•  
  •  
 

Abstract

This article provides an estimate of a model of student loan defaults using a rich panel data file. The file was constructed by merging administrative data on student loans, higher education enrollment and performance, and ACT test data for a large cohort of first time, full-time, degree-seeking students who entered Missouri two- and four-year public higher education institutions in the Fall 1992 semester. These loan recipients were tracked forward to December 1999 to determine which ones defaulted on their loans. The authors identify a variety of individual characteristics associated with loan defaults, however, the variable with the largest effect on the default odds ratio is continuous enrollment Within windows ranging from four to eight semesters, students who are continuously enrolled or who complete their program are far less likely to default than are students who drop out during the same period. The authors also assess the predictive power of their statistical model "out of sample” on a subsample of student borrowers and illustrate the potential use of the model in targeting default prevention resources to students most at risk of default.

Share

COinS