Akwum Onwunta

Essays on Credit Risk Modeling

Gutachter: Prof. Dr. Peter Winker, Prof. Dr. Bessler, Prof. Dr. Overbeck (FB 07)

Abstract: In a typical bank, risk capital for credit risk far outweighs capital requirements for any other risk class. Key drivers of credit risk are concentrations in a bank's credit portfolio caused by systematic factors that affect the credit quality of multiple borrowers. The objective of the first part this work is to analyze techniques for specifying default correlations through systematic factors. Our analysis is performed in the framework of a structural credit portfolio model. Here, we use asset correlations derived from default and rating time series for validating the calibration results based on equity data. In addition, we investigate whether non-Gaussian factor models replicate historical rating and default data better than normally distributed models. Next, we study the stability of the derived correlations under stressed market conditions. Moreover, we analyze the impact of the two popular rating philosophies - through-the-cycle and point-in-time ratings – on the asset correlations. The final part of the thesis studies clustering of obligors in a credit portfolio using Threshold Accepting, a local search optimization technique, which has recently performed reliably in credit risk clustering especially when considering several realistic constraints. Using a relatively large real-world retail credit portfolio, we propose a new computationally less complex technique to validate ex-post the precision of the grading system.