nonacademicresearch.org Editorial
The economic return to completing a college degree — measured as the wage premium relative to workers with only a high school diploma — has been one of the most studied questions in labor economics for 50 years. The central finding is robust: each additional year of schooling is associated with approximately 8–10% higher wages in the United States, with similar magnitudes in other high-income countries. But the aggregate figure conceals substantial variation: returns differ by field of study, institution type, and individual characteristics.
The economic return to completing a college degree — measured as the wage premium relative to workers with only a high school diploma — has been one of the most studied questions in labor economics for 50 years. The central finding is robust: each additional year of schooling is associated with approximately 8–10% higher wages in the United States, with similar magnitudes in other high-income countries. But the aggregate figure conceals substantial variation: returns differ by field of study, institution type, and individual characteristics. The evidence also remains genuinely divided on how much of the wage premium reflects human capital (skills acquired in school) versus selection (more capable individuals sorting into higher education). Both matter, and the policy implications differ.
In 1958, Jacob Mincer published what would become one of the most cited papers in economics, establishing a formal relationship between years of schooling, experience, and earnings. The "Mincer earnings equation" framework has anchored six decades of subsequent research. The basic finding — that more education predicts higher earnings — has replicated in virtually every country studied.
The question is whether this association reflects causation, and if so, through what mechanism. If higher-educated workers earn more because education makes them more productive (human capital theory), then expanding access to education should raise productivity and wages broadly. If they earn more primarily because employers use credentials as signals of underlying ability (signaling theory), then expanding education redistributes credential benefits without creating commensurate economic value. Real-world returns to education are likely a mixture of both, with the composition varying by context.
This distinction matters for policy. The US has approximately $1.75 trillion in outstanding student loan debt. Public subsidies for higher education are enormous. If the returns are concentrated in certain fields, institutions, or individual characteristics, then who captures those returns — and who takes on debt to pursue them — has substantial distributive consequences.
The most widely cited estimates of the "return to schooling" — the wage premium per additional year of education — come from instrumental variable estimates using natural experiments that create exogenous variation in educational attainment.
Card (1995), using proximity to colleges as an instrument for educational attainment in the National Longitudinal Survey of Young Men, estimated returns to schooling of approximately 9–14%. Angrist and Krueger (1991), using quarter of birth as an instrument (affecting compulsory schooling completion via school entry cutoff dates), found returns of approximately 7–9%. These IV estimates tend to be at least as large as OLS estimates, suggesting that if selection bias exists, it does not substantially inflate the estimated returns.
For the college wage premium specifically — the earnings advantage of a bachelor's degree holder over a high school graduate — the Federal Reserve Bank of New York's analysis of Census data shows the premium rising from approximately 40% in the 1970s to approximately 80% today, with the gap stabilizing in recent years. The average annual earnings advantage in 2022 was approximately $30,000 for young workers with a bachelor's degree versus a high school diploma.
The college wage premium is real, large, and has increased over time. This basic finding is not contested.
The aggregate return masks enormous variation by field of study. Georgetown University's Center on Education and the Workforce has published detailed analyses of earnings by major:
The premium for a computer science degree relative to a high school diploma is substantially larger than the premium for an arts degree. This has become more pronounced over time as technology-sector wages have increased relative to other sectors.
Mountjoy and Hickman (2021), in a paper using a regression discontinuity design around admission cutoffs, found that the marginal return to attending a selective institution (versus not being admitted) was positive and economically significant, but that the returns were heterogeneous: students who were marginally admitted to selective institutions gained substantially more from attending than average students who were admitted.
The question of whether institution selectivity independently predicts earnings was famously addressed by Stacy Dale and Alan Krueger in a 2002 paper in the Quarterly Journal of Economics. Using a "self-revelation" method that compared students who applied to the same set of colleges (adjusting for application ambition), they found that graduating from a more selective institution had little independent effect on earnings — with one important exception: the premium was substantial for students from lower-income families. This suggests that elite institutions provide access to networks and opportunities that matter most for students who would not otherwise have those connections.
