Wavelet analysis of financial contagion

Abstract

The aim is to estimate a factor model fitted to financial returns to disentagle the role played by common shock and idiosincratic shocks in shaping the comovement between asset returns during periods of calm and financial turbulence. For this purpose, we use wavelet analysis and, in particular, the Maximum Overlapping Discrete Wavelet Transform, to decompose the covariance matrix of the asset returns on a scale by scale basis, where each scale is associated to a given frequency range. This decomposition will give enough moment conditions to identify the role played by common and idiosincratic shocks. A Montecarlo simulation experiment shows that our testing methodology has good size and power properties to test for the null of no contagion (that is, of absence of an increasing role of idiosincratic shocks during turmoil). Finally, using Full Information Maximum Likelihood, we fit our model to test first for the presence of contagion whithin the East Asian region stock msarkets during the 1997-1998 period of financial turbulence, and, then, whether there is contagion from an index of financial distress in the US to East Asia during the recent sub-prime crisis.