VaR方法在沪深股市风险测量中的应用研究

【作者】 刘毅；

【导师】 吴润衡；

【作者基本信息】 北方工业大学 ， 数量经济， 2006， 硕士

【摘要】 VaR风险价值方法是上世纪90年代以后发展起来的一种新型风险管理工具,作为一种余融风险测量和控制的模型,它简单易操作,应用范围广,相比于传统的金融风险管理模型,具有更高的实用价值。 中国证券市场从始至今只有十多年时间,与发达国家较为成熟的证券市场相比,尚处于市场发展的初期阶段,也处于一个复杂的、多变的风险环境中。此外,中国已于2001年加入WTO,根据我国政府对WTO的承诺,中国证券市场要逐步对外开放。因此对风险的管理和控制也显的更为重要。 目前,VaR的度量方法有很多,而且,由不同方法计算出的VaR值往往相差很大,因此,寻找与中国股票市场特点相适应的风险度量方法成为一个难题。本文给出了一种将理论分析与实证分析相结合的VaR模型精确性评价方法。 本文应用了VaR计算的方差——协方差方法中的三种典型的模型,分析了在不同置信水平下和不同分布假设下模型的精确程度。所选取的三种模型分别是GARCH、EGARCH和TARCH模型,通过其对波动率的估计来计算VaR。另外,还针对每种模型采用了正态分布、t分布以及广义误差分布的假设,共计九种方法。本文应用这些模型对上证180指数和深圳成份指数的日收益率序列的VaR估计进行了实证研究。 文中所用的检验方法有两种,分别是Christoffersen区间预测法和损失函数检验法。 最终研究结果表明,股指收益率序列的分布形态假设、样本数据量以及选取的置信度都对VaR模型的精确程度有很大的影响。更多还原

【Abstract】 VaR technique is a new risk management method that has been developed in 1990’s. As a quantitative model to measure and control financial risk, compared with traditional models, it is easy to understand and apply so as to have more practical significance.It has only more than ten years since the stock market of China came into being. Compared with the more complete developed countries’ stock market, it is in the earlier developing period, also in the complicated and mercurial risk environment. In additional, since China has joined WTO in 2001, its securities business will be open to foreign investors in a short time. So it is an important and urgent task to pay attention to risk management and controlling.Nowadays, we can use a lot of models to estimate VaR, but different model often deduces different VaR value. So it is difficult to find appropriate methods to manage the risk of China stock market. This paper analyzes the accuracy of different evaluate method by combined theoretical analysis and empirical analysis together.This paper compares three typical models of variance-covariance methods, and analyze the accuracy of different evaluate method with certain distribution and confidence degrees. We evaluate VaR with GARCH, EGARCH and TARCH models and assume certain distribution, such as normal distribution, Student-t distribution and generalized error distribution (GED), 9 methods in all. We compare the estimations of VaR in an application to daily returns on the Shanghai 180 index and Shenzhen Component index.There are two tests methods to evaluate the accuracy of the VaR model in this paper. One is Christoffersen Test, and the other is Loss Function Test.The finding of this paper indicated that the distribution assumption of the return series, Sample data quantity and confidence degree are all have the tremendous effect on the accuracy of the VaR models.更多还原