NotesFAQContact Us
Search Tips
ERIC Number: ED490995
Record Type: Non-Journal
Publication Date: 2004
Pages: 35
Abstractor: Author
A Multinomial Logit Model of Attrition that Distinguishes between Stopout and Dropout Behavior
Stratton, Leslie S.; O'Toole, Dennis M.; Wetzel, James N.
Online Submission, Paper presented at the Annual Forum of the Association for Institutional Research (AIR) (44th, Boston, MA, May 28-Jun 2, 2004)
College attrition rates are of substantial concern to policy makers and economists interested in educational attainment and earnings opportunities. This is not surprising since nationwide, almost one-third of all first-time college students fail to return for their sophomore year. There exists a substantial body of literature seeking to model this attrition using simple logit specifications to differentiate between those enrolled and those not enrolled in the particular term of interest. Results from such analyses have been used to design policy responses to reduce attrition. While this literature assumes that all attrition is permanent, there is a growing body of evidence indicating that it is not. A substantial fraction of students who leave one institution do return to some institution of higher education within a short time. If this short-term stopout behavior is different from longer-term dropout behavior, the usual analyses will confound the determinants of these two outcomes and the policy responses currently in place will be poorly targeted. Our goal is to determine whether this is indeed a problem, whether the factors associated with stopout behavior are statistically different from the factors associated with dropout behavior. By providing a more realistic description of attrition, this paper makes a substantial contribution to the college enrollment literature. We explicitly recognize that there are two possible types of withdrawals: short-term stopout and long-term dropout. We use longitudinal data from the 1990 Beginning Postsecondary Survey to differentiate amongst those students who remain continuously enrolled through the first calendar year, those who stop out for a short period of time, and those who stay out on a long-term basis. Most importantly, we use a multinomial logit specification to model these three outcomes. This specification allows us to test whether the factors associated with stopout behavior are statistically different from the factors associated with dropout behavior. If they are not, then the standard approach to attrition analysis is appropriate. If they are, the approach tested here will provide a more accurate picture of attrition and will better predict enrollment behavior. This analysis can then help to better identify at-risk populations and design intervention programs. (Contains 5 tables and 10 footnotes.)
Publication Type: Reports - Research; Speeches/Meeting Papers
Education Level: Higher Education
Audience: N/A
Language: English
Sponsor: N/A
Authoring Institution: N/A