Multivariate contagion and interdependence

This paper proposes a multivariate test to measure the statistical and economic significance of contagion through analysis of extreme unobserved common shocks. Contagious episodes are endogenously determined with no need, but the possibility, to specify the source country. Application to a panel of equity returns during the Asian crisis of 1997–1998 finds that interdependencies are substantially more important than contagion. However, the periods of contagion evident show that it is short-lived, split between positive and negative movements and reverses quickly. In comparison to other Asian crisis countries, Hong Kong is the main driver of contagion in the crisis. The proposed methodology and the empirical findings provide a more detailed picture of contagion than commonly applied tests.Fry gratefully acknowledges funding from ARC grant DP0985783