Bank Lending Policy, Credit Scoring and the Survival of Loans

Abstract

To evaluate loan applicants, banks increasingly use credit scoring models. The objective of such models typically is to minimize default rates or the number of incorrectly classified loans. Thereby they fail to take into account that loans are multiperiod contracts for which reason it is important for banks not only to know if but also when a loan will default. In this paper a bivariate Tobit model with a variable censoring threshold and sample selection effects is estimated for (1) the decision to provide a loan or not and (2) the survival of granted loans. The model proves to be an effective tool to separate applicants with short survival times from those with long survivals. The bank's loan provision process is shown to be inefficient: loans are granted in a way that conflicts with both default risk minimization and survival time maximization. There is thus no trade-off between higher default risk and higher return in the lending policy.