半参数C-Vine Copula模型理论及其金融风险结构测度研究

发布时间：2017-12-28 01:03

本文关键词：半参数C-Vine Copula模型理论及其金融风险结构测度研究　出处：《天津财经大学》2016年博士论文　论文类型：学位论文

【摘要】：2016年,作为第十三个五年规划的开局之年,我国经济已经进入"新常态"。"新常态"不是一个固定不变的状态,而是一个动态优化的过程。在此过程中,经济和金融市场的各个指标都呈现出了新的特点,其内在风险关联更加复杂。这使得传统的经济计量模型难以有效地测度风险相关结构,需要研究新的理论工具解决"新常态"下的风险结构测度问题。由于半参数C-VineCopula具有稳健灵活易于估计的特性,可以很好地解决此方面的问题,所以对该函数的研究具有重要的理论意义与应用价值。Copula函数模型是一种新型多元联合分布建模工具。其显著特点是两步法构造,使得各个变量的个体特征体现在其边缘分布中,而相关性结构体现在Copula函数中。与传统的多元正态、多元t分布比,该模型放松了假设条件,构造更为灵活简便,应用更为广泛,能够更为真实地拟合经济变量的联合分布。鉴于以上优点,根据目前金融风险结构的特征,论文对Copula函数模型进行了以下几个方面的研究:论文首先对Copula函数的中外文献进行了梳理和总结。其次,对相关性测度指标及Copula函数建模的边缘分布、函数模型、估计和检验等相关步骤中所用工具的理论性质和应用范围进行了详细而深入的研究。再次,就非参数核密度估计方法与较为流行的Vine Copula模型的结合进行了研究,构造了半参数C-Vine Copula模型。最后,在以上研究的基础上,针对美、中、日、德、英五国股票市场的风险相关性程度及结构进行了测度研究。并就其经济意义进行了讨论和分析。论文的主要创新体现在以下几个方面:第一,提出了在数据样本数量较少或存在内在非线性转换机制的情况下,可以使用STAR模型拟合边缘分布,构造STAR-Copula模型。第二,在非参数核密度方法研究部分,对二阶核、高阶核和非对称核进行了理论研究及在各种分布环境下的模拟,表明了正态核和非对称核具有良好的估计性质,双角核、三角核和Epanechnikov核存在较小的边界效应。第三,结合非参数核密度方法对数据限制较少的优点和Vine Copula模型构造灵活的优点,首创了半参数C-Vine Copula模型,并就其在各种环境下的估计性质进行了 MonteCarlo模拟,表明了该模型具有良好的稳健性,适用于数据分布不规律以及数据异常波动情况下的多元联合分布建模。论文创新给"新常态"金融风险相关结构测度研究提供了有力的定量分析工具。
[Abstract]:In 2016, as the thirteenth year of the opening of the five year plan, China's economy has entered a "new normal". The "new normal" is not a fixed state, but a process of dynamic optimization. In this process, the various indicators of the economic and financial markets have shown new features, and their internal risk association is more complex. This makes the traditional econometric model difficult to effectively measure the risk related structure, and we need to study new theoretical tools to solve the risk structure measurement problem under the new normal. Because semi parametric C-VineCopula has the characteristics of robust, flexible and easy to estimate, it can solve this problem well. Therefore, the study of this function has important theoretical significance and application value. The Copula function model is a new multicomponent joint distribution modeling tool. Its remarkable feature is the two step structure, which makes the individual characteristics of each variable in its edge distribution, and the correlation structure is embodied in the Copula function. Compared with the traditional multivariate normal distribution and multivariate t distribution, the model relaxes the hypothetical condition, and it is more flexible, simple and widely applied, which can more accurately match the joint distribution of economic variables. In view of the above advantages, according to the characteristics of the current financial risk structure, the paper studies the following aspects of Copula function model: first, the paper summarizes and summarizes the Chinese and foreign literature of Copula function. Secondly, we have done detailed and in-depth research on the theoretical nature and application scope of the tools used in the related steps, such as the correlation measure index and the marginal distribution, function model, estimation and inspection, which are modeled by the Copula function. Thirdly, the combination of the non parametric kernel density estimation method and the more popular Vine Copula model is studied, and a semi parametric C-Vine Copula model is constructed. Finally, on the basis of the above research, we measure the degree and structure of risk correlation in the stock market of the five countries of the United States, China, Japan, Germany and Britain. And its economic significance is discussed and analyzed. The main innovations of the paper are as follows: first, it is proposed that when the number of data samples is small or there is an inherent nonlinear transformation mechanism, we can use the STAR model to fit the edge distribution and construct the STAR-Copula model. The second part, in the study of non parametric kernel density method, the two order nuclear, nuclear and non nuclear high order symmetry was studied and the distribution in various simulation environment, show the normal nuclear and non symmetric kernel estimation with good properties of double angle triangular nucleus, nucleus and the Epanechnikov nucleus in the boundary effect small. Third, combined with the non parametric kernel density method for less data limitations and the advantages of Vine Copula model to construct the advantages of flexible, pioneered the semiparametric C-Vine model of Copula, and the MonteCarlo simulation of the environment in a variety of estimation properties, the results show that this model has good robustness and is suitable for the data distribution rule and multivariate data modeling the joint distribution under the condition of abnormal fluctuations. The paper provides a powerful quantitative analysis tool for the study of the structure measurement of the "new normal" financial risk related structure.
【学位授予单位】：天津财经大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：F224;F831.51

【相似文献】