Accession Number:



Retention Analysis Model (RAM) For Navy Manpower and Personnel Analysis

Descriptive Note:

Technical Report,01 Oct 2017,31 Dec 2018

Corporate Author:


Report Date:


Pagination or Media Count:



This report addresses deficiencies in our understanding of service members career trajectories. The insights generated will be used to construct more sophisticated and useful models of long run manpower projections, allowing complex simulations to predict the impact of personnel policy changes. This will allow Navy leadership to avoid unanticipated shocks to service member supply and quality. This report proceeds along two lines. First, we collect a dataset of Navy officers and examine their career trajectory, paying particular attention to their educational background and sociodemographic characteristics. Using long-term trend, as well as regression analysis, we find significant retention rate differences over the long run across gender, marital and dependent status, race, and education level. While the long run trends and regression results are illuminating, we should be wary of drawing definite conclusions about the innate ability or desire of officers to stay or separate based on these analyses. Without a formal model to distinguish between correlation and causation, we should recognize that the findings in this study primarily help direct our modeling efforts in subsequent years. Second, we provide an in-depth description of dynamic programming models, demonstrating their usefulness and internal consistency for predicting rational, forward-looking agents making choices that affect their future. We provide a detailed technical description of the model, defining value functions, Bellmans equations, and other concepts necessary to program, estimate, solve, and simulate a dynamic programming model. We then propose the path forward to examine how service members in different communities may make different career choices.

Subject Categories:

  • Personnel Management and Labor Relations

Distribution Statement: