Multiscale characterization of volatility of main stock indexes

Abstract

We show that nonstationary behavior of volatilities of stock markets can be succinctly described in terms of the non-decimated wavelet transform, and we indicate how this characterization can be used to improve clustering of the markets. The methodology combines multiresolution analysis with independent component modeling thus addressing the inherent nature of volatility as being multiscale and convolution. The clustering results provide more information related to complex nature of market volatility, show robust behavior and agree with the results obtained in literature.