An alternative method to test for contagion with an application to the Asian financial crisis

Abstract

This paper investigates the size properties of a test for contagion based on an asymptotic t-distribution. The simulations show that this asymptotic test does not have correct size properties. An alternative test method based on case-resampling bootstrapping is introduced to improve on the correctness of inference. The simulations show that this new test has much better size properties. It also has quite high power properties and it is robust to ARCH effects. The method is applied to testing for contagion from Thailand to Indonesia during the Asian financial crisis.

Notes

¹Forbes and Rigobon (2002) note the straightforwardness of using cross-correlation coefficients and provide a review of earlier articles using this method and the other methods mentioned here. Some of the articles we mention in this paragraph come from their review. Another method for empirical investigation of contagion is to examine whether Granger causality changes during the crisis period, as in Granger et al. (2000) and Hatemi-J and Roca (2005). For additional publications on contagion see also, Yamori (1999), Collins and Gavron (2004) and Boschi (2005).

²Forbes and Rigobon (2002) show that their adjustment is algebraically equivalent to one suggested by Boyer et al. (1999) and Loretan and English (2000).

³A formal derivation of this equation is given in Hatemi-J (2004).

⁴A version of the program for applications of the case-resampling bootstrap method is available from the authors by request.