基于Copula贝叶斯估计的行业风险差异分析  

Copula Bayesian Approach to Evaluating the Dynamic Effect of the System Risk at Industry Level

作　　者：赵宁[1] Winston T Lin 孙雪卿[1] 

机构地区：[1]东北财经大学金融学院,辽宁大连116025 [2]325A Jacobs Management Center, Buffalo, New York 14260, USA

出　　处：《数学的实践与认识》2015年第10期17-27,共11页Mathematics in Practice and Theory

基　　金：国家自然科学基金面上项目(71273042);国家自然科学天元基金(11226250);国家教育部一般项目(12YJA790078);国家博士后基金(2013M541236);辽宁省教育厅重点实验室基础研究项目(LZ2014048);东北财经大学青年培育项目(2014114,批准号DUFE2014Q22)

摘　　要：通过对常替代弹性资本资产定价模型中投资标度问题的分析,提出了Copula贝叶斯估计方法用以获得系统风险β与投资标度比λ的联合后验分布.Copula贝叶斯估计方法针对数据非正态特征及强相关性特征而构建,采用Copula函数取代原有普通贝叶斯估计方法中的正态假设.传统贝叶斯估计方法假设了正态的似然函数,忽略了数据可能存在尖峰后尾等在金融实证数据分析中普遍存在的非正态情况.Copula贝叶斯估计算法采用半相依回归法处理数据的强相关性问题,将原有函数依照数据形式假设为非正态结构.针对来自6个工业产业24组公司数据的系统风险参数β与其对应的投资标度参数比λ进行估计,获得不同行业中系统风险参数与投资标度之间的动态关系并进行分析,为业界投资及相关研究提供有效参考建议.We propose the copula Bayesian estimation approach to get the posterior distribution of the parameters of the system risk beta and the investment horizon ratio lambda in constant elasticity of substitution capital asset pricing model （CES-CAPM） with the investment horizon. The copula Bayesian approach instead of the traditional Bayesian estimation is built to consider the pattern of the data with the strong correlation and the non-normal dis- tributions. The potent problem of the traditional Bayesian estimation is that the assumption of normal likelihood function ignores some fluctuations such as high peak and fat tail relative to kurtosis and skewness, which have been frequently reported in financial data analyses. We choose the Seemingly Unrelated Regressions method which is concerning the affect of the correlation of residuals. The reason we chose the copula function is to fit the pattern of the data. The empirical work analyzes the interaction of the system risk and the investment horizon in 24 companies from six different industries. Under this situation, the conclusion and remarks is given.

关 键 词：资本资产定价模型 投资标度 系统风险值 Copula贝叶斯 

分 类 号：F272.3[经济管理—企业管理] O212.8[经济管理—国民经济]