摘要
本文以上证综指的高频数据样本为例,全面探讨了各类历史波动率模型以及实现波动率模型的构建方法,同时采用滚动时间窗的样本外预测法,实证计算了不同模型假定下的指数波动率预测值,并运用基于自举法的SPA检验,评估了各种波动率模型对上证综指波动的预测精度。结果显示,基于ARFIMA的实现波动率模型以及随机波动模型(SV)具有最高的波动率预测精度,但在加入实现波动率作为附加解释变量以后,并未显著提升标准SV和GARCH模型的预测能力。
In this paper, one high-frequency dataset of the most important stock index in Chinese stock market is used to construct historical volatility models and realized volatility models. Based on Out-of-sample predicting results using rolling time windows method, we compare the predicting performance of different kinds of volatility models using bootstrapping SPA test. The empirical results show that, realized volatility models and stochastic volatility models are the best models for volatility forecasts in Chinese stock market. Furthermore with RV as additional explanatory variables, no improvements of predicting performance are found in SV and GARCH models.
出处
《金融研究》
CSSCI
北大核心
2007年第07A期138-150,共13页
Journal of Financial Research
基金
国家自然科学基金(70501025)的资助