Can exposed graduates’ worse first jobs be attributed to employers adopting AI?

Research brief, 17 October 2026. Question: do the observed initial employment and earnings gaps for graduates of AI-exposed majors track actual employer adoption, or contemporaneous remote work and industry demand? A good answer observes graduate major, applications and later earnings along with employer rollout timing and remote-work policy, plus comparable pretrends and market-wide spillovers. Our career-path synthesis has linked Census graduate outcomes, a national unemployment null, and a competing remote-work explanation, but no employer-to-applicant linkage.

1. What does the Census graduate paper’s original design measure, and which outcomes are conditional on in-state/full-quarter employment?
2. Does the new Census study of exposure and adoption among US firms independently measure take-up, and can it link to graduate cohorts?
3. What does the remote-work study’s original design rule out, and what do the authors say it cannot rule out?
4. Do independently checked national graduate outcomes overturn a concentrated entry-quality finding?

Search each subquestion once, skim originals, then search only identification gaps. Recheck load-bearing claims against the Census and authors’ originals; record disagreements and what cannot be inferred. A weekend Census paper post must be readable from its summary alone; the already-unread UK defender case may instead argue for leaving the feed empty.