基于加权双核局部线性估计的市场风险分析

发布时间：2018-05-07 10:24

本文选题：CVaR + CES　；参考：《广西师范大学》2015年硕士论文

【摘要】：最近三十年,受到经济全球化、信息技术以及金融理论等因素的影响,全球金融市场得到了迅速发展。这使得全球的金融市场变得更加开放,全球范围内的资本流通速度加快而且更加自由化。在全球金融市场中不同风险特性的资本得到重新配置和组合,这导致全球金融市场的运作方式和风险的表现产生了很大程度的改变,因而金融市场出现了前所未有过的波动。与此同时,金融机构为了规避金融风险,提升市场竞争力,开展了一系列的金融创新活动,在技术进步和管制放松的刺激情况下,这些活动就显得异常活跃[2]。随着全球经济的快速发展,金融全球化、自由化逐渐加强。随之而产生的市场风险也成为人们的关注热点,全球范围内的专家学者都对其展开了研究。怎样去量化市场风险,也即怎样去测定市场风险,就成为摆在我们面前急需解决的一个大问题。标准差(Standard Deviation)、绝对离差(Absolute Deviation、偏差(Deviation)、下端部分矩(Lower Partial Moment)、风险价值VaR、条件风险价值CVaR、条件期望损失CES等度量方法是目前正在使用或已经被提出来作为度量风险的主要工具。其中,由于巴塞尔委员会对于VaR的认可,VaR开始受到全球金融分析方面专家的青睐,被选用作为金融机构风险管理的国际统一标准。随着VaR模型及其计算方法的不断发展和优化,以及Artzner在1999年初次提出一致性公理后,将VaR作为风险度量的标准遭到了质疑,因为有研究者在理论和实证分析这两方面都证实了VaR对于次可加性并不满足,因而得出VaR不是一致性风险度量的结论[3]。在此情况下,人们为了弥补VaR的不足,于是就开始构造和设计一个既容易估计和计算又满足一致性公理的风险度量。Rockafellar[5]和Uryasev提出了CVaR;Scaillet[6]提出了CES的非参数估计；Artzner[7]等提出了最坏条件期望WCE；Acerbi[4]提出了谱风险度量等等,并证明了它们既是一致性风险度量又能够方便计算。其中CES和CVaR应用的较为广泛,因为它们相对于VaR度量更有优势。本文介绍了国内外对VaR、CES以及CVaR的研究情况,并且简单介绍了四种窗宽选择的方法：主观选择法、参照标准分布法、经验法则和无偏最小平方交叉实证法。在模拟研究中,窗宽通过经验法则选出,使用R软件编程,生成随机数,运用局部线性估计模拟出CVaR和CES的估计值,列出不同分位数p下两者的数值,并进行相应的比较,分析CVaR和CES的变化趋势。然后,以深成指数和上证指数为研究样本,用Eviews软件画出两只股票的日收益率图像,同时对股票日收益率进行ADF检验。从ADF检验和日收益率图像可以得出,深成指数序列和上证指数日收益率序列都是平稳的。然后运用模拟中的局部线性估计对条件风险价值CVaR和条件期望损失CES进行估计,计算出股票市场数据的(:VaR和CES值,统计分析出超过真实VaR的百分比,以此来验证模拟中的结论。
[Abstract]:In the last three decades, the global financial market has developed rapidly under the influence of economic globalization, information technology and financial theory. This makes global financial markets more open, global capital flows faster and more liberalized. In the global financial market, the capital with different risk characteristics has been reconfigured and combined, which has led to a great change in the operation mode and risk performance of the global financial market, resulting in unprecedented fluctuations in the financial market. At the same time, in order to avoid financial risks and enhance market competitiveness, financial institutions have launched a series of financial innovation activities. With the rapid development of the global economy, financial globalization and liberalization are gradually strengthened. The market risk has become the focus of attention, which has been studied by experts and scholars all over the world. How to quantify the market risk, that is, how to measure the market risk, has become a big problem that needs to be solved in front of us. Standard deviation, absolute deviation, lower Partial moment, risk value, conditional expectation loss, CES and so on, are currently used or proposed as the main tools to measure risk. Among them, due to the Basel Committee's recognition of VaR began to be favored by experts in the field of global financial analysis, it is chosen as the international standard of risk management of financial institutions. With the continuous development and optimization of the VaR model and its calculation methods, and after Artzner first proposed the consistency axiom in 1999, the use of VaR as the criterion of risk measurement has been questioned. Because some researchers have proved that VaR is not satisfied with subadditivity in both theoretical and empirical analysis, it is concluded that VaR is not a consistent risk measure [3]. In this case, in order to make up for the deficiencies of VaR, So we began to construct and design a risk measure which is easy to estimate and calculate and satisfy the consistency axiom. Rockafellar [5] and Uryasev proposed Cvar Rn scale [6], and CES's nonparametric estimation Artzner [7], and so on, and proposed the worst-case expectation WCEE _ Cer _ bi [4] and proposed spectral risk measurement, and so on. It is proved that they are not only consistent risk measurement but also easy to calculate. CES and CVaR are widely used because they have more advantages than VaR metrics. This paper introduces the domestic and foreign research on CVaR and CVaR, and briefly introduces four methods of window width selection: subjective selection method, reference standard distribution method, empirical rule and unbiased least square cross empirical method. In the simulation study, the window width is selected by the rule of thumb, the random number is generated by using R software, the local linear estimation is used to simulate the estimated values of CVaR and CES, and the values under different quantiles p are listed and compared accordingly. Analyze the change trend of CVaR and CES. Then, using Shencheng Index and Shanghai Stock Exchange Index as the research samples, the daily yield image of two stocks is drawn by Eviews software, and the ADF test of the daily return rate of stock is carried out at the same time. From the ADF test and the daily yield image, it can be concluded that both the Shenzhen exponent series and the Shanghai stock index daily yield series are stable. Then the local linear estimator is used to estimate the conditional risk value (CVaR) and the conditional expectation loss (CES). The ratio of CES and CES of the stock market data is calculated and the percentage above the real VaR is statistically analyzed to verify the conclusion in the simulation.
【学位授予单位】：广西师范大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：F224;F830

【相似文献】