Predicting enlisted reenlistment rates

Abstract

Manpower management and retention has been an issue for the military since the military became an all-volunteer force in 1973. Annually, the Bureau of Personnel Metrics and Analytics Branch (BUPERS-34) predicts Navy reenlistment rates and sets numeric reenlistment goals for the upcoming fiscal year. These goals ultimately take into account end strength considerations as well as Enlisted Community Manager requirements. BUPERS-34 uses linear regression to forecast what the expected reenlistment rate will be, given current conditions; if no force shaping actions (e.g., reduce accessions, change personnel policies) are taken. If the forecasted reenlistment rate is different than requirements from an end strength/community management perspective, then the force shapers in the Manpower, Personnel, Training and Education Policy Division (N13) have a signal that steps may need to be taken to bring the two in line. In this thesis, the current BUPERS-34 Navy reenlistment prediction method is evaluated and alternative models to improve the prediction accuracy are suggested. Results of the analysis suggest the removal of several variables from the current model, due to lack of statistical significance, and the addition of Selected Reenlistment Bonus as a predictive variable for reenlistment.