Multivariate Volatility Estimation with High Frequency Data Using Fourier Method

Availability of high frequency data has improved the capability of computing volatility in an efficient way. Nevertheless, measuring volatility/covariance from the observation of the asset price is challenging for two main reasons: observed asset prices are generally affected by noise microstructure effects and tick-by-tick returns are asynchronous across different assets. In this paper we review the definition and the statistical properties of the so called Fourier estimator of multivariate volatility, with particular focus on using high frequency data. Exploiting the fact that the method allows to compute both the integrated and the instantaneous volatility, we show how to obtain estimators of the volatility of the volatility and the leverage as well. Further, we study the performance of the estimator in forecasting and in terms of portfolio utility in the presence of microstructure noise contaminations