Topics in portfolio credit risk and operational risk: dependence/stress modeling and robust estimation

提出的研究课题是我在信用风险和操作风险建模方面的持续研究项目。最近的经验证据表明，在压力时期，资产和违约的相关性都急剧上升。我建议研究一类依赖的默认模型，它们在压力下可能包含过度的相关性。将研究纳入信用衍生品价格和风险措施的依赖程度的含义。这一领域的研究非常重要，因为典型的模型是建立在条件独立性和正态性假设的基础上的，无法解释市场的极端运动，从而引入了对风险的严重低估。基础资产的联合动力学建模是信用风险建模的基础。我打算研究一个包含随机波动率和随机相关性的多资产模型。参数估计和诊断工具与一个特定的应用到欧洲主权信用风险的建模将被考虑。除信用风险外，还将研究操作风险建模方面的几个问题。投资组合问题包括开发一种用于外部损失数据缩放的稳健回归方法。将研究多变量损失模型，以纳入业务单元和事件类型之间的任何相互依赖关系。

The proposed research topics are aspects of my continuing research programs in credit risk and operational risk modeling.The recent empirical evidence shows that both asset and default correlations violently increase during stress periods. I propose to study a class of dependent default models, which can incorporate excessive correlations under stress. The implication of the level of dependence incorporated to the credit derivative prices and the risk measures will be studied. This area of research is quite important as the typical models which are built on the conditional independence and the normality assumptions, fail to explain the extreme movement in the market and, thus, introduce a severe underestimation of risk.The modeling of the joint-dynamics of underlying assets is fundamental in credit risk modeling. I intend to study a multi-asset model that incorporates stochastic volatility and stochastic correlations. Parameter estimation and diagnostic tools with a particular application to the modeling of European sovereign credit risk will be considered. In addition to credit risk, a few issues will be studied on operational risk modeling. The portfolio question includes the development of a robust regression method for external loss data scaling. Multivariate loss models will be studied to incorporate any interdependence between business units and event types.