Associations between majors of graduating seniors and average SATs of incoming students within higher education in the U.S.

摘要

While Scholastic Assessment Test (SAT) scores are becoming less widespread than in previous years, they continue to be used as an input by many higher education institutions in the United States to select which students to accept among applicants. This paper explores the association between average SAT scores of incoming undergraduate cohorts and major completions of graduating student cohorts. College Scorecard data from 2019 is collected from all U.S. undergraduate degree-granting, higher education institutions reporting average SAT scores of incoming cohorts (n=1,389). A multivariate beta regression approach, which allows for overdispersion and unit-interval responses, is proposed to explore associations between graduation rates by major (explanatory variables) and SAT percentiles of new student cohorts (response). Forty-nine percent of the variability in average SAT percentiles of incoming cohorts can be explained by the graduation proportions by major within institutions. Results show strong concurrent positive associations between average SAT percentiles of incoming cohorts and proportions of students graduating in: STEM fields; ethnic, cultural, and gender studies; social science; or languages, among others (p<0.01). A negative association is found between average SAT percentiles of incoming cohorts and graduating cohorts in degrees like security law enforcement or parks & recreation and fitness, as well as some traditional major choices, such as theology and psychology (p<0.01). Results are consistent by institution size, as well as public versus private, across most clusters. A statistical framework is introduced for analysis of the expected impact on average SAT percentiles of future student cohorts derived from changes in proportions by major of graduating student cohorts. Higher education institutions can benefit from the proposed methodology by adjusting their degree offerings to their target cohorts. While illustrated using SAT scores due to their historical prevalence and availability across institutions, the proposed approach can utilize any alternative quantitative measure of student preferred characteristics.