摘要
本文分别在Artzner等(1999)提出的一致风险测度理论框架内和随机占优理论框架内比较VaR和CVaR的优劣,指出CVaR在性质上要优于VaR。但在椭圆分布假定下,VaR依然保持子可加性和二阶随机占优的一致性。由于椭圆分布包含诸如t分布以及帕累托分布等能够反映厚尾特征的分布,因此VaR依然可以刻画金融时间序列数据的尾部特征。此外,本文探讨了CVaR在风险管理和监管实践中遇到的问题,指出CVaR模型的事后检验不易实施。
In this paper, we compare merits and demerits of VaR and CVaR based on the theory of coherent risk measure proposed by Artzner et al. （1999） and the theory of stochastic dominance. We conclude that CVaR is superior to VaR as far as some properties are concerned. VaR, however, remains sub-additive and is consistent with second-order stochastic dominance under the assumption of elliptical distributions. Since elliptical distributions include fat-tailed distributions such as Students＇s t-distribution and Pareto distribution, VaR can still depict the tail characteristics of financial time series data. Moreover, we study the problems with CVaR when applying to financial management and supervision. We point out that the ex-post test of CVaR is difficult.
出处
《数量经济技术经济研究》
CSSCI
北大核心
2007年第3期125-133,共9页
Journal of Quantitative & Technological Economics
