机构地区: 中山大学
出 处: 《数学的实践与认识》 2009年第5期34-41,共8页
摘 要: RiskMetrics是当今最为流行的风险度量模型,然而其基础假设-标准化收益服从正态分布,却备受置疑.放宽此假设,以更灵活的t分布,广义误差分布,混合正态分布,Johnson Su-正态,Pearson IV分布代替,建立了五种扩展的RiskMetrics模型.我们用沪深股市日收益数据进行实证比较分析,回测结果表明,扩展模型明显优于标准模型,而基于非对称分布假设的模型优于基于对称分布的模型. RiskMetrics is the most popular risk management model. But one of its basic assumptions, the standardized return following normal distribution, is in discussion. In this paper, the standard RiskMetrics model is expanded. The normal distribution is replaced by some more flexible distributions such as t, generalized error distribution, mixed normal, Johnson Su-normal and Pearson Ⅳ distribution. Our empirical comparison using stock daily return of china shows, the expanded models perform better than the standard, and the expanded models based on asymmetric distribution perform better than that based on symmetric distribution.
关 键 词: 风险价值 模型 分布 广义误差分布 混合正态分布 正态 分布 回测检验
领 域: [理学] [理学] [自动化与计算机技术] [自动化与计算机技术]