论文部分内容阅读
基于高频数据的波动率矩阵估计可有效解决传统低频估计面临的种种瓶颈问题。然而,由于受非同步和微观结构噪声等的影响,传统的高频波动率矩阵估计会产生艾普斯效应,并偏离其理论值。本文主要考虑非同步逐笔高频数据的三种同步化方法和五种传统已实现波动率矩阵的纠偏降噪方法,并从数值模拟和沪深股市的实证分析两个角度对两类方法分别展开了全面深入的比较研究。结果表明:更新时间同步化法最大程度地保留了数据信息,传统未纠偏的已实现波动率矩阵具有艾普斯效应,其偏差较大,多变量已实现核估计、双频已实现波动率矩阵估计、调整的已实现波动率矩阵估计的纠偏降噪效果较好,事先平均HY估计和HY估计相对表现较差。研究结果可为相关领域工作者进一步的研究与应用提供方法上的参考与指导。
The volatility matrix estimation based on high frequency data can effectively solve the bottleneck problems that traditional low frequency estimation faces. However, due to the influence of asynchronous and microstructure noise, the traditional estimation of high frequency volatility matrix will produce Epps effect and deviate from its theoretical value. In this paper, we mainly consider three synchronization methods of asynchronous high-frequency data and five traditional methods of correction of the volatility matrix to correct the noise, and from the numerical simulation and empirical analysis of Shanghai and Shenzhen stock markets two methods respectively Launched a comprehensive and in-depth comparative study. The results show that the update time synchronization method preserves the data information to the maximum extent. The traditional uncorrected realized volatility matrix has the Epstein effect, the deviation is large, the multivariate kernel estimate is achieved, the dual frequency has realized the volatility matrix It is estimated that the adjusted RMS of the realized volatility matrix estimation is better, and the prior average HY estimation and the HY estimation are relatively poor. The research results can provide methodological reference and guidance for further research and application of workers in related fields.