A conditional Gini: measure, estimation, and application

Abstract

The Gini measure of inequality can be written as the ratio of means. By extension, we define the conditional Gini as the ratio of conditional means. This conditional Gini shows how inequality varies with population characteristics, both levels and differences, and creates novel opportunities to trace inequality as differences in observables vary across the population. We propose a regression-based method for estimating the conditional Gini, and a method for inference using recent techniques for clustered data. The conditional Gini is estimated using a large, nationally representative household survey from Thailand. We find that wealth differences are associated with significantly less income inequality among households that use the financial sector than among those that do not, consistent with the idea that financial access relaxes self-financing constraints and broadens economic opportunity.