Ethnic and sex differences in the predictive validity of the scholastic achievement test for college grades

The predictive validity of SAT-Verbal and SAT-Mathematics scores for college grades was examined for males and females and for four ethnic groups in a nationally representative American sample. SAT scores "overpredicted" the grades of males and of Asians, Hispanics, and blacks, i.e., these groups did not obtain as good grades as would be predicted from their SAT scores. Conversely, SAT scores "underpredicted" the grades obtained by females and whites. i.e., these groups obtained better grades than would be predicted from their SAT scores. The possible explanations for these differential predictive validities are discussed.