Breaking down the gender pay gap in it through a machine learning model

WAISBROT, SEBASTIÁN; EDELSZTEIN, VALERIA

Conferencia; I Women in Bioinformatics & Data Science LA conference; 2020

Women in Bioinformatics & Data Science LA

In recent years, interest in reducing the gender pay gap (GPG), that is to diminish the imbalances in earnings between men and women, has gained room in economic discussions all around the globe. Accurately measuring the GPG is important to assess how far we are from equality. In Argentina, women earn, on average, 27% less than men (1). This difference is observed in all occupational categories and the gap becomes greater when analyzing the hierarchical positions.However, the unadjusted GPG is a complex indicator. Although it provides an overall picture of the difference between men and women salaries, it does not take into account that this difference can be attributed not only to direct discrimination through ?unequal pay for equal work? but also related to many other factors, including the concentration of one gender in certain activities ('segregation'), the ease of access to higher paid hierarchical positions (?glass ceiling?), the total working experience and the number of working hours. Therefore, being able to decompose the GPG and to determine the contribution of its different components is important in order to design appropriate policies for reducing it. In this regard, some efforts have been made (2-3).In this work, we propose a decomposition approach based on a machine learning model in order to find out the value for the GPG among a population of 5742 IT-related workers and how much of that value can be exclusively attributed to direct gender discrimination.According to our analysis, there is a GPG of 20%, 8% of which can be explained exclusively by direct discrimination, while 12% can be ascribed to other factors, among which total years of experience, educational level and number of people in charge are the main contributors.