Volatility co-movements: a time scale decomposition analysis

Abstract

In this paper we investigate short-run co-movements before and after the Lehman Brothers’ collapse among the volatility series of US and a number of European countries. The series under investigation (implied and realized volatility) exhibit long-memory and, in order to avoid missspecification errors related to the parameterization of a long memory multivariate model, we rely on wavelet analysis. More specifically, we retrieve the time series of wavelet coefficients for each volatility series for high frequency scales, using the Maximal Overlapping Discrete Wavelet transform and we apply Maximum Likelihood for a factor decomposition of the short-run covariance matrix. The empirical evidence shows an increased interdependence in the post-break period and points at an increasing (decreasing) role of the common shock underlying the dynamics of the implied (realized) volatility series, once we move from the 2-4 days investment time horizon to the 8-16 days. Moreover, there is evidence of contagion from the US to Europe immediately after the Lehman Brothers’ collapse, only for realized volatilities over an investment time horizon between 8 and 16 days.