Differential Prediction of FAA Academy Performance on the Basis of Race and Written Air Traffic Control Specialist Aptitude Test Scores

The written air traffic control specialist (ATCS) aptitude test battery was evaluated for evidence of predictive bias within the framework of the Uniform Guidelines on Employee Selection Procedures (29 CFR 1607) in a retrospective analysis. Step-down hierarchical regression analysis (Lautenschlager & Mendoza, 1986) was used to investigate differential prediction of performance in initial ATCS training at the Federal Aviation Administration (FAA) Academy in a sample of 282 African-American and 8,542 white first-time competitive entrants. Analysis based on correlations without corrections for restriction in range found significant differences in the intercepts, but not slopes, for African Americans and whites. Analysis based on correlations, corrected for explicit and implicit restriction in range, found significant differences in slopes and intercepts by race, suggesting that separate regression equations were appropriate to predict Academy performance for the groups. The two analyses indicated that the composite score on the written ATCS test battery exhibited predictive bias as defined by the Uniform Guidelines on Employee Selection Procedures (29 CFR 1607) and Cleary (1968). Specifically, the composite score TMC over-predicted the performance of African Americans in initial training at the FAA Academy. As a consequence of the over-prediction, significantly more of the African Americans that were accepted into training for the ATCS occupation on the basis of their aptitude test scores went on to fail training than would have been expected on the basis of the common or majority (white) regression line. An alternative explanation is considered that the observed differential prediction reflected criterion bias or other group differences in factors such as educational achievement and age. A path analytic approach is outlined for investigating the complex interactions between test score, the criterion, race, education, and age.