There is substantial variability in the degree to which the SAT and high school grade point average (HSGPA) predict first-year college performance at different institutions. This paper demonstrates the usefulness of multilevel modeling as a tool to uncover institutional characteristics that are associated with this variability. In a model that includes both HSGPA and SAT scores as predictors of FYGPA, the results suggest that higher mean SAT score at an institution is associated with slightly smaller validity coefficients for both HSGPA and SAT scores. The validity coefficients for HSGPA were larger at institutions with a larger percentage of white freshmen; and the predictive validity of the SAT was found to be higher at institutions with a larger percentage of students submitting SAT scores, and at smaller institutions. [Slides presented at the Annual Meeting of the American Educational Research Association (AERA) in 2010.]