Census Bureau finds AI-exposed grads facing recession-like job losses
A working paper from three US Census Bureau economists finds that college graduates from the majors most exposed to artificial intelligence have faced job and earnings setbacks comparable in scale to those of graduates who entered the workforce during the Great Recession, and the researchers date the shift to the arrival of ChatGPT in late 2022.
The paper looks at the top decile, the ten percent of college majors most exposed to AI, which includes computer science, mathematics, statistics, several engineering disciplines, networking and even journalism. For graduates in these fields, the likelihood of finding a job right after graduation fell by five percentage points, and full-quarter initial earnings dropped 13 percent. "The most AI-exposed decile of college majors saw their likelihood of initial employment decline by five percentage points, while full-quarter initial earnings declined by 13 percent," the economists wrote in the paper. "This earnings decline is comparable in magnitude to the earnings losses associated with graduating into a large recession."
About half of that earnings decline traces to graduates shifting into lower-paying industries such as restaurants and retail as they struggle to land jobs in their own field. On the employment side, changes in self-employment and graduate school enrollment together could account for as much as half of the five-percentage-point drop in initial employment.
The pattern is uneven: graduates from less AI-exposed majors, including nursing and many education fields, have not seen the same deterioration in early-career prospects. The economists also weighed a straightforward oversupply of graduates as an alternative explanation and found it does not fully account for the pattern. Growth in AI-exposed majors was rapid before ChatGPT's release and began leveling off quickly afterward, survey data suggest enrollment in some of the most exposed programs has started to decline, and the number of graduates from those majors who land jobs in their own field has fallen faster than graduate numbers overall.
The authors, speaking to The Register, stressed that AI's effect on hiring may not run only through companies that actually use the technology themselves. "Firms may have chosen to hire fewer new graduates for reasons that are related to AI's existence, but aren't related to firms' use of the technology themselves," they said, pointing to employers who may be less certain about the future or worried that AI is making it harder to screen candidates. The researchers cautioned against extrapolating the short-term findings into a long-term forecast for the wider labor market: "We should be very cautious in speculating about how AI will impact the economy as a whole based on shorter-term effects on only a subset of workers."
Key facts
- Census Bureau economists find that likelihood of initial employment for the top decile of AI-exposed college majors fell five percentage points since ChatGPT's late 2022 release.
- Full-quarter initial earnings for those graduates dropped 13 percent, a decline the paper calls comparable in magnitude to graduating into a large recession.
- About half of the earnings decline comes from graduates shifting into lower-paying sectors like restaurants and retail.
- Changes in self-employment and grad school enrollment combined could explain as much as half of the five-percentage-point employment decline.
- Less AI-exposed majors, including nursing and many education fields, have not shown the same deterioration in early-career job prospects.
Why it matters
This is one of the more concrete pieces of evidence tying generative AI to a measurable, recession-scale hit on a specific slice of the youth labor market rather than a vague warning about future job loss. It comes from government economists using Census employment and earnings data, not a survey or a vendor-commissioned report, and it puts numbers on a shift that has mostly circulated as anecdote: computer science and similarly AI-exposed graduates struggling to find first jobs at anything like the rate they used to.
Who it affects
New graduates from the top decile of AI-exposed majors: computer science, mathematics, statistics, several engineering disciplines, networking, and even journalism. Graduates from less exposed fields such as nursing and most education programs are the explicit contrast group in the paper and have not seen the same drop in early-career employment or earnings.
How to use it
For graduates and career advisors in the exposed majors, the paper is a data point to weigh against the still-strong demand in less AI-exposed fields, and a reason to look past headline hiring numbers to the field-specific outcomes. For employers and policymakers, the authors' own caveat matters: firms may be cutting junior hiring for reasons connected to AI's existence without themselves using the technology, so the effect cannot be read off a company's own AI adoption alone.
How solid is it
The paper comes from three Census Bureau economists (unnamed in the article) and draws on the agency's employment and earnings data rather than a survey. The researchers explicitly tested and set aside a simple oversupply explanation, showing that growth in AI-exposed majors slowed after ChatGPT's release rather than continuing to swell the applicant pool. It is a working paper, not described in the article as peer-reviewed, and the article gives no title, publication venue or release date for it.
Risks and caveats
The effects are measured shortly after graduation and, per the authors, tend to ease the further graduates get from their graduation date, though the paper notes that Great Recession-era graduates suffered lasting earnings effects from a similarly rocky start. The researchers themselves caution against extrapolating these short-term, subset-of-workers findings into a broad forecast for the whole economy, and the article does not break out figures for computer science majors specifically, only for the AI-exposed top decile as a group.
“The most AI-exposed decile of college majors saw their likelihood of initial employment decline by five percentage points, while full-quarter initial earnings declined by 13 percent. This earnings decline is comparable in magnitude to the earnings losses associated with graduating into a large recession.”
— the paper's authors, three US Census Bureau economists