擔保債券憑證評價模型之比較分析

Abstract

The objective of this thesis is to provide a comparative analysis of four representative models in pricing CDOs. The models compared in this thesis are the Gaussian copula, one-factor Gaussian copula, normal inverse Gaussian copula, and the defaultable-CRR models. The first three models are based on the factor copula pricing framework, but the last model is a characteristic combination of structural models and reduced form default models implemented on binomial trees. The effectiveness of these four models during the subprime crisis will be examined by comparing the fitness of these models with the market data and by evaluating the stability of parameter values over time. The market quotes of the tranche iTraxx Europe with five-year time to maturity are considered as examples for numerical evaluation in the thesis. The results indicate that the normal inverse Gaussian copula model outperforms other models before the crisis, especially for equity tranche, the 3~6 % tranche, and the most senior tranche, but after the crisis, the Gaussian copula model performs better. The performance of the defaultable-CRR model is poor due to the some fundamental problems.