基于Copula的商业银行风险综合度量实证研究

发布时间：2018-01-03 08:04

本文关键词：基于Copula的商业银行风险综合度量实证研究　出处：《山西财经大学》2013年硕士论文　论文类型：学位论文

【摘要】：随着经济一体化和全球化的不断发展，各国经济与全球金融市场也随之紧密的联系在一起，一旦某一个经济体陷入困境，它就会牵扯到其它经济体，甚至达到经济崩溃的边缘。例如1997年的亚洲金融危机迅速在东南亚乃至全世界蔓延，而2008年由美国引起的次贷危机演变成了一场世界级的金融危机。可见，金融市场的局部波动可能会波及到其它金融市场，甚至放大，，最终演变成一场灾难性的金融危机。商业银行等金融机构作为世界经济的重要参与者，其面临的风险主要包括信用风险、市场风险和操作风险等。而各类风险之间的相依结构是研究商业银行整体风险大小的一个重要环节，这是因为风险之间可能包含相同的风险因子或者风险之间存在联动效应，因此确定风险之间的相依结构是综合度量商业银行风险的重中之重。而目前由于风险之间的复杂结构及研究工具的局限性等原因，致使风险的综合度量在一定程度上缺乏准确性，所以需要引进新的技术和方法来研究风险之间的相互依赖关系。为了更有效的描述风险之间的相关结构，弥补传统风险度量方法的局限性，需要探索更能真实反映各类风险特性及相关结构的风险度量工具。很多研究结果表明许多金融风险的边际分布存在明显的厚尾性(HeavyTailed)，并不能使用正态分布这样的假设，线性相关系数对于方差不存在的厚尾分布也不能准确反映变量间的相关性。但是Copula函数自身有很好的性质，它对非线性相关结构可以准确地反映，从而提高了度量风险的准确性。商业银行作为重要的金融机构之一，面临着多种多样的风险，而风险间又存在非线性的相关结构，因此采用Copula函数这种技术手段可以为商业银行的风险综合度量提供更准确的方法。本文对我国9家上市商业银行进行了实证分析，首先研究了三大风险的边际分布，利用线性因子模型、ARMA-GARCH模型分别建立了信用风险及市场风险边际分布；其次利用POT模型等方法得到了操作风险的边际分布；最后通过Copula函数描述三种风险之间的相关结构，利用Eviews和Matlab技术求得Copula函数的参数，并使用SPLUS软件求解ES。最终得到结论：利用Copula函数所算得的结果要比不同风险值简单加总小很多，可见不同风险值简单加总高估了整体风险值，因此采用Copula函数这种技术手段可以为商业银行风险的综合度量提供更准确的方法。
[Abstract]:With the continuous development of economic integration and globalization, the economies of various countries and the global financial markets are closely linked with each other. Once one economy is in trouble, it will involve other economies. Even to the brink of economic collapse. For example, in 1997, the Asian financial crisis spread rapidly in Southeast Asia and around the world. In 2008, the subprime mortgage crisis caused by the United States turned into a world-class financial crisis. It can be seen that the local volatility of financial markets may spread to other financial markets, or even magnify. It turned out to be a catastrophic financial crisis. As an important participant in the world economy, commercial banks and other financial institutions mainly face credit risks. Market risk and operational risk and so on. And the structure of dependence between various types of risks is an important link in the study of the overall risk of commercial banks. This is because the risk may contain the same risk factors or there is a linkage between the risk. Therefore, determining the dependent structure between risks is the most important part of the comprehensive measurement of commercial bank risk. But at present, due to the complex structure between the risks and the limitations of research tools and other reasons. As a result of the lack of accuracy in the comprehensive measurement of risk, it is necessary to introduce new techniques and methods to study the interdependence of risk. In order to describe the related structure of risk more effectively. To make up for the limitation of traditional risk measurement methods, we need to explore risk measurement tools that can reflect all kinds of risk characteristics and related structures more truthfully. Many studies have shown that the marginal distribution of many financial risks has obvious thick tail, so it is not possible to use the hypothesis of normal distribution. Linear correlation coefficient can not accurately reflect the correlation between variables for the thick tail distribution where variance does not exist, but the Copula function itself has good properties, it can accurately reflect the nonlinear correlation structure. Commercial banks, as one of the important financial institutions, are faced with a variety of risks, and there are nonlinear correlation structures among the risks. Therefore, the use of Copula function can provide a more accurate method for comprehensive risk measurement of commercial banks. This paper makes an empirical analysis of 9 listed commercial banks in China. Firstly, it studies the marginal distribution of the three major risks and makes use of the linear factor model. The ARMA-GARCH model establishes the marginal distribution of credit risk and market risk respectively. Secondly, the marginal distribution of operational risk is obtained by means of POT model. Finally, the correlation structure between the three kinds of risks is described by Copula function, and the parameters of Copula function are obtained by Eviews and Matlab techniques. Finally, it is concluded that the results calculated by using Copula function are much smaller than the sum of different risk values. It can be seen that the sum of different risk values simply overestimates the overall risk value, so the use of Copula function can provide a more accurate method for the comprehensive measurement of commercial bank risk.
【学位授予单位】：山西财经大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F830.3;F224;F832.33

【参考文献】