Dynamic correlation analysis of financial contagion in Asian markets in global financial turmoil

Abstract

This article investigates the dynamics of correlation between 11 Asian stock markets and the US stock market. By utilizing the method of ‘principal components’, we identify a single latent factor that can explain a major portion of variation in the weekly returns of these 11 markets from 1993 to early 2009. We employ the asymmetric Dynamic Conditional Correlation (DCC) model to estimate the correlation between this Asian factor and the US stock market. We find that there is a mean shift in the estimated DCC in the period from late of 2007. We refer this finding as contagion from the US to the Asian markets. However, we find no such evidence of having contagion between the US and individual markets in Asia during the Asian financial crisis.

Acknowledgements

The views and analysis expressed in this article are those of the authors, and do not necessarily represent the views of the Hong Kong Monetary Authority. The authors express their thanks to Lu Jin for her excellent research assistance provided to this article. The authors thank the useful comments and discussion from Hans Genberg and Cho-Hoi Hui. All errors, of course, are ours.

Notes

¹ The EMEAP Central Banks consists of Australia, China, Hong Kong (of China), Indonesia, Japan, Korea, Malaysia, New Zealand, the Philippines, Singapore and Thailand.

² Yang (2005) conducted a DCC analysis on the stock market correlations between Japan and four Asian emerging markets and test the hypothesis of volatility contagion with a dummy variable during the crisis period in a regression equation of the correlation dynamics, after controlling for an upward trend. Chiang et al. (2007) investigated the evolution of the correlation dynamics of nine Asian markets with a control for the US market and test the existence of contagion in an Autoregression (AR) setting with dummy variables for the periods before, during and after the Asian financial crisis and an indicator variable for the changes of sovereign credit-rating changes.

³ They estimated the DCC model over the weekly return of the 11 equity markets in the EMEAP economies together with the weekly return of the US equity market.

⁴ This component approach is also used by Cheung et al. (2009) in their study of the trade integration in the 13 Asian and Oceanic economies.

⁵ For a good introduction to the theory and applications of principal component analysis, see Jolliffe (2002).

⁶ We do not report the VAR filters here since the filters are mainly used to obtain serially uncorrelated residuals for the DCC models. The estimated order of lags of the VARs is six and our tests show that no serial correlation remains in the obtained residuals.

⁷ They also find that the coefficient of the innovations in dynamics of Q _t is also generally insignificant. Again, please refer to Table 6a in their paper.