Click on Paper Titles to View Abstracts
STEM Employment Resiliency During Recessions: Evidence from the COVID-19 Pandemic with James C. Davis, Gerald R. Marschke, and Andrew J. Wang; (Latest Version); R&R @ Research Policy; Based on NBER Working Paper No. 29568
STEM occupational employment suffered smaller peak-to-trough percentage declines than non-STEM employment during both the Great Recession and COVID-19 recession, suggesting a relative resiliency of STEM employment during recessions in the digital age. We exploit the sudden peak-to-trough declines in STEM and non-STEM employment during the COVID-19 recession to measure STEM recession-resiliency during the pandemic, decomposing our difference-in-differences estimate into parts explained by various sources including differences in demographics, educational attainment, job tasks, remote work capability, industry, and STEM knowledge importance on the job. We find that STEM knowledge importance on the job explains the greatest share of STEM employment resiliency, and that workers in non-STEM occupations who nonetheless use STEM knowledge experienced higher employment rates during the pandemic. We show that R&D expenditures and employment also remained resilient, suggesting only a mild effect of the COVID-19 pandemic on innovative activity. Altogether, our findings suggest that increasing opportunities for STEM training—including outside the college-track—may help improve the employment resiliency of workers during future recessions.
Green Card Quotas and the Misallocation of Talent: Evidence from the STEM Doctoral Labor Market; (Latest Version); Submitted
The rates of startup formation, business dynamism, and productivity growth in the US economy have declined since the early 2000s, and a growing literature seeks to identify common forces driving these trends. In this paper, I show that binding green card quotas may contribute to these declines by diverting the most highly-skilled workers in the economy away from entrepreneurial ventures resulting in a misallocation of talent. Using a simple job choice framework, I show that the sudden emergence of country-specific green card delays in October 2005 incentivized Chinese and Indian STEM doctorates to seek employment at established firms over startups as the latter are more likely to shut down prior to the resolution of delays. A difference-in-differences analysis reveals that STEM doctorates who faced green card delays reduced their likelihoods of working in US startups over established firms in the first decade of their careers by 42%. This suggests that policies enabling foreign-born STEM doctorates to avoid green card delays or maintain green card eligibility in the face of job destruction are likely to increase the share of such doctorates working at startups early in their careers.
A Machine Learning Approach to Identifying Postdocs in LEHD Data with James C. Davis, Gerald R. Marschke, and Andrew J. Wang (Latest Version)
This paper details the creation of the ACS-LEHD Doctorate Panel—a new linked employer-employee longitudinal dataset of the doctoral workforce enabling researchers to analyze the quarterly labor market outcomes of STEM doctorates and postdocs within the secure environment of a Federal Statistical Research Data Center (FSRDC). To impute the quarterly postdoc employment status of doctorates in matched ACS-LEHD data, we train a machine learning algorithm on the small share of data for which quarterly postdoc employment status is known, yielding an out-of-sample imputation accuracy of over 97%. We include a preliminary analysis of the earnings disparity between postdoc-trained and nonpostdoc-trained biomedical doctorates in the ACS-LEHD Doctorate Panel, finding that postdoc-trained biomedical doctorates tend to earn less than their nonpostdoc-trained counterparts, and that this difference in pay narrows, but does not disappear, when including firm and occupation fixed effects.
