Is volatility lognormal? Evidence from Italian futures

We study the unconditional volatility distribution of the Italian futures market, measuring it via Fourier analysis. Our data set consists of all tick-by-tick transactions in 2000 and 2001, a period characterized by unusually high volatility levels in its final part, because of the dramatic events following 11 September 2001. Our results show that the standard assumption of lognormal unconditional volatility has to be rejected for such a turbulent sample, since it is unable to capture the tail behavior of the distribution; a much better description is provided by a Pareto tail.