Purpose: Berry et al.'s (J Appl Psychol 96:881-906, 2011) meta-analysis of cognitive ability test validity data across employment, college admissions, and military domains demonstrated that validity is lower for Black and Hispanic subgroups than for Asian and White subgroups. However, Berry et al. relied on observed test-criterion correlations and it is therefore not clear whether validity differences generalize beyond observed validities. The present study investigates the roles that range restriction and criterion contamination play in differential validity. Design/Methodology/Approach: A large dataset (N > 140,000) containing SAT scores and college grades of Asian, Black, Hispanic, and White test takers was used. Within-race corrections for multivariate range restriction were applied. Differential validity analyses were carried out using freshman GPA versus individual course grades as criteria to control for the contaminating influence of individual differences between students in course choice. Findings: Observed validities underestimated the magnitude of validity differences between subgroups relative to when range restriction and criterion contamination were controlled. Analyses also demonstrate that validity differences would translate to larger regression slope differences (i.e., differential prediction). Implications: Subgroup differences in range restriction and/or individual differences in course choice cannot account for lower validity of the SAT for Black and Hispanic subgroups. Controlling for these factors increased subgroup validity differences. Future research must look to other explanations for subgroup validity differences. Originality: The present study is the first differential validity study to simultaneously control for range restriction and individual differences in course choice, and answers a call to investigate potential causes of differential validity.
Bibliographical noteFunding Information:
Acknowledgments This project was supported in part by the Meredith P. Crawford Fellowship from the Human Resources Research Organization. We thank the College Board for providing the data for this project. We also thank Richard Landers and Haoyu Yu for invaluable computer programing support.
Copyright 2013 Elsevier B.V., All rights reserved.
- Cognitive ability tests
- Differential validity
- Range restriction