Are Asian Stock Market Returns Predictable?

ABSTRACT

We conduct predictability tests for selected Asian stock markets using monthly data from the period March 2001–April 2012. Asian market bears and returns are predicted using the U.S. stock market bears and returns. A two-state Markov-switching model is employed to distinguish between the bull and bear regimes in the U.S. and Asian stock markets. The in-sample predictability analysis suggests that the U.S. market returns and bears are important predictors of Asian market returns and some Asian bears. The out-of-sample predictability exercise is not able to reinforce the in-sample results, which is in large part due to the small forecasting sample size.

KEY WORDS:

• Asian stock markets
• bear market
• predictive models
• U.S. stock market
• two-state Markov-switching model

Notes

^1. Testing the predictability of bear markets is not new. Chen (2009), for instance, examines the predictability of the U.S. bear market using macroeconomic variables. However, evidence surrounding the predictability of market bears using international stock market bears and returns could not be found.

^2. Evidence contrary to the above also exits. For example, Brockman, Liebenberg, and Schutte (2010) mention that the comovement between the U.S. stock markets and international markets has been on a decline since the second half of the twentieth century.

^3. Martens and Poon (2001, p. 1805) report a similar finding that “… daily correlations between international stock markets increases under extremely adverse conditions when a large negative return has been registered.” (Also see Longin and Solnik 2001).

^4. As shown by several studies, knowledge of stock market cycles is important for market timing for those investors whose portfolio rebalancing is dictated by the turning points of the stock market cycles (see Straetmans et al. 2012; Chen 2009). In this regard, information about whether market returns (or a bear market) can predict another market will further improve investors’ ability to time the predictor market and the international markets that the predictor market can predict.

^5. We followed Psaradakis and Spagnolo (2003) to determine the state dimension for the data. On the basis of the Akaike information criterion, Schwarz’s Bayesian information criterion, and Hannan and Quinn’s criterion, it was found that all the return series are best characterised by two regimes rather than three regimes (see also Psaradakis and Spagnolo 2006). These statistics are available upon request.

^6. For details on the adjusted OLS and GLS, refer to Westerlund and Narayan (2012, 2014).