A non-random walk revisited

"In this paper, we test for short and long memory in asset prices across 44 emerging and industrialized economies. Using methodology from Lo and MacKinlay (1988) and Lo (1991), we find that markets with a poor Sharpe ratio are more likely to reject the random walk than better performing markets. We also make a methodological contribution. Contrary to the Baillie (1996) criticism, our long memory analysis suggests that the choice of a truncation lag is not as important as one might initially believe. Tests that reject the null hypothesis tend to do so across any reasonable choice in lag"--Federal Reserve Board web site.