中国金融市场波动率模型预测能力比较探究.docVIP

中国金融市场波动率模型预测能力比较探究摘 要:使用高频数据构造的实现类波动率估计量已经越来越多的用于波动率模型预测精度的衡量与比较。本文采用来自沪深两市的主要股指日内高频数据,构建了两类日内波动率的非参估计量实现波动RV与实现极差RR,分别用于GARCH和CARR类模型的比较与扩展;并使用Mincer-Zarnowitz (MZ)回归方程和基于此基础之上的一个渐进正态检验统计量对各模型的相对优劣及统计显著性进行对比研究。结果显示,标准CARR表现最好,而GARCH类扩展以及CARR类扩展模型均未能显著提升模型的预测能力,从实证上说明了CARR模型使用每日价格极差信息对波动率建模是充分有效的。 关键词:实现极差;实现波动; CARR;GARCH;高频数据 中图分类号:F830.9 文献标识码:A 文章编号:1003-5192(2009)05-0020-07 A Forecast Comparison of Volatility Models in Chinese Financial Markets YIN Lian-qian, SHAO Xi-dong (Jinhe Center for Economic Research, Xi?an Jiaotong University, Xi?an 710049, China) Abstract:The realized volatility estimators constructed by high-frequency data have gained popularity in measuring and comparing the forecast abilities of different volatility models. We employ intra-day high frequency data of two main indexes in Shanghai and Shenzhen Stock Markets in constructing two kinds of non-parametric estimators realized volatility and realized range, which are used to compare GARCH and CARR families and build their extensions. Mincer-Zarnowitz regression equation and an asymptotic normal statistics based the equation are used to judge the satistical significance of their superiority. The results show that standard CARR model performs the best, and GARCH extensions as well as those CARR extensions do not improve the standard model?s forecasting ability in the sense of statistical significance, which demonstrates empirically that modeling volatility with daily price ranges of CARR model is highly efficient. Key words:realized range; realized volatility; CARR; GARCH; high frequency data 1 引言 金融资产的波动率估计一直都是金融经济学多个领域关注的核心内容。如何选择恰当的波动率模型以对资产的价格波动进行尽可能精确的估计和预测,对于金融定价模型,资产组合配置,风险的描绘和管理都有重要的理论和实际意义[1]。 尽管实证研究的证据很久以前就明白地显示条件方差既是时变的又是高度持续的,由于相关的研究工具的缺乏,直到70年代学术研究中一直都使用恒常波动率,如BS期权定价模型。自从90年代起,研究波动率的工具才被大量地发展出来。目前估计金融资产收益率日波动性的模型大致有以下几种类别[2]:第一类方法是充分利用衍生工具市场的价格发现功能。衍生工具提供了市场出清时的价格信息,其中也包括揭示波动性的信息。期权是一种价格受到诸多因素影响的资产,然而通常期权市场上除了其标的资产的波动性以外,其余的因素都是可以观测到的。这样通过某个期权定价公式,就可以