amplified by a federal effort to increase enrollment—leveraging the differential timing and size of enrollment increases across institutions in a complementary accounting exercise, I find that these expansions account for around one-third of women’s gains from 1970 to 1980.

Women in Law and the Draft

(with Benjamin Pyle)

Between 1964 and 1973, women’s representation in full-time law school programs rose fivefold, from 3.7% to 20.1%. This paper examines whether Vietnam War draft policy contributed to this increase. In 1968, men enrolled in law school lost eligibility for 2-S student deferments, threatening law schools’ tuition revenues and incentivizing schools to admit more women to stabilize enrollment. To test this mechanism, we construct a school-by-year dataset of enrollment counts split by women/men and full-time/part-time status. Using a uniform adoption difference-in-differences design, we find that women’s representation rises by 2 percentage points in full-time programs relative to part-time programs (which were far less exposed to draft risk). This effect represents a 45% increase over women’s baseline representation of 4.5% in 1967. We further show that, after the draft, law schools more connected to undergraduate institutions transitioning from all-male to coeducational experienced larger increases in women’s enrollment. A shift-share analysis indicates that a one-standard-deviation increase in exposure to these coeducation transitions raised women’s first-year enrollment shares by 1-2 percentage points, highlighting the joint role of draft policy and expanding educational opportunity in accelerating women’s entry into law.

Works in Progress

Addressing Racial Disparities in Medical Education and Science: The Role of the Civil Rights Movement (with Francesca Truffa and Ashley Wong)

Medical Schools, Physician Maldistribution, and Mortality (with Caitlin Carroll, Tyler Radler, and Yue Xu)

Difference-in-Differences with Census Data (with Brenden Timpe)

Non-Peer-Reviewed Publications

Changes in the US Gender Gap in Wages in the 1960s

(with Martha Bailey and Bryan Stuart, AEA Papers and Proceedings, 111: 143-148, 2021)

COVID-19 and stay-at-home orders: identifying event study designs with imperfect testing

(with Jaedo Choi, Elird Haxhiu, Nishaad Rao and Taeuk Seo, COVID Economics, 76: 23 April 2021)

This paper estimates the dynamic effect of Stay-At-Home (SAH) orders on the transmission of COVID-19 in the United States. Identification in this setting is challenging due to differences between real and reported case data given the imperfect testing environment, as well as the clearly non-random adoption of treatment. We extend a Susceptible-Infected-Recovered (SIR) model from Epidemiology to account for endogenous testing at the county level, and exploit this additional structure to recover identification. With the inclusion of model-derived sufficient statistics and fixed effects, SAH orders have a large and sustained negative effect on the growth of cases under plausible assumptions about the progression of testing. Point estimates range from a 44% to 54% reduction in the growth rate of cases one month after a SAH order. We conclude with a discussion on extending the methodology to later phases of the pandemic.

Resting Papers

Continuous Treatment Difference-in-Differences with Unknown Controls: A Data-Driven Approach

(with Elird Haxhiu)

This paper studies difference-in-differences (DiD) research designs where all observations receive a continuous treatment (or dose) in response to an aggregate policy, so there is no group that is ex post unexposed. We evaluate the common practice of dichotomizing the dose space into “treated” and “control” groups using an empirical percentile, then running a standard DiD design. We interpret this design as attempting to estimate a Binned Average Treatment Effect on the Treated (BATT), where the treatment group is specified by the researcher. We develop a framework to estimate the BATT when the dose takes effect only after a cutoff value, the Minimum Effective Dose (MED), and propose a first stage model selection procedure to determine a suitable control group. To conduct valid inference after model selection, we use a bootstrap aggregated estimator of the BATT, where a confidence interval is generated by a second-order bootstrap of this procedure, following Efron (2014). To conclude, we provide simulation evidence that suggests that this procedure delivers marked improvements in mean- squared error relative to common practice.