Multiple Prediction of College Graduation From Pre-Admission Data

Abstract

THE PURPOSE of this study was to investigate the practicality of predicting graduation of college students on the basis of information available prior to acceptance and matriculation. The subjects in the study were 659 men entering the University of Geor gia in September, 1959, without previous college ex perience. The data used included SAT scores, high school average, third of high school graduating class, number in high school graduating class, and the num ber of high school units earned in English, m at he matics, science, social studies, and foreign lan guages. The criterion used was whether or not the subjects graduated within five calendar years, that is, by August, 1964. Also used in a supplementary phase of the study were three variables which would not be obtainable until some time after the student entered college. These variables were first quarter grade average, first year grade average, and pro bationary status. The fact that the University of Georgia bases freshman admissions upon a predicted average ob tained through a multiple regression equation utiliz ing as predictors SAT-Verbal score, SAT-Mathema tics score, and high school average suggested to the writer the desirability of investigating a similar ap proach using probability of graduation instead of pre dicted first quarter grade average. The writer is not aware of any published studies which utilize the mul tiple correlation technique to predict college gradu ation or non-graduation. Ikenberry (4), Prediger (8), and Vorreyer (10) have carried out discriminant analysis studies in college level prediction. The ma jority of the predictive studies reported in the liter ature, however, deal with: either the prediction of college grade averages (1, 3, 6, 11); or a compari son of means or proportions of persisters and non persisters in which the predictor variables are dealt with one at a time (1, 5, 7, 9). There would be strong arguments in favor of ad mitting students at least partly on the basis of like lihood of graduation, if such could be done with rea sonable accuracy. This is not to suggest that grad uation or non-graduation represents complete attain ment or failure of attainment of personal or institu tional goals; then on the other hand, grade average at any particular point often tells us much less than we would like to know about a student's educational development. Any dichotomous criterion has certain disadvan tages including the inevitability of large errors in prediction. However, college graduation is a factor which influences both educators and students in an either-or fashion. For example, the person who al most graduated may find that society ascribes to him roles and status more closely resembling those of the person not attending college than those of the col lege graduate. A degree serves as a union card for many positions.