Improving the Classification Efficiency of the Armed Services Vocational Aptitude Battery Through the Use of Alternative Test Selection Indices

Abstract

Abstract : This study examines procedures for selecting tests for a battery as one means of improving the classification efficiency and potential economic benefit of the Army's current operational test battery. Five test selection methods, each representing competing measures of selection and classification efficiency, are compared under varying conditions expected to affect classification efficiency. The research approach adopted involves a simulation of the Army selection and classification process based on the coverage of Army jobs provided by the Project A data. Comparisons of the classification efficiency produced under each test selection methodology are reported in terms of the average mean predicted performance (MPP) produced by each simulation. The results confirm the predictions of Zeidner and Johnson (1989) that the use of a classification efficient test selection procedure can improve the utility of the Army assignment process. The use of a method which maximizes potential classification efficiency results in as much as twenty percent gain in MPP over use of a method which maximizes predictive validity when used for a five-test battery. In addition, doubling the number of jobs used in the simulation results in a performance gain of ten percent. This gain is an underestimate of what we would predict from increasing the number of classification efficient predictor composites or aptitude areas (AAs) corresponding to effectively reconstructed job families. Capitalization on disparate means and variances across jobs (hierarchical classification efficiency) provides no greater benefit for classification efficiency than is possible from a pure allocation efficient assignment strategy.