摘要
本研究提出直觀分析法，解析標的資產的違約對雙層擔保債權憑證的損失情形，透過此方法可以直接引用單因子聯繫結構模型建構雙層擔保債權憑證的損失分配。另外，為了模型能夠符合真實市場經濟狀況，我們考量Normal Inverse Gaussian (NIG)分配，加入隨機回復率、隨機相關性的假設，建構成不同型態的模型。求得可靠的一般擔保債權憑證之評價模型，套用至評價雙層商品仍然可得合理的信用價差，進而分析商品真實的風險特徵。隨即為了改善直觀分析法的效率，以及提升適用性，本研究提出抽樣方法來解析多資產下的雙層擔保債權憑證的損失情形。 研究結果顯示直觀分析法可以準確掌握雙層擔保債權憑證的損失情形，依此可求得準確的損失分配，以及合理的信用價差。另一方面，使用抽樣方法取得部分的損失情形，用來估計一個損失分配，進而求得的信用價差與直觀分析法的結果相當接近，我們透過數學歸納法驗證，推測使用抽樣方法評價雙層擔保債權憑證仍然可得合理的信用價差。
並列摘要
This study proposed an intuitive analysis method to calculate the loss distribution of the underlying assets for CDO Squared. This method can be derived from one factor Gaussian copula model which was used to construct the loss distribution of regular CDO. In addition, in order to fit the market quotes for CDO products, we extend the model to consider the Normal Inverse Gaussian distribution, stochastic recovery, and stochastic correlation. The reasonable spreads of CDO Squared were obtained by using the best parameter estimates in pricing regular CDOs. And then, we analyze the real risk characteristics of CDO Squared. Followed by propose sampling method for constructing the loss distribution for higher dimensional asset pool in CDO squared so that computational efficiency can be enhanced. On the other hand, we used partial cases of all possible loss situations to derive the loss distribution for CDO squared, and then obtained credit spreads for it. The result of sampling method is very close to that for the intuitive analysis. This can be verified by mathematical induction and speculated that using sampling method can still obtain a reasonable credit spreads.