The well-known Blinder–Oaxaca [Blinder, J. Hum. Resour. 8(4), 436–455 (1973); Oaxaca, Int. Econ. Rev. 14(3), 693–709 (1973)] decomposition divides the wage differential between men and women into a part, which can be explained by differences in individual characteristics, and another part, which is usually interpreted as discrimination. This decomposition neglects any distributional issues in evaluating discrimination, thus permitting undesirable compensation between positively and negatively discriminated women. Jenkins [J. Econ. 61(1), 81–102 (1994)] has criticized this aspect, instead, preferring a distributional approach, where the entire distribution of experienced discrimination is evaluated. Following Jenkins [J. Econ. 61(1), 81–102 (1994)], Del Río et~al. [J. Econ. Inequal. 9(1), 57–86 (2011)] use a distributional approach, adapting the Foster–Greer–Thorbecke [Econometrica 52(3), 761–766 (1984)] class of poverty indices to the study of discrimination.
Studies adopting this approach merit little attention as regards the issue of the separate measuring of wage discrimination and occupational discrimination. Alternatively, we have used the Foster–Greer–Thorbecke indices for measuring wage discrimination and occupational discrimination separately. Similar to the technique employed in the Brown–Moon–Zoloth decomposition [J. Hum. Resour. 15(1), 3–28 (1980)], we have employed a multinomial model for estimating the theoretical distribution of women in occupations, which would result in the absence of occupational discrimination.