Estimation of volatility functionals: the case of a square root n window

Abstract

We consider a multidimensional Ito semimartingale regularly sampled on [0,t] at high frequency 1/Δ_n, with Δ_n going to zero. The goal of this paper is to provide an estimator for the integral over [0,t] of a given function of the volatility matrix, with the optimal rate 1/\sqrt{Δ_n} and minimal asymptotic variance. To achieve this we use spot volatility estimators based on observations within time intervals of length k_nΔ_n. In [5] this was done with k_n tending to infinity and k_n\sqrt{Δ_n} tending to 0, and a central limit theorem was given after suitable de-biasing. Here we do the same with k_n of order 1/\sqrt{Δ_n}. This results in a smaller bias, although more difficult to eliminate.

arXiv admin note: substantial text overlap with arXiv:1207.3757