Predicting the paycheck: using machine learning to understand determinants of income

Abstract

Income is a variable of interest in industrial/organizational psychology due to its relationship with outcomes like turnover, motivation, and psychological well-being. However, current research on income has generally assumed a linear relationship between predictors and income, not accounting for potential curvilinear effects or variable interactions. Further, studies on income indicate that large amounts of variance are unaccounted for, suggesting there are predictors yet to be identified. This study addresses those gaps in the research by using machine learning techniques and a large archival data set to investigate the strength and nature of how variables contribute to predicting income. Results demonstrate the effectiveness of machine learning techniques over traditional OLS regression and identifies variables not found currently in the literature. Findings from this research can be used both to create more effective organizational compensation systems as well as indicate targets for interventions to address income inequality.