Analysis of Navy Delayed Entry Program and recruit training center attrition

Abstract

Attrition from the Navy's Delayed Entry Program (DEP) and attrition from Bootcamp are costly phenomena. The Commander of Naval Recruiting (CNRC) and Center for Naval Analysis (CNA) have periodically modeled both DEP and Bootcamp attrition with logistic regression. This thesis analyzes current data provided by CNRC and CNA. Both DEP and Bootcamp attrition are modeled using logistic regression and tree-structured classification. For DEP, the logistic model indicates that individuals who accept incentives prior to enlistment (i.e., Navy College Fund or Enlisted Bonus Program) and individuals who change enlistment programs (while in DEP) have a significantly lower propensity to attrite from DEP than others. The DEP tree model indicates that an individual with a low Armed Forces Qualification Test (AFQT) score, no high school diploma and a long scheduled DEP duration has a 97% probability of attriting. For Bootcamp, the logistic model indicates that individuals who use tobacco products, individuals who do not exercise, and individuals that have criminal waivers have a significantly higher propensity to attrite than others. The Bootcamp tree model shows that smokers and individuals with low AFQT scores have higher propensities to attrite than others. The models are tested using random partitions and this analysis shows that all of the models predict poorly at the individual level, despite strong statistical significance.