A new CESifo working paper finds no evidence that AI has displaced recent US college graduates from the job market, with summer 2026 unemployment for that cohort landing at 7.3% — squarely inside the 6.3% to 7.8% range seen every summer since 2022. The Munich-based researchers, Robert Fairlie and Jane Wu, examined US Census Current Population Survey microdata for Bachelor's-degree holders aged 22 to 25 who are not pursuing further study.
The paper, titled 'The Early Impacts of AI on Employment Among Recent College Graduates,' focused on new entrants because 'changes in labor demand may first appear through reductions in hiring' rather than through layoffs of tenured staff. Fairlie and Wu built statistical tests comparing recent graduates to non-college graduates in the same age band and to older college graduates aged 30 to 49, and layered in job-role AI exposure scores from a 2023 study.
“there is no evidence of any significant, widespread displacement or reduction in hiring of recent college graduates in absolute or relative levels.”— Robert Fairlie and Jane Wu, CESifo economics researchers
In almost every comparison across the 2022 to 2026 window, the researchers report, the differences between groups were not statistically significant. The 2026 number stays unremarkable even after expanding the definition to include graduates who told the CPS they 'want a job' but were not actively searching — a group excluded from the headline unemployment rate.
Key facts
- 01Summer 2026 unemployment for recent US college graduates aged 22–25 was 7.3%, within the 6.3% (2022) to 7.8% (2024) range.
- 02CESifo researchers Robert Fairlie and Jane Wu found no statistically significant AI-driven hiring drop, using US Census Current Population Survey microdata from 2022 to 2026.
- 03The finding contradicts a recent Stanford study using ADP payroll data that showed entry-level 'AI-impacted' occupations lagging.
- 04The paper flags rising ChatGPT Enterprise token use and per-employee AI spending over the past 12 months as reasons the 2027 class could see different results.
The result cuts directly against a Stanford University study released last month that found entry-level employment in AI-impacted occupations lagging other fields. That study used payroll data from HR firm ADP; the CESifo paper argues the ADP dataset measures the supply of filled roles, while the CPS unemployment rate also reflects aggregate demand for workers — a distinction that can move in opposite directions if the pool of job seekers shrinks alongside the pool of jobs.
The class of 2026 was flagged in advance as the one AI might finally hit. The CESifo authors themselves note a sharp Census-recorded uptick in firms 'replacing a large number of employee tasks with AI,' along with broad increases in AI spending per employee and ChatGPT Enterprise token use over the last 12 months.
“AI literally until December [2025] was not actually good enough to do any of the jobs that they're actually cutting.”— Marc Andreessen, venture capitalist
BlackRock CEO Larry Fink said in March that 'the speed at which AI is changing' had him worried 'we could see the highest unemployment rate among them in years — even without a recession.' Venture capitalist Marc Andreessen made a related claim earlier this year, arguing that only in the last several months has the technology been capable enough to actually take over the entry-level tasks employers are cutting.
The numbers so far do not bear that thesis out. Fairlie and Wu write that the data 'tell a consistent story in which unemployment among recent college graduates in summer 2026 was not unusually high relative to earlier summers,' across every comparison group they tested.
The paper is careful about what it does and does not show. The authors describe summer 2026 as a 'useful first test' of AI's impact on the labor market for new entrants, not a verdict. They explicitly warn that deeper workplace AI adoption could show up in later cohorts.
“if the intensity of AI use in the workplace continues to increase, the graduating classes of 2027 and later might be more affected than the class of 2026, and additional years of data will be needed to hone in on whether effects emerge as workplace use of AI deepens.”— Robert Fairlie and Jane Wu, CESifo economics researchers
For AI companies selling into the enterprise, the CESifo finding complicates a narrative that has quietly become part of the sales pitch: that models are already good enough to absorb the work of a junior analyst or associate. If graduating classes are still being hired at 2022-through-2024 rates while ChatGPT Enterprise token consumption climbs, one of two things is true. Either the productivity gains are being reinvested into more output rather than fewer workers, or the actual task-replacement rate for entry-level knowledge work is running well behind the marketing. Either way, the class of 2027, and the payroll data that follows it, is the next real test.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
