Returns To Education
Returns to education is the part of economics that asks a deceptively simple question: what does a person get back from going to school?
9 frameworks, from 1960 to today · Education Economics · Economics
- Human capital 1960-Present
- Ability Bias and Selection Models 1970-Present
- Signalling and Screening Theory 1973-Present
- Mincer Earnings Function 1974-Present
- Heterogeneous Returns and Marginal Returns 1990-Present
- Non-Pecuniary Returns 1990-Present
- Credibility revolution 1991-Present
- Skill-Biased Technological Change 1992-Present
- Local Average Treatment Effect 1994-Present
Overview
Returns to education is the part of economics that asks a deceptively simple question: what does a person get back from going to school? The "return" is usually measured in money—how much more a worker earns for each additional year of schooling or for a degree—but it also includes things money does not buy directly: better health, a longer life, more stable employment, and the satisfaction of a more interesting job. The question matters because education is one of the largest investments a family or a government ever makes. If schooling reliably raises productivity and earnings, then subsidising it can be both fair and efficient1. If the extra earnings mostly reflect something else—raw talent, family connections, or a credential that merely sorts people—then the case for spending public money on it looks different1. Almost every debate in this subfield is a version of that tension.
Classical economists and the early idea of human capital
The idea that skills are a form of capital is old1. Adam Smith, in The Wealth of Nations (1776), argued that the "acquired and useful abilities" of a country's inhabitants are a real part of its capital stock, and that the cost of acquiring them is repaid with profit. For most of the nineteenth and early twentieth centuries, though, economists treated labour as a fairly homogeneous input and education as consumption or a public good rather than an investment. The tools to measure returns did not yet exist, and the data—large surveys linking schooling to earnings—were scarce.
The human capital revolution of the late 1950s and 1960s
The modern subfield begins with a cluster of work that turned education into an investment problem. Theodore Schultz's 1961 article "Investment in Human Capital" argued that spending on schooling, health, and training is capital formation, and that the rapid growth of economies could not be explained by physical capital and labour alone. Gary Becker's 1964 book Human Capital gave the theory its systematic form: individuals weigh the costs of schooling (tuition plus forgone earnings) against the stream of higher earnings it produces, and invest up to the point where the marginal return equals the cost of funds. Becker also distinguished general training, useful to any employer, from firm-specific training, useful only to the current one—a distinction that explained why firms pay for some training and workers for other kinds.
The empirical breakthrough came from Jacob Mincer1. His 1958 paper and especially his 1974 book Schooling, Experience, and Earnings produced what became the workhorse of the field, the Mincer earnings function: a regression of the logarithm of earnings on years of schooling and years of work experience (usually entered as a quadratic to capture the fact that earnings rise steeply early in a career and then flatten). The coefficient on schooling approximates the percentage increase in earnings from one more year of school2. In practice, researchers found returns of roughly 5 to 10 percent per year in many countries—a figure that has proved remarkably stable across decades and places. The Mincer equation's appeal was that it needed only three variables and could be estimated on any household survey, which is why it spread around the world and became the standard way to compare returns across countries.
The sorting problem: ability bias and signalling
Almost as soon as returns were being estimated, critics pointed out a flaw. People who stay in school longer differ from those who leave early in ways that also raise earnings—ability, motivation, family background. If the able both stay in school and earn more, the Mincer coefficient overstates what schooling itself does. This is ability bias, and it set the agenda for decades. Early responses tried to control for measured ability using IQ scores or sibling comparisons; the estimates typically fell but did not vanish.
A deeper challenge came from signalling theory13. Michael Spence's 1973 paper "Job Market Signaling" argued that education might raise earnings without raising productivity at all: if more able workers find it cheaper to obtain a degree, the degree can serve as a credible signal of ability to employers, who then pay more for it. In the extreme version, schooling is pure sorting. The screening hypothesis, developed by Kenneth Arrow and others around the same time, made a related point from the employer's side: firms use credentials to screen applicants, so the private return to a degree can exceed its social return. These theories did not deny that skills matter; they insisted that the observed earnings gap bundles together productivity gains and information about workers, and that the two cannot be separated by looking at earnings alone.
The credibility revolution
The decisive change came from a shift in method. From the 1990s onward, economists stopped trying to control statistically for unobserved ability and instead looked for situations where schooling changed for reasons unrelated to a person's talent—natural experiments. The approach became known as the credibility revolution, because a study's claim to causal knowledge now rested on a credible source of exogenous variation rather than on a long list of control variables4.
