偏肥尾分布的APARCH模型研究

金融收益率时间序列分布的不对称性已被视为金融市场的一个＂典型事实＂,但现有文献对金融时间序列偏度的关注相对较少。由于受到估计方法和计算效率的约束,波动率模型通常假定条件分布为正态分布或对称肥尾分布,如t分布和广义误差分布,偏度建模比肥尾建模要复杂得多。通过运用芬南德兹—斯蒂尔方法,在对称分布的基础上构造出有偏肥尾分布。在假定APARCH模型的残差项服从这种偏肥尾分布的情况下,研究波动率的拟合效果与预测能力。实证结果显示,与残差项服从对称分布的波动率模型相比,有偏性对波动率模型的估计和预测都有显著影响,向前预测的步数越多,影响越明显。

Reseach on APARCH Models with Skewed & Fat-Tailed Distributions

The asymmetry of financial return distribution has been regarded as a ＂typical fact＂ ,while less attention is relatively paid to the skewness of financial time distribution in the current literature. For the limitation of estimation methodology and computation performance,volatility models assume that the distribution of residuals is normal distribution,or symmetrical fattailed distribution,such as t or generalized error distribution. Modeling skewness is much complicated than modeling fat tail. The author constructs the skewed and fat-tailed distributions based on symmetric distributions by using Fernandez-Steel method. And the author studies the prediction of volatility together with APARCH model. Compared with volatility mod-els with symmetrical distribution,empirical results show that there are significant influences of the skewness on simulation and prediction of volatility models.