基于非参数GARCH-EVT模型的上证市场风险度量

发布时间：2017-12-31 14:16

本文关键词：基于非参数GARCH-EVT模型的上证市场风险度量　出处：《南京财经大学》2014年硕士论文　论文类型：学位论文

【摘要】：随着金融市场在我国不断的发展,作为其重要组成部分的股票市场的风险也愈发的受到关注和重视。本文拟研究上海证券市场的市场风险,对度量市场风险的传统方法进行了改进和补充,力图得到一种可以广泛使用且更为准确的测度模型。对于风险的测度,当今的主要方法是估算风险价值(VaR),这是应用最为广泛的一种度量风险的工具。该方法将原本抽象、不易描述的风险变成了直观的数字,具有简洁明了的特点。但传统VaR计算方法都需要较多的前提假定,很多情况下这些假设与现实是不相符的,这就降低了模型的可信性。本文将非参数GARCH模型和极值模型(EVT)相结合,充分利用这两个理论的特点,达到优化VaR估计方法的目的。利用非参数GARCH模型可以较好的克服原始收益率数据的群聚波动性,并且获得优于其他GARCH族模型的拟合结果,从而为极值模型提供理想的原始数据；极值理论只研究市场数据分布的尾部特征,进而避免了对整个数据进行分布假设；非参数GARCH族模型和极值模型的结合,估计出的动态VaR值也优于传统方法的静态VaR。本文首先介绍了当前金融市场的发展情况和金融市场风险的重要意义,总结了风险度量理论在国内外的研究现状,介绍了VaR的基本概念、计算原理以及方法。其次本文建立了非参数GARCH模型,通过理论和实证检验证明了其优于其他GARCH族模型,特别在计算的VaR时,其优点可以得到更大的发挥；最后,在极值理论的基础上,得到VaR结果,并将极值模型计算的静态VaR值和非参数GARCH-EVT模型计算的动态VaR值的准确率进行了对比,通过比较发现,利用非参数GARCH-EVT模型明显的提高了VaR的计算准确率。本文的创新之处主要是将非参数GARCH模型引入到股票市场的风险度量中,证明了其在估计VaR时优于其他GARCH族模型。同时,针对VaR的估计,本文引入了极值理论,并通过建立一系列的方法确定较为客观的阈值；最后将非参数GARCH模型与极值模型相结合,组成非参数GARCH-EVT模型,扩大了传统方法的应用范围,放松了传统假定,得到了更为准确稳定的估计结果。
[Abstract]:With the development of financial market in the continuous development of our country, as an important part of the risk of the stock market has become more and more concern and attention. This paper intends to study the Shanghai stock market risk, the traditional method for measuring market risk is improved and added, force diagram which can be widely used and more measure model accurate. For the risk measure, the main method is the estimation of value at risk (VaR), which is a risk measurement tool widely used. The method of the abstract, the risk is not easy to describe a visual digital, has the characteristics of simple. But the traditional VaR calculation method is the premise we need more assumptions, in many cases, these assumptions are not consistent with the reality, which can reduce the reliability of the model. In this paper, the parameters of GARCH model and extreme value model (EVT) combined with full use of the two The characteristics of the theory, to optimize the VaR estimation method. Using the nonparametric GARCH model can overcome the original rate of return data clustering volatility, and obtain the fitting results is better than the other GARCH models, the original data so as to provide an ideal model of extreme value; extreme value theory only on the tail characteristics of market data distribution. To avoid the data distribution hypothesis; combined with non parametric GARCH models and extreme value model, dynamic VaR estimated value is superior to the traditional method of static VaR., this paper firstly introduces the important significance of the development and the current financial market risk of financial market, summarizes the current situation of research on risk measurement theory at home and abroad. The basic concept of VaR is introduced, the calculation principle and method. Secondly, this paper established the parameters of GARCH model, through theoretical and empirical test proves that it is superior to the He GARCH model, especially in the calculation of VaR, its advantages can be a greater role; finally, based on the extreme value theory, get the results of VaR, and the accuracy of dynamic VaR calculation model for the calculation of the static VaR extreme value and non parametric GARCH-EVT model value were compared, by comparison obviously, improve the calculation accuracy of VaR by non parametric GARCH-EVT model. The main innovation of this paper is the non parametric GARCH model is introduced to the stock market risk measurement, prove the estimation of VaR is superior to the other GARCH models. At the same time, according to VaR estimates, this paper introduces the extreme value theory, and more objective to determine the threshold value by setting up a series of methods; finally, combined with the non parametric GARCH model and extreme value model, non parametric GARCH-EVT model, expanding the scope of application of the traditional method, the traditional assumption was relaxed, The results are more accurate and stable.

【学位授予单位】：南京财经大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：F832.51;F224

【相似文献】