中国市场债务抵押债券定价研究

发布时间：2018-05-14 04:14

本文选题：债务抵押债券 + KMV模型　；参考：《湖南大学》2014年硕士论文

【摘要】：债务抵押债券（CDO）是一种基于信用资产池现金流为支撑，以分券的方式发行的结构性衍生金融产品。由于其复杂的内在结构和交易流程，使得对其进行定价尤为困难。本文基于国内二级市场流动性严重不足，市场缺乏信用违约数据的现状，以及国内市场所发行的CDO普遍存在资产评级高、集中度高等特点。运用KMV模型，利用国内市场数据估算出资产池中各资产的违约概率。针对违约事件导致资产价值发生跳跃的情形，在模型中引入“跳跃”过程进行描述。然后，针对信用资产收益序列的“尖峰厚尾”的特征，信用资产的相关结构是动态的，非线性相关的。本文引入Gaussian Copula和Student-t Copula函数估算出各债务人相互的违约相关结构。最后，利用无套利定价原理求解出各分券的收益面和损失面，并运用Monte Carlo模拟估算出各个分券的合理信用价差。实证结果显示，针对资产的突变过程，通过在模型中引入“跳跃”过程加以刻画，运用KMV模型估算“债务人”的违约概率是有效的。并且，在高级券层的定价精度优于次级券层。其次，，引入Copula函数能够很好的拟合信用资产的尾部非线性风险特征，并且Student-t Copula函数相对于Gaussian Copula函数的模拟结果更优，平均能缩小3-5个bps。最后，本文发现KMV模型中，运用虚拟“债务人”进行分类，会提高资产的信用，进而使得次级券层的定价误差加大。
[Abstract]:CDO is a kind of structured derivative financial product based on the cash flow of credit asset pool. Because of its complex internal structure and transaction process, it is particularly difficult to price it. This paper is based on the lack of liquidity in domestic secondary market, the lack of credit default data in the market, and the high asset rating and high concentration of CDO issued in domestic market. The default probability of each asset in the asset pool is estimated by using KMV model and domestic market data. In view of the situation that the default event leads to the jump of asset value, the "jump" process is introduced into the model. Then, the correlation structure of credit assets is dynamic and nonlinear. In this paper, Gaussian Copula and Student-t Copula functions are introduced to estimate the relative structure of each debtor's default. Finally, the yield and loss surfaces of each coupon are calculated by using the no-arbitrage pricing principle, and the reasonable credit spread of each coupon is estimated by Monte Carlo simulation. The empirical results show that it is effective to use KMV model to estimate the default probability of "debtor" by introducing "jump" process to describe the sudden change process of assets. Moreover, the pricing accuracy in the premium bond layer is better than that in the secondary bond layer. Secondly, the Copula function can fit the tail nonlinear risk characteristics of the credit assets well, and the Student-t Copula function is better than the Gaussian Copula function, and it can reduce 3-5 bps on average. Finally, this paper finds that in the KMV model, the use of virtual "debtor" to classify assets will improve the credit of assets, and then make the pricing error of sub-securities layer increase.
【学位授予单位】：湖南大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：F832.51

【参考文献】