The performance of structural credit risk models on commercial mortgages: An empirical investigation

Depended on the option-pricing framework, this study provides a theoretical approach to estimate the default probabilities of commercial mortgages and an empirical test on the relation between default triggers and actual default behavior. In particular, this study focuses on the model comparison between the performance of a single trigger model and a double trigger model. Distinct from previous studies on commercial mortgages, this study introduces two new methodologies to real estate finance: the first-passage-time approach and the Receiver Operating Characteristic (ROC) approach. The merit of the first-passage-time approach lies in that it can capture the fact that default occurs at any time before maturity. On the other hand, the ROC approach helps risk managers not only to visualize the model performance but also to provide a rigorous test on model prediction accuracy.; Using 17,616 lockout commercial loans issued between 1995 and 2001, this study finds that the property value model provides the most accurate default prediction among all three structural models. This evidence implies that the loan-to-value (LTV) ratio is the major variable to explain the occurrence of default in commercial mortgages. In addition, this study does not find definite evidence to reject the null hypothesis that the ROC ratio of the double trigger model is significantly different from that of the single trigger model. Namely, the cash flow hypothesis is not fully supported based on the output of structural models.; In sum, this research investigates the validity of structural models and tests the cash flow proposition. The results show that the structural models do have better discriminative power than a random model. On the other hand, despite the fact that the double trigger model outperforms the cash flow model, the evidence that the insignificant difference between the prediction accuracy of the double trigger model and the property value model does not support the cash flow proposition.