Market microstructure noise, integrated variance estimators, and the limitations of asymptotic approximations: a solution ⁄

Abstract

A growing literature has advocated consistent kernel-based estimators of integrated variance in the presence of flnancial market microstructure noise. We flnd that the asymptotic results derived for these estimators do not satisfactorily represent the estimators’ flnite sample properties and provide little guidance about how to efiectively apply them in practise. We show how to optimize the flnite sample properties of these estimators. Empirically, we flnd that their suboptimal implementation can lead to little or no flnite sample gains when compared to the classical realized variance estimator. However, signiflcant gains can be recovered using our approach.