Recent college graduates in computer science and other fields highly exposed to artificial intelligence are seeing their starting salaries drop by 13 percent, according to a new working paper from three US Census Bureau economists published this week. The research examines how graduates from the majors most affected by AI have fared since ChatGPT launched in late 2022 and triggered the generative AI boom. The earnings decline rivals the losses experienced by older Millennials who graduated during the Great Recession, the economists found.
Graduates in the top decile of majors with the highest AI exposure saw their likelihood of finding initial employment fall by five percentage points, the paper reports. Full-quarter starting earnings dropped 13 percent for this group. The most affected fields include computer science, mathematics, statistics, certain engineering disciplines, networking, and journalism. About half of the earnings decline stems from graduates shifting into lower-wage industries like restaurants and retail as job prospects in their fields deteriorate. Meanwhile, graduates from less AI-exposed majors—including nursing and many education fields—haven't seen the same decline in early-career employment.
"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 write. The paper notes this earnings decline "is comparable in magnitude to the earnings losses associated with graduating into a large recession." The researchers told The Register that while the impacts tend to ease as young graduates move further from graduation, previous research on Great Recession graduates suggests years of poor job prospects can leave workers behind the curve on earnings for years to come.
The paper explores whether an oversupply of new graduates might explain the pattern, but the authors say that's unlikely. The number of graduates in AI-exposed majors grew rapidly in the years before ChatGPT's release, then leveled off quickly afterward, with survey data suggesting enrollments in some of the most exposed programs have started falling. However, the number of those graduates landing jobs in their field has dropped even faster. Changes in self-employment and graduate school enrollment could account for as much as half of the five-percentage-point employment decline, the research concludes. The authors told The Register that firms may be hiring fewer new graduates for reasons related to AI's existence but not necessarily to their own use of the technology—for instance, businesses not embracing AI may simply feel less certain about the future or worry that AI makes it harder to find suitable candidates.
The researchers say they're concerned about the magnitude of these results but hesitate to turn this short-term trend into a long-term forecast for the labor market. Historical evidence shows many jobs performed today didn't exist decades ago, and it's not easy to predict what future jobs might be or what skills they'll require, the authors note. Still, the evidence to date suggests AI is reshaping the labor market and having a disproportionate impact on young people who went to school for jobs now being affected by the technology, like software development and writing. The paper's bottom line: new computer science graduates slinging coffee or waiting tables have a ready explanation—AI is making their degrees less valuable right out of the gate. Employers facing talent shortages in technical fields may need to rethink how they evaluate fresh graduates versus investing in automation, while universities will likely face pressure to redesign curricula around skills that complement rather than compete with AI tools.