债务抵押契约模型市场重构与违约损失率分布

Calibration of CDO Model by Market Data and LGD Distribution

在违约损失率(LGD)是随机变量的假设下,应用推广的两因子高斯-Copula债务抵押契约(CDO)定价框架,通过极小化相对熵,讨论了利用市场公开报价数据进行系统违约因子和LGD分布的重构问题.计算结果验证了违约具有偏态性的特点.所建立的模型可看成是对高斯-Copula模型的修正.在计算该问题时,采用迭代方法,避免了处理非线性问题以及非光滑优化问题的求解困难.数值计算结果表明,算法是稳定和收敛的.

This paper deals with the calibration of collateralized debt obligation（CDO） model and loss given default（LGD） distribution on the basis of the market data by minimizing the relative entropy.Stochastic recovery and generalized two-factors Gauss-Copula model are supposed.The computational result,which shows the skewed distribution of the default,can be considered as an modification of classical Gauss-Copula model.The iteration techniques are adopted to avoid the nonlinearity and nonsmooth of target function.The numerical results prove the stability and convergence of the algorithm.