Trends in the relative distribution of wages by gender and cohorts in Latin America

Ana Hermeto, Cedeplar, UFMG

Most of previous works regarding gender wage differentials in Latin America are based on parametric methodologies for estimation of wage differentials by gender, using traditional mincerian regressions or quantile regression estimates, with classical or bayesian statistics. The aim of our paper is to analyze and decompose changes in earnings relative distribution between men and women in different cohorts, using the relative distribution framework. This methodology considers non-parametrical tools which allow an exploratory analysis that is independent of parametric assumptions on the mathematical form of the response-variable probabilities. We use density estimates of the kernel probability for each sex and cohort and decompositions of the relative distribution to get substantive evidences for gender differentials and relative mobility in Latin American countries (Brazil, Argentina, Uruguay, Mexico and Chile), from the 1980s to 2010s. We use microdata from the National Household Sample Surveys to analyze the wage differentials between male and female workers. We found out that, for each age group, women in more recent cohorts are better off in terms of the gender gap. Besides, we verified that each cohort experienced an improvement in the gender gap over the life cycle. For the oldest cohort, changes in the structure of the wage gap were preponderating, and for the youngest cohort, the change in the median gap was the main determinant of the upward mobility over its working life cycle. Finally, results point that changes in the marginal density of education, related to the wage gap (adjusting the relative distribution), were not crucial for the observed shifts in the wage gap between the cohorts, by age group, even when we consider that the more recent cohort is more educated than the earliest.