A Logistic Regression Model to Predict Graduate Student Matriculation

Abstract

Higher education institutions invest a significant amount of resources every year to recruit new students. However, higher education administrators have been continuously facing challenges in enrollment management due to the demographic shifts, dramatic increases in educational costs, intense competition among institutions, and the uncertain nature of human selection patterns (Baum, Kurose, &McPherson, 2013).[3] Today's post-baccalaureate applicants are more knowledgeable than in previous years, because they can access information on a specific graduate program, in a given college, at any time. As reported in numerous studies, the number of graduate students switching out of their universities continues to be an essential issue. A useful prediction model of matriculation that uses available student data is highly desirable to assist the graduate students with timely advising early in their universities. This study was designed to build a predictive model for the probability that a specific admitted graduate student will matriculate. The results indicated that ten predictive variables were statistically significant at the .05 level. Getting an assistantship made the most substantial positive contribution in predicting student matriculation, followed by FAFSA, experience with the university, campus, degree level, college, gender, age, the number of days between application and admission, and distance to the university. This study's results could be beneficial for improving marketing efforts aimed toward individuals with characteristics most likely to enroll. Administrators could calculate the predictive score (or percentage) for each prospective student based on the predictive model. Marketing efforts could then concentrate on those applicants whose predictive score is high and eliminate the low qualifying students from their recruitment plan.