High personnel loss rates among recruits who have signed up for the Army's Delayed Entry Program (DEP) are becoming an increasing problem for DEP program managers. Therefore, a research project was conducted to examine DEP loss as a function of sociodemography and policy variables at the microdata level. Two DEP loss models were created. The first included high school graduates and nongraduates, and the second included high school seniors. In each model, several scenarios were staged to measure different combinations of the following variables: age, gender, Armed Forces Qualification Test (AFQT) score, education, contracted DEP length, military occupational specialty (MOS), region of the country, and enlistment and incentive options. A logistic regression (logit) model that estimates loss probabilities on an individual basis was specified, and categories of high- and low-risk individuals were identified. The multivariate analysis revealed that the following variables were associated with higher DEP loss: signing up for a longer DEP length, being female, choosing to participate in the Army College Fund, and being a high school graduate or nongraduate as opposed to a high school senior. The model was found to present a significant improvement over previous research because it permits measurement of the effects of changing several parameters simultaneously, thereby allowing for the prediction of DEP loss probability on an individual basis. (MN)