基于Copula模型的股票与债券投资组合策略研究

发布时间：2018-01-05 06:11

本文关键词：基于Copula模型的股票与债券投资组合策略研究　出处：《首都经济贸易大学》2017年硕士论文　论文类型：学位论文

【摘要】：金融市场风云变幻,2008年以来中国经历了两次比较大的经济波动,也反射出证券市场的系统性风险和指标监控的匮乏。对于投资者来说,如何在这样的市场环境中规避风险,获取收益是最关注的问题,在我国股票市场和债券市场的分割较为明显,但是两个市场又同时组成了资本市场的两大支柱,所以研究两者之间的关联关系,并且从投资组合的角度对投资者做出建议是十分具有现实意义的。随着研究的加深,研究者发现市场之间并不是简单的Granger因果关系或者线性相关关系,存在着“非正态性”“非对称性”“尖峰厚尾”和“波动聚集性”等特点,而Copula函数由于其对边缘分布没有要求,可以直接对相关关系进行建模等优势,在对金融时间序列相关性的刻画上发挥着重大的作用。在理论方面,本文描述分析了五种静态Copula函数、一种混合Copula函数,三种藤结构Copula函数以及三种动态Copula函数的定义和性质,并且介绍了风险度量的相关理论。在实证方面,本文选取上证指数、深证综合指数和中证债券综合指数为研究对象,上证指数和深证综合指数代表股票市场趋势和波动,中证债券综合指数是衡量债券市场趋势和波动的重要指数,由于影响的滞后效应和保证数据的连续性,本文未对数据非同期进行删减,首先对两个市场2008年来的发展现状和波动情况进行了定性的描述,之后用GARCH(0,4)-t模型来拟合上证指数的边缘分布,用GARCH(3,0)-t模型来拟合中证债券指数的边缘分布,用GARCH(1,1)-t模型来拟合深证指数的边缘分布,在验证股票市场和债券市场相依性方面,本文选取三种指数通过五种静态模型和三种藤结构Copula函数来验证,并对模型拟合效果进行分析;在对动态模型性质的研究中,我们选择上证指数和深证综合指数,并用三种时变模型并和一种静态混合Copula模型来拟合,对比分析选择出拟合效果最好的模型,最后实现Copula函数和风险价值的结合,根据历史收益率,通过蒙特卡罗方法随机产生服从GARCH-t边缘分布和Copula联合分布的随机数列,计算等比重下的Va R和CVaR的值,通过失败率检验的方法进行检验拟合的效果,并通过10000次的模拟进行投资组合的优化,选择出最优的投资组合。通过研究发现,在静态Copula中,相比椭圆类Copula函数,阿基米德Copula函数更能拟合收益率序列的“非对称性”和“尖峰后尾”的特征,而三种藤结构的Copula函数拟合的效果优于任何单一的静态模型。在对上证指数和深证指数进行静态混合Copula模型和时变SJC Copula模型的拟合后,发现时变SJC Copula模型更能衡量出时间序列之间不断变化的相依关系;基于藤结构Copula模型研究了股票和债券市场的最优配比,并对比了Copula-mean-Variance模型和mean-CVa R模型,验证了CVa R在衡量风险上的优越性。同时可以看到随着对于期望收益率的提高,最优的投资组合会向股票市场发生偏移,而且具有厚尾特征的股票的占比就会增大。
[Abstract]:The financial market since 2008 has experienced China amidst the winds of change, two large economic fluctuations, reflecting the lack of systematic risk of the securities market and index control. For investors, how to avoid the risk in such a market environment, income is the most concern in China's stock market and bond market segmentation the more obvious, but the two markets and formed the two pillars of the capital market, so the study of relationship between the two, is of great practical significance and from the perspective of portfolio for investors to make recommendations. As the research deepened, the researchers found that between the market is not a simple causal relationship or linear correlation Granger there is a relationship, "non normal" "non symmetry" "fat tail" and "volatility clustering" and other characteristics, and the Copula function because of its marginal distribution No, it can directly model the advantages of correlation, play a significant role in the portrayal of the correlation of financial time series. In theory, this paper describes and analyzes the five kinds of static Copula function, a hybrid Copula function, three kinds of rattan structure Copula function and three kinds of dynamic Copula function definition and the nature, and introduces the theory of risk measure. In the empirical analysis, this paper selects the Shanghai index and Shenzhen Composite Index and comprehensive index card bonds as the research object, the Shanghai Composite Index and Shenzhen composite index represents the stock market trend and volatility, CSI bond index is an important index to measure the bond market trend and volatility, because the influence of continuous lag effect and ensure the data, the data of non simultaneous deletion, firstly, the development status and the fluctuation of the two markets of 2008 A qualitative description, then use GARCH (0,4) -t model to fit the marginal distribution of the Shanghai index, GARCH (3,0) -t model to fit the marginal distribution of CSI bond index, GARCH (1,1) -t model to fit the marginal distribution of Shenzhen stock index, in the verification of the stock market and bond market dependence. This paper selects three index by five static models and three kinds of rattan structure Copula function to verify, and the model fitting effect is analyzed; in the research on the dynamic properties of the model, we select the Shanghai index and Shenzhen composite index, and with three kinds of time-varying model and a static Copula model to fit. Comparative analysis of selection of the best fitting effect of the model, finally using the Copula function and the value at risk, according to the historical return, by the method of Monte Carlo random edge obeys GARCH-t distribution and Copula joint distribution of random numbers The calculation of the proportion of Va, R and CVaR, to test the fitting effect through the method of failure rate test, and optimize the investment portfolio through 10000 times of simulation, select the optimal portfolio. The study found that in static Copula, compared to Copula elliptic function, Archimedes function Copula the fitting characteristics of the return series of "non symmetry" and "tail", and the effect of three kinds of Copula function fitting structure of the vine is superior to any single static model. SJC Copula model in fitting of Shanghai stock index and Shenzhen stock index for static Copula model and the time varying SJC Copula, found the model can measure more dependent changing relationship between time series; the optimal proportion of rattan structure based on Copula model of stock and bond market, and compares the Copula-mean-Variance model and mean-CVa R The model validates the superiority of CVa R in measuring risks. At the same time, we can see that with the increase of expected yield, the optimal portfolio will shift to the stock market, and the proportion of stocks with thick tail will increase.

【学位授予单位】：首都经济贸易大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F224;F832.51

【相似文献】