Predicting College Outcomes Using High School GPA and SAT Scores

• In Event: Methodology and Application in Structural Equation Modeling

Abstract

This paper advances the conversation about the use of assessments in the college admissions process. We have access to a novel dataset that allows us to follow 11th grade California public high school students to either the UCs or the CSUs. We reexamine the limitations of traditional models and propose new estimators based on statistical learning. We then use these new methods to show how much uncertainty remains in any prediction of college outcomes even after common explanatory factors are controlled for. We find that adding SAT scores to predictive models of college outcomes only minimally improves the accuracy of these models over models with high school grades alone.

Authors

• Kramer Anketell Dykeman, University of California - Davis

• Michal Kurlaender, University of California - Davis