Abstract
The most appropriate approach to test for weak-form market efficiency is to examine whether the stock returns are Martingale Difference Sequence (MDS). However, the MDS tests have been largely ignored by previous studies, as the empirical analysis is dominated by Variance Ratio (VR) tests and Independent and Identically Distributed (IID)-based nonlinearity tests. This article re-examines the weak-form efficiency of 14 Asian stock markets using the generalized spectral martingale test. The result shows that all the return series are not MDSs, indicating the presence of return predictability and hence market inefficiency.
Acknowledgement
The authors thank Juan Carlos Escanciano and Jae Kim for their generosity in sharing their computer codes.
Notes
1 A white noise process is not necessarily a MDS because it may have a nonzero conditional mean. Examples are the bilinear autoregressive and nonlinear moving average processes which have zero autocorrelation, yet they are not MDS because both processes can be predicted nonlinearly using their own past history.
2 Stock price follows a random walk if is an IID sequence with respect to information set A random walk is a martingale process, but a martingale process may not be a random walk. For instance, the volatility of an Autoregressive Conditional Heteroskedasticity (ARCH) process is predictable and this serial dependence in variance implies that the return series are not an IID sequence. However, they are MDSs because the conditional mean is 0.