Population check: no broad recent-graduate unemployment spike through summer 2026

Robert W. Fairlie and Jane Wu, “The Early Impacts of AI on Employment among Recent College Graduates,” NBER Working Paper 35796, September 2026; IZA-hosted full original and data description. Career-path question. They analyze nationally representative Current Population Survey monthly microdata, comparing recent college graduates with older college graduates and younger non-college workers; inspect June–August 2026 against prior summers, include people outside official unemployment who want work, and check occupational AI exposure and remote-work feasibility.

There is no statistically detectable relative summer-2026 unemployment spike for recent college graduates under their comparisons, including the expanded “wants work” measure; the expanded measure itself is appreciably higher than official unemployment. Exposure-interaction estimates do not give a clean broad AI unemployment effect; some remote-work association appears. Their inference is a check on the popular national mass-unemployment claim, not evidence that the most exposed graduates are unharmed. Small recent-graduate monthly subsamples mean uncertainty and limited occupational detail; CPS does not observe employer AI take-up, job applications rejected, graduate field in comparable linked form, within-occupation job quality, or long-term career trajectories. In some classifications a substantial fraction of unemployed graduates lacks an occupation, frustrating exposure tests. National null can coexist with five-point initial-job gaps and lower earnings among the most-exposed majors in linked Census study.