Correlation Dynamics in East Asian Financial Markets

Abstract

We examine the dynamic relationship between stock returns and exchange rate changes using daily data from January 1994 to September 2013 for six East Asian countries. We use the multivariate GARCH-DCC model in order to disclose the relationship between stock markets and foreign exchange markets which is important for understanding financial stability. The estimation results reveal time varying correlations in the pre- and post-Asian crisis and the Global Financial Crisis periods for all countries. The correlations are stronger when the crisis intensifies. The degree of interdependence between both markets reflects a mutual markets response to shocks and changes in policy.

KEYWORDS:

• stock market
• foreign exchange market
• currency crisis
• dynamic conditional correlation
• interdependence

Notes

1. The Malaysian exchange market is an exception in these tests: the Q-test in levels shows autocorrelation, while the Q-test in squared series reveals the opposite; the ARCH-test suggests no ARCH effect in both financial markets. The pegged exchange rate system in Malaysia following Asian crisis may be responsible for these test outcomes.

2. Hansen and Lunde (2005) conclude that the relatively simple GARCH(1,1) performs extremely well compared to the more advanced alternative models in terms of in-sample performance as well as its predictive ability.

3. We set lags up to five and check what value of order minimizes the AIC (Akaike Information Criterion) and SC (Schwarz criterion) criteria. The SC suggests that we choose order of one whereas the AIC suggests a longer lag length of two. We retain to the AR(2) model. In addition, coefficient of $r_{t-2}$ is significantly different from zero, suggesting we do need at least two lags of r. See Lestano and Kuper (2014) for more information.

4. The exceptions are the Q tests for the residuals for the foreign exchange rate changes in Indonesia and Thailand for 1, 12 and 24 lags, for stock returns for Thailand for 1 lag, and for Malaysia and Singapore for 12 lags and 24 lags.