ANALYSIS OF THE EFFECT OF CONDUCT WAIVERS ON U.S. ARMY ENLISTED SOLDIER FIRST-TERM ATTRITION

Abstract

In fiscal year (FY) 2018 the Army failed to meet its recruitment goals for the first time in over a decade. The missed recruiting goal suggests that the Army is facing an increasingly challenging recruiting environment. Compounding the Army's recruitment challenge is a first-term attrition problem. The purpose of our research is to gain insights on recruitment policies, specifically enlistment conduct waivers, and their effect on retention in order to best support the U.S. Army achieving its Department of Defense mandated end-strength. Our null hypothesis is that there is no difference in the first-term attrition rates of soldiers with conduct waivers compared to soldiers without conduct waivers. This study analyzes U.S. Army solder entry data from FY 2009 through FY 2013. Our five preliminary models are logistic regression, support vector machine, classification and regression trees, random forest, and adaptive boosting. We use confusion matrices and the Area Under the Curve metric to determine the model that most accurately predicts attrition. The random forest and logistic regression models are the best predictors. We fail to reject the null hypothesis and recommend the Army consider eliminating conduct waivers or making them easier to obtain in order to increase the pool of eligible candidates. Our analysis finds that Tailored Adaptive Personality Assessment System scores are important predictors of first-term attrition.