Using available non-demographic data for default probability estimation in automobile loans: Case of an Indian NBFC

The objective of this paper is to test the possibility of developing a model to calculate the default probability in car loans in the context of Indian Non Banking Financial Institutions (NBFCs) where detailed customer demographics are usually not maintained or not freely available to line managers. The study attempts to check the utility of the non-demographic variables already available with the organization with an objective to immediately apply the model developed in a scenario when relationship lending has reached its limits and transactional lending is the need of the hour. Said data for borrowers of an Indian NBFC was analysed using the Binary Logistic Regression Model for a cohort of 10,137 customers who availed loans from April 2003 to March 2007. Customer loan performance was observed from April 2008 to ensure at least 12 month seasoning. The results suggest that ten out of eleven recorded variables significantly impact the loan defaults. Using the proposed variables to develop a credit model has potential benefit to smaller companies where data capturing is still not up to the mark and only data available is that pertaining to loan characteristics.