Estimating proportional hazards in default and prepayment of personal loans with unobserved borrower heterogeneity

This paper utilizes a statistical model of competing risk proportional hazards to study default and prepayment in unsecured personal loans. The model allows for time-varying covariates and accounts for interval censoring and unobserved borrower heterogeneity. A simulation experiment with sixteen scenarios shows that the estimator yields nearly unbiased estimates. Application to a dataset of unsecured personal loans provides time-dependent prediction in default and prepayment, especially when payment history is incorporated.