Predictive Analytics Approach to Evaluate Wage Inequality in Engineering Organizations

Abstract

Wage inequality is a source of many social and economic problems, and is the target of mitigating programs both nationally and internationally. The primary step toward developing effective programs to reduce or eliminate wage inequality is identifying employees at risk of such inequalities. This study used 17,889 data points from USDOT workforce demographic information and salary data to analyze wage inequality and develop a novel framework to identify employees at risk of wage inequality. The evaluation framework includes (1) a salary prediction model, developed using artificial neural networks (ANNs), to estimate employees’ salaries based on demographic information and identify underpaid employees; (2) a minority index, which is defined to score the underrepresentation of each employee regarding gender, ethnicity, and disability, based on the current status of employee diversity in the organization; and (3) a decision model, which uses the salary prediction model and minority index based on historical data to determine if new employees are at risk of wage inequality. The analysis showed that although women are underrepresented among USDOT employees, there was no significant wage inequality between men and women. Furthermore, the lowest minority index was for White men without disability, and the highest for American Indian/Alaska Native women with disability. In addition, the results of evaluating the proposed framework had an accuracy of 98%, with a harmonic mean (F1) score of 81.8%. The framework developed in this study can enable any engineering organization to establish an unbiased wage rate for its employees, resulting in reduction or elimination of wage inequality and its consequent challenges among its employees.