OaxacaBlinder

The Oaxaca-Blinder decomposition is a statistical method used to analyze differences in average outcomes between two groups, typically by separating the gap into an explained portion and an unexplained portion. Developed independently by Alberto Oaxaca and Ronald Blinder in the early 1970s, it has become a standard tool in labor economics and other fields for studying wage gaps, educational achievement gaps, and similar disparities.

The method begins by estimating separate regression models for each group, regressing the outcome on a common set of covariates (such as education, experience, and occupation for wages). By comparing the expected outcomes from these models, the overall difference between groups is decomposed into two parts. The explained part attributes differences to variations in average characteristics across groups, while the unexplained part attributes differences to the way those characteristics are rewarded (the coefficients), which can reflect discrimination, unobserved factors, or model misspecification. In practice, a reference set of coefficients is chosen (often a pooled or non-discriminatory coefficient), and the gap is split accordingly.

Extensions of the basic approach accommodate nonlinear models, such as log-linear specifications for wages, and use alternative reference standards. The decomposition is widely applied to gender, racial, or educational gaps and can be adapted to multiple outcomes or to quantify sensitivity to the choice of reference coefficients.

Limitations include sensitivity to the choice of reference coefficients, potential omitted-variable bias, and the interpretation of the unexplained component, which can mix discrimination with unmeasured variables. Despite these caveats, the Oaxaca-Blinder decomposition remains a foundational, widely used tool for understanding disparities across groups.