中国股票市场投资组合风险测度及其实证研究

发布时间：2018-01-12 06:02

本文关键词：中国股票市场投资组合风险测度及其实证研究　出处：《成都理工大学》2014年硕士论文　论文类型：学位论文

【摘要】：随着国际金融市场的动荡不定,各国投资者们都急切需要加强金融风险管理技术；金融市场风险一直以来就是各投资者们密切关注的重要因素,而有效进行金融风险的度量则为各投资者进行科学合理投资提供了理论依据；由于金融市场特有的波动特征,如收益率的尖峰厚尾性、波动聚类性、波动的溢出效应、杠杆效应以及波动的长记忆性和持续性等,使得金融风险准确测度存在较大的的困难。 1952年Markowitz建立均值—方差理论首次对风险和收益进行量化,获得了风险最小化的计算方法,开启了金融风险度量的大门。但是均值—方差理论存在一定的前提假设与实际金融事实不符,不能很好的描述金融特征。VaR(Value atRisk)在险价值技术是上世纪90年代以后发展起来的一种新型风险度量方法,具有较广的应用范围,和传统的风险度量方法相比,具有更高的实用价值。然而VaR技术也存在很多问题,比如它没有考虑尾部风险、容易使投资者采用实际风险更高的的投资、不具有次可加性违背风险分散化原理,很难应用到投资组合优化等。针对VaR技术的不足,理论界提出了许多的改进方法,ES(也称C-VaR)风险度量法就是其中之一。ES风险度量法具备有效度量损失在VaR水平之上的损失值,能充分的测量尾部损失、具有次可加性能、能很容易被采用在资产组合中等优势。金融序列的波动预测模型主要分为两类：GARCH族模型和SV族模型。Engle(1982)提出ARCH模型能有效描述金融时间序列的波动聚类特征。Bollerslev(1986)进一步提出广义ARCH模型(即GARCH),它不仅能很好地描述时间序列的波动聚类特征,而且还能很好的捕捉金融数据分布的厚尾特性。由于SV模型不依赖于历史信息,许多学者证实SV模型很适合引入于金融领域的实际研究。并且国内外有较多的研究表明SV模型具有比GARCH模型更好的金融时间序列的刻画能力。接下来探讨金融市场间的相关性关系,金融市场间的相关性研究非常重要,尤其是涉及到资产定价、资产组合,波动的传导和溢出效应等问题。常用的相关分析方法有线性相关系数、Granger因果分析法,但它们都存在很大的缺陷：(1)金融时间序列的特性表明简单的线性相关关系不符合金融事实,无法分析变量间存在的非线性关系,而且各变量间主要表现为非线性关系；(2)Granger因果关系法一般只能得出是否存在因果关系的简单定性的结论,无法进行定量描述。为了解决以上难题我们引入Copula函数,它最早由Sklar(1959)提出。Copula函数作为一种新的统计方法来分析变量间的相关关系,被广泛应用于非参数统计领域,尤其对随机变量间的相关性研究非常可靠。他的特点在于能直接对变量的边缘分布函数作为变量进行分析,而不限制边缘分布的选择。Frees和Valdez(1998)、 Embrechts (1999) Mcneil和Straumann (1999)等人将Copula技术引入到金融领域,为金融领域模型的进一步研究的提供了新思路。本文主要针对金融市场的波动特征,采用GARCH族模型和SV族模型对上证综指和深圳成指的金融数据进行波动率预测和在险价值VaR、期望损失ES的度量,有效刻画我国股市波动的主要特征：尖峰厚尾性、波动聚集性、杠杆效应、溢出效应等；同时证实ES度量法相比VaR更有优势。通过波动率模型对收益率序列的边缘分布进行准确的描述,接下来引入Copula函数对沪深两指数收益率的联立分布准确刻画他们的相依结构,然后分别计算投资组合VaR和ES值,并进行对比,证实Copula函数下的ES风险度量方法对我国投资组合风险测度有较好的适用性,并获得了在组合风险值最小情形下的最佳投资比例。本文为指导我国投资者在理性投资理念下,科学进行投资决策提供了理论依据。
[Abstract]:Along with the volatility of international financial market , investors are eager to strengthen financial risk management technology ; financial market risk has been an important factor that investors pay close attention to , while the effective financial risk measurement provides theoretical basis for investors to make scientific and reasonable investment . Because of the unique fluctuation characteristics of financial markets , such as the peak thickness of yield , the volatility clustering , the spillover effect of fluctuation , the lever effect and the long memory and persistence of fluctuation , make the financial risk accurate measure have great difficulty . The theory of VaR ( Value atRisk ) is a new type of risk measurement method developed in the 1990s . However , VaR technology has many problems , such as the fact that it does not consider the tail risk , it is difficult to apply to portfolio optimization . The volatility forecasting model of the financial sequence is divided into two categories : ARCH model and SV family model . Engle ( 1982 ) proposed ARCH model to describe the volatility cluster feature of the financial time series . Bollerslev ( 1986 ) further proposed the generalized ARCH model ( that is ) , it can not only describe the volatility clustering feature of time series well , but also can capture the thick tail characteristic of financial data distribution well . Because SV model is not dependent on historical information , many scholars confirm SV model is suitable for the real research in financial field . In order to solve the above problems , we introduce Copula function as a new statistical method to analyze the relationship between variables , which is widely used in the field of non - parametric statistics . Based on the volatility characteristics of financial markets , this paper forecasts the volatility of Shanghai Composite Index and Shenzhen Chengzhi financial data by using the ARCH model and SV family model . It also proves that the ES metric method is more advantageous than VaR . It also proves that the ES measurement method has better applicability to the risk measure of Chinese stock market . Then , we prove that the ES risk measurement method under Copula function has better applicability to the risk measure of portfolio risk .

【学位授予单位】：成都理工大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：F832.51;F224

【参考文献】