Oaxaca-Blinder Decomposition
Also known as: Blinder-Oaxaca Decomposition, Wage Gap Decomposition, Threefold Decomposition, Detailed Decomposition
The Oaxaca-Blinder decomposition is a regression-based technique that splits the difference in a mean outcome between two groups — classically the average wage gap between men and women or between racial groups — into a part explained by differences in observable characteristics (endowments such as education and experience) and an unexplained part attributed to differences in how those characteristics are rewarded (the coefficients). Introduced independently by Ronald Oaxaca and Alan Blinder in 1973, it became the standard tool for studying wage discrimination and group disparities.
Key highlights
- Intuitively partitions a group gap into 'differences in characteristics' versus 'differences in returns' in the units of the outcome.
- Provides a detailed decomposition attributing portions of the gap to individual covariates, aiding interpretation.
- Built on familiar OLS regressions and widely implemented, so it is easy to apply and communicate.
- Extends to nonlinear models, quantiles, distributional gaps, and pooled-reference specifications for richer analysis.
Intuition
This section is available to Pro members. Upgrade to Pro
How it works
This section is available to Pro members. Upgrade to Pro
When to use it
Use the Oaxaca-Blinder decomposition when you want to explain a mean outcome gap between two groups and separate the part due to differences in measurable characteristics from the part due to differences in how those characteristics are valued. It is the standard method in labor economics for studying gender, racial, and other wage gaps, and is widely applied to health, education, and other outcomes. It assumes a correctly specified linear outcome model and treats the unexplained component as a residual, so the result is descriptive, not causal: the unexplained part absorbs both genuine differential treatment and any omitted productivity-relevant variable. For non-mean statistics or distributional gaps, quantile and reweighting extensions (RIF regression, DiNardo-Fortin-Lemieux) are used instead.
Strengths & limitations
- Intuitively partitions a group gap into 'differences in characteristics' versus 'differences in returns' in the units of the outcome.
- Provides a detailed decomposition attributing portions of the gap to individual covariates, aiding interpretation.
- Built on familiar OLS regressions and widely implemented, so it is easy to apply and communicate.
- Extends to nonlinear models, quantiles, distributional gaps, and pooled-reference specifications for richer analysis.
- Descriptive, not causal: the unexplained component conflates differential treatment with all omitted relevant variables, so it is at best an upper bound on discrimination.
- Suffers the index-number problem — results depend on the choice of reference coefficient vector β*.
- The detailed decomposition of the unexplained part is sensitive to the choice of base category for categorical variables (identification problem).
- Assumes a correctly specified, usually linear, outcome model; misspecification biases both components.
Common pitfalls
This section is available to Pro members. Upgrade to Pro
Applications
This section is available to Pro members. Upgrade to Pro
Frequently asked
What is the difference between the twofold and threefold decompositions?
The threefold decomposition splits the gap into an endowments effect, a coefficients effect, and an interaction term that captures the simultaneity of differences in characteristics and returns, all relative to one group as the base. The twofold decomposition instead chooses a single reference (nondiscriminatory) coefficient vector β* and folds the interaction into the explained and unexplained components, yielding just 'explained' and 'unexplained' parts. The twofold form is more common in discrimination studies because it maps cleanly onto the explained-versus-unexplained dichotomy.
How does Oaxaca-Blinder relate to the Kitagawa decomposition?
Both decompose a difference between two groups into a component due to differing 'rates' (effects/returns) and a component due to differing 'composition' (characteristics/endowments). Kitagawa's 1955 demographic decomposition splits a difference between two summary rates into rate and composition effects using cell-level data. Oaxaca-Blinder generalizes this logic to a regression setting with continuous and multiple covariates, which is why the family is sometimes called the Kitagawa-Oaxaca-Blinder decomposition.
Can the unexplained component be interpreted as discrimination?
Only with great caution. The unexplained component is a residual: it captures the part of the gap not accounted for by the included characteristics, which contains any genuine differential treatment but also every productivity-relevant factor omitted from the regression. If important determinants of the outcome are unmeasured, the unexplained part overstates discrimination. It is therefore best read as an upper bound or a descriptive quantity to be probed further, not a clean causal estimate of discrimination.
Sources
- 1.Oaxaca, R. (1973). Male-female wage differentials in urban labor markets. International Economic Review, 14(3), 693–709.
- 2.Blinder, A. S. (1973). Wage discrimination: Reduced form and structural estimates. The Journal of Human Resources, 8(4), 436–455.
You have read it. What now?
Cite this page
ScholarGate. (2026, June 22). Oaxaca-Blinder Decomposition. ScholarGate. https://scholargate.app/economics/oaxaca-blinder-decomposition