Abstract: The financial asset's yield distribution has some typical characteristics such as "leptokurtic", "fat tail","skewed" and "Asymmetry", but the traditional normal distribution, t distribution, SKST distribution cannot fully describe these characteristics, which has influenced the efficiency of parameter method of VaR models based on them. In recent years, the theoretical circle has proposed AEPD, AST, ALD and other distributions to improve the description of the financial asset's yield distribution. The CSI 300 index is choosed to analysis and compare the description on the typical characteristics of financial asset's yield distribution and also the measurement differences of VaR models based on them. The empirical results show that the more typical characteristics of financial asset's yield distribution the model captures, the better to measure VaR. It only proves that the measurement effects of the VaR models based on AEPD, AST distribution are absolutely better than the model based on normal distribution, but have no obvious difference with model based SKST distribution; Model based on ALD performs even worse when measuring the short VaR but performs best when measuring the long VaR under the low quantile.

Key words: VaR, AEPD, AST distribution, SKST distribution, asymmetric Laplace distribution