基于因子Copula模型的我国大型上市公司股票收益关联性及风险分析

发布时间：2018-04-19 00:10

本文选题：单因子Copula模型 + 嵌套Copula模型　；参考：《吉林大学》2017年硕士论文

【摘要】：在改革开放进一步深化和经济发展的不断推动下,我国金融市场逐步发展健全和完善,金融市场之间的依赖性和金融资产的价格协同效应愈来愈显著,其中股票市场作为金融市场的重要组成部分,不同市场、不同板块、不同行业以及不同股票之间常常存在着联动效应,某一市场或资产的波动,经常会引起其他市场或资产的波动,导致风险会迅速波及、传染、放大至其他市场或资产。随着我国股票市场的深入发展,不同上市公司之间的联系和依赖越来越强,公司股票之间的关联性也越来越明显,对我国大型上市公司股票收益之间的关联性和投资风险进行分析,对投资组合构建、市场风险管理乃至股市的健康发展都有着十分重要的意义。本文基于Copula理论基础,利用因子Copula模型和结构因子Copula模型中的嵌套Copula模型,分析了以沪深300成分股为代表的我国大型上市公司股票的收益率序列,计算得到了不同行业内每对股票收益之间的Spearman秩相关系数、相依尾部加权测度和不同资产组合的VaR和ES,以此分析了不同行业内各公司股票收益的关联性和投资组合风险,以及以全部沪深300成分股为代表的整个市场的投资组合风险。本文选取了沪深300成分股近5年的日对数收益率序列,剔除上市时间不满5年的股票,利用两阶段极大似然估计法,首先采用GARCH(1,1)-Gaussian模型、GARCH(1,1)-t模型分别对每只股票收益率序列进行拟合,并用AIC信息准则选择拟合效果较好的模型,经过对标准残差序列的K-S检验和Ljung-Box自相关检验发现,GARCH(1,1)-Gaussian模型、GARCH(1,1)-t模型可以较好的拟合各收益率序列的边缘分布,并且利用单因子Copula模型对各公司股票收益的标准残差序列进行拟合,发现在所有17个二级行业中,保险、材料、地产、能源、汽配、食品饮料、银行、运输、资本市场等9种行业的股票收益序列拟合效果较好的为单因子BB1 Copula模型,公用、零售、媒体、耐用服装、软件、硬件、制药生物、资本品等8种行业的股票收益序列拟合效果较好的为单因子Rotated Gumbel Copula模型;同时本文利用结构因子Copula模型中的嵌套Frank Copula模型,对17个行业的全部股票收益残差序列进行了拟合,并得到了相关模型参数。通过极大似然估计得到相关参数后,本文利用单因子Copula模型计算了每个行业内不同股票收益序列之间的Spearman秩相关系数、相依尾部加权测度,分别分析其相关性和尾部相关性,发现大部分配对股票收益的秩相关系数在0.3到0.8的区间内,呈明显的中度正相关,每个行业内部的整体相关性也呈明显的中度正相关,其中资本市场、保险、能源、运输等行业的内部整体关联性较强,食品饮料、硬件等行业整体关联性较弱;在尾部相关性方面,通过对各行业内所有配对成分股相依尾部加权测度的均值分析,发现各行业内部均呈正的尾部相关性,除银行业的上下尾部加权测度均值一致外,其余行业均表现为下尾相关性强于上尾相关性,其中银行业和资本市场业的上下尾部加权测度均值均高于0.6,说明两个行业内部的整体尾部相关性均较强。最后,本文利用单因子Copula模型和嵌套Frank Copula模型,采取蒙特卡洛仿真技术,计算每个行业等权重投资组合和整个市场等权重投资组合的VaR和ES,分析其投资风险,发现在置信水平分别为99%、97.5%、95%的情况下,资产组合的VaR和ES较大的行业为零售业、银行业、硬件业、资本市场业、制药生物业、保险业等,其中零售行业由于只有两只成分股,分散程度低,整体风险明显高于其他行业,而对于整个市场的资产组合,由于分散程度较高,在不同的置信度下,VaR和ES均明显较低,投资风险明显低于单个行业的资产组合。
[Abstract]:Continue to promote the further deepen reform and opening up and economic development, China's financial market is gradually improving and perfecting the financial market development, the dependence between financial asset prices and increasingly significant synergies, including stock market as an important part of the financial market, the different markets, different sectors, different industries and between different stocks often there is a linkage effect, a market or asset volatility, will often lead to other markets or asset volatility, risk will quickly spread, infection, enlarged to other markets or assets. With the further development of China's stock market, listed companies between different links and rely on more and more strong, the company stock Association the more and more obvious, to China's large stock return correlation and investment risk analysis, portfolio construction, market risk management Have a very important significance and the healthy development of the stock market. Based on Copula theory, using the nested Copula model Copula model factor and structure factor Copula model, analyzes the CSI 300 stocks as the representative of China's large-scale listed stock return series, the Spearman rank correlation coefficient between different industries within each of the stock return calculation, dependent tail weighted measure and different portfolio of VaR and ES, analyzed the different industries within the company stock return correlation and portfolio risk, and the whole market with full CSI 300 stocks as the representative of the investment portfolio risk. This paper chooses the CSI 300 stocks daily log return rate series of nearly 5 years, excluding the listed stock time of