The effects of number of predictors, predictor intercorrelations, vali
dity, and level of subgroup difference on composite validity adverse i
mpact ratios, and mean subgroup difference associated with various pre
dictor composites, including and excluding a ''high impact'' ability m
easure, were assessed. The size of subgroup differences is substantial
ly smaller when low-impact predictors are combined with a high-impact
predictor, but hiring ratios for majority and minority groups still in
dicate a prima facie case of discrimination, using the fourth-fifths r
ule for most predictor-criterion combinations. However, the validity o
f a composite of alternate predictors and cognitive ability may exceed
the validity of cognitive ability alone and reduce the size of subgro
up differences.