Several designs did the work. In the United States, Joshua Angrist and Alan Krueger (1991) used the fact that men born in different quarters were subject to different compulsory-schooling and child-labour laws, and that those born earlier in the year could leave school at a younger age; the resulting variation in schooling was used to estimate returns. In the UK, researchers exploited a 1947 school-leaving-age reform. In Indonesia, Esther Duflo (2001) used a massive school-construction programme to estimate the return to an extra year of schooling, finding returns around 7–9 percent. In the 2000s, a series of studies used changes in compulsory schooling laws across European countries—often associated with the work of Philip Oreopoulos—and consistently found returns in the range of 5 to 15 percent, with larger effects for disadvantaged students. The general finding was that the Mincer estimates were not wildly inflated: schooling does raise earnings, though the causal effect is often somewhat smaller than the naive correlation.
A parallel literature used twin studies, comparing identical twins with different amounts of schooling, to difference out family and genetic background; these too generally found positive returns, though with wide uncertainty.
The college wage premium and skill-biased technological change
While the microeconometric work was refining the return to a year of school, a macro-level puzzle emerged. From the late 1970s, the earnings gap between college graduates and high-school graduates in the United States widened sharply, even as the supply of college graduates grew. If returns were falling with supply, the premium should have narrowed. The explanation that took hold was skill-biased technological change: new technologies, especially computing, raised the demand for skilled workers faster than the education system could supply them. The canonical statement is the 1992 paper by Lawrence Katz and Kevin Murphy, "Changes in Relative Wages, 1963–1987," which showed that the college premium rose when the growth of the college-educated workforce slowed. The framework was later extended to account for the polarisation of the labour market, in which middle-skill jobs were automated away while both high-skill and low-skill service jobs grew.
This line of work tied the returns literature to the economics of inequality. If technology keeps raising the demand for skill, then the return to education depends on how many people get educated—a race between technology and schooling. It also helped explain why returns differ so much across countries and over time: they reflect the balance of supply and demand in each labour market, not a fixed law of nature.
Heterogeneity and marginal returns
A further refinement was the recognition that "the" return to education is an average that hides enormous variation. Returns differ by field of study, by institution, by gender, by family background, and by the level of schooling. The marginal return—the return to the next year of school for a particular person—is what should guide individual and policy decisions, and it can differ sharply from the average. The classic statement is the "ability bias" debate's successor: the heterogeneous returns literature associated with James Heckman and others, which used econometric models of selection to estimate returns for people at different points in the distribution. A recurring finding is that returns are often higher for those who would otherwise have dropped out—the marginal student—which strengthens the case for policies that keep disadvantaged students in school.
The local average treatment effect framework, formalised by Guido Imbens and Joshua Angrist in 1994, clarified what the natural-experiment estimates actually measure: the return for the "compliers," those whose schooling changed because of the instrument. This is a real return, but it is not necessarily the return for everyone, and interpreting it requires knowing who the compliers are.
Non-pecuniary returns
Alongside the earnings literature, a body of work has measured what schooling does beyond wages. Better-educated people report better health, live longer, are less likely to be unemployed, and are less likely to commit crimes. Some of this reflects the same selection problem—healthier people may also stay in school—but studies using the same natural experiments have found causal effects of schooling on health behaviours and on civic participation. The non-pecuniary returns matter for policy because they mean that earnings-based estimates understate the full social return to education.
The present
Today the field works with a mature toolkit. The Mincer equation remains the standard descriptive device, used to compare returns across countries and to track the college premium over time. Causal estimates from compulsory-schooling reforms, school-construction programmes, and lottery-based admissions are the benchmark for how much schooling actually raises earnings. The signalling and human-capital interpretations coexist, and a growing literature tries to separate them by looking at what employers actually learn about workers and what students actually learn in school. The heterogeneity agenda has moved to the centre: researchers now ask not whether education pays but who it pays for, at what margin, and in what field. Open questions remain about the returns to specific skills and credentials, the role of non-cognitive traits, the returns in low- and middle-income countries where data are thinner, and how automation and artificial intelligence will reshape the demand for skill in the coming decades.
References
- Human capital (Background, Criticism, introduction, History), Wikipedia
- Mincer earnings function, Wikipedia
- Signalling (economics), Wikipedia
- Credibility revolution, Wikipedia
The frameworks of Returns To Education
Each framework is a way of working the field has used: when it held, and the question it answers.
