An assessment of admissions strategy effects and self-selection issues in an enrollment study of Kettering University.

College administrators should understand the effects their unique admissions and financial aid strategies have on enrollment at their institution. Administrators should also determine whether the effect of a strategy is related to self-selection, meaning the effect may not be related to the student's experience of the strategy, but instead to some student characteristic that makes the student choose to experience the strategy. College administrators must also learn whether their recruitment strategies have differential effects across student types. In this dissertation, I study whether and how recruitment strategies affect enrollment behavior at Kettering University, whether those effects are related to the students' choices to participate in those strategies, and whether the effects differ by demographic and academic characteristics. I use human capital theory as a conceptual framework and estimate logistic regression models for the student choices of completing the application, visiting campus, submitting the FAFSA, and enrolling, and institutional choices of offering admission and awarding scholarships. I also estimate a multinomial logit model (MNLM) on the three enrollment choices---enrolling in July, enrolling in October, and not enrolling. The inclusion of a set of interaction terms between student attributes and the significant strategy variables suggests these activities do not have differential effects on enrollment by demographic or academic qualities. However, my review of the MNLM results suggests one type of campus visit might be related to self-selection, so I estimate an instrumental variable probit model using the number of chances a student has to attend that program as the instrument. The results indicate attendance is related to self-selection, meaning attendees are already likely to enroll prior to visiting. The enrollment model I create has considerable practical utility for the institution. Although the results are not generalizable due to the single-institution context, the conceptual framework and empirical approach are generalizable at many types of institutions. And, the analytic strategies I propose can be adapted for other types of higher education research, including alumni giving behavior, student retention, inquiry-to-application models, or any study in which participants make choices associating them with a particular group.Ph.D.EducationHigher educationUniversity of Michigan, Horace H. Rackham School of Graduate Studieshttp://deepblue.lib.umich.edu/bitstream/2027.42/127065/2/3328969.pd