Subgroup differences in predictor and criterion variances and differential validity.

Investigated whether findings of differential validity could be explained by taking account of subgroup differences in predictor and/or criterion variances. The Thorndike equations for restriction in range were used to "correct" subgroup validities for both predictor and criterion for all race and sex subgroup data sets in which subgroup validities and standard deviations were available. The data indicate that corrections for differences in predictor variances had little effect on differences in subgroup validities. Equating criterion variances in those cases in which female validities exceeded male validities had the effect of virtually eliminating the difference in validity. It is suggested that findings of differential validity may be due to methodological problems associated with the criteria used, although in the data reported here corrections for differences in variances would increase the difference in male and female validities. (12 ref) (PsycINFO Database Record (c) 2010 APA, all rights reserved)