Optimal Army Officer Retirement

Abstract

We model an optimal stopping problem describing the decision of an individual officer to retire from the Army. Our model incorporates the High Three retirement system and current compensation. We solve the resulting Markov decision process model by dynamic programming and investigate model sensitivities. Previous results from similar models suggested that the due course officer would reach an optimal retirement state at 23 years if passed over for promotion to Colonel, and at 26 years if selected for promotion. Our results suggest that the change in compensation structures and the retirement system have shifted the first optimal retirement state for most officers to be as soon as vested in retirement benefits, currently at 20 years. We explore manpower planning and policy implications of shifting retirement behavior and sensitivity of our model to altering model assumptions.

Keywords: Markov Decision Processes, Dynamic Programming

JEL Classification: C44, C61, J26