On time-varying predictability of emerging stock market returns

Abstract

The two recent studies of Cajueiro and Tabak (2004b) and Hull and McGroarty (2014) investigate the predictability of emerging stock market returns based on the Hurst coefficient—a simple but powerful measure of long-range dependence. Unfortunately, the insights gained in these studies are limited because they (i) present conflicting evidence on the time-varying nature of the estimated Hurst coefficients and (ii) incorrectly equate random walk behaviour with market efficiency. In this note, we revisit the issue of time-varying predictability for a rich sample of 21 emerging markets in the 27-year period from 1988 to 2015. Extending the two aforementioned studies by various alternative fractal estimators of the Hurst coefficient, trend regressions and several robustness checks, our analysis reveals significant downward trends in the local Hurst coefficients of almost all markets. Specifically, we document vanishing predictability over time, which indicates that profitable emerging market investment strategies based on past returns may not continue their good performance in the future. Furthermore, we explicitly point out why a random walk is neither a necessary nor a sufficient condition for rationally determined security prices, and thus signs of predictability (randomness) should not be interpreted as evidence for market inefficiency (efficiency).