Graduating Into Disruption: Labor Market Outcomes for AI-Exposed College Majors with Cody Orr and Lee Tucker
How are new college graduates affected by the rise of artificial intelligence, and what can this tell us about the mechanisms behind AI's overall labor market effects? We use administrative records on college graduates to observe how economic outcomes among college majors with differing levels of labor market AI exposure have evolved since large language models became available. We find that post-graduation employment, earnings, and job switching patterns among the most AI-exposed college majors began to diverge immediately following the introduction of ChatGPT in late 2022. In regression-adjusted estimates, the most AI-exposed decile of college majors saw their likelihood of initial employment decline by five percentage points, while full-quarter initial earnings declined by thirteen percent. This earnings decline is comparable in magnitude to the earnings losses associated with graduating into a large recession. Roughly half of the decline in earnings is attributable to lower earnings within the industry sectors that employ these graduates, with the remainder resulting from a shift in the industry mix into lower-wage sectors such as restaurants and retail. The effects attenuate as graduates move further from labor market entry but remain substantial for the most exposed majors.
Press: Bloomberg, Fortune, Epoch Times, The Register, LinkedIn, Inside Higher Ed, Palo Alto Online, Chronicle of Higher Ed.
Unemployment Insurance, Wage Pass-Through, and Endogenous Take-Up [Slides] with Martin Gervais and Roozbeh Hosseini
We study how endogenous take-up shapes wage responses among unemployment insurance recipients during the pandemic-era benefit expansions. Benefit Accuracy Measurement data show a sharp shift in the composition of claimants alongside modest increases in reservation wages in states with greater Pandemic Unemployment Assistance utilization: a benefit increase of roughly 170 percent is associated with a 7.6 to 12.3 percent rise in reservation wages, implying an elasticity of 0.067 to 0.108. In the Current Population Survey, take-up rose from about 27 percent before the pandemic to 41 percent in 2020; a prediction model estimated on pre-pandemic data reproduces this rise only when expected benefits are included. Wages thus responded modestly while participation responded sharply. We interpret these facts through a directed search model in which workers differ in the cost of claiming. Higher benefits raise the wages of existing recipients but also draw in high-cost marginal workers, who choose lower wages, so their entry attenuates the increase in recipient-average wages. Quantitatively, endogenous take-up reduces the recipient-average wage response by about 30 percent.
The Churn Ladder [Slides] with David Wiczer
In job-ladder models, wage dispersion arises because firms trade off higher wages against the risk of losing workers to competitors. We validate the empirical relevance of a ``churn ladder"—a systematic pattern in which workers move from low-wage, high-turnover firms to high-wage, low-turnover firms through job to job transitions. However, the canonical model cannot generate the steep churn gradient we observe in the data: poaching alone produces too little cross-firm dispersion in turnover. What it misses is that lower-wage firms also lose more workers to nonemployment, substantially amplifying churn differences across the ladder. We introduce heterogeneity in both firm and worker separation rates, generating endogenous sorting and allowing the model to match observed joint patterns of wages and churn. Worker-level heterogeneity alone accounts for half of the variation in firm-level churn. Ignoring wage-dependent separations to nonemployment biases estimates of firms' separation elasticities and overstates labor market power.
Part-time Employment and Firm-level Labor Demand Over the Business Cycle
Part-time employment for economic reasons (PTE) is countercyclical, volatile, and transitory. Workers in PTE are nearly three times more likely than the unemployed to return to full-time work in a given month, and seven times more likely than fulltime workers to become unemployed. Using household survey data, I demonstrate that cyclical fluctuations in PTE come from changes in the transition rates between fulltime and part-time employment rather than between part-time and unemployment. Moreover, these movements are primarily due to within-job changes in hours. Accordingly, I model part-time work focusing on a firm’s decision to hire, fire, or partially utilize its labor force. Firms in the model are heterogeneous in size and productivity, and are subject to search frictions. The model produces firm-level utilization of parttime employment which is consistent with observed worker flows, and varies across the size and age distributions of firms. Over the business cycle, the model matches the observed relative volatility of unemployment and PTE. Part-time labor utilization by firms increases the volatility of vacancies and unemployment in the model relative to the case with only an extensive margin.
Hiring with a Quantity-Skill Tradeoff with Fil Babalievsky and Henry Hyatt
Recalls, Occupational Mobility, and Reservation Wages in the UI Claimant Population with Martin Gervais and Roozbeh Hosseini
The Composition and Re-employment Outcomes of UI Recipients with David Wasser and Caelan Wilkie-Rogers
It's All a Matter of Degrees: Comparing Survey and Administrative Educational Attainment Data with Andrew Foote and Cody Orr
Walras–Bowley Lecture: Market Power and Wage Inequality with Shubhdeep Deb, Jan Eeckhout and Aseem Patel Econometrica, Vol 92, no. 3 (2024), 603-636.
Comments from John Van Reenen and Gianluca Violante, Replies, and Supplemental Materials
Earnings Growth, Job Flows and Churn with Satoshi Tanaka and David Wiczer - Journal of Monetary Economics Vol. 135, (2023), 86-98.
What Drives Wage Stagnation: Monopsony or Monopoly? with Shubhdeep Deb, Jan Eeckhout and Aseem Patel Journal of the European Economic Association 20, no. 6 (2022), 2181-2225.
Optimal Unemployment Insurance in a Directed Search Model with Reza Boostani and Martin Gervais - Economic Inquiry (2022), 1– 24
Predicting the Effect of Adding a Citizenship Question to the 2020 Census with J. David Brown, Suzanne M. Dorinski, Misty L. Heggeness, and Moises Yi, Demography (2019) 56: 1173
Working paper version: Understanding the Quality of Alternative Citizenship Data Sources for the 2020 Census - CES Working Paper version, 2018
Supreme Court Decision on the Citizenship Question
Working paper version: Estimating the Potential Effects of Adding a Citizenship Question to the 2020 Census - IZA Discussion Papers 12087
Hard to count: How Survey and Administrative Records Modeling can Enhance Census Nonresponse Followup with Melissa Chow, Hubert Janicki, Mark Kutzbach, and Moises Yi, Statistical Journal of the IAOS (2019)
Real-time 2020 Administrative Record Census Simulation: A New Design for the 21st Century with J. David Brown, Samuel R. Cohen, Genevieve Denoeux, Suzanne Dorinski, Misty L. Heggeness, Carl Lieberman, Linden McBride, Marta Murray-Close, Hongxun Qin, Allen E. Ross, Danielle H. Sandler, and Moises Yi - CPEX Report, 2023
Estimating the U.S. Citizen Voting-Age Population (CVAP) Using Blended Survey Data, Administrative Record Data, and Modeling Technical Report with J. David Brown, Genevieve Denoeux, Misty L. Heggeness, Carl Lieberman, Lauren Medina, Marta Murray-Close, Danielle H. Sandler, Joseph L. Schafer, Matthew Spence, and Moises Yi - CES Working Paper 23-21, 2023
Determination of the 2020 U.S. Citizen Voting Age Population (CVAP) Using Administrative Records and Statistical Methodology with John Abowd, William R. Bell, J. David Brown, Michael B. Hawes, Misty L. Heggeness, Andrew D. Keller, Vincent T. Mule Jr., Joseph L. Schafer, Matthew Spence, and Moises Yi - CES Working Paper 20-33, 2020
Resting Papers:
Matching Labor Flows in Search Models with Labor Force Participation
Labor Market Cycles and Business Cycles with Peter M. Summers