担保债权凭证之评价——多因子模型和KMV模型之探讨

探讨担保债权凭证商品之评价，包含缩减式模型及结构式模型两种研究方法；前者以多因子相关性模型，而後者以KMV模型为方法探讨之主轴。多因子相关性模型中资产的违约分配函数分别假设为指数、韦伯及Burr分配；再分别结合Gauss Copula或t5 Copula函数，估计商品的信用价差。实证分析以台湾“玉山银行债券资产证券化特殊目的信托2005—1受益证券”为例。实证研究结果发现，指数分配之信用价差估计值偏大，Burr分配估计值最小；t5 Copula函数之信用价差估计值都较Gauss Copula函数之估计值大。此外，将数据作适当调整後应用KMV模型之信用价差估计值比多因子相关性模型之估计值大。

Pricing collateralized debt obligation via Multi-Factor Model and KMV Model

This paper investigates the pricing premium of Collateralized Debt Obligation, via the Reduced Form Model and the Structural Form Model; the former focused on the multi-factor correlation model, and the latter focused on the KMV model, both are the core investigations. The asset default distribution functions for the multi-factor correlation model are assumed to follow Exponential, Weibull or Burr distributions; combining with the Gauss Copula or the t5 Copula functions, the credit spreads of the CDO are estimated. An empirical analysis is applied on the Taiwan ＂E. SUN bank debt securities assets of the Trust 2005-1 beneficiary securities＂. The empirical results indicate that the credit spread estimates under the Exponential distribution are the largest and that from the Burr distribution are the smallest; results of the t5 Copula function are greater than that obtained from the Gauss Copula function. Besides, after re-arranging data, the credit spread estimates under the KMV model are larger than that under the multi-factor correlation model.