Generative AI as Seniority-Biased Technological Change: Evidence from U.S. Résumé and Job Posting Data
Study finds generative AI is seniority biased, shrinking entry roles while senior jobs hold or grow. Using 245 million postings and records at 285,000 firms, a classifier tags adopters using AI integrator vacancies. After 2023 Q1, junior headcount drops 7.7% and 12% relative to seniors. Hit is slower entry, not layoffs: 3.7 fewer junior hires per quarter. Wholesale and retail cut junior hiring 40%. Midtier grads fare worst. Limits include nonrandom adoption and short horizon.
What it examines
The paper tests if generative AI is “seniority‑biased,” hurting juniors more than seniors. Using LinkedIn résumé--posting data for 62 million workers in 285,000 U.S. firms (2015--2025), it flags firm adoption via LLM‑classified “AI integrator” postings, then applies difference‑in‑differences and triple‑differences to track junior vs. senior employment within firms.
What it concludes
Since early 2023, AI‑adopting firms cut junior headcount mainly by slowing hiring; senior employment held or grew. Effects are largest in wholesale/retail; mid‑tier graduates are most affected; promotions of incumbents rise. Uses: workforce planning, hiring and training, curriculum design, and policy targeting. Limits: nonrandom adoption, short horizon, adoption mismeasurement.
Evidence objects
Generative AI emerges as senioritybiased change: using 245 million postings and rsums for 62 million workers across 285,000 firms, an LLM flags AI integrator adoption; impacts inflect sharply in 2023Q1.
key_findings bullet 1 · key_findings · validation V0
Differenceindifferences finds adopters junior headcount down 7.7% after six quarters and 12% relative to seniors; seniors rise. Mechanism is slower entry: 3.7 fewer junior hires quarterly, fewer separations, more promotions.
key_findings bullet 2 · key_findings · validation V0
Wholesale/retail sees the steepest pullbackabout 40% fewer junior hiresand education effects are Ushaped, with midtier graduates hardest hit. Contributions: behaviorbased adoption, withinfirm identification. Limits: selection, undercounted adopters, short 2023--2025 window.
key_findings bullet 3 · key_findings · validation V0
Extending $SBTC$ to within-firm seniority, this paper uniquely examines generative AIs seniority-biased effects using LinkedIn/Revelio and LLM-detected AI integrator postings. Employing $DiD$ and $DDD$, it decomposes hires, separations, promotions, showing post-$2023$ adopter declines in junior employment via hiring slowdowns, varying by sector and education tier. Policy-relevant, investor-informative, yet measurement-challenged, preliminary.
key_findings bullet 4 · key_findings · validation V0
Raw abstract and provenance
We study whether generative artificial intelligence (AI) constitutes a form trade. Heterogeneity by education reveals a U-shaped pattern: mid-tier
Source row: 974 · abstract type: snippet