Show Me the Money: A Machine Learning Analysis of Army and Marine Corps Continuation Pay for Optimizing Incentive Timing

Abstract

Continuation Pay (CP), introduced in 2018 under the Blended Retirement System (BRS), is a mid-career retention incentive that may be offered between 8 and 12 years of service (YOS), yet most services offer it only at 12 YOS. Despite this flexibility, little empirical research evaluates the optimal timing of CP, and no studies use observed CP take-up behavior to assess alternative timing policies. This thesis addresses that gap by applying machine learning (ML) methods to historical Army and United States Marine Corps (USMC) CP data to estimate the effects of expanding CP eligibility from a single 12 YOS offer to a discretionary 8–12 YOS window. Using Army CP decision data to train an ML model, projections are generated for USMC personnel under both 8–12 YOS and current 12 YOS policies. Assuming comparable decision behavior across services, the 8–12 model increases projected end strength by 5,709 Marines within a fully BRS population between 8 and 16 YOS. Annual take-up is projected at 4,247 Marines, with costs of $58 million at a 2.5 multiplier and $121.9 million at a 5.0 multiplier. Retention gains exceed CP costs on a per-dollar basis at both multiplier levels. If retention shortfalls exist within the 8–12 YOS window, adoption of an 8–12 YOS CP policy is recommended, particularly before fully auto-enrolled BRS cohorts dominate the mid-career population. The 8–12 model mitigates mid-career retention drop-offs by aligning incentives with individual career decision points.