EWMA-GARCH模型与GARCH模型在估计收益率波动上的差异的实证及理论分析

Empirical and Theoretical Analysis of the Difference between EWMA-GARCH Model and GARCH Model on the Estimation of Fluctuation of Yield Rate

VaR作为衡量风险的指标,其核心则在于对波动,亦即方差的估计。基于时间序列,关于条件方差的经典模型是GARCH模型,尽管后来又衍生出了EGARCH,PARCH等复杂模型,但在实务中GARCH模型仍占有重要的地位。文章分析了一种比较新的结合了EWMA模型的GARCH模型(以下称为EWMA-GARCH模型)计算VaR的参数估计方法,以检验其在估计波动上的实用性,并对实证检验结果做了理论分析。分析结果表明,尽管该结合模型缺乏完整的理论支持,但是其计算效果仍比较良好,当然这样良好的结果是建立在因缺乏理论依据而导致的对模型的其他要求之上的.至于是采用受理论支持的模型还是并不输实践价值的模型,文章也给出了一定的建议。

As an indicator of risk measurement, the kernel of VaR is to estimate the fluctuation, or the variance. Based on time series, the classical model regarding conditional variance is GARCH model, which is playing an important role even though other more complicated model such as EGARCH and PARCH were proposed. A relatively new GARCH model combining with EWMA model （hereafter called EWMA-GARCH model） is analyzed as a tool for estimating parametem in calculation of VaR. Its validity of estimating fluctuation is examined and the results of the empirical examination are also analyzed theoretically~ Results suggest that this model is efficient for such estimation although it is not supported perfectly by theory, and this efficiency is after all based on the other hypothesis on the model due to the imperfect theoretical support. Some suggestions are proposed for adopting whether the theoretically well supported model or the one bearing certain practical values.