Graduating Into Disruption: Labor Market Outcomes for GenAI-Exposed College Majors

We construct a measure of AI exposure at the level of college major by combining task-level exposure scores with the occupational composition of recent graduates observed in the American Community Survey (ACS). For each major, we first recover the distribution of occupations in which graduates are employed shortly after completion and map these occupations to task-based AI exposure indices. This yields a pre-period, shift-share style exposure measure that is predetermined with respect to subsequent technological adoption. We then examine how occupational employment patterns for each major evolve around the widespread introduction of large language models (LLMs) in late 2022, when tools such as ChatGPT became broadly accessible, using data from Post-Secondary Employment Outcomes (PSEO). Specifically, we look at employment rates and earnings at fixed points in time relative to graduating, time-to-first job, and job switching rates. Finally, we consider whether outcomes vary by type of exposure (augmentative versus automative) and dispersion of exposure (majors that feed into a narrow set of exposed occupations versus majors with less exposed occupations as viable alternatives).