长记忆特征下国债市场的风险价值研究

The Research of the VaR in Chinese Government Bond Market with the Property of Long Memory

本文从定量的角度考察了我国交易所国债市场和银行间国债市场的市场风险。首先考察了国债收益率的尖峰厚尾性、异方差性和平稳性;接着估计了在t分布和偏t分布下GARCH模型与HYGARCH模型,利用失败率检验与动态分位数测试检验了模型预测的VaR值的精确程度。实证结果表明,银行间国债市场波动的长记忆性大于交易所国债市场;当使用参数法估计VaR值时,波动率的长记忆性是重要的影响因素;偏t分布并没有较大改善交易所国债市场预测的VaR值的准确性,然而偏t分布下的HYGARCH模型却更适合我国银行间国债市场。

This paper empirically studies the value-at-risk of the exchange government bond market and the Interbank government bond market in china. Firstly, We investigate the well-known fat-tail phenomenon and the properties of the heteroscedasticity and stability in the return series. Secondly, We estimate the value-at-risk using the GARCH model and the HYGARCH model with Student-t and skewed Student-t distribution and check the one-step-ahead forecasting VaR by employing failure rate test and dynamic quantile test. The empirical results show that the long-memory property of the Interbank government bond market is larger than the one in the exchange bond mark ; the long-memory properties of the volatility very important effect on calculating VaR by using the parameter method; we also get the result that HYGRACH with Student-t distribution performs no better for the exchange government bond market, But HYGARCH with skewed Student-t distribution model performs better for the Interbank government bond market.