Multipower Variation Under Market Microstructure Effects

The asymptotic theories used to estimate the integrated variance using realised variance or multipower variation suggest that returns should be sampled at the highest possible frequency. This leads to a bias problem due to market microstructure effects that can completely invalidate the theory. There is a trade-off between bias and variance when choosing the sample frequency. There is an urgent need for estimators of integrated variance that are unbiased and efficient under these effects. In this paper, multipower variation is studied under this perspective and alternative estimators are defined using the subsampling and averaging method.