less than 5 years, the maximum likelihood estimation method using two stages, firstly using GARCH (1,1) -Gaussian model, GARCH (1 1, -t) model respectively for each stock return series fitting, choose a good fitting effect model and AIC information criterion, through the test of K-S and Ljung-Box on the standard residual autocorrelation test showed that GARCH (1,1) -Gaussian GARCH (1,1) model, -t model can better fit the rate of return the sequence of marginal distribution, and using single factor Copula model standard error on the stock returns of each company sequence fitting, found in all 17 two industries, insurance, energy, materials, real estate, auto parts, food and beverage, banking, transportation, capital market and other 9 kinds of industry stock returns better fitting effect single factor BB1 Copula model, the public, media, retail, durable clothing, hardware, software, bio pharmaceutical industry, capital goods and other 8 kinds of stock return series of good fitting effect for the single factor Rotated Gumbel Copula model; at the same time Using nested Frank Copula model structure factor Copula model, all the stock return residuals of 17 industry fitting, and relevant model parameters. The relevant parameters obtained by maximum likelihood estimation, using the single factor Copula model Spearman rank correlation coefficient between different stock returns within each industry were calculated. The tail dependence weighted measure to analyze the correlation and tail dependence, respectively, found that most pairs of stock yield rank correlation coefficient in the range of 0.3 to 0.8, showed a moderate positive correlation showed a moderate positive correlation between the overall each within the industry, including capital markets, insurance, energy, transportation and other industries inside the overall association is strong, food and beverage, hardware and other industries overall weak associations; at the end of the correlation, the industry in a paired component Analysis of mean weighted measure of tail dependence, found the tail correlation within the industry positively, in addition to the banking industry on the lower tail weighted measure mean the same, the rest of the industry showed a lower tail correlation is stronger than the upper tail dependence, including banking and capital markets in the tail of weighted measure was higher than that in 0.6. The overall internal tail correlation of two industries are strong. Finally, using the single factor Copula model and the nested Frank Copula model, adopt the Monte Carlo simulation, calculate the weight of each industry portfolio and the market portfolio weights of VaR and ES, to analyze the investment risk, found in the confidence level were 99%, 97.5% 95%, under the condition that the assets of the combination of VaR and ES is larger for the retail industry, banking industry, hardware industry, pharmaceutical industry capital market, property insurance, health, the retail industry as a result of Only two constituent stocks, low dispersion, the overall risk is significantly higher than other industries, and for the entire market portfolio, due to the higher dispersion, VaR and ES are significantly lower under different confidence levels, and the investment risk is significantly lower than the single industry portfolio.

【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F832.51

【参考文献】