Credit Risk Assessment Using Survival Analysis For Progressive Right-Censored Data: A Case Study in Jordan

In credit risk management, the Basel Committee provides a choice of three approaches for financial institutions to calculate the required capital; standardized approach, Internal Ratings-Based (IRB) approach, and Advanced IRB approach. The IRB approach is usually preferred compared to the standard approach due to its higher accuracy and lower capital charges. The objective of this study is to use several parametric models (exponential, log-normal, gamma, Weibull, log-logistic, Gompertz) and non-parametric models (Kaplan-Meier, Nelson-Aalen) to estimate the probability of default which can be used for evaluating the performance of a sample of credit risk portfolio. The models are fitted to a sample data of credit portfolio obtained from a bank in Jordan for the period of January 2010 until December 2014. The best parametric and non-parametric models are selected using several goodness-of-fit criteria, namely MSE, AIC and BIC for parametric models and SE and MAD for non-parametric models. The estimated default probability is then applied to forecast the credit risk of a corporate portfolio at 99.9% confidence level and several time horizons (3 months, 6 months, 9 months, 1 year). The results show that the Gompertz distribution is the best parametric model, whereas the Nelson-Aalen estimator is the best non-parametric model for predicting the probability of default of the credit portfolio.