Testing For Contagion: A Conditional Correlation Analysis

In this paper we test for contagion within the East Asian region, contagion being defined as a significant increase in the degree of co- movement between stock returns in different countries. For this purpose we use a parameter stability test and, following Rigobon (2004), we control for three types of bias, resulting from heteroscedasticity, endogeneity and omitted variable respectively. The null of interdependence against the alternative of contagion is then tested as an over-identifying restriction. Unlike other studies, our approach is based on full-sample estimation, and hence avoids the power problems arising from the typical situation of a large “non-crisis” and a small “crisis” sample. We also select endogenously the breakpoints corresponding to the beginning of the contagion period, and finally we impose more plausible restrictions in order to identify the system. Our findings suggest the existence of contagion within the East Asian region, consistently with crisis-contingent theories of asset market linkages.