金融资产波动率估计的最优内生抽样方案的设计与应用  

The Optimal Endogenous Sampling Scheme for Financial Asset Volatility Estimation

作　　者：周辰月 崔文昊 ZHOU Chenyue;CUI Wenhao(School of Economics and Management,Beihang University,Beijing 100191,China)

机构地区：[1]北京航空航天大学经济管理学院,北京100191

出　　处：《计量经济学报》2023年第1期238-258,共21页China Journal of Econometrics

基　　金：国家自然科学基金(72103014);国家自然科学基金重点项目(72033001)。

摘　　要：在基于高频数据的金融资产波动率研究领域,相较于外生抽样方案,当抽样时间为内生时通常能够更有效地捕捉到价格波动是一个共识.目前被广泛采用的内生抽样方式为当价格变化超过某一给定门槛值时进行一次抽样,然而更为具体的方案,例如门槛值如何选定,仍未有定论.本文在原有研究基础上提出了一种能够使波动率估计量渐近方差达到最小的内生抽样方案,该最优内生抽样方案下的已实现波动率将不再存在渐近偏差,且其置信区间宽度至多为5-min抽样方案的1/√3.文章使用蒙特卡罗模拟研究了最优内生抽样方案的有限样本性质,同时在实证研究中采用多种常见的波动率预测模型对不同抽样方案所得波动率估计量的预测能力进行对比,结果显示使用最优内生抽样方案所得波动率估计量作为解释变量时的预测精度更高.It is well-established in the field of return volatility estimation with highfrequency data that the price fluctuation could be captured more effectively when sampling times are endogenous compared to the case where sampling times are exogenous.Currently,the most commonly adopted endogenous sampling scheme is to sample whenever the change in price process exceeds a given threshold.However,there is no consensus on how to choose a specific sampling scheme,such as choosing a specific threshold.In this study,we propose an optimal endogenous sampling scheme that minimizes the asymptotic variance of the volatility estimator based on the existing literature.The realized volatility obtained under the optimal endogenous sampling scheme has no asymptotic bias and the width of its confidence bound is at most 1/√3 of that of 5-min exogenous sampling scheme.We study the finite sample property of the optimal sampling scheme using Monte Carlo simulation.We also illustrate the usefulness of the optimal sampling scheme in an empirical application and find that the proposed optimal sampling scheme leads to a significant increase in forecasting accuracy of future volatility across various commonly used volatility forecasting models.

关 键 词：已实现波动率 高频金融数据 抽样时间内生性 波动率预测 

分 类 号：F832[经济管理—金融学]