Fairness of Army ASVAB Test Composites for MOS and Job Families
Interim rept. Apr 2002-Aug 2003
ZEIDNER (J) AND ASSOCIATES BETHESDA MD
Pagination or Media Count:
Major goals of this research are to evaluate the fairness to minorities of the new operational ASVAB seven test composites and to compare the seven test composites with the previously utilized nine test composites. The experiment evaluates fairness in predicting performance outcomes as measured by the actual SQT Skill Qualification Test scores of individuals assigned to MOS by the operational system. Fairness is measured by the difference between predicted performance and SQT scores across gender and race. Complete fairness is indicated by very small mean difference and fairness to minorities is present when mean difference in the minority group is zero or has a positive sign overprediction. The sample uses ARIs data set of ASVAB scores limited to 66 MOS and SOT scores obtained during FY 1987-1989. The total sample size of first-term enlistees was 83,132. Findings indicate that the proposed new seven least squares estimate LSE test composites are comparatively fair to minorities. The results using seven test composites are comparable to the previous study using nine test composites.