Measuring wage discrimination according to an expected utility approach

Following on from the seminal works by Blinder (1973) and Oaxaca (1973), many methods have been proposed to measure wage discrimination against women. Some of these methods focus on the entire distribution of the discrimination experienced by each woman, underlining a common aspect of poverty and discrimination analysis: the latter two are both based on an idea of deprivation which originates from a poverty line (in the case of poverty) and from the expected wage in the absence of discrimination (in the case of wage discrimination) (Jenkins, 1994; Del Río et al., 2011). These approaches hinge on conditional-to-individual-characteristics expected wages, lacking in any focus regarding the entire conditional wage distribution faced by each woman. In this paper we will discuss an expected utility approach to the study of wage discrimination. Adjusted and unadjusted for discrimination conditional-to-individual-characteristic wage distributions are evaluated for each woman by means of a utility function. And, in order to evaluate the presence and the discrimination intensity, these distributions will be compared on the basis of the respective certainty equivalent wages. As the choice of the utility function affects the results of the analysis, we will also evaluate the share of women for which the adjusted for discrimination conditional wage distribution second-order stochastically dominates the un-adjusted distribution. Finally, an empirical analysis will be performed for the Italian labour market.