AI, work from home, and the young-graduate job market

Open original

The article argues that the drop in employment for young college graduates since 2023 is better explained by work from home than by AI, using CPS occupation data and Stripe spending proxies. The authors frame it as an accounting exercise, not a causal finding, and note that the research on both questions is still mixed.

  • Young grad (22–26) employment-population ratio fell to 88% in September 2026, a point below three years earlier, while mid-career workers (30–49) have fully recovered to pre-pandemic levels.
  • For every 10 percentage points of WFH adoption in an occupation, young grad employment growth lagged middle-career workers by 3.9 to 6.6 points; the authors estimate WFH explains 56% to 88% of the young-grad FTE decline.
  • AI exposure shows a negative relationship that is much weaker and statistically indistinguishable from zero once WFH is included, and Stripe data shows young workers lost ground in regions where WFH-proxy spending grew faster.

Young US college grads (ages 22–26 with a bachelor's) are doing worse in the job market. Their employment-population ratio hit 88% in September 2026, down a full point from three years earlier and nearly 1.5 points below pre-pandemic. Unemployment for this group was near 6% at the end of 2025, versus about 4% before COVID, and the typical unemployed grad has been searching 12 weeks, a third longer than a few years ago. The author, writing for Stripe Economics, takes seriously the idea that AI is squeezing entry-level hiring, but argues that blaming AI alone doesn't fit the data. Young non-graduates have also lost ground since 2023, and they're less exposed to AI. Mid-career workers aged 30–49 have fully recovered to pre-pandemic levels. Other shocks landed around the same time as ChatGPT in late 2022, including high inflation, interest rates, a collapse in job switching after the "Great Resignation," and big swings in immigration, so isolating any single cause is hard. The piece's central claim is that work from home (WFH) may be a bigger factor than AI for young grads. Using Current Population Survey data, the author compares the change in employment between recent grads and middle-career workers from 2023 to the last twelve months, across occupations and Census divisions. For every 10 percentage points of WFH adoption in an occupation, young grads' employment fell 3.9 to 6.6 points further behind middle-career workers. As a rough accounting exercise, WFH could explain between 0.69 and 1.08 points of the 1.23-point decline in young grads' full-time-equivalent employment ratio, or 56% to 88% of it. The author is careful to say this is not a causal estimate. The mechanism they point to is that remote work weakens the in-person mentoring, feedback, and networking young workers need, so employers facing remote-friendly roles lean toward experienced hires. The research around this is mixed, and the author lays out the disagreement fairly. Some studies find that AI exposure has hurt young workers' hiring and earnings, including Brynjolfsson, Chandar, and Chen (2026) and Orr, Tucker, and Warren (2026), who found about a five-point drop in initial employment for grads in the most AI-exposed majors. Others, like Fairlie and Wu (2026) and Kauhanen and Rouvinen (2026), find little sign of AI displacement. Emanuel, Harrington, and Pallais (2026) find that junior software engineers benefit from being physically near the office, and that remotable occupations have seen relative rises in young-grad unemployment. The evidence on WFH is also not one-sided, since remote work has helped disabled and older workers. For anyone deciding whether to read the full piece, it's worth it if you care about the labor-market debate, because the author's regression details and appendix are where the real case for WFH over AI gets made.

Ask about this article