Degree Type


Date of Award


Degree Name

Doctor of Philosophy


Educational Leadership and Policy Studies

First Advisor

Larry H. Ebbers


The study examined the efficacy of three independent variables (ACT, ACT-index, and SAT-index) in predicting the first-to-second-year retention of undergraduate college students. The data were gathered by, and obtained from, ACT, Inc. and represented all students who sat for the ACT Assessment during the 1999 test administration period and whose higher education institution participated in a retention study conducted by ACT, Inc. A large sample size (n = 39,216), composed of approximately equal representation of retained and not retained students, was selected from an original database of 87,915 cases.;The primary purpose of this inquiry was to determine if the computed variables, ACT-index and SAT-index, predicted between-year retention as well as the ACT composite score. Each index measure represented how well each student performed on the standardized assessment compared to the average performance of his or her high school classmates. These indices allowed for quick identification of those students who performed better than average compared to their peers. Some higher education policy makers propose the use of merit-indices in predicting retention as a way to increase diversity and combat anti-Affirmative Action trends. These policy makers believe merit-indices account for differences in academic background that hinder the success of some students.;The researcher employed logistic regression analysis as the primary statistical technique in this research. Logistic regression procedures regressed the dichotomous dependent variable, retention, onto several predictor variables. Simple regression models, each including only the three measures of interest (ACT, ACT-index, and SAT-index), and backward stepwise logistic regression models, including all identified variables, were utilized to explore the data. Z-score transformations and change-in-chi-square analyses were used to compare across regression models.;All three variables of interest were significant predictors of retention. Odds ratio measures indicated that the ACT variable more powerfully predicted retention, followed by ACT-index. The influence of the SAT-index measure on retention was negligible. When the model was split by racial category, however, the ACT and ACT-index variables significantly predicted retention of Caucasian and African-American students, but failed to significantly predict retention for the Asian American, Hispanic, and Multiracial/Other racial groups. The SAT-index variable was not a significant predictor of retention when the model was split by racial categories.;Findings indicate that the merit-indices should not replace the ACT composite score as a predictor of retention, especially for non-Caucasian and non-African-American students. The ACT composite score appeared to predict retention better than either of the computed merit-indices.



Digital Repository @ Iowa State University,

Copyright Owner

Robert Dean Reason



Proquest ID


File Format


File Size

92 pages