Managing Student Loan Default Risk: Evidence from a Privately Guaranteed Portfolio

Abstract

This research models default development for a large proprietary dataset of private (nonfederally guaranteed) education loans extended to law school students in the early 1990s. Employing the statistical techniques of survival analysis and credit scoring, the study documents a pronounced seasoning effect for such loans and demonstrates the robust predictive power of credit bureau scoring of student borrowers. Other constructs found to be statistically predictive of default include school-of-attendance (or, alternatively, a measure of perceived school reputation), geographic location of attended school, and new attorney unemployment rate within certain regions. Although statistically predictive, these last constructs are of far less substantive importance in assessing credit risk than are the effects of portfolio seasoning and scoring (an ordinal measure of the risk of extending credit to an individual based upon their past credit behavior). The article challenges the prevailing approach to modeling student loan default (one that searches for “institutional” as well as “borrower” explanations) and suggests a return to the older, simpler banking paradigm of borrower willingness and borrower ability to repay.