More recent work by Raj Chetty and colleagues (Opportunity Insights) using administrative tax records finds similar results: for most students, graduating from a selective versus less selective institution matters less than the field studied and the individual's prior characteristics. The exception is for Ivy Plus institutions, where attendance significantly increases access to high-income careers even controlling for prior characteristics.
A specific test of the signaling hypothesis examines the "sheepskin effect" — whether the credential itself (completing a degree) produces a wage premium beyond what would be predicted by years of schooling alone. If education is pure human capital, then each year of schooling should produce proportional returns and there should be no discontinuity at degree completion. In fact, most studies find a credential premium: completing a degree produces higher wages than completing all but the last year of coursework.
Jaeger and Page (1996) estimated sheepskin effects of 17–28% for bachelor's degrees in the United States, controlling for years of schooling. This is consistent with signaling playing a meaningful role alongside human capital accumulation. Neither pure theory fits the data; both mechanisms appear to operate.
The focus on four-year college degrees has led to underestimation of returns to vocational and associate degree programs. Jepsen, Troske, and Coomes (2014), examining administrative data from Kentucky community colleges, found substantial earnings returns to vocational credentials in technical fields — including healthcare, manufacturing, and construction — often exceeding the returns to academic associate degrees.
For many workers, particularly those entering healthcare support occupations (radiologic technicians, respiratory therapists, dental hygienists), vocational credential programs of 18–24 months produce earnings gains comparable to four-year programs in lower-return fields.
The strongest challenge to inflated claims about college returns is heterogeneity. The average return to a bachelor's degree is positive and substantial, but the distribution is wide. Approximately 28% of associate degree holders earn more than the median bachelor's degree holder, according to Georgetown CEW data. A substantial fraction of bachelor's degree holders — particularly in lower-earning majors at lower-selectivity institutions — earn wages similar to or lower than high school graduates in certain trade occupations.
The debt burden associated with higher education increasingly affects the net return. A graduate with $100,000 in student loans and a $45,000 starting salary may be worse off than a high school graduate who entered the workforce four years earlier with no debt. The gross wage premium is often cited without accounting for foregone earnings and debt service costs.
Finally, the signaling interpretation has provocative implications. If a meaningful portion of the college wage premium reflects signaling rather than skill acquisition, then marginal expansion of college attendance — particularly into populations for whom completion is uncertain — may produce debt without commensurate human capital gains.
The average return to completing a college education is large, has increased over time, and is well-documented by multiple methodologies across multiple countries. The wage premium for a bachelor's degree over a high school diploma in the United States is approximately 80%, representing an advantage of roughly $30,000 annually in early-career earnings.
This aggregate obscures crucial heterogeneity. Returns vary enormously by field of study, institution type, and individual characteristics. For some students in some programs, the return clearly justifies the cost and debt. For others, particularly in lower-earning fields at high-cost institutions, the calculus is less favorable.
The policy implication is not that higher education is overvalued but that uniform credential subsidies and debt financing may not distribute costs and benefits optimally. Incentivizing high-return programs, strengthening vocational pathways, and improving completion rates (which affect both the credential premium and the debt burden) are all evidence-supported levers. The question is not whether education has returns — it does — but who captures them and at what cost.
nonacademicresearch.org Editorial (2026). The Returns to Education: What Does a Degree Actually Buy?. nonacademicresearch.org. nar:t2yq7vx43z4ss63gda
@misc{699qsfde,
title = {The Returns to Education: What Does a Degree Actually Buy?},
author = {nonacademicresearch.org Editorial},
year = {2026},
howpublished = {nonacademicresearch.org},
note = {nar:t2yq7vx43z4ss63gda},
}Temporary identifier. This paper carries a temporary nar:* identifier valid for citation within the independent research community. A permanent DOI will be minted via DataCite once the platform completes nonprofit registration.
Log in to join the discussion.
Loading…