基于Copula理论的沪港股市相关性及尾部相关性研究

发布时间：2019-05-17 00:33

【摘要】：随着金融全球化进程的发展，各金融市场之间的相互依存性不断加强，单一股票或市场的分析越来越不能满足金融市场研究的需要，相关性分析在金融应用中变得越来越重要，，已经成为金融市场风险度量的关键。但是基于线性相关关系的传统相关性度量只专注于线性相关的程度,忽视了金融市场结构之间的关系,特别是尾部相关特征的研究。Granger因果分析方法只能做定性分析，无法给出定量的结论。本文将非线性相关分析工具——Copula方法用于金融市场的相关性研究中，分析了上证指数和恒生指数之间的常相关、时变相关的相关性和尾部相关性。本文的内容结构安排如下：第一部分介绍选题背景及意义、国内外研究现状；第二部分介绍关于Copula的一些基础知识；第三部分介绍建模、参数估计和检验；第四部分介绍本文用到的静态和时变Copula函数；第五部分实证研究；第六部分为结论。关于研究的方法,目前的研究多是假设边缘分布服从t分布或正态分布,采用静态Copula方法研究两个城市之间的相关性或尾部相关性,一般采用参数估计方法(Parametric Approach),参数估计方法一般会对边缘分布函数作相关的假设（假定服从t分布或服从正态分布等）,但假设很难符合实际的情况,因此边缘拟合效果往往不是很好,这往往影响参数估计效果。本文用基于秩的极大似然估计的方法,通过静态C opula和时变C opula对沪港股票市场相关性及尾部相关性进行研究。结果表明,沪港股市的相关性总体呈现增长趋势；尾部相关性方面，下尾相关程度略大于上尾，按两阶段分析，第一阶段几乎不存在下尾相关，上尾相关亦不明显；第二阶段的尾部（尤其是下尾）关联程度明显高于第一阶段，并且下尾相关程度大于上尾。
[Abstract]:With the development of financial globalization and the interdependence of financial markets, the analysis of single stock or market can not meet the needs of financial market research, and correlation analysis is becoming more and more important in financial application. It has become the key to the measurement of financial market risk. However, the traditional correlation measurement based on linear correlation only focuses on the degree of linear correlation, neglecting the relationship between financial market structures, especially the tail correlation characteristics. Granger causality analysis method can only do qualitative analysis. It is impossible to give a quantitative conclusion. In this paper, Copula method, a nonlinear correlation analysis tool, is used to study the correlation between Shanghai Stock Exchange Index and Hang Seng Index, and the correlation between Shanghai Stock Exchange Index and Hang Seng Index, time-varying correlation and tail correlation are analyzed. The content structure of this paper is as follows: the first part introduces the background and significance of the topic, the research status at home and abroad; the second part introduces some basic knowledge about Copula; the third part introduces modeling, parameter estimation and test. The fourth part introduces the static and time-varying Copula functions used in this paper; the fifth part is empirical research; the sixth part is the conclusion. With regard to the research methods, most of the current studies assume that the edge distribution obeys t distribution or normal distribution. The static Copula method is used to study the correlation or tail correlation between the two cities, and the parameter estimation method (Parametric Approach), is generally used. The parameter estimation method generally makes relevant assumptions about the edge distribution function (assuming that it obeys t distribution or normal distribution, etc.), but it is difficult to conform to the actual situation, so the edge fitting effect is often not very good. This often affects the effect of parameter estimation. In this paper, the correlation and tail correlation of Shanghai and Hong Kong stock markets are studied by static C opula and time varying C opula by using the method of maximum likelihood estimation based on rank. The results show that the correlation of Shanghai and Hong Kong stock markets as a whole shows an increasing trend, and the correlation degree of the lower tail is slightly higher than that of the upper tail in terms of tail correlation. According to the two-stage analysis, there is almost no lower tail correlation in the first stage, and the upper tail correlation is not obvious. The correlation degree of the tail (especially the lower tail) of the second stage was significantly higher than that of the first stage, and the correlation degree of the lower tail was greater than that of the upper tail.
【学位授予单位】：重庆大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：F224;F832.51

【参考文献】