VaR和ES的非参数估计的模拟研究和实证分析

发布时间：2018-01-02 06:10

本文关键词：VaR和ES的非参数估计的模拟研究和实证分析　出处：《广西师范大学》2014年硕士论文　论文类型：学位论文

【摘要】：近二十年来,在经济全球化、现代金融理论以及信息技术等因素的影响下,全球金融市场迅速发展.这致使各国金融市场的开放程度不断加剧,资本在世界范围内的流动加速且自由化.不同风险特性的各种资本在全球金融市场内重新配置和组合下,很大程度地转变了全球金融市场的运作方式和风险表现,金融市场表现出前所未有的波动性.同时,金融机构通过展开一系列的金融创新活动来回避金融风险、提高市场竞争力,这些活动在管制放松和技术进步的刺激下异常活跃.在全球化和自由化的金融背景下,金融风险管理己成为人们关注的话题,而怎么把这些金融风险量化,即怎么测定金融风险,就成为了一个急需解决的问题. 标准差、绝对偏差、在险价值VaR,条件风险值CVaR,最坏条件期望WCE,期望损失ES等是当前正在使用或己经提出来的度量风险的主要工具.其中,因VaR在经济和金融领域的风险管控上发挥了极其重要的作用,因此巴塞尔委员会在1996年明确了用VaR方法来度量银行面临的市场风险的规定.从此以后,世界各地的金融分析方面的专家纷纷开始选用VaR作为金融机构风险管理的国际标准.随着VaR模型及其计算方法的不断发展和改进,自1999年Artzner et al初次提出一致性公理后,用VaR作为风险度量的标准遭到了质疑,由于有研究学者从理论和实证分析两个方面都证实了VaR不满足次可加性这个性质,从而得出VaR不是一致性风险度量的结果.为了弥补VaR的不足,人们开始构造设计一个不仅满足一致性公理而且容易估计和计算的风险度量Acerbi et.al提出了期望损失ES (Expected Shortfall)这一风险度量工具,并证明了它既是一致性风险度量,又能够方便计算. 深受各金融机构及管理人员重视的VaR有着致命的弊端—不具备次可加性,而ES风险测度满足一致性条件,使得ES在金融风险管理中成为越来越重要的风险度量工具. 在不同的窗宽条件下,窗宽越大,ES的核估计量越小.ES的两个非参数估计：一个是过度损失大于VaR的样本估计量；另一个是对前一个估计进行平滑化处理(Scaillet,2004Mathematical Finance)的核估计量.我们期望通过平滑处理,可以得到更精确的估计量.结果显示,平滑处理并没有得到损失更精确的估计量.这样的结果不同于VaR估计,对VaR进行平滑处理后可以使得估计量的方差和均方根误差均减小.因此,ES基于过度损失的样本平均的估计量,是ES的一个很好的估计. 全文分为四章：第一章,介绍了金融风险度量发展背景,以及风险管理的意义. 第二章,给出了VaR和ES的理论模型以及非参数估计量,以及给出VaR和ES估计量的标准误. 第三章,数值模拟.运用四种常见的时间序列模型,对ES的核估计和样本估计进行数值模拟,并比较它们的优劣,得到用ES核估计的方法计算的结果并没有比用ES样本估计得到的结果更精确. 第四章,是实证分析部分.用ES的样本估计方法对上证指数和深成指数的每日收益的对数序列进行了实证研究,得到的结果显示深成指数的波动率高于上证指数的波动率.
[Abstract]:Under the influence of economic globalization , modern financial theory and information technology , global financial markets have developed rapidly over the past two decades . Based on VaR , VaR is used to measure the market risk faced by banks . As a result of VaR model and its calculation methods , VaR is used as the standard of risk measurement . As a result of VaR model and its calculation method , VaR is used as the risk measure . The VaR of financial institutions and managers has a fatal defect - - no subadditivity , and ES risk measure meets the consistency condition , so that ES becomes more and more important risk measurement tool in financial risk management . The larger the window width is , the larger the width of the window , the smaller the kernel estimators of ES . Two non - parameter estimates for ES : One is that the excessive loss is greater than the sample estimate at VaR , and the other is that the estimation of the previous estimate is more accurate . The results show that the smoothing process does not get more accurate estimates . The result is different from the VaR estimation . The result is different from the VaR estimation . Therefore , it is a good estimate of ES based on the estimation of the sample average of over - loss . The full text is divided into four chapters : Chapter one introduces the background of financial risk measurement and the significance of risk management . In chapter 2 , the theoretical models of VaR and ES and the non - parameter estimation are given , and the standard error of VaR and ES estimator is given . In chapter 3 , the numerical simulation is carried out using four common time series models , the kernel estimation and sample estimation of ES are simulated numerically , and their advantages and disadvantages are compared , and the results obtained by the method of ES kernel estimation are not more accurate than the results obtained with ES samples . The fourth chapter is the empirical analysis part . By using ES ' s sample estimation method , the logarithmic series of the daily earnings of the Shanghai index and the deep index number is studied empirically . The results show that the fluctuation rate of the index is higher than the fluctuation rate of the Shanghai index .

【学位授予单位】：广西师范大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：O212.7;F832.51

【参考文献】