Understanding Organizational Transitions: Interpretable Top Manager Turnover Prediction using a Heterogeneous Graph Neural Network
Year: 2024
Type: conference-paper
Abstract: Managerial turnover has garnered increasing attention with the growth of talent mobility and global competition. Top managers, being key figures in leadership groups, can significantly affect organizational performance, highlighting the importance of predicting turnover amongst them and understanding their transition patterns. Although extensive predictive modeling methodologies have been employed in previous research, most do not utilize the historical career trajectories of top managers and neglect the effects of transitions of co-workers. Additionally, works utilizing stronger predictive techniques typically do so at a trade-off of being able to extract interpretable insights from the constructed models. In an effort to enrich our understanding of top manager transitions, this work starts from the construction of a heterogeneous graph consisting of managers and the companies for which they have worked. Then, a graph neural network based model is utilized to predict potential turnover behavior. A novel interpretation framework capable of assigning attribution in terms of influence on the predicted outcome to all elements in the graph is then introduced to explain the factors affecting turnover in practice. The complete methodology is executed on a large data set stemming from the China Stock Market & Accounting Research Database (CSMAR).
Authors: Yameng Guo, Qiuzhen Ren, Seppe vanden Broucke
Institutions: Ghent University, KU Leuven
Language: English
Topic: AI and HR Technologies
Domain: Social Sciences
Open Access status: green
Funder: China Scholarship Council
Award: G1102879388
OpenAlex ID: W4407601635