时序金融数据的VaR分析

发布时间：2018-03-18 06:36

本文选题：VaR　切入点：非线性分位数回归　出处：《天津科技大学》2013年硕士论文　论文类型：学位论文

【摘要】：在2001年加入WTO之后,我国的外汇储备量已经攀升到了世界第一的位置,而在2005年采用浮动汇率以来,人民币汇率一直承受着外界政治、经济的多方压力,导致人民币汇率处于不断增值的趋势中,加之2008年金融危机对金融市场的巨大冲击使得很多金融巨头纷纷倒闭,金融行业得以重新洗牌,在新挑战和新机遇并存的形势下,加快我国金融行业监管理论与实践方面的水平变得尤其重要。目前,在经过多年的研究与探索之后,基于VaR的金融市场风险度量方法业已成为国外诸多金融机构的重要风险控制工具。而国内的汇率管理制度和汇率风险监管方法仍然比较落后,不足以形成稳定的汇率水平和先进的风险控制系统。在这种形势下,对国际先进金融风险测度方法的学习与研究显得尤为重要,本文在目前广泛使用的VaR方法的基础上进行了拓展与深入研究,主要从基于非线性分位数回归的VaR估计方法角度展开。希望能够为汇率风险监管提供一种有力的工具。本文通过对人民币/美元收益率序列进行统计分析发现该序列存在着非正态性、尖峰厚尾、非对称性、波动集聚、条件方差性等特征。使用传统的时间序列模型ARMA模型或者非线性最小二乘回归模型对序列进行拟合,经过检验发现这两种模型不能刻画该序列的这些特征。由于GARCH族模型充分利用了ARCH模型的条件异方差性和学生t分布和GED分布的尖峰厚尾的特性,经过检验发现各种GARCH类模型均能很好的描述收益率序列的异方差性。经过实证分析可以得出这样的结论,非线性分位数回归模型估计时序金融数据的VaR时,能够全面刻画金融数据之间不同相关结构下的相关关系从而能够得出具有灵活性的度量方法和有多重选择的结论。
[Abstract]:After joining the WTO in 2001, China's foreign exchange reserves have climbed to the first position in the world. Since 2005, when the floating exchange rate was adopted, the RMB exchange rate has been subject to external political and economic pressures. The RMB exchange rate is in the trend of increasing value. In addition, the huge impact of the financial crisis on the financial market in 2008 has caused many financial giants to close down one after another, and the financial industry has been able to reshuffle, under the situation of the coexistence of new challenges and new opportunities. It is particularly important to speed up the theoretical and practical level of financial industry supervision in China. At present, after many years of research and exploration, The financial market risk measurement method based on VaR has become an important risk control tool for many foreign financial institutions, while the domestic exchange rate management system and exchange rate risk supervision method are still relatively backward. It is not enough to form a stable exchange rate level and an advanced risk control system. In this situation, it is particularly important to study and study international advanced financial risk measurement methods. Based on the VaR method which is widely used at present, this paper has carried on the development and the thorough research. From the point of view of VaR estimation method based on nonlinear quantile regression, this paper hopes to provide a powerful tool for the supervision of exchange rate risk. The sequence is nonnormal, The characteristics of peak thick tail, asymmetry, fluctuation agglomeration, conditional variance, etc. The traditional time series model ARMA model or nonlinear least square regression model are used to fit the sequence. It is found that these two models can not describe these characteristics of the sequence. Because the GARCH family model makes full use of the conditional heteroscedasticity of the ARCH model and the characteristics of the student t distribution and the GED distribution, It is found that all kinds of GARCH models can describe the heteroscedasticity of the return series well. The conclusion can be drawn from the empirical analysis that the nonlinear quantile regression model can estimate the VaR of time series financial data. It can comprehensively describe the relationship between financial data under different correlation structures, so that we can get the flexible measurement method and the conclusion of multiple choices.
【学位授予单位】：天津科技大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F832.5;F224

【相似文献】