Determining market categorization of United States ZIP codes for purposes of Army recruiting

Abstract

The Army relies on Zone Improvement Plan (ZIP) codes to assign recruiters and to track recruit production. ZIP codes have different densities of potential recruits; the Army uses commercial market segmentation data to analyze markets and past accessions to assign recruiters and quotas to maximize production. We use 347 variables from publicly available United States government agencies for each of 34,007 ZIP codes to cluster ZIP codes into similar groups. We use between 2 and 18 clusters for each of five categories of data, using three dissimilarity calculation methods, and three clustering algorithms. Using national recruiting leads as a proxy for market potential, we find the best cluster assignment by fitting Poisson regressions predicting leads from ZIP code cluster membership. Economic cluster assignments predict leads with a pseudo R-squared value of 0.69, reducing the need for United States Army Recruiting Command to rely on proprietary data with 66 market segments per ZIP code for market analysis and predicting recruiting potential. These 18 clusters provide an easier tool for recruiting commanders. Additionally, these clusters offer a new method of identifying potentially high-production ZIP codes without using previous accessions and the highly correlated number of recruiters assigned as predictor variables.