极值理论在高频数据中的VaR和CVaR风险价值研究

Studies on VaR and CVaR for High Frequency Data Based on Extreme Value Theory

高频数据具有与低频数据明显不同的特征。本文引入广义帕雷托分布代替传统的正态分布等,精确描述金融高频数据收益的厚尾特征;并且计算高频数据下的VaR和CVaR,然后利用深成A指数据进行返回检验。两种返回检验方法的结果表明,极值理论方法可以比较精确地度量VaR和CVaR。

High frequency data has entire different characteristics with lower frequency data. In the paper, replacing some conventional distributions such as normal distributions, a generalized pareto distribution is introduced to finely describe the characteristics of the tail of return for high frequency data, and further to estimate the VaR and CVaR. The two statistical tests for the data of Shenzhen index A are conducted and the results indicate that VaR and CVaR can be precisely estimated by the method of extreme value theory.