DEVELOPING A MARKOV MODEL FOR FORECASTING END STRENGTH OF SELECTED MARINE CORPS RESERVE (SMCR) OFFICERS

Abstract

This thesis develops Markov models for prior service (PS) and non-prior service (NPS) Selected Marine Corps Reserve (SMCR) officers. Data were collected from the Total Force Data Warehouse (TFDW) for all SMCR officers who served between September 30, 1998 and October 31, 2012. Determining SMCR officer end strength is necessary for reserve manpower planners to balance the force structure, minimize personnel excesses and shortages that impact training and labor costs, and to manage career progression. The PS model validation and analysis show that an aggregate monthly rate and unique monthly transition rates produce similar results. Both models perform well and they are consistent and accurate. Consistency and accuracy are important because budget planners and recruiting command rely on manpower estimates during the fiscal year. Overall, the aggregate monthly rate models perform slightly better than the unique monthly transition rate models with respect to end strength prediction, average strength prediction, and cost. More importantly, all four PS models performed better than the current Reserve Affairs model. We are unable to validate the NPS officer model. Since there are so few observations, the transition rates are suspect because they have a very high variance.