Predicting Different Levels of Academic Success in College Using High School GPA and ACT Composite Score

esearck Report Semes 200Predicting Different Levels of Academic Success in College Using High School GPA and ACT Composite ScoreJulie Noble Richard SawyerACTAugust For additional copies write: ACT Research Report Series RO. Box 168Iowa City, Iowa 52243-0168© 2002 by ACT, Inc. All rights reserved. Predicting Different Levels of Academic Success in College Using High School GPA and ACT Composite ScoreJulie Noble Richard Sawyer AbstractThis study compared the effectiveness of ACT Composite score and high school GPA for predicting different levels of first-year college GPA. Logistic regression models were estimated for predicting first-year GPA levels ranging from 2.00 to 3.75 for a sample of postsecondary institutions. The prediction models estimated from one year’s data were then applied to data from the next year (crossvalidation over time). The resulting statistics on prediction accuracy were summarized across institutions.Both high school GPA and ACT Composite score were effective in predicting success at the2.00, 2.50, and 3.00 levels of first-year GPA; high school GPA was somewhat more accurate than ACT Composite score at these levels. High school GPA was not an effective predictor of success at higher levels of first-year GPA, however. For example, even a 4.00 high school GPA corresponded to very low probabilities of success at the 3.25, 3.50, and 3.75 levels of first-year GPA. Moreover, high school GPA values below 3.00 provided little differentiation among students across first-year GPA levels. ACT Composite score predictions, in contrast, were effective at all first-year GPA levels.ii Predicting Different Levels of Academic Success in College Using High School GPA and ACT Composite ScoreCollege admissions officials typically use both high school GPA and scores on college entrance tests (such as the ACT Assessment) to predict, formally or informally, an applicant’s probability of academic success in the first year of college (Breland, H., Maxey, J., Gemand, R., Cumming, T. and Trapani, C., 2002). Academic success is typically measured by first-year college GPA.Although high school GPA and college GPA both measure educational achievement, they also include other personal characteristics such as effort, attendance, conformity, and motivation (Goldman & Widawski, 1976; Stiggins, et al., 1989). In contrast, ACT scores primarily measure educational achievement in college-preparatory courses (ACT, 1997c). One might therefore expect high school GPA and ACT scores to be related to first-year GPA in different ways: High school GPA likely relates to both the cognitive and the noncognitive components of college GPA. ACT scores, on the other hand, likely relate only to the cognitive components of college GPA.Goldman's and others’ research in the 1970s found that the presence of non-achievement components in college grades was related to average student ability (e.g., Goldman & Hewitt, 1975; Goldman, Schmidt, Hewitt, & Fisher, 1974; Goldman & Widawski, 1976). They found that high college grades were more likely to reflect cognitive achievement and less likely to reflect noncognitive factors. One might therefore expect that• Predictions of moderate levels of college GPA (e.g., 2.00) based on high school GPA are more accurate than predictions based on ACT scores, but• Predictions of high levels of college GPA (e.g., 3.00) based on ACT scores would be more accurate than predictions based on high school GPA. The research comparing predictions of first-year GPA from ACT scores or high school GPA has largely been correlational in nature (e.g., ACT, 1998d). For example, recent summary statistics for institutions participating in ACT's Research Services showed that, across 129 colleges, the median multiple correlation (across institutions) between first-year GPA and the four ACT scores (in English, mathematics, reading, and science reasoning) was .43 (Maxey, in press). The corresponding median correlation between first-year GPA and four high school average grades was .48. The eight-predictor ACT score/high school grade average multiple correlation with first-year GPA was .53. These results were not corrected for range restriction caused by selection on ACT scores and/or high school grades.Other research has examined prediction accuracy at first-year GPA levels of 2.00 or higher or 3.00 or higher (ACT, 1997c). This study showed that high school GPA was slightly more accurate that ACT scores in predicting whether students earn a 2.00 or higher GPA in the first year of college. For predicting whether students earn a 3.00 or higher college GPA, however, ACT Composite score and high school GPA had the same accuracy. The typical percentage of accurate predictions was 79% using either