The Effect of Labor Market Information on Community College Students’ Major Choice

GAICIL Insights

Choosing a college major is a bet on a labor market most students cannot see.¹ The Effect of Labor Market Information on Community College Students' Major Choice shows just how foggy the view is. Studying community college students in California, Rachel Baker, Eric Bettinger, Brian Jacob and Ioana Marinescu find that less than 15% of a sample rank broad categories of majors accurately by labor market outcomes. Students believe salaries are 13% higher than they actually are, on average, and they underestimate the probability of being employed by almost 25%.

Read those two errors together and a strange picture emerges: students are too optimistic about pay and too pessimistic about getting hired. The fog distorts in both directions. And it is thickest for those with the least room for error: lower income students are more likely to make mistakes in estimating the probability of employment. Information gaps become an engine of inequality, layered on top of the disadvantages students already carry.

What actually drives the choice? Course enjoyment and grades come first. Expected labor market outcomes matter too, and more than non-experimental data suggest. Experimental estimates exceed OLS estimates: a 10% increase in salary is associated with a 14 to 18% increase in the probability of choosing a specific category of majors. The authors find evidence consistent with the larger experimental effect being driven by students' lack of information at baseline. When outcomes are uncertain, it can be rational to discount them.

That is the paper's quiet bombshell. Students who look indifferent to earnings may be behaving sensibly in the face of noise. Clear the noise and earnings start to count. The authors conclude that better information about labor market outcomes has the potential to significantly affect major choice, and that raising the salience of those outcomes may help students improve them.

Artificial intelligence turns this from a classroom puzzle into an urgent design problem. AI is reshaping the labor market itself, shifting which skills command a premium and which occupations grow. Returns to majors are likely to move faster than before. A student relying on folk wisdom, or on data a few years stale, is betting on a labor market that may no longer exist. The fog documented here thickens unless information keeps pace.

The same technology can clear it. Personalized guidance that draws on current labor market data, explains uncertainty and compares options can deliver at scale what an overloaded counselor cannot. The study's experimental design proves that information interventions move beliefs and choices. The task is to put accurate, current, understandable information in front of every student, starting with lower income students who make the largest errors.

There is a trap to avoid. Students already overestimate salaries. An AI advisor that cheerfully amplifies optimistic numbers would make the problem worse. Accuracy and honesty about uncertainty are the product, and the flattering forecast is a defect.

Public data hold the key. Earnings and employment outcomes by field of study are public goods whose value rises as the economy changes faster. Making them open, timely and machine-readable lets developers build advising tools and lets students compare programs on outcomes. A modest investment in data plumbing could power a whole ecosystem of guidance.

Respect for preferences remains essential. Students care about what they study and how well they expect to do, and the evidence says enjoyment and grades dominate. Better information leaves those preferences intact and lets students weigh them against realistic expectations of employment and pay. Correcting systematic errors while honoring choice expands freedom.

The economic lesson is about prices. Skill markets work when participants can see the price signals. When those signals are obscured, students invest in human capital partly blind, and the costs fall on those with the fewest resources. As AI reshapes work, clearer signals delivered through accessible tools may be among the cheapest, most effective adaptation policies on offer. The California evidence suggests such signals would change decisions, and that the students who need them most would gain the most.

One provocation follows. A society that spends heavily on retraining workers after the fact might get more for its money by fixing the information students use before they enroll.