预测VaR:高阶矩可行域未必越广越好  

Predicting VaR: the Wider Feasible Zone of Higher Order Moments May not Imply Better Performance

作　　者：方立兵[1] 郭炳伸 

机构地区：[1]电子科技大学经济与管理学院 [2]台湾政治大学国际贸易系

出　　处：《数量经济技术经济研究》2011年第11期124-137,共14页Journal of Quantitative & Technological Economics

摘　　要：高阶矩可行域反映了分布函数对样本高阶矩特征(非对称和尖峰、厚尾等)的适应能力,是影响VaR预测绩效的重要因素之一。本文以中、美、英、日四个股市的收益为样本,实证比较了五种高阶矩可行域各不相同的分布函数的VaR预测绩效。结果发现,可行域过于宽广(狭窄)的分布易于高(低)估样本的偏度和峰度,并高(低)估VaR,而可行域适中的广义偏斜-t和偏斜-t分布的预测绩效相对较好。此结果可为后续研究改进VaR的预测绩效提供有益借鉴。The feasible zone of higher order moments reflects the suitability of a distribution for various samples＇ characteristics of higher order moments, such as the asymmetry, leptokurtosis and heavy tails. Thus, it is also an important factor on the performance of VaR prediction. This paper empirically compares the performance of VaR prediction conditional on five types of non-normal distributions which have different feasible zone of higher order moments. Sampling daily returns from the main stock markets of the world, such as China, US, UK and Japan, we document the following empirical results that the widest （narrowest） feasible zone of higher order moments may over- （under-） estimate the skewness and kurtosis, and the same to the VaR prediction. Skewed t and Generalized Skewed t distribu- tions perform better. The conclusion will benefit the following studies on improving the performance of VaR prediction.

关 键 词：风险价值 条件分布 高阶矩 

分 类 号：F830.91[经济管理—金融学]