基于VaR模型的金融市场风险测量方法的探析

发布时间：2018-01-18 01:27

本文关键词：基于VaR模型的金融市场风险测量方法的探析　出处：《东北财经大学》2013年硕士论文　论文类型：学位论文

更多相关文章： 风险价值(VaR) Copula-GARCH Monte Carlo

【摘要】：随着科技和信息的不断发展,全球经济和金融一体化的步伐不断加快,金融市场的波动性和风险也在不断增加,对国家和企业的影响也越来越大,从而使金融风险管理已为越来越多的金融机构和投资者所重视。风险的度量、分析技术发展得非常迅速。风险价值VaR (Value at Risk)方法起源于20世纪80年代,是一种衡量金融市场风险统计的方法,它被广泛应用于银行,证券公司,大宗商品和其他贸易组织。这项研究方法的主要特点是度量和分析市场风险在金融市场中的风险价值,在现代金融风险管理中具有重要的地位。我国对VaR方法的应用也在逐渐发展完善之中,对其进行研究的内容也有很多。其主要内容是(1)预测未来可能出现的金融损失风险值计量。(2)识别危险因素的分布如何影响投资组合的分布。(3)为投资组合优化和风险管理提供可靠的信息。传统的VaR计量方法存在着一些缺点,如正态分布的假设,不满足一致性公理,缺乏对极端事件和金融资产间的尾部相关性的考虑等。这些都会影响投资组合的投资风险和效果,影响VaR度量的精确度。而Copula的理论相对传统方法有着理论优势,Copula模型可以将相关关系和相关模式的研究有机地结合在一起。通过Copula函数,可以捕捉到变量间非线性、非对称和尾部的相关关系。Copula理论为分析多变量金融时间序列问题提供了崭新的思路。本文主要的研究思路是考虑到金融市场中的波动聚集现象,建立GARCH模型来测度金融市场风险,并针对VaR方法的一些缺点,引入Copula函数,在边际分布上针对金融数据中出现的尖峰厚尾现象分别假定变量服从正态分布,t分布,GED分布。通过失败率检验和K-S检验确定了最优的边际分布模型和最优Copula函数。将最优边际分布GARCH模型与最优Copula函数相结合,通过Monte Carlo模拟法度量沪深指数的投资组合的VaR。本文总共分成六部分,主要内容如下：第一章为引言,简要地概述课题研究的背景和意义,综述国内外有关VaR的研究现状及国内外利用Copula方法度量VaR的情况,最后对本文的研究内容、方法及创新与不足点做了简单说明。第二章是金融风险知识。简单概括了金融风险的定义和对金融风险度量的传统方法。第三章是VaR的理论基础。首先,详细介绍了VaR理论产生的现实背景,概念,VaR的持有期和置信水平的选择。其次,重点阐述三种传统的度量VaR的方法,主要包括德尔塔-正态法,历史模拟法,Monte Carlo模拟法。然后,给出了计算VaR的模型和检验方法。最后对VaR的优缺点做了简单的介绍。第四章为Copula函数相关知识。详细解释了Copula函数的定义及相关性质,介绍了Copula的常见函数和相关定理。简单叙述运用Copula函数的相关性度量即Kendall's tau和Spearman's同时,对尾部相关性做了简要说明。最后详细阐述了利用基于Copula-GARCH模型的Monte Carlo模拟法计算VaR的方法。第五章为论文的实证分析部分。本文构造的投资组合为上证指数和深证综指按等权组合,时间为2008.1.1-2013.4.3,总共是1278组数据。采取对数收益率进行实证分析,首先估计边际分布模型,通过分析数据特征,检验波动性,平稳性,自相关-偏自相关以及ARCH效应,分别在正态分布,t分布,GED分布下进行回归,通过进一步的计量分析,确定最优的模型为GARCH(1,1)-t模型。然后通过K-S检验认为t-Copula函数为描述变量间相关性最合适的Copula函数,并利用秩相关系数进行验证。本文将t-Copula函数结合GARCH-t模型来对沪深指数的投资组合进行Monte Carlo模拟,以度量证券市场的风险。从实证结果可以看出,用t-Copula-GARCH-t函数度量VaR是有效的。第六章为论文的结束语。首先总结全文内容,包括论文的原理、方法和模型,然后对针对未来的相关研究提出了几种政策建议。本文的创新之处是建立计算风险价值VaR的Copula-GARCH模型,并应用到分析沪深指数投资组合的风险测度,使VaR的计量精度大大提高,Monte Carlo模拟法从历史数据出发,利用一系列检验和统计方法,找到能较好得刻画数据特征的分布函数,据此进行模拟的效果比较好。本文的不足之处在于：上证指数和深证综指组成的投资组合中,它们相应的投资比例是固定的,没有进行动态投资优化；由于Copula的估计相对较难,二元的Copula模型相对来说最多,二元的Copula显然不能很好对多个资产进行分析。
[Abstract]:With the continuous development of information technology and the global economic and financial integration is accelerating the pace of financial market volatility and risk are also increasing influence on the state and enterprises is also increasing, so that the financial risk management has been valued by financial institutions and investors more and more. To measure the risk analysis technology to develop very quickly. The risk value of VaR (Value at Risk) method originated in 1980s, is a statistical method to measure the risk of financial market, it is widely used in banks, securities companies, commodities and other trade organizations. The main features of this study is to measure and method of value at risk in the financial market risk in the market, has an important position in modern financial risk management in China. The application of VaR method is also gradually developed, there are also many of its research content Its main contents are: (1) predict the possible financial loss risk measurement in the future. (2) identify the distribution of risk factors and how to affect the distribution of portfolio. (3) provide reliable information for portfolio optimization and risk management.
The traditional method of VaR measurement has some shortcomings, such as the normal distribution assumption, does not meet the consistency axiom, lack of extreme events and the tail correlation between financial assets is considered. These will affect the portfolio investment risk and effect, effect of VaR measurement accuracy. And the theory of Copula relative to the traditional method there is a theoretical advantage, Copula model can study the organic correlation and the correlation pattern together. Through the Copula function can capture the correlation between nonlinear, asymmetric and tail correlation theory of.Copula for multivariate financial time series analysis provides a new way of thinking.
