PREDICTIVE VALIDITY OF THE ACT TESTS AT SELECTIVE COLLEGES ABSTRACT Three studies, each dealing with an aspect of comparative validity of ACT and SAT at selective colleges, are included. The first study considered the predictive efficiency of the ACT test scores and ACT test scores plus high school grades at 120 colleges, separated into three groups according to average college ACT Composite. Predictive efficiency was not found to vary appreciably at various points on the score scale. For 40 colleges where the mean ACT Composite was 24.5 or higher, the median multiple R using the ACT test scores was .46 and using the ACT test scores plus high school grades was .58. By contrast, at the 40 colleges where the mean ACT Composite was 15.5 to 20.0, the median multiple R using the ACT test scores was .46 and using the ACT test scores plus high school grades was 56. The difference is negligible. The median correlation of .58 using ACT test scores and high school grades at selective colleges compares favorably with .54 reported in CEEB materials as the median correlation using SAT test scores and high school rank at colleges defined by the same level of selectivity. Further, when the 40 colleges with the high mean ACT Composites are analyzed more closely, there is not a tendency for the predictive efficiency of ACT test scores and high school grades to decline as college mean ACT Composite goes up. The second study concerned four selective colleges where all students had taken both the ACT and the SAT. At all four of these colleges, the mean ACT Composite score was above 24.5; the mean SAT Total score was above 1200. In all four cases, the ACT test scores gave a better prediction of freshman overall Grade Point Average than did SAT. The median R with ACT was 407, with SAT, .316. The third study took place at the U.S. Air Force Academy, a highly selective institution. A focus of concern in this study was to correct for selection on one of the tests, in this case the SAT, by designing the study in such a way that prior selection of students by the SAT would not be an extraneous factor in the analysis of the comparative predictive validity of ACT and SAT. Using the conventional formulas for correction, the comparison of correlation coefficients again favored the ACT over the SAT, .56 to 52 for Sample 1, and .51 to.43 for Sample 2. Finally, when CEEB Achievement tests in English and mathematics were added to the two SAT scores in a multiple, the R obtained was comparable to or below that achieved by the ACT tests alone. The evidence in the three studies points to the conclusion that ACT and SAT scores typically yield similar results at selective colleges, and where they do not the ACT is usually favored with higher correlation coefficients. PREDICTIVE VALIDITY OF THE ACT TESTS AT SELECTIVE COLLEGES Oscar T. Lenning' When The American College Testing Program (ACT), with its college admissions and guidance battery, was introduced at the beginning of the last decade, some college educators were concerned about whether this new battery would predict freshman grades as well as older examinations such as the Scholastic Aptitude Tests (SAT). The consensus of studies published since that time, however, has been that the ACT Assessment predicts grades for typical college populations generally as effectively as or better than the SAT battery (Boyce & Paxson, 1965; Burns, 1964; Chase, et al., 1963a, 1963b; Lins, Abell, & Hutchins, 1966; Lenning & Maxey, 1972; Munday, 1965; Passons, 1967; Zimmerman & Michael, 1967). Furthermore, by 1971 more than 2,000 higher education institutions were participating in the ACT Program (The American College Testing Program, 1971, p. 3). In spite of the general acceptance of the ACT Assessment, one question has not been answered to everyone's satisfaction. Although they would agree that ACT is as efficient a predictor as SAT for typical colleges, it has been the subjective contention of some that ACT should not predict as well as SAT for highly selective colleges having a preponderance of students with exceptional academic ability. There has been no objective evidence to support such a belief; rather it has been based on the fact that, unlike the SAT, the ACT was not specifically designed for use by highly selective colleges. It was the purpose of this project to collect all available objective information bearing on the question of comparative predictive validity of the two tests at selective colleges. Three separate studies provide such data, and they are summarized here. The first study presents correlations typically obtained with ACT data at colleges having quite different ability levels. This not only addresses the question as to whether validity varies appreciably at various points on the score scale, but also permits comparison with validity figures on the SAT reported by Angoff (1971). The second deals with four selective colleges where all students had taken both the ACT and SAT, and comparison of predictive validities was thus possible. The third is a case study of one selective institution, the U.S. Air Force Academy. In this study corrections were made for selection, so that this extraneous factor would not bear on the results. 'The author gratefully acknowledges several people who helped with this report. First is Risdon J. Westen of the United States Air Force Academy, coauthor of the third study reported here. Second are Nancy S. Cole, Leo A. Munday, and E. James Maxey, all of the ACT Research and Development Division. Formerly a member of the ACT Research and Development staff, the author is now Senior Staff Associate with the National Center for Higher Education Management Systems/Western Interstate Commission on Higher Education. |