Realized Variance in the Presence of Non-IID Microstructure Noise

Abstract

It is a well known fact that at high sampling frequencies, the contamination of microstructure noise causes the Realized Variance to be a biased measure of the Integrated Variance. Recent developments in this field propose sampling on lower frequencies, sub-sampling techniques, or bias corrections using the autocorrelation patterns in the data. In this paper we propose a structural decomposition of the efficient price process and the microstructure noise. At the highest sampling frequency, we allow for potential correlation between the efficient price and the microstructure noise. For 20 actively traded stocks at Nasdaq, we find that the method provides a lower bound on Realized Variance. Applying a recently introduced bias correction reveals a very long persistence in transaction by transaction returns corrects the Realized Variance upwards to a level equal to low frequency Realized Variance. It remains questionable, however, whether this long persistence should be seen as microstructure noise, or is an inherent feature of the price process.