Dealing with a potential bias in estimating the share of discriminated women

The Blinder-Oaxaca [1, 6] decomposition neglects any distributional issues of discrimination. Instead, Jenkins [5] has argued the importance of a distributional approach in evaluating wage discrimination, focusing on the entire distribution of discrimination experienced by each woman. In their distributional approach, Del Río et al. [3] have adapted the Foster, Greer and Thorbecke (FGT) [4] poverty indices in studying wage discrimination. These discrimination indices depend on a parameter which can be interpreted as a measure of aversion to discrimination. When the aversion parameter is zero, the index measures the share of discriminated women. In this paper we will demonstrate that the naïve approach to the estimation of the share of discriminated women – similar to that used by Del Río et al. [3] – could be considerably biased. We propose testing the significance of the discrimination experienced by each woman, using appropriate statistical tests.