基于GARCH-stable模型的原油市场风险度量
发布时间：2019-01-13 10:28
【摘要】：随着原油价格的暴跌,原油的市场价格风险越来越受到人们的关注。对波动率的衡量最常用的方法是GARCH模型。但是大量的研究表明传统的GARCH的残差仍然存在明显的尖峰厚尾性,也就是说GARCH模型的一个基本假设：残差服从独立同分布标准正态分布,是不成立的；而且GARCH有低估风险的倾向,模型风险不容忽视。解决这些问题的一个办法是用某个厚尾分布作为GARCH模型的条件分布。本文采用stable分布作为其条件分布。近年来的分形热让stable分布重新受到关注,stable分布是一种具有尖峰厚尾性的分布,通过四个参数：特征指数、偏斜指数、尺度参数和位置参数,可以灵活的调节分布的尾部、峰度、尺寸,甚至偏斜度。但是由于它的分布函数不存在显式的表达式,只有通过数值法才能实现其价值,计算机技术的发展极大的促进了stable分布的应用。本文以WTI和Brent两个世界上最大的原油品种为例,研究原油市场的价格风险。首先证明了原油价格变化可以用独立同分布stable分布拟合,这时价格波动率能够用stable分布的尺度参数σ度量,但是这里的波动率停留在静态的层次上,不具有时变性。然后本文将stable分布作为GARCH模型的条件分布,提出了GARCH-stable模型的概念,并用来预测原油市场的价格波动率,把stable分布的使用扩展到了动态的情形。本文使用极大似然估计法作GARCH-stable模型的参数估计,得到了模型的条件波动率σt,并且采用图检验法对模型的残差进行检验,发现stable分布对模型残差的拟合度很高,有效地解决了GARCH模型的残差与条件分布不吻合的问题,用它作为GARCH模型的条件分布非常合适。进一步地,本文在前面得到的条件波动率σt的基础上,采用最著名的风险度量方法--VaR模型,度量原油市场风险。为了对比模型的优劣,本文对VaR模型做了失败率检验。95%和99%两个置信度下的检验结果表明了GARCH-stable模型是合适的,相比之下,GARCH-normal等模型虽然通过了95%置信度下的失败率检验,但是却没有通过却没有通过99%置信度下单失败率检验。作为补充,本文还简要介绍了分形理论与分形分析方法,并对stable分布做了具体的介绍。
[Abstract]:With the collapse of crude oil price, people pay more and more attention to the market price risk of crude oil. The most commonly used method for measuring volatility is the GARCH model. But a large number of studies show that the residual of traditional GARCH still has obvious spike and thick tail, that is to say, a basic assumption of GARCH model: the standard normal distribution of residual clothing from independent same distribution, is not true; And GARCH has the tendency to underestimate the risk, model risk can not be ignored. One way to solve these problems is to use a thick tail distribution as the conditional distribution of the GARCH model. In this paper, stable distribution is used as its conditional distribution. The fractal heat in recent years has refocused the stable distribution. The stable distribution is a kind of distribution with sharp peak and thick tail. It can adjust the tail of the distribution flexibly through four parameters: characteristic index, skew index, scale parameter and position parameter. Kurtosis, size, even skew. However, because its distribution function does not have explicit expression, it can realize its value only by numerical method. The development of computer technology has greatly promoted the application of stable distribution. This paper takes WTI and Brent as examples to study the price risk of crude oil market. Firstly, it is proved that the price change of crude oil can be fitted by independent and distributed stable distribution, and the price volatility can be measured by the scale parameter 蟽 of the stable distribution, but the volatility stays at the static level and does not have time variability. Then, the stable distribution is taken as the conditional distribution of the GARCH model, and the concept of GARCH-stable model is put forward, which is used to predict the price volatility of crude oil market, and the use of stable distribution is extended to the dynamic case. In this paper, the maximum likelihood estimation method is used to estimate the parameters of the GARCH-stable model. The conditional volatility 蟽 t of the model is obtained, and the residual error of the model is tested by the graph test method. It is found that the stable distribution has a high fitting degree to the model residual. The problem that the residual error of GARCH model is not consistent with the conditional distribution is effectively solved, and it is very suitable to use it as the conditional distribution of GARCH model. Furthermore, on the basis of the conditional volatility 蟽 t, this paper uses the most famous risk measurement method, VaR model, to measure the market risk of crude oil. In order to compare the advantages and disadvantages of the model, the failure rate of the VaR model is tested. The test results under 95% and 99% confidence level show that the GARCH-stable model is suitable. Although GARCH-normal and other models pass the 95% confidence test, they fail to pass the 99% confidence test. As a supplement, fractal theory and fractal analysis method are briefly introduced, and stable distribution is introduced in detail.
【学位授予单位】：浙江工商大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：F416.22
本文编号：2408348
[Abstract]:With the collapse of crude oil price, people pay more and more attention to the market price risk of crude oil. The most commonly used method for measuring volatility is the GARCH model. But a large number of studies show that the residual of traditional GARCH still has obvious spike and thick tail, that is to say, a basic assumption of GARCH model: the standard normal distribution of residual clothing from independent same distribution, is not true; And GARCH has the tendency to underestimate the risk, model risk can not be ignored. One way to solve these problems is to use a thick tail distribution as the conditional distribution of the GARCH model. In this paper, stable distribution is used as its conditional distribution. The fractal heat in recent years has refocused the stable distribution. The stable distribution is a kind of distribution with sharp peak and thick tail. It can adjust the tail of the distribution flexibly through four parameters: characteristic index, skew index, scale parameter and position parameter. Kurtosis, size, even skew. However, because its distribution function does not have explicit expression, it can realize its value only by numerical method. The development of computer technology has greatly promoted the application of stable distribution. This paper takes WTI and Brent as examples to study the price risk of crude oil market. Firstly, it is proved that the price change of crude oil can be fitted by independent and distributed stable distribution, and the price volatility can be measured by the scale parameter 蟽 of the stable distribution, but the volatility stays at the static level and does not have time variability. Then, the stable distribution is taken as the conditional distribution of the GARCH model, and the concept of GARCH-stable model is put forward, which is used to predict the price volatility of crude oil market, and the use of stable distribution is extended to the dynamic case. In this paper, the maximum likelihood estimation method is used to estimate the parameters of the GARCH-stable model. The conditional volatility 蟽 t of the model is obtained, and the residual error of the model is tested by the graph test method. It is found that the stable distribution has a high fitting degree to the model residual. The problem that the residual error of GARCH model is not consistent with the conditional distribution is effectively solved, and it is very suitable to use it as the conditional distribution of GARCH model. Furthermore, on the basis of the conditional volatility 蟽 t, this paper uses the most famous risk measurement method, VaR model, to measure the market risk of crude oil. In order to compare the advantages and disadvantages of the model, the failure rate of the VaR model is tested. The test results under 95% and 99% confidence level show that the GARCH-stable model is suitable. Although GARCH-normal and other models pass the 95% confidence test, they fail to pass the 99% confidence test. As a supplement, fractal theory and fractal analysis method are briefly introduced, and stable distribution is introduced in detail.
【学位授予单位】：浙江工商大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：F416.22
【共引文献】
相关期刊论文 前1条
1 伍笑萍;李忠民;;基于GARCH模型的WTI原油现货市场的风险分析[J];合肥工业大学学报(自然科学版);2013年09期
相关博士学位论文 前1条
1 陈磊;石油市场的内外部联系、价格发现与风险管理研究[D];电子科技大学;2012年
，本文编号：2408348
本文链接：https://www.wllwen.com/kejilunwen/shiyounenyuanlunwen/2408348.html
