Credit risk model of commercial banks.

• Main Authors: Lim, Serena Xiu Ling., Foong, Amy Qian Yu., Low, Xi Ting.
• Other Authors: Leon Chuen Hwa
• Format: Final Year Project
• Language: English
• Published: 2009
• Subjects: DRNTU::Business::Finance::Banking
• Online Access: http://hdl.handle.net/10356/15077
• Institution: Nanyang Technological University

In year 2008, the number of bank defaults in the U.S was 25, the highest recorded in nine years. In just the first two months of year 2009, the number of bank defaults was 16. The detrimental and domino-like repercussions of these defaults cannot be undermined. Common credit risk models to predict default probability are mainly tailored towards the manufacturing industry. A test of the more commonly used Edward I. Altman’s Z-score model on the banking industry gave faulty results, and further proved the unsuitability of these models for the banking industry. There is thus a need to develop a model that can accurately predict bank defaults in a stipulated time period, which was chosen as five years. By adopting the methodology of Altman’s Z-score model, the CAMELS framework for the selection of variables and 195 observations of default and non-default banks, a model was subsequently developed. The scores developed by the model were segregated into three zones; bankrupt, grey and non-bankrupt for better interpretation, where each zone has an accompanying probability of default. The validity of the model was tested through its variables with a Z-test and correlation test, and the final scores produced by the model with an in-sample test, out-sample test and a test of the adequacy of the five-year prediction period. Results showed that the variables were relatively independent and had significant discriminating abilities between the default and non-default banks. The in-sample test revealed a Type I error of 9.32% and Type II error of 0%. The out-sample test exhibited a Type I and II error of 10% each. A subsequent test using an alternative three-year prediction period was slightly superior with higher accuracy. However, because priority was accorded to Type II error, which was relatively unchanged from the five-year period, there was no sufficient need to revise the model. Through this model, banks can practice better risk management to safeguard its credit position, and effectively lower the number of bank defaults in years to come. Nonetheless, the inadequacy of the current sample paves the need for future research that also encompasses confirming the accuracy of current predictions and the aptness of the model across countries.