Using Predictive Analytics to Understand Factors Affecting Transfer Student Persistence and Graduation

Abstract

It is the norm for institutions to report on their retention and graduation rates only for first-year student cohorts. Colleges and universities that report their first-year retention rates in the 90% range often do not account for their newly admitted transfer students. Much of the nuance in reporting retention comes from unaccounted transfer student registrations and enrollments. Reporting transfer retention is also much harder, since many transfer students do not have predictable patterns of enrollment. This study examined factors that contribute to graduation, dropout and persistence and how they differ by race, socioeconomic class, and gender. Based on a new student questionnaire conducted in 2015, 2016, 2017 by a large research institution in the Mid Atlantic, an exploratory statistical technique CHAID (Chi-Squared Automatic Interaction Detection) designed for a categorical dependent variable, was employed to establish the characteristics of transfer students who had a high probability to drop out after transferring to their new institution. Examining the dendrogram, one can easily classify the various “at-risk” student groups by tracing each of the terminal groups to the root of the tree. The results of this study provide context and information for developing transfer-friendly programming and interventions at both community colleges and four-year institutions. The results will be valuable to senior-level staff, front line student support staff, faculty, and community organizations focused on helping students who seek re-enrollment after an extended academic leave period. Additionally, this study will demonstrate how modeling techniques can be used to develop predictive models for different populations, across different colleges.