ASSESSING CREDIT DEFAULT USING MULTINOMIAL LOGISTIC REGRESSION: EMPIRICAL EVIDENCE FROM A JOINT STOCK COMMERCIAL BANK

Abstract

This article has an aim to assess credit default prediction on the joint stock commercial bank using data collected from 120 bank credit profiles. The multinomial logistic regression is used to estimate the factors that influence the credit risk. The results show that polynomial logit performs better than binary logit. At the level of credit risk 1, the effects on risk signals include Collateral, financial capability of customers, diversified business operations, the experience of bank's staff and loan inspection and supervision. At the level of credit risk 2, the factors that affect the credit risk of the commercial bank are only four factors related to the signal, less than one such factor with the degree of credit risk 1, the collateral does not affect the level of credit risk 2. It provides several suggestions for risk management and policy implications to help mitigate credit risk.

Keywords: Credit risk; Multinomial logit; doubtful debt and risky debt.