No broad AI job surge for recent graduates, paper finds

Recent graduates did not show a broad, statistically significant rise in unemployment linked to artificial intelligence by summer 2026, according to a working paper by Robert Fairlie and Jane Wu. The paper was released as CESifo Working Paper No. 12994 and as National Bureau of Economic Research Working Paper 35796.

The authors used microdata from the Current Population Survey, the Census Bureau survey that tracks employment and unemployment in the United States. They looked at June, July and August 2026 for recent graduates, defined as bachelor degree holders ages 22 to 25 who were not enrolled in further education.

For this cohort, the unemployment rate was 7.3 percent in summer 2026. That figure sat above the 6.3 percent recorded in summer 2022 and below the 7.8 percent recorded in summer 2024, placing it within the range of recent summers, the paper and reports on it said.

How recent graduates compared with earlier summers

Summer matters in this comparison because unemployment among new graduates predictably rises as graduates enter the labour market. Comparing one summer with another helps separate that seasonal pattern from an unusual change, reports on the paper explained.

The paper puts the summer 2026 figure in context with earlier summers. Graduate unemployment was 6.3 percent in summer 2022 and 7.8 percent in summer 2024, so the 7.3 percent figure for summer 2026 sat between those two points. The Current Population Survey data behind those figures cover the same summer months each year, which makes the comparison a like for like one.

On that range, graduate unemployment in summer 2026 did not stand out as a spike. The authors found no statistically significant increase in unemployment for the cohort relative to earlier summer periods, according to a summary of the paper.

How recent graduates compared with other groups

Fairlie and Wu also compared the cohort with two other groups: older college graduates and young people without a college degree. The CESifo analysis used Current Population Survey records for each comparison, so the groups were measured on the same basis. They found no meaningful divergence in trends against either group, reports on the paper said.

The authors also tested whether jobs with higher AI exposure fared worse. They divided jobs by potential AI exposure using an earlier assessment of tasks that AI systems could perform. Across almost all comparisons, differences in trends from 2022 to 2026 were not statistically significant.

An expanded measure added people who said they wanted a job but were not actively searching. That measure added nearly two percentage points to the rate for recent graduates, to about 9.3 percent, but it also showed no statistically significant relative increase in summer 2026, the paper abstract said.

Why recent graduates are a test case for AI hiring

The authors focused on recent graduates because hiring changes may show up first in new openings. Employers can limit new openings for routine office work without laying off experienced staff, reports on the paper explained. Junior roles often involve draft writing, document summaries, information processing and basic data analysis, tasks that match the capabilities of large language models.

That logic made recent graduates a useful group to examine. It did not, by itself, prove that firms had broadly made that choice by summer 2026. The paper notes a sharp increase in firms replacing tasks with AI, along with rising AI spending and token use, in one account, and those are leading signs that may matter in later hiring rounds.

The result is best read as a snapshot of early data, as one summary put it, and the paper presents it in those terms. It does not claim that AI cannot affect entry level work in the future, and it does not erase findings from other studies that use different groups, measures or time periods.