The main idea of this paper is to consider the phenomenon in the financial market volatility, establish GARCH model to measure the risk of financial market, and some disadvantages of VaR method, introducing the Copula function, the marginal distribution of the fat tail phenomenon in financial data variables are assumed to obey normal distribution, t distribution, GED distribution through the failure rate test and K-S test to determine the marginal distribution model and the optimal Copula function optimal. The optimal marginal distribution of GARCH model and the optimal combination of Copula function, through the Monte Carlo simulation method to measure the Shanghai and Shenzhen index portfolio VaR.
This article is divided into six parts, the main contents are as follows:
The first chapter is the introduction. It briefly summarizes the background and significance of the research, summarizes the research status of VaR at home and abroad, and the situation of using Copula to measure VaR at home and abroad. Finally, it gives a brief description of the research contents, methods, innovations and shortcomings.
The second chapter is the knowledge of financial risk. The definition of financial risk and the traditional methods for measuring the financial risk are briefly summarized.
The third chapter is the theoretical basis of VaR. Firstly, introduces the background, the concept of VaR VaR theory, the holding period and the confidence level. Secondly, focuses on three kinds of traditional measuring methods of VaR, including the delta normal method, historical simulation, Monte Carlo simulation method. Then, the model and method of calculation of VaR test is given. Finally the advantages and disadvantages of the VaR to do a simple introduction.
The fourth chapter is the Copula function related knowledge. A detailed explanation of the definition of Copula function and related properties, introduces the common functions and related theorems of Copula. The simple description of correlation using Copula function measurement Kendall's tau and Spearman's at the same time, the tail correlation are briefly described. Finally expounded the calculation method VaR Copula-GARCH model Monte based on the method of Carlo simulation.
The fifth chapter is the empirical analysis part of this paper. This paper constructs a portfolio for the Shanghai index and Shenzhen composite index according to the right combination, time is 2008.1.1-2013.4.3, a total of 1278 sets of data.
Take the empirical analysis log returns, first estimates the marginal distribution model, through the analysis of test data characteristics, volatility, stationarity, autocorrelation and partial autocorrelation and ARCH effect were in normal distribution, t distribution, GED distribution was measured by regression, further analysis, to determine the optimal model for GARCH (1,1) -t model.
Then the t-Copula function to describe the correlation between variables of the appropriate Copula function through the K-S test, and the result is verified by using rank correlation coefficient. In this paper the t-Copula function based on the GARCH-t model of the Shanghai and Shenzhen index portfolio of Monte Carlo simulation, to measure the risk of securities market. The empirical results show that using the t-Copula-GARCH-t function to measure VaR effective.
The sixth chapter is the concluding remarks of the paper. First, we summarize the contents of the paper, including the principles, methods and models of the paper, and then put forward several policy recommendations for future research.
The innovation of this paper is to establish the Copula-GARCH model to calculate the risk value of VaR, and applied to the analysis of the Shanghai and Shenzhen index portfolio risk measure, the measurement accuracy of VaR is greatly improved, Monte Carlo simulation method based on historical data, using a series of test and statistical method, the distribution function can find better characterization of data characteristics, accordingly the simulation result is good.
The inadequacies of this article are: Shanghai index and Shenzhen composite index portfolio, their corresponding investment ratio is fixed, there are no dynamic investment optimization; because the Copula estimation is relatively difficult, the Copula model is relatively the most two yuan two yuan Copula, obviously not good for multiple assets analysis.

【学位授予单位】：东北财经大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F830.9;F224

【相似文献】