The Early Impacts of AI on Employment among Recent College Graduates (w35796)
A · Strong evidence An NBER working paper using nationally representative CPS microdata with difference-in-differences and event-study designs, and the first look at the 2026 graduate cohort, though descriptive and limited by small samples.
Summary based on the full text·Original: about 70 minutes to read
In short
An NBER working paper (No. 35796, September 2026) by Robert W. Fairlie and Jane Wu (UCLA). It asks whether the class of 2026, the first cohort entering a labor market with widespread workplace AI use, saw higher unemployment. Using Current Population Survey microdata through August 2026, it finds no spike in summer 2026 relative to earlier summers and no relative rise against older graduates or young non-graduates.
Key numbers
- 7.3%summer 2026 unemployment of college graduates aged 22-25, within the 6.3-7.8% range of 2022-2025
- 10.4%expanded rate including graduates who want a job but are not searching, highest of five summers but not significantly
- 4.5 ppgap to older graduates (30-49), unchanged from summer 2025
As reported by the source.
English·Web page·opens nber.org
What it reports
- Summer (June-August) unemployment for graduates aged 22-25: 7.1% (2022), 6.3% (2023), 7.8% (2024), 7.2% (2025), 7.3% (2026).
- Monthly 2026: 5.4% in May, 7.8% in June, 7.4% in July, 6.8% in August; summer rates run nearly 2 points above the roughly 5% base in other months.
- Expanded "sidelined" rate adding those who want a job: 10.4% in summer 2026 vs 9.4%, 9.3%, 10.1% and 9.4% in 2022-2025, beating the previous high by 0.3 points.
- Against older graduates (30-49): 2.8% vs 7.3%, gap stable at 4.5 points; on the expanded measure the gap widened from 5.3 to 6.2 points.
- Against young non-graduates: 6.8% vs 7.3% in 2026; the expanded-measure advantage of graduates narrowed from 2.2 to 0.3 points.
- Interactions with occupational AI exposure are positive but not significant; teleworkability shows a positive, marginally significant link.
- Fewer than 17% of unemployed recent graduates lost jobs through layoffs.
How they got it
CPS basic monthly files (more than 130,000 people per month), restricted to 22-25-year-olds with a bachelor's degree as highest degree and not enrolled. Comparisons of summer 2026 against 2022-2025, plus difference-in-differences and event-study models with two comparison groups. Occupation-level interactions use theoretical and observed AI-exposure measures and a teleworkability index.
Limits
- Small CPS samples cannot rule out increases of 2-3 percentage points.
- Many unemployed graduates report no occupation (15.5% of the unemployed, 35.8% for the expanded measure), so the most affected group drops out of occupation analysis.
- A sharp drop confined to a few exposed majors would barely move the average; AI exposure and remote work are correlated.
Why it matters for companies
Headline claims that AI is already shutting graduates out are not visible in national unemployment data for 2026. That does not settle hiring policy: other studies using payroll and postings data find reduced junior hiring in exposed roles, and the expanded measure is drifting up. For employers the safer reading is that any effect so far runs through slower hiring, not layoffs. Companies cutting graduate intake on the assumption that AI replaces juniors are acting ahead of the evidence.
Summary based on the full text. Read the original for the full argument and data.Read the original on nber.org