The Mincer earnings equation — log(wage) = alpha + beta*S + gamma1*X + gamma2*X^2 + epsilon — is the foundational empirical specification in labor economics, where S is years of schooling, X is potential experience, and beta estimates the percentage return to an additional year of education. The equation fits observed earnings data remarkably well across countries, but the OLS estimate of beta (typically 8-13%) is biased by ability bias (high-ability individuals get more education AND earn more, inflating the apparent return) and selection bias. Instrumental variable approaches — using compulsory schooling laws, quarter of birth, or distance to college as instruments — generally find returns similar to or slightly above OLS estimates, suggesting that ability bias may be roughly offset by measurement error or that the local average treatment effect for compliers exceeds the average return.
The Mincer equation is one of the most successful empirical specifications in all of economics. Jacob Mincer's 1974 formulation — modeling the log of earnings as a linear function of years of schooling and a quadratic in potential experience — fits the data so well across countries, time periods, and demographic groups that it has become the default starting point for any empirical analysis of wages. Its success comes from grounding the specification in human capital theory: the log-linear relationship between earnings and schooling arises from a model where each year of schooling raises productivity by a constant percentage, and the concave experience profile reflects declining human capital investment over the career.
The coefficient on schooling (beta) is typically interpreted as the percentage increase in earnings from one additional year of education. OLS estimates across developed countries generally fall between 8% and 13% — a remarkably high return compared to most physical investments. But interpretation of this coefficient is fraught with endogeneity concerns. The most important is ability bias: individuals with higher innate ability (cognitive ability, motivation, family advantages) tend to both acquire more education and earn more regardless of education. If ability is correlated with schooling and earnings but omitted from the regression, the schooling coefficient absorbs some of the ability effect and is biased upward.
The instrumental variables revolution in labor economics was motivated precisely by this problem. Angrist and Krueger (1991) used quarter of birth as an instrument: compulsory schooling laws mean that students born earlier in the year reach the legal dropout age at a younger grade, leading to slightly less schooling. Since quarter of birth is plausibly random (unrelated to ability), it provides exogenous variation in schooling. Their IV estimate of the return to education was about 7-8%, close to OLS. Other instruments — changes in compulsory schooling laws, distance to college, tuition policy changes — have also been exploited, with IV estimates often slightly exceeding OLS estimates.
The finding that IV estimates are similar to or above OLS estimates was initially surprising, since ability bias was expected to inflate OLS. Two explanations have been proposed. First, measurement error in self-reported years of schooling creates attenuation bias that pulls the OLS estimate downward, potentially offsetting the upward ability bias. Second, the LATE interpretation — IV identifies the effect for compliers, who in the case of compulsory schooling instruments are marginal students who would have dropped out without the law. If these students have particularly high returns to the marginal year (perhaps because dropping out carries a large stigma or because the last years of required schooling are particularly valuable), the LATE exceeds the average return.
The Mincer equation's simplicity is both its strength and its limitation. It assumes a constant percentage return per year of schooling (each year has the same proportional effect), which may not hold — the return to the 12th year (completing high school) may differ from the return to the 16th year (completing college) due to sheepskin effects (credential bonuses at degree completion). Extensions include allowing for nonlinear returns, controlling for observable ability measures, adding covariates for demographic and family background, and allowing returns to vary across the distribution of earnings. Despite these refinements, the basic Mincer specification remains the discipline's workhorse.