- 1960-Present
- 1970-Present
Ability Bias and Selection Models
Does schooling raise earnings or just pick able people?
- 1973-Present
Signalling and Screening Theory
Can a degree raise pay without raising skill?
- 1974-Present
Mincer Earnings Function
How much more do workers earn per year of school?
- 1990-Present
Heterogeneous Returns and Marginal Returns
Who gains most from an extra year of school?
- 1990-Present
- 1991-Present
- 1992-Present
Skill-Biased Technological Change
Why has the college wage premium risen?
- 1994-Present
Local Average Treatment Effect
Whose return does an experiment actually measure?
How the ideas connect
Each arrow runs from an earlier idea to the later one it shaped. The list below says what passed between them.
Ability Bias and Selection Models grew out of Human capital
Once Becker and Mincer framed schooling as an investment whose return could be estimated from earnings, the immediate question became whether the estimated return captured schooling or unobserved ability. The ability-bias literature grew directly out of human capital theory as its internal critique, using IQ and sibling controls to test the Mincer coefficient.
Carried by Gary Becker; Human Capital (1964); Jacob Mincer.
Signalling and Screening Theory reacted against Human capital
Spence's signalling model defined itself against the human capital assumption that schooling raises productivity, arguing that a degree can raise earnings purely by certifying pre-existing ability. The screening hypothesis made the same challenge from the employer's side, so signalling arose as an attack on the human capital interpretation of the schooling-earnings gradient.
Carried by Michael Spence; Job Market Signaling (1973); Kenneth Arrow.
Mincer Earnings Function grew out of Human capital
Mincer's earnings function operationalised Becker's human capital investment model, turning the theory of optimal schooling into a regression of log earnings on years of schooling and experience. The Mincer equation became the standard empirical implementation of human capital theory.
Carried by Jacob Mincer; Schooling, Experience, and Earnings (1974); Gary Becker.
Credibility revolution replaced Ability Bias and Selection Models
The credibility revolution replaced the ability-bias strategy of controlling for unobserved talent with the search for exogenous variation in schooling. Angrist and Krueger's quarter-of-birth instrument and later compulsory-schooling and school-construction designs reframed the question as one of causal identification rather than statistical control.
Carried by Joshua Angrist; Alan Krueger; Esther Duflo.
Local Average Treatment Effect grew out of Credibility revolution
The local average treatment effect was developed to give a precise interpretation to the instrumental-variable estimates produced by natural experiments. Imbens and Angrist showed that an instrument identifies the return for compliers, making LATE the theoretical foundation of the credibility revolution's estimates.
Carried by Guido Imbens; Joshua Angrist.
Non-Pecuniary Returns grew out of Human capital
The non-pecuniary returns literature extended Becker's human capital framework beyond earnings to health, longevity, job satisfaction and employment stability. It treated these outcomes as additional components of the return to schooling investment.
Carried by Gary Becker.
Skill-Biased Technological Change influenced by Human capital
Skill-biased technological change took the human capital notion that schooling raises productivity and embedded it in a supply-and-demand framework for relative wages. Katz and Murphy's 1992 paper explained the rising college premium as demand for human capital outpacing its supply.
Carried by Lawrence Katz; Kevin Murphy; Changes in Relative Wages, 1963–1987 (1992).
Heterogeneous Returns and Marginal Returns grew out of Ability Bias and Selection Models
Heterogeneous-returns models grew out of the selection framework developed to handle ability bias, generalising it from a single correction term to person-specific returns. The econometric machinery of selection on unobservables was repurposed to estimate marginal returns for different types of students.
Carried by James Heckman.
Credibility revolution influenced by Signalling and Screening Theory
Signalling theory sharpened the question of whether the earnings return reflects productivity or sorting, which the credibility revolution then tried to settle with natural experiments. The causal-effects literature is explicitly framed as testing the human capital versus signalling interpretations of the schooling coefficient.
Carried by Joshua Angrist; Alan Krueger.
Heterogeneous Returns and Marginal Returns grew out of Mincer Earnings Function
The heterogeneous-returns literature attacked the Mincer function's single average coefficient, arguing that the return to the next year of school differs across people and that the average is not the relevant parameter for policy. Heckman and others built selection models to estimate returns at different points in the distribution.
Carried by James Heckman.
