Predicting Success in College: The Importance of Placement Tests and High School Transcripts

This paper uses student-level data from a statewide community college system to examine the validity of placement tests and high school information in predicting course grades and college performance. We consider the ACCUPLACER and COMPASS placement tests, using two quantitative and two literacy tests from each battery. We find that placement tests do not yield strong predictions of how students will perform in college. Placement test scores are positively—but weakly—associated with college grade point average (GPA). When we control for high school GPA, the correlation disappears. Placement test scores are positively associated with college credit accumulation even after controlling for high school GPA. After three to five semesters, a student with a placement test score in the highest quartile has on average nine credits more than a student with a placement test score in the lowest quartile. In contrast, high school GPAs are useful for predicting many aspects of students’ college performance. High school GPA has a strong association with college GPA; students’ college GPAs are approximately 0.6 units below their high school GPAs. High school GPA also has a strong association with college credit accumulation. A student whose high school GPA is one grade higher will have accumulate approximately four extra credits per semester. Other information from high school transcripts is modestly useful; this includes number of math and English courses taken in high school, honors courses, number of F grades, and number of credits. This high school information is not independently useful beyond high school GPA, and collectively it explains less variation in college performance. We also calculate accuracy rates and four validity metrics for placement tests. We find high “severe” error rates using the placement test cutoffs. The severe error rate for English is 27 to 33 percent; i.e., three out of every ten students is severely misassigned. For math, the severe error rates are lower but still nontrivial. Using high school GPA instead of placement tests reduces the severe error rates by half across both English and math.