基于Copula-EVT模型对股指在险价值的计量

发布时间：2018-03-01 03:28

本文关键词： 风险管理 VaR、CVaR、极值理论 Copula函数　出处：《华中师范大学》2012年硕士论文　论文类型：学位论文

【摘要】：随着金融全球化进程的加快,金融市场面临的风险日益复杂化和多样化,有效的进行风险管理成为金融行业的重中之重。而风险管理的关键在于对风险价值的测量,如何精确的度量不同形式的风险成为学术界和金融界关注的热点和难点。本文搜集了全球主要的六支股指,从不同的投资决策出发,全而系统的运用不同的方法针对不同形式的资产价值估算风险价值。本文有两条主线：一是资产形式,单一资产资产形式如沪深300,如果观察时变风险,本文采用基于GARCH类模型的参数估计法来估计其时变风险,如果投资者侧重于关注极值风险,本文运用极值理论来求其极值风险；对于资产组合形式,投资者或者投资机构不仅关注每项资产间的相关关系,还重视其相关模式,本文结合极值理论与Copula函数理论来求其在险价值。二是估计方法,本文按照从参数估计到半参数估计到非参数估计的思路对针对不同的投资决策来估计风险价值。看似是对风险测度方法的讨论,实则以研究方法为技术支撑来达到精确测度风险价值的目的。大量的研究成果表明：金融时间序列收益率的分布呈现尖峰厚尾、波动的集聚性等特征,而不是传统假设的正态分布。针对这一特点本文的第三三章以讨论了沪深300的收益特征并在估算在险价值。本文采用偏态t分布下的FIGARCH模型来估计时变VaR。检验结果表明偏态t分布下的VaR估计优越于传统的正态分布、学生t分布、GED分布下的估计值。在金融风险的管理中,投资者往往更关注极值事件,所以如何合理的刻画极值分布,求出极值情况下的风险价值尤为重要。本文第四章利用极值理分别描述所搜集的六支股指的尾部分布情况并估计出VaR、CvaR。根据风险分散化原理,大多数投资者都会进行多元化的投资,以降低风险。这牵涉到多元极值的问题,第五章在极值理论的基础之上引入Copula函数理论,来简化多元极值问题。采用Copula-EVT模型分析由搜集到的六支股指等权重组成的资产组合的风险价值。失败率检验的结果表明：在95%和99%的显著性水平下,失败率和显著性水平都很接近,说明多元t-Copula模型能较好的描述多资产的相依性结构,Copula-EVT的模型选择适宜。本文的最后对全文做了进行了总结,分析了本文研究的不足之处,在极值理论和Copula理论的研究现状之上对未来的研究方向做了展望。
[Abstract]:With the acceleration of the process of financial globalization, the risks faced by financial markets are becoming increasingly complex and diversified. Effective risk management has become the most important part of the financial industry, and the key to risk management lies in the measurement of risk value. How to accurately measure different forms of risk has become a hot and difficult issue in academic and financial circles. This paper collects the six major stock indexes in the world and starts from different investment decisions. The whole system uses different methods to estimate the risk value for different forms of asset value. There are two main lines in this paper: one is asset form, one is single asset form, such as Shanghai and Shenzhen 300. If time-varying risk is observed, this paper uses parameter estimation method based on GARCH model to estimate time-varying risk, if investors focus on extreme value risk. In this paper, extreme value theory is used to calculate the extreme value risk. For the form of portfolio, investors or investment institutions not only pay attention to the relationship between each asset, but also attach importance to its related model. In this paper, the extreme value theory and Copula function theory are combined to find the value in danger. In this paper, according to the idea from parameter estimation to semi-parameter estimation to non-parametric estimation, we estimate the value of risk for different investment decisions. It seems to be a discussion of the method of risk measurement. In fact, the research method is taken as the technical support to achieve the purpose of accurately measuring the value of risk. A large number of research results show that the distribution of financial time series returns shows the characteristics of peak and thick tail, agglomeration of volatility, and so on. The 33th chapter of this paper discusses the income characteristics of CSI 300 and estimates its value at risk. In this paper, the FIGARCH model under skewed t distribution is used to estimate the time-varying VaR. The results show that the VaR estimation under skew t distribution is superior to the traditional normal distribution. Student t distribution is estimated under GED distribution. In financial risk management, investors tend to pay more attention to extreme value events, so how to describe the extreme value distribution reasonably, It is very important to find out the value of risk under extreme value. Chapter 4th describes the tail distribution of the six stock indexes collected and estimates the tail distribution of the six stock indexes collected in Chapter 4th. According to the principle of risk decentralization, Most investors make diversified investments to reduce risk. This involves the problem of multivariate extremum. Chapter 5th introduces Copula function theory on the basis of extreme value theory. To simplify the multivariate extremum problem. The Copula-EVT model is used to analyze the risk value of the portfolio composed of six stock indexes with equal weights. The results of the failure rate test show that: at the significance level of 95% and 99%, The failure rate and the significance level are very close, which indicates that the multivariate t-Copula model can better describe the dependence structure of multiple assets and the model selection of Copula-EVT is suitable. At the end of this paper, the author summarizes the full text, analyzes the deficiency of this paper, and looks forward to the future research direction on the basis of the research status of extreme value theory and Copula theory.
【学位授予单位】：华中师范大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：F830.91;F224

【引证文献】