When the college class of 2026 picked up their diplomas, many entered the job hunt under a cloud of anxiety. Over the prior year, corporate executives and prominent technology leaders warned that early-career professionals would face unprecedented hiring freezes as generative artificial intelligence assumed responsibility for entry-level tasks such as writing code, compiling reports, and conducting research. According to an NBER working paper by economists Robert W. Fairlie and Jane Wu, however, the predicted wave of early-career joblessness did not appear in the national data through the summer of 2026.
The researchers evaluated federal employment survey data covering the first full summer cohort to graduate after corporate spending on artificial intelligence accelerated. They found that while unemployment among new degree holders was elevated during the summer, the increase aligned with ordinary seasonal patterns rather than a sudden technological disruption.
Competing Forecasts for Early-Career Hiring
Leading up to mid-2026, economists and corporate executives voiced sharply conflicting expectations about how automated tools would influence hiring for new degree holders. Some high-profile investors and chief executives warned of an impending crisis, arguing that software tools could rapidly eliminate entry-level white-collar positions. In contrast, other industry figures dismissed those concerns as exaggerated, suggesting that expanded computing power would stimulate labor demand and make junior staff more productive.
To examine whether early-career opportunities were shrinking, Fairlie and Wu, both researchers at the University of California, Los Angeles, focused on individuals ages 22 to 25 with a bachelor’s degree who were no longer enrolled in school. Junior staff typically perform standardized analytical and administrative work that current software systems can handle reasonably well. Because employers facing economic or technical adjustments often pause new hiring before dismissing existing staff, the authors reasoned that new graduates seeking their first career positions would serve as an early indicator of employer demand.
Indicators of corporate adoption made 2026 an informative test period. Enterprise software usage expanded rapidly between mid-2025 and 2026, while median business spending on automated tools per employee more than doubled on major financial platforms. Furthermore, federal surveys showed that the share of businesses automating significant numbers of worker tasks grew substantially over the preceding two years. If technology was actively displacing junior white-collar workers, summer 2026 was the first labor market entry window where those consequences should have been visible.
Accounting for Seasonality and Sidelined Workers
The researchers drew microdata from the monthly Current Population Survey, an extensive survey conducted jointly by the U.S. Census Bureau and the Bureau of Labor Statistics that samples more than 130,000 people each month. Over the period spanning January 2022 to August 2026, their analysis evaluated records for more than 45,000 young college graduates.
A primary hurdle in analyzing early-career labor data is the seasonal nature of graduation. Long-term trends show that unemployment among young degree holders routinely rises by nearly two percentage points every June, July, and August. As tens of thousands of students finish classes at the same time, the sudden influx of job seekers naturally pushes up unemployment rates before easing in the fall. Evaluating changes without adjusting for this seasonal rhythm can lead observers to mistake ordinary post-graduation job searching for a wider economic crisis.
The authors also addressed a limitation in standard government employment statistics. To be officially classified as unemployed, a jobless person must have actively looked for work within the preceding four weeks. Many recent graduates are relocating, taking short personal breaks, or reviewing job listings without submitting applications, meaning they are officially classified as out of the labor force. To capture unmet employment desire more completely, Fairlie and Wu established an expanded measure that includes what they termed “sidelined unemployment,” adding individuals who report that they want a job even if they have not yet met the formal criteria for an active search.
Tracking the 2026 Employment Numbers
The study found that the official unemployment rate for recent college graduates averaged 7.3 percent across June, July, and August of 2026. This figure was slightly above the summer averages in 2022 (7.1 percent), 2023 (6.3 percent), and 2025 (7.2 percent), but remained lower than the 7.8 percent recorded in the summer of 2024. Across the entire post-pandemic timeline, the summer 2026 rates remained well within the normal historical range.
When the authors incorporated the sidelined graduates who reported wanting work, the expanded unemployment rate averaged 10.4 percent during summer 2026. While higher than the 9.4 percent recorded in summer 2025, it stood only 0.3 percentage points above the 10.1 percent recorded in summer 2024. Statistical models that controlled for seasonal variations, demographic traits, and longer-term baseline trends revealed that neither the standard unemployment rate nor the expanded metric showed a statistically significant increase in summer 2026.
The authors also compared recent graduates to two separate benchmark groups: college graduates ages 30 to 49, and young adults ages 22 to 25 without a college degree. When compared with older, experienced graduates, the young cohort experienced no statistically significant decline in relative employment outcomes. When compared with young workers without degrees, college graduates exhibited slightly higher relative unemployment, though the difference was not statistically significant, and one factor behind it was lower joblessness among non-college youth rather than an unusual spike in graduate unemployment.
Remote Work and Occupational Exposure
To examine whether specific sectors were quietly shedding junior positions, the researchers tested whether unemployment concentrated in occupations with high exposure to automated software. Using both theoretical task evaluations and observed AI usage data based on Claude usage patterns, they tracked graduate outcomes across different job categories. In both cases, the statistical relationships between software exposure and 2026 graduate unemployment were positive but small and not statistically distinguishable from zero.
In contrast, the researchers identified a positive, statistically significant relationship between graduate unemployment and an occupation’s ability to be performed remotely. Recent graduates seeking entry into teleworkable roles experienced higher unemployment in 2026 compared to those in roles requiring on-site presence. Prior research suggests that remote working environments can weaken informal workplace mentorship and hands-on guidance, leaving companies more hesitant to onboard inexperienced junior workers.
Because occupations that rely heavily on digital automation are often the same jobs that permit remote work, cleanly separating the influence of technology from workplace location remains difficult. The authors noted that a considerable fraction of unemployed graduates do not report an occupation in survey records due to a lack of prior work experience, which complicates occupational tracking.
The researchers emphasized that their findings provide a descriptive picture rather than causal proof, finding no evidence of a widespread AI-driven displacement shock for new college graduates through mid-2026. At the same time, they cautioned that adoption is an evolving process. If corporate integration of automated tools deepens, future graduating classes may face different conditions, making ongoing evaluation of early-career entry patterns an ongoing priority for labor economists.
