Wisdom from the Crowd: Can Recommender Systems Predict Employee Turnover Decisions and Destinations?
Abstract
Can algorithms that predict your movie and shopping preferences also predict turnover likelihood and destination? Situated in a larger program of research on HR analytics applied to employee turnover, this research applies a type of machine learning technique, collaborative filtering (CF) recommender systems, to examine a key component of turnover theorizing, the interplay between an employee’s evaluation of their current job and that of the alternatives, as well as relationships with future turnover behaviors. Leveraging CF recommender systems, we adopt two operationalizations of the interplay, the quantity and quality of more desirable alternatives. The current study also applies CF recommender systems to predict the turnover destinations of employees prior to turnover. To achieve these two goals, we applied recommender systems on the National Longitudinal Survey of Youth 1979 dataset. Our results show that there is low error (i.e., residual) between the anticipatory satisfaction ratings estimated by the algorithms and employee self-reported job satisfaction, providing support to the construct validity of recommender system estimated ratings. The results also indicate that both the quantity and the quality of more attractive alternatives positively correlated with employees’ future turnover behavior. Moreover, the current study demonstrates that, in addition to helping researchers understand why employees quit their jobs to pursue other opportunities, CF recommender system algorithms can also facilitate researchers and practitioners predict where dissatisfied employees are likely to move.
Zoom link: https://eur-nl.zoom.us/j/91097113940
