A New Measure of Vector Dependence, with an Application to Financial Contagion

Abstract

We propose a nonparametric measure of association between any number of random vectors that is based on the empirical copula process. The measure is insensitive to the dependence of components within vectors and only captures association between vectors as a whole. We calculate approximate values of the new measure for several common multivariate copula families, provide a non-parametric estimator to enable statistical inference, and derive its asymptotic properties under general conditions. To illustrate the applicability of the new measure, we use it to assess the degree of contagion between several regional groups of financial markets during and after the financial crisis of 2008, where the new measure is used to net out contagion within each region. We find strong evidence of contagion between regions as a whole, with some regions showing little interdependence before and after crisis, but becoming closely-connected during the crisis years.

Keywords: Copula, measures of vector dependence, financial crises, financial contagion, nonparametric statistics

JEL Classification: C14, G01, G15