【作者】 程芳；

【导师】 王红；

【作者基本信息】 华中科技大学 ， 金融学， 2015， 硕士

【摘要】 随着我国利率市场化进程的加快,银行间同业拆借利率作为我国货币市场基准利率的地位也更加稳固。2010年以来,银行间同业拆借交易量大幅上涨,并且拆借利率的波动愈发频繁,波动的幅度也大幅增加,同业拆借风险更为显著。本文以2010年1月至2014年12月的隔夜拆借利率与七天拆借利率为研究对象,使用Copula-GARCH-VaR模型对我国银行间同业拆借利率的波动及风险进行了实证研究。首先,在对数据进行描述性统计分析的基础上,进行了平稳性、相关性和ARCH效应检验。结果表明,隔夜、七天拆借利率的波动率序列不服从正态分布,并且存在明显的尖峰厚尾、波动集聚现象。然后,选择GARCH类模型来构建隔夜拆借与七天拆借的边缘分布,发现GARCH模型能够很好地消除数据的异方差性,并且正态GARCH)1,1(模型比较适合描述隔夜拆借利率与七天拆借利率波动率的边缘分布。随后,对数据进行概率积分变换,使用Copula函数来刻画两者之间的相关结构。结果显示,积分变换之后的数据分布基本对称,再根据平方欧式距离最小化原则,使用t-Copula函数更好地描述了分布的尖峰厚尾特征。最后,通过Monte Carlo模拟法计算出隔夜拆借与七天拆借在t-Copula-GARCH模型下的95%VaR值为0.2491。同时,对计算结果进行了Kupiec回测检验以评估模型的仿真预测效果,检验结果为LR统计量小于卡方分布对应的临界值,说明Copula-GARCH-VaR模型适用于我国银行间同业拆借的风险管理方法。更多还原

【Abstract】 With the progress of marketization of interest rates in China, the position of the interbank lending rate as the benchmark interest rate in money market is more secure. Since 2010, interbank lending transactions have increased dramatically, with fluctuations of interbank lending rates being more frequent and volatilities being greatly increasing, therefore, interbank interest rate risk is now more significant.Based on data of overnight interest rates and lending rates for seven days from January, 2010 to December, 2014, we use Copula-GARCH-Va R model to analyze China’s interbank interest rate volatility. Firstly, we carry on descriptive statistics analysis as well as stability, correlation and the ARCH effect test. The results show that the data of overnight lending rates and lending for seven days do not obey a normal distribution, and there is an obvious peak thick tail and volatility clustering phenomenon.Then, we select GARCH class models to build the marginal distribution of overnight and seven days lending rates, and find that the GARCH model can effectively eliminate the heteroscedasticity of data, and normal GARCH(1,1) model is suitable for describing the marginal distribution. Then, we use Copula function to depict the correlation structure between the two. Results show that the distribution of the integral transformed data is almost symmetric, and according to the minimization European square distance principle, using t-Copula function is more suitable to describe the peak thick tail characteristics of the distribution.Finally, we calculate 95%Va R value of portfolio of the two by Monte Carlo simulations under t-Copula-GARCH model, and the result is 0.2491. At the same time, we go through Kupiec back-test to evaluate the simulation prediction effect of the model, the test results show that the LR statistic value is less than the corresponding critical value of the chi-square distribution, which means that Copula-GARCH-Va R model is suitable as one of the risk management methods of China’s interbank lending rates.更多还原