Modeling Employee Flight Risk for American International Group, Inc.
Student Researchers: Ana Godonoga, Matthew Lambert and Allie Webb,
Faculty Advisors: Lisa Verdon (Business Economics) and John Ramsay (Mathematics)
The goal of this project was to develop a model that will allow AIG to predict, based on key variables, an employee's likelihood of leaving the company. The work toward this goal resulted in a variety of deliverables provided for AIG. First, the AMRE team cleaned and recoded several databases of uncombed data which were then merged into one master data file to be used for statistical analysis. In addition to this new "statistics-friendly" database, a data cleaning program was created in Microsoft Excel. This program gives AIG a tool that will speed the data cleaning process as they add new data for analysis. Third, some static snapshot analysis was performed on the new master data file. Key variables in flight risk prediction were determined through literature research and statistical tests. A few noteworthy findings were shared with AIG with recommendation to pursue this analysis further as more data becomes available. Finally, a mathematical model for predicting employee flight risk was created. The model is based on Probit analysis and assigns a flight risk probability to each employee based on key profile variables. AIG was provided a final set of recommendations for ways to improve their systems and data gathering techniques as they continue with their predictive analytics efforts.