Loading field map
Why do people with more education earn higher wages? The answer seems obvious—schooling teaches useful skills—but a deeper puzzle has animated the economics of education for decades. If education simply signals pre-existing ability rather than building productive skills, then the observed earnings premium might reflect sorting, not learning. This tension between human capital and signaling theories has driven the evolution of the subfield, pushing researchers to develop ever more refined empirical tools and to confront the possibility that both mechanisms are at work.
Before the theoretical debate took shape, Jacob Mincer provided a simple empirical workhorse. In his 1974 book Schooling, Experience, and Earnings, Mincer proposed a regression equation that relates log earnings to years of schooling and a quadratic in work experience. The coefficient on schooling—the Mincerian return—estimates the average percentage increase in earnings for an additional year of education. This function is deliberately atheoretical: it describes a correlation without explaining why it exists. Yet it became the standard tool for measuring returns across countries and time periods, setting the stage for the theoretical frameworks that would follow. The Mincerian equation remains in wide use today, not as a theory but as a descriptive baseline that every subsequent framework must either explain or challenge.
In the early 1960s, Gary Becker and Theodore Schultz articulated a full theoretical account of the education–earnings relationship. Human capital theory treats education as an investment: individuals forgo current earnings to acquire skills that raise their future productivity. The Mincerian return is reinterpreted as the rate of return on that investment, reflecting the market's valuation of the skills learned. This framework dominated policy discussions for decades, justifying public spending on education as a way to boost national productivity and reduce inequality. Human capital theory provided a coherent, optimistic story: more schooling means more skills, which means higher wages. But it left a crucial question unanswered: what if employers reward the diploma, not the learning?
In 1973, Michael Spence published "Job Market Signaling," which offered a radically different interpretation. In his model, education does not increase productivity; instead, it serves as a costly signal that reveals a worker's pre-existing ability. Employers cannot observe ability directly, so they use educational attainment as a screening device. Because more able individuals find schooling less costly, they acquire more education, and employers rationally pay them more. The observed return to schooling could therefore reflect signaling rather than human capital accumulation. This created a fundamental ambiguity: the same empirical pattern—higher earnings for the more educated—could be explained by either theory. The screening hypothesis did not deny that education might also build skills; it argued that the signaling component could be substantial, and that the two mechanisms were empirically difficult to separate.
To break the stalemate, researchers looked for a prediction that distinguished the two theories. Human capital theory implies that returns should be roughly linear: each additional year of schooling adds a similar increment to earnings. Signaling theory, by contrast, predicts a premium for completing a degree—the "sheepskin effect"—because the credential itself conveys information. Starting in the late 1980s, studies such as Jaeger and Page's "Sheepskin Effects in the Returns to Education" found that wage gains are indeed larger at degree completion than at other years, supporting the signaling view. Yet the returns to years of schooling remained substantial even after controlling for credentials, suggesting that human capital also matters. The sheepskin framework did not resolve the debate; instead, it showed that both mechanisms coexist. Credentialism—the idea that employers use educational credentials as a sorting device—became a distinct lens, but it did not replace human capital theory. The two frameworks entered a period of live disagreement, each with empirical support.
A deeper problem had been lurking since the Mincerian era: the observed correlation between schooling and earnings might reflect selection bias, not causation. People who choose more education may differ in unobserved ways—ability, motivation, family background—that also raise their earnings. In the 1990s, a new generation of researchers turned to quasi-experimental methods to isolate causal effects. Using instrumental variables such as compulsory schooling laws (Angrist and Krueger, 1991) or college proximity (Card, 1995), they found that the causal return to schooling is often larger than the ordinary least squares estimate, not smaller. This finding surprised many: if selection bias were positive, the causal estimate should be lower. The larger IV estimates suggested that the return is heterogeneous—it varies across individuals—and that those who are induced to stay in school by policy changes (the "compliers") may have higher returns than the average person.
This framework transformed the subfield. Instead of estimating a single return to education, researchers now routinely estimate distributions of treatment effects. The finding that returns are higher for disadvantaged groups—those with less education or from poorer backgrounds—has direct policy implications: it suggests that interventions to increase schooling for these groups could yield large earnings gains. But it also speaks to the human capital versus signaling debate. If returns are higher for the marginalized, human capital theory would predict that they have more to gain from skill acquisition. Signaling theory could also explain the pattern if credentials are more valuable for workers with weaker signals. The causal revolution sharpened the empirical ground rules but did not settle the theoretical question. Modern research combines detailed data on curriculum content, skill assessments, and labor market outcomes to trace the mechanisms through which education affects earnings.
Today, the five frameworks coexist with a clear division of labor. The Mincerian earnings function remains the standard descriptive tool for international comparisons and basic summaries. Human capital theory provides the default interpretation for most policy analysis, while signaling and credentialist perspectives continue to inform research on labor market sorting and the value of degrees. The causal identification framework dominates empirical work, with researchers using randomized experiments, regression discontinuity designs, and instrumental variables to estimate credible effects. The leading frameworks agree that education has a causal impact on earnings, but they disagree sharply on the relative importance of skill acquisition versus signaling. They also agree that the return is not a single number—it varies by context, by person, and by the type of education. The central disagreement persists because both mechanisms are likely at work, and their relative weight may differ across labor markets, levels of schooling, and historical periods. The subfield has moved from asking "does education cause higher earnings?" to asking "through which channels, and for whom?"—a question that keeps the human capital and signaling debate very much alive.