Spot volatility estimation using delta sequences (Q2339119)

The authors study a class of estimators for a spot volatility of a price process modelled by a univariate semimartingale. The volatility may be a deterministic function of the observable variable, or may be presented as càdlàg process with possible jumps. The price process may have Poisson jumps or microstructure noise.\newline The spot volatility estimator is presented as the convolution of squared price returns with a delta sequence of functions which converges to the Dirac delta function. The class of considered estimators includes kernel estimators as a special case as well as estimators based on non-kernel delta sequences.\newline A full asymptotic theory for the estimators within the class is provided. The work makes different estimation approaches (kernels, wavelets, Fourier) comparable.