Contextual Factors Associated With the Validity of SAT Scores and High School GPA for Predicting First-Year College Grades

Abstract Prior research has shown that there is substantial variability in the degree to which the SAT and high school grade point average (HSGPA) predict 1st-year college performance at different institutions. This article demonstrates the usefulness of multilevel modeling as a tool to uncover institutional characteristics that are associated with this variability. The results revealed that the predictive validity of HSGPA decreased as mean total SAT (i.e., sum of the three SAT sections) score at an institution increased and as the proportion of White freshmen increased. The predictive validity of the three SAT sections (critical reading, mathematics, and writing) varied differently as a function of different institution-level variables. These results suggest that the estimates of validity obtained and aggregated from multiple institutions may not accurately reflect the unique contextual factors that influence the predictive validity of HSGPA and SAT scores at a particular institution. Notes aSAT is the composite (sum) of SAT–CR, SAT–M, and SAT–W. aThe correlations are corrected for restriction of range. bSAT is the composite (sum) of SAT–CR, SAT–M, and SAT–W. aThe correlations are corrected for restriction of range. *p < .05. *p < .05. aModel estimated with restricted maximum likelihood estimation; the remaining models were estimated under maximum likelihood. bThe robust standard errors for the final model are presented because some of the Level-2 variables showed departures from a normal distribution. cVariable was group-mean centered. dVariable was grand-mean centered. eVariable was divided by 100. *p < .05. **p < .01. *p < .05. **p < .01.