Predicting student’s loan default in Kenya: fisher’s discriminant analysis approach

Abstract

The current high student enrollments in Kenyan universities has outstretched HELB in terms of loan disbursement. High default rate which stood at 43% as at 2012/2013 financial year have been the major challenge to HELB in meeting its core mandate of disbursing loans, scholarship and bursaries to needy students who have qualified to join local universities. The goal of this study was to develop a student loan default model that can predict if a new loan applicant is likely to be a defaulter or non-defaulter. This study examines characteristics of 7,354 loan borrowers from HELB between year 2009 and 2013. The study predictors were; age, gender, marital status, dependence, degree major, employment, loan awarded, family income, and bursary application, while the outcome variable was loan status (default or non-default). The findings showed that, employment status had the greatest discriminatory power in classifying the borrowers. This was followed by age, degree major (education), bursary application and gender in that order. The predicted model explained 36 percent of the variance in the discriminant function. In addition, the developed model was able to correctly classify 77 percent of the loan borrowers as either defaulters or non- defaulters. Interventions that would focus on the success of the student after college were seen as the main actions that would curb loan default.