金融高频数据的偏差校正“已实现”双幂次变差  

Bias corrected bipower variation based on high-frequency financial data

作　　者：李胜歌[1,2] 张世英[2] 

机构地区：[1]天津财经大学经济学院,天津300222 [2]天津大学管理学院,天津300072

出　　处：《系统工程学报》2008年第4期392-397,共6页Journal of Systems Engineering

基　　金：国家自然科学基金资助项目(70471050)

摘　　要：金融高频数据中微观结构噪声的存在严重影响了金融波动率估计量的准确性.为了消除微观结构噪声给波动率估计量带来的偏差,构建更准确的金融波动率估计量,选取具有稳健性的"已实现"双幂次变差对其做了偏差校正,提出偏差校正的"已实现"双幂次变差,通过定理证明了其渐近无偏性与有效性,并用深证成指的金融高频数据验证了这一理论成果.因此偏差校正的"已实现"双幂次变差是具有稳健性、渐近无偏性与有效性等良好统计性质的金融波动率估计量,它为金融应用研究领域奠定了基础.The existence of microstructure noise in high-frequency financial data affects the accuracy of financial volatility estimator seriously. To get rid of the bias of financial volatility estimator brought by the microstructure noise and construct more precise estimator, this paper corrects the bias of real- ized bipower variation which is robust and presents a bias-corrected realized bipower variation. Then the paper proves that the estimator is asymptotic unbiased and effective. The practical study of high- frequency data of Shenzheng stock market also demonsstrates this result. So the bias-corrected real- ized bipower variation has many good statistieal properties such as robustness, asymptotie unbiasedness, effectiveness etc and it lays a foundation for financial applied research.

关 键 词：高频数据 微观结构噪声 偏差 “已实现”双幂次变差 有效性 

分 类 号：F830[经济管理—金融学]