Paper detail

Optimal Kernel Estimation of Spot Volatility of Stochastic Differential Equations

Kernel Estimation is one of the most widely used estimation methods in non-parametric Statistics, having a wide-range of applications, including spot volatility estimation of stochastic processes. The selection of bandwidth and kernel function is of great importance, especially for the finite sample settings commonly encountered in econometric applications. In the context of spot volatility estimation, most of the proposed selection methods are either largely heuristic or just formally stated without any feasible implementation. In this work, an objective method of bandwidth and kernel selection is proposed, under some mild conditions on the volatility, which not only cover classical Brownian motion driven dynamics but also some processes driven by long-memory fractional Brownian motions or other Gaussian processes. We characterize the leading order terms of the Mean Squared Error, which are also ratified by central limit theorems for the estimation error. As a byproduct, an approximated optimal bandwidth is then obtained in closed form. This result allows us to develop a feasible plug-in type bandwidth selection procedure, for which, as a sub-problem, we propose a new estimator of the volatility of volatility. The optimal selection of kernel function is also discussed. For Brownian Motion type volatilities, the optimal kernel function is proved to be the exponential kernel. For fractional Brownian motion type volatilities, numerical results to compute the optimal kernel are devised and, for the deterministic volatility case, explicit optimal kernel functions of different orders are derived. Simulation studies further confirm the good performance of the proposed methods.