Determinants of voluntary turnover: A data-driven analysis for blue and white collar workers

Abstract

BACKGROUND: There is a growing interest in HR-analytics because of its ability to analyze employee behavior based on HR data. Predicting voluntary turnover of employees is an important topic of study, both in academia and industry. OBJECTIVE: The current study analyzes determinants for turnover, distinguishing between blue and white collar workers. The turnover analyses are based on a dataset from a payroll company, in contrary to previous turnover studies that used survey and interview data. METHODS: The studied dataset contains demographic and work specific factors for more than 380000 employees in 15692 Belgian corporations. Logistic regression is used to estimate individual turnover probabilities, the goodness of the model is tested with the AUC method. RESULTS: The study confirms turnover determinants and differences between blue and white collar workers that were described in previous work based on survey and interview data. Additionally, the study exposes so far unstudied turnover determinants and differences between blue and white collar workers. Confirmed determinants are among others age, seniority, pay and work distance. New determinants are company car, meal vouchers, night work and sickness. Different relationships to turnover are revealed for blue and white collar workers based on gender, number of children, nationality and pay. CONCLUSIONS: The presented dataset-based approach has its merit in analyzing turnover: it enables to study actual turnover instead of turnover intentions, and reveals new turnover determinants and differences between blue and white collar workers.