A Model of Air Force Enlisted Retention: Theoretical Background, Implementation, and Estimates Using Conditional Choice Probability Estimation : RAND Corporation , September 30 , 2025
A Model of Air Force Enlisted Retention: Theoretical Background, Implementation, and Estimates Using Conditional Choice Probability Estimation : RAND Corporation , September 30 , 2025
From the report: “Discrete dynamic choice models are useful for analyzing military retention and the factors that affect retention, including compensation, but are difficult to estimate. This difficulty comes from the need to repeatedly solve stochastic, dynamic programming problems in the course of estimating the parameters of the model. This study addresses whether there might be a way to make it easier to estimate these models.
Research for this project involved exploring methods of analyzing retention that avoid having to solve dynamic programming problems, building on an approach pioneered by Hotz and Miller (1993), who showed that, under suitable assumptions, there is a well-defined relationship between the value that an individual assigns to the alternatives that they face (in this case, staying or leaving the active component) and the empirical probability that they choose one option over another (sometimes called a conditional choice probability [CCP]). This relationship allows expression of the expected value of the maximum of choosing to stay or leave as a simple function of the empirical probability of leaving. This stands in contrast to the RAND dynamic retention model (DRM) (Gotz and McCall, 1984), in which the calculation of the expected value of the maximum of the choice of staying or leaving in the next period is based on fully solving a dynamic program. However, the DRM requires less stringent assumptions than the Hotz-Miller approach, so it can be applied more widely.”
Authors - Mattock, Michael G.Related Resources
