David Díaz-Villarejo

Universidad Carlos III de Madrid · Economics · 2026–27 job market

Fields

Econometrics

Job market paper

Gini Index Regression

Many economic questions concern how and why inequality varies between populations with different characteristics. Comparisons of average outcomes do not reveal such variation. This paper develops Gini Index Regression (GIR), a framework in the spirit of mean or quantile regression that directly models the Gini of the conditional distribution of a nonnegative outcome, without choosing a parametric functional form for that distribution. In the linear case, the parameter of interest is shown to have a closed-form solution and, under model misspecification, the fitted function is a weighted best linear approximation to the conditional Gini. An omitted variable bias formula shows that changing controls affects both the projection and the ranking of outcomes. The estimator combines nonparametric series estimation of conditional ranks with weighted least squares, and reproduces empirical group Ginis in saturated group-indicator specifications. We establish √n-consistency, asymptotic normality, and validity of the bootstrap, accounting for both stages. A prediction characterization of the conditional Gini is also used to propose an R2-type goodness-of-fit measure. An application to U.S. data finds that the wage inequality gap by union status remains significant even after controlling for relevant characteristics, and that college-educated workers indeed possess the lowest levels of inequality.