Understanding the Impact of Socio-Economic Factors on Navy Accessions.

摘要

In a fiscally constrained environment, Navy Recruiting Command (NRC) must assign its recruiters to maximize the annual number of accessions by each recruiting station. Our thesis built on research in this area and made use of open source socio-economic data from several sources, including the Internal Revenue Service (IRS) and the Federal Bureau of Investigation (FBI). Beginning with a response variable of annual Navy accessions and a set of 71 independent predictor variables populated from ZIP code-level data, we fit and validated six predictive regression models. Models were fit using multiple linear regression (MLR) at the station level and zero-inflated negative binomial (ZINB) regression at the ZIP code level. We identified average number of recruiters, adjusted gross income (AGI) < $25,000, and total veterans as the principal drivers of accession production. We identified AGI > $200,000, unemployment compensation, and total number of universities in a ZIP code as the principal inhibitors to accessions. With out-of-sample data and using 95% prediction intervals, we tested the performance for each of the MLR models and validated them using the five assumptions of linear models. We tested the ZINB models against an out-of-sample subset using Mean Absolute Deviation (MAD) and true negatives, which verify the prediction rate of structural and random zeros. MAD and true negatives demonstrated improvement from previous zero-inflated Poisson models developed in 2011 by Y. K. Pinelis, E. Schmitz, Z. Miller and E. Rebhan, of the Center for Naval Analysis (CNA), in An Analysis of Navy Recruiting Goal Allocation Models.