【作者】 任仙玲；

【导师】 张世英；

【作者基本信息】 天津大学 ， 技术经济及管理， 2008， 博士

【摘要】 经济全球化和金融市场国际化导致了金融市场之间的联系越来越紧密,彼此的关系更加复杂。由于准确刻画金融市场之间的相依结构是研究投资组合以及风险管理的基础,在金融决策中,为了提高决策的准确性,降低决策风险,研究分析金融市场之间的相依结构是非常必要的。Copula函数是将随机向量的联合分布函数与其对应各分量的边缘分布函数连接在一起的函数,是描述多个金融市场之间相依结构的重要工具。运用它构造联合分布函数时不受边缘分布函数的限制,可以将随机向量的边缘分布函数及其相依结构分开研究。本文运用Copula函数研究金融市场相依结构的建模问题,探讨关于Copula函数的一些理论问题及其在风险管理、投资组合中的应用。运用Copula函数来构建金融资产相依结构的基础是在众多的Copula函数族中选择一个合适的Copula函数,常用的方法是AIC或BIC准则,该准则在Copula函数密度函数存在的条件下有效。为了解决Copula函数的密度函数不存在或密度函数没有显式表达式时Copula函数的选择问题,本文提出了基于非参数核密度估计的Copula函数选择原理,并用蒙特卡罗模拟方法,在金融资产边缘分布函数属于不同类型的情况下,将AIC准则与基于非参数核密度估计的Copula函数选择原理进行了系统的比较。系统地研究了目前存在的一些基于Copula函数的相依性测度,并在此基础上,利用门限Copula函数,分别给出了下门限相依性测度与上门限相依性测度的概念。由于现实金融市场间的相依结构通常随市场的调节不断变化,而且市场“利好消息”与“利坏消息”对金融市场相依结构的影响通常具有不对称性,因此,本文基于门限GARCH模型思想,提出了具有门限结构的时变Copula模型,并利用该模型研究了世界及地区股票市场对我国股票市场间相依结构的影响问题。在金融风险管理及投资组合方面,系统研究了基于VaR和ES的金融风险管理方法,构建了一元APD-GARCH模型,并将该模型与多元Copula相结合,研究了不同相依结构下的均值—ES有效前沿问题。实证结果表明,在研究投资组合以及风险管理问题时,考虑金融市场间的非对称尾部相依结构是十分必要。本论文是国家自然科学基金资助项目《多变量矩序列长期均衡关系及动态金融风险规避策略研究》(No:70471050)的组成部分。更多还原

【Abstract】 Dependencies among financial markets have significantly increased and relationships among them become more complex. This phenomenon is a direct consequence of economic globalization and the globalization of the financial market. As the foundation of risk management and portfolio investment, description on dependence structure correctly in financial markets is very important. So it is necessary to research on the dependence structure among financial markets in financial decision-making in order to increase the correction of decision and to decrease the risk of decision.Copula functions, which are important tool to describe the dependence structure of financial markets, are functions which connect the joint distribution function with its marginal distributions. It is possible to study the dependence structure of random vectors irrespective of their margins by use of the Copula function to construct the multivariate joint distribution. This paper study dependence structure among financial markets based on the Copula theory and discuss some academic question of Copula function and its application in risk management and portfolio investment.The foundation of modeling the dependence among the financial assets by Copula function is the choice of an appropriate parametric copula in a given set of copulas. The commonly used approach to selecting copula is AIC or BIC criterion which relies on the density function of each Copula in the given set of Copulas. However, the explicit expressions of density function of some Copulas may be very complicated or very difficult to obtain, which cause great inconvenience in the application of AIC or BIC criterion. In this sense, a new Copula selection method based on nonparametric kernel density estimation is proposed. Under different marginal distributions, the Copula selection method based on nonparametric kernel density estimation is compared with AIC by means of Monte Carlo simulation Methods.Dependence measures based on Copula functions, which are used commonly, are systemically studied in this paper. From the analysis, the definitions of upper threshold dependence measures and lower threshold of dependence measures are proposed by threshold Copula functions.The dependence structures of financial markets are incessantly changing with the accommodation of market. And asymmetry lies in the influence of positive and negative innovations on the dependence structures of financial market. So, a time-varying Copula model with threshold structure is proposed based on the idea of the threshold GARCH model. By this model, the influence of international and regional stock markets on the dependence structures of domestic financial market is studied.In the aspect of risk management and portfolio investment, financial risk management methods based on VaR and ES are studied. Furthermore, the APD-GARCH model is established. With this model and multivariate Copula functions, the efficient frontier of portfolio optimization based on mean-ES is studied. The experiential result indicates that it is very necessary to consider the asymmetry tail dependence structure in research on risk management and portfolio investment.The research is sponsored by National Natural Science Foundation of China: Research on Long Run Equilibrium in Multivariate Moments Series and Avoiding Tactics of Dynamic Financial Risk (No.70471050).更多还原