Credibility revolution influenced by Mincer Earnings Function
The Mincer equation supplied the baseline estimate that the credibility revolution set out to test and, in many cases, to correct. Natural-experiment studies report their instrumental-variable estimates against the OLS Mincer coefficient as the benchmark to be validated or revised.
Carried by Joshua Angrist; Alan Krueger; Esther Duflo.
Skill-Biased Technological Change influenced by Mincer Earnings Function
The college wage premium studied by Katz and Murphy is the Mincer schooling coefficient estimated separately by education group. The supply-demand framework was built on the same log-earnings regressions that the Mincer function had made standard.
Carried by Lawrence Katz; Kevin Murphy.
Beyond the field
Where Returns To Education's ideas came from in other fields, and what they led to there: the strongest 12 of 24 links.
Human capital led to Skill-Biased Technological Change in Higher Education Economics
Skill-biased technological change is a direct extension of human capital theory: it applies the idea that schooling is investment in skills to explain shifts in relative wages via technology-driven demand for skill.
Carried by Gary Becker; Jacob Mincer.
Signalling and Screening Theory led to Signaling Theory in Corporate Finance Theory
The education-economics signalling and screening literature, especially Spence's model, was the direct source of the corporate-finance signalling models of Ross and Myers.
Carried by Michael Spence.
Skill-Biased Technological Change influenced by Human capital in Human Capital And Skill Formation
The SBTC literature takes the human capital framework's notion of heterogeneous worker skill and asks how technology changes its relative return, a direct extension of Becker's and Mincer's work.
Carried by Gary Becker; Jacob Mincer.
Skill-Biased Technological Change influenced by Human capital in Wage Theory
Skill-biased technological change was built directly on human capital theory: it treats schooling as investment in a skill endowment whose relative price shifts with technology, the framework Becker and Mincer supplied.
Carried by Gary Becker; Jacob Mincer.
Local Average Treatment Effect grew out of Rubin causal model in Microeconometrics
LATE was derived within the Rubin causal model: Angrist, Imbens and Rubin added monotonicity and an exclusion restriction to potential outcomes to identify a complier average causal effect.
Carried by Joshua Angrist; Guido Imbens; Donald Rubin.
Human capital led to General versus Firm-Specific Training in Human Capital And Skill Formation
The training distinction is a branch of the human capital account of returns to education and training, specifying how returns are split between worker and firm.
Carried by Gary Becker.
Human capital led to Causal Identification of Returns to Schooling in Human Capital And Skill Formation
Causal identification of returns to schooling is a refinement of the human capital account of returns to education, adding credible research designs to estimate the return. It grew directly out of the human capital returns framework.
Human capital influenced Human Capital Index in Human Capital Development
The index converts schooling into productivity using estimated returns to education, the core quantity of the returns-to-education literature. Its methodology explicitly relies on Mincerian wage regressions to translate human capital into output terms.
Carried by World Bank; Jacob Mincer.
Human capital influenced Health-Nutrition-Human Capital Production Process in Human Capital Development
Returns-to-education analysis supplied the empirical framework later extended to estimate returns to health and nutrition investments.
Carried by Jacob Mincer.
Ability Bias and Selection Models grew out of Rates of Return to Education in Human Capital Development
Ability bias and selection models arose directly from returns-to-education estimation, addressing the omitted-ability problem in Mincer-style earnings regressions.
Ability Bias and Selection Models influenced Signaling and Screening Theory in Education Economics
Selection models and ability-bias concerns about returns to schooling framed the question signalling theory answered, namely whether education raises productivity or merely reveals ability.
Mincer Earnings Function influenced Rates of Return to Education in Human Capital Development
Rates of return to education were computed by fitting Mincer-type earnings equations, so the returns literature built directly on the earnings function.
Carried by Jacob Mincer.

Returns To Education is a planet in Nousfarer
Its city has a tower for each of these 9 frameworks. Land, walk in, and climb one concept at a time.
Land on Returns To EducationRelated fields
Fields its ideas are linked to
- Human Capital Development 8 links
- Human Capital And Skill Formation 5 links
- Higher Education Economics 3 links
- Education Economics 2 links
- Impact Evaluation 1 link
- Microeconometrics 1 link
- Corporate Finance Theory 1 link
- Labor Economics 1 link
- Labor Supply Theory 1 link
- Wage Theory 1 link
More in Education Economics
See where Returns To Education sits on the Map of Human Knowledge