Research On Pricing Synthetic Collateralized Debt Obligation (SCDO) Under Financial Crisis

In the summer of 2007, subprime crisis broke out in the United States, which gave rise to the subsequent global financial crisis. During this crisis, numerous investment banks, hedge funds and other financial institutions have posted significant loss. One main reason for this is that these financial institutions don't have appropriate models to price structured credit derivatives in the market. They have paid big prices for their inability. For example, some of them were acquired by others and some went into bankruptcy directly. The current pricing models for credit derivatives include structural model, reduced-form model, incomplete information model and factor model. The last two are based on the two former models. These models have showed their demerits during this crisis. The most popular model for pricing Synthetic Collateralized Debt Obligation (SCDO) is One Factor Gaussian Copula model. Nevertheless, some implied correlation coefficients calibrated from One Factor Gaussian Copula model are larger than 1, which makes no sense in mathematics. This dissertation improves the One Factor Gaussian Copula model and comes up with a new model called One Factor Gaussian Copula model Considering Lag Factors. We take into account lag factors which have impact on the quality of collateral in the asset pool. The new model better captures the quality of collateral and gives reasonable implied correlation coefficients between assets in the asset pool.In this dissertation, the first chapter introduces the transaction process, motivation, classification and the impact on the current crisis of SCDO. Chapter two discusses the two basic models for pricing credit derivatives: structural model and reduced-form model. Additionally, chapter two introduces Copula function and Credit Default Swap (CDS). In chapter three, we present SCDO and how to calculate tranche spread through One Factor Gaussian Copula model. Chapter four discusses the sensitivity of spread to correlation using One Factor Gaussian Copula model and introduces the two approaches to calculate correlation: compound correlation and base correlation. In Chapter five, we introduce the new model: One Factor Gaussian Copula considering Lag Factors and calibrate the new model to market spreads. Additionally, we list the strengths of this new model over standard One Factor Gaussian Copula model in chapter five.