基于ARCH模型族的VaR方法在商业银行利率风险管理中的应用

发布时间：2018-01-03 03:16

本文关键词：基于ARCH模型族的VaR方法在商业银行利率风险管理中的应用　出处：《山东大学》2014年硕士论文　论文类型：学位论文

【摘要】：自从1996年我国正式启动利率市场化改革以来,我国的利率市场化改革稳步推进,先后放开了同业拆借市场利率、债券市场利率、银行间市场国债、政策性金融债的发行利率、境内外币贷款等各种利率。利率放开后,利率对金融环境波动的敏感性增加,波动的频率和幅度都显著提高。商业银行的资产主要是金融资产,利率变动会导致其资产价值的变动,商业银行在利率的波动中承受的利率风险会增加。而此前我国商业银行的风险管理重点主要集中与信用风险,对利率管理的经验不足,研究利率风险管理对我国商业银行在利率市场化背景下积极应对利率风险、建立全面完善的风险管理体系至关重要。本文共分为五章。第一章主要介绍利率市场化背景下本文的研究意义、国内外的研究现状和本文的研究思路与框架。第二章从外在宏观因素和内在微观因素两大方面详细分析了商业银行的利率风险的原因；简要介绍了敏感性缺口分析法、久期缺口分析法、凸度缺口分析法等利率风险的度量方法,并重点介绍了VaR方法包括其原理、计算方法、优缺点等。第三章从金融序列的波动性入手,简要介绍了ARCH模型及其变形GARCH模型、GARCH-M模型、EGARCH模型和IGARCH模型,并分析了他们的模型特征、适用条件和优缺点。另外由于正态分布的在描述尖峰厚尾性方面的局限性,在模型中引入了两种分布：t分布和GED分布。第四章以2008年10月8日至2013年12月31日间上海银行间隔夜拆借利率的1812个数据为样本,用Eviews、Excel等数据软件分析了隔夜拆借利率对数收益率的平稳性、自相关性、正态性和尖峰厚尾性、ARCH效应等基本统计特征。第五章分别使用AR(1)-GARCH(1,1), AR(1)-GARCH-M(1,1)-EGARCH(1,1).AR(1)-IGARCH(1,1)等模型对数据进行了拟合,每种模型均基于正态分布、t分布、GED分布三种情况作了比较分析；根据极大似然函数值、AIC、SC值及模型系数显著性检验结果等因素,本文确定了拟合最优的为基于GED分布的模型；在95%、99%两种置信水平下计算了基于各模型的VaR值并对模型拟合效果和VaR值进行了结果分析。结果表明,基于GED分布的AR(1)-GARCH(1,1).AR(1)-GARCH-M(1,1).. AR(1)-EGARCH(1,1).AR(1)-IGARCH(1,1)模型能较好地拟合上海银行间隔夜拆借利率对数收益率,反映数据的尖峰厚尾性。且模型拟合结果显示,隔夜拆借利率具有显著地反杠杆效应和长记忆性。95%的置信水平下,只有基于AR(1)-EGARCH(1,1)一GED模型的VaR值通过了Kupiec检验,说明能显著反映同业隔夜拆借利率不对称波动的EARCH模型也能准确可靠的计算VaR值且优于其他模型。在99%的置信水平下除AR(1)-IGARCH(1,1)模型外,其他三个模型计算出的VaR值都能通过Kupiec检验,风险度量效果良好。
[Abstract]:Since 1996, our country officially launched the interest rate marketization reform, our country interest rate marketization reform has promoted steadily, has successively released the interbank borrowing market interest rate, the bond market interest rate, the interbank market national debt. Interest rates on the issuance of policy-oriented financial bonds, domestic foreign currency loans and other interest rates. After the liberalization of interest rates, the sensitivity of interest rates to fluctuations in the financial environment has increased. The frequency and amplitude of volatility have increased significantly. The assets of commercial banks are mainly financial assets, and the change of interest rate will lead to changes in the value of their assets. The interest rate risk that commercial banks bear in the interest rate fluctuation will increase. But the risk management of our country commercial bank mainly focuses on the credit risk, and the experience of interest rate management is not enough. The study of interest rate risk management is very important for Chinese commercial banks to actively deal with interest rate risk under the background of interest rate marketization and to establish a comprehensive and perfect risk management system. This paper is divided into five chapters. The first chapter mainly introduces the research significance of this paper under the background of interest rate marketization. The second chapter analyzes the reasons of interest rate risk of commercial banks from two aspects of external macro factors and internal and micro factors. This paper briefly introduces the measurement methods of interest rate risk, such as sensitivity gap analysis, duration gap analysis and convexity gap analysis, and focuses on the VaR method, including its principle and calculation method. The third chapter introduces the ARCH model and its modified GARCH model, starting with the volatility of financial series, and introduces the GARCH-M model. EGARCH model and IGARCH model are analyzed, and their characteristics, applicable conditions, advantages and disadvantages are analyzed. In addition, due to the limitations of normal distribution in the description of peak thick tail. Two distributions are introduced into the model:. T distribution and GED distribution. Chapter 4th is based on 1 812 data of Shanghai inter-bank overnight offered rate from October 8th 2008 to December 31st 2013. The stability, autocorrelation, normality and peak and thick tail of the logarithmic yield of overnight interest rate are analyzed by means of Eview Excel and other data software. ARCH effect and other basic statistical characteristics. In Chapter 5th, we used ARGARCH1, ARGARCH-MU 1, ARP1GARCH-MU 1 and EGARCH1, respectively. (1) the data were fitted by the models such as (1). The data were fitted with each model. Each model was compared and analyzed based on the normal distribution and the GED distribution. Based on the maximum likelihood function (MLF) SC and the significance of the model coefficients, the optimal fitting model based on GED distribution is determined in this paper. The VaR values based on each model are calculated at 95% and 99% confidence levels, and the results of model fitting and VaR are analyzed. The results show that the GED distribution is based on the GED distribution. The model can fit the logarithmic yield of Shanghai inter-bank overnight offered interest rate and reflect the peak and thick tail of data. The fitting results show that the model can fit well the logarithmic yield of Shanghai inter-bank interest rate. The overnight lending rate has a significant anti-leverage effect and long memory. 95% confidence level, only based on AR(1)-EGARCH(1. 1) the VaR value of a GED model passed the Kupiec test. It shows that the EARCH model, which can reflect the asymmetric fluctuation of interbank interest rate, can calculate the VaR value accurately and reliably and is superior to other models. -IGARCH1. 1) except the model, the VaR calculated by the other three models can pass the Kupiec test, and the effect of risk measurement is good.
【学位授予单位】：山东大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：F224;F832.33;F822.0

【参考文献】