THE DEVELOPMENT AND APPLICATION OF AN INSTITUTIONAL DATA BASED RETENTION AND VOLUNTARY ATTRITION MODEL (COLLEGE STUDENT, WITHDRAWAL, DROPOUTS)

This study focused on the development and application of a model for studying college student persistence using readily available institutional data. Five groups of variables based on Tinto\u27s (1975) theoretical explanation of college persistence were studied: (1) background variables: high school rank, SAT scores, size of high school graduating class; (2) commitment variables: average credit hours completed per semester, number of times academic major was changed, evidence of college credit earned before full-time enrollment, pattern of enrollment as continuous or discontinuous and, if discontinuous, semesters completed before the student left and elapsed semesters before the student returned; (3) integration variables: ratio of semesters for which academic distinction was awarded, ratio or semesters for which academic probation was earned, ratio of semesters for which neither probation nor distinction was earned, and cumulative grade index; (4) control variables: school or initial enrollment and gender; and (5) the dependent variable disposition: completer or voluntary withdrawer. Institutional data for approximately 48,000 full-time, degree seeking students were collected at the West Lafayette campus of Purdue University. Data were collected cross-sectionally and longitudinally. Cross-sectional data examined student disposition as a withdrawer or completer at the beginning of the sophomore year. Longitudinal data examined student disposition after a period of at least five years and up to 15 years following initial enrollment. Multivariate statistical analysis of the data showed that significant differences existed between completers and withdrawers. Those differences supported Tinto\u27s explanation of college persistence: completers possessed background variables that demonstrated better academic skills (e.g., high SAT scores), were committed more firmly to the goal of earning a degree (e.g., averaged over 12 credit hours completed per semester) and were integrated more highly within the institutional environment (e.g., strong academic performance) compared to withdrawers. Cross-validation of discriminant analysis results demonstrated that second year status as a completer or withdrawer was predicted with 71% to 86% accuracy (M = 80%). Prediction of final status as a completer or withdrawer resulted in 63% to 83% accuracy (M = 75%). The model should be applied to data from other institutions to assess external validity. Continued research based on institutional data was suggested as a means of understanding better the process of college student persistence