Forecasting Marine Corps enlisted losses

Abstract

The Marine Corps has recently been authorized to increase end strength by approximately 20,000 Marines over the next 3 years. This has made forecasting of attrition an even more vital part of manpower planning. In order to successfully plan accessions to build the force we must be able to predict yearly attritions within the Marine Corps as accurately as possible. Because the enlisted force makes up the largest portion of the Marine Corps it is the most critical piece in accurately forecasting attritions. This research compared end of active service (EAS) losses to non-EAS losses (excluding retirement). It used logit regressions to forecast losses with some success. It is not the final word in forecasting but rather a proof of concept in predicting such losses. All three of the models that were used to predict losses for fiscal years 2005, 2006, and 2007 had misclassification rates below 22 percent. This logit technique uses the attributes found in the models to predict a Marine's probability of becoming an NEAS loss. This logit technique does not take averages across years to predict losses; rather, it finds the attributes that are more likely to be associated with NEAS loss according to the data. This research is the beginning stage of what can ultimately be a model that looks at entry level recruits' attributes with an eye toward predicting if they will become NEAS losses in the future.