Has AI impacted the labor market yet?
A · Strong evidence A structured review of more than 20 recent labor-market studies by Chicago Booth economist Alex Imas and Jacob Schaal, reporting each study's data, effect size and identification problems.
Summary based on the full text·Original: about 25 minutes to read
In short
A September 2026 Substack review by Alex Imas (University of Chicago Booth) and Jacob Schaal. It asks whether AI has visibly changed the labor market yet. Their answer is narrow: there is contested evidence that AI is reducing hiring for junior white-collar roles most exposed to it, but no sign yet of aggregate displacement in unemployment or layoffs. Remote work is the main rival explanation for the junior decline.
Key numbers
- 19%employment of 22-25-year-olds in high-exposure US roles below trend by June 2026 (Canaries dashboard, descriptive)
- 8-10%junior employment decline at GenAI-adopting firms vs non-adopters after 2 years (Hosseini and Lichtinger)
- 1%laid-off US workers who attribute their dismissal to AI (Gallup)
- 88% / 21%share of US employment vs share of tasks covered by AI use in Google's ATLAS data
As reported by the source.
English·Web page·opens aleximas.substack.com
What it reports
- Brynjolfsson et al. (ADP payroll): 13% fall in junior employment in high-exposure roles by July 2025; dashboard shows 19% below trend by June 2026, roughly halving with education controls.
- Tucker (Census): hires of 22-24-year-olds in the most exposed fifth of industries fell about 9% after ChatGPT; 12.4% relative employment fall by mid-2025.
- Orr, Tucker and Warren: 6.7 million graduates; most exposed tenth of majors 5 pp less likely employed, initial earnings about 13% lower.
- Nordic data: Finland and Norway find no significant effect; Norwegian exposed occupations grew 0.1% vs 0.3% since October 2022.
- Lambert and Schindler: with remote work added, the AI-exposure coefficient on junior hiring drops to zero; Brynjolfsson et al. rerun it on payroll data and find the opposite.
- Humlum and Vestergaard (Denmark, 25,000 workers): null effects on earnings and hours, ruling out effects above 2%, despite 93% usage in firms with enterprise chatbots.
- Kharazian et al. (21,000 US firms): high-intensity AI spenders grew employment 10% and entry-level 12% vs later adopters.
- Fairlie and Wu: graduate unemployment 7.3% in summer 2026, within the 2022-25 range.
- Yotzov et al. (nearly 6,000 executives): 69% use AI; about 90% report no significant effect on employment or productivity so far.
How they got it
A narrative literature review of working papers, administrative data studies, surveys and usage data available through October 2026, grouped by method: exposure-based designs, adoption-based designs, Nordic population data and remote-work tests. The authors summarise each study's sample and identification threats.
Limits
- Not a systematic meta-analysis; study selection is the authors'.
- Imas contributed to Google's ATLAS, which the review cites.
- Most cited studies are unrefereed working papers.
Why it matters for companies
The evidence says any AI effect so far shows up in fewer junior hires, not layoffs; in a UK survey subsample, firms expecting cuts put about two-thirds of them on reduced hiring. That makes graduate and entry-level intake the decision point. The review also flags a long-run cost: fewer juniors means fewer future seniors, with one model (Ide) estimating 0.05-0.35 pp lower long-run US GDP growth. Companies trimming junior roles should decide explicitly how they will train the next layer of experts.
Summary based on the full text. Read the original for the full argument and data.Read the original on aleximas.substack.com
- Tags
- Future of work