The Adaptive Markets Hypothesis: Testing for Variable Efficiency and Cyclical Profitability in the South African Market

Abstract

This research attempts to discover whether the Adaptive Market Hypothesis theory is applicable in the South African financial market and explores the innovation and cyclical profitability implications of the Adaptive Market Hypothesis theory. This is achieved in two parts: first by determining if returns follow a random walk or not and second by analysing the consistency of technical and fundamental factors to explain the cross-section of equity returns between 1 January 1998 to 31 December 2017. The tests of stock return dependency include a total of five tests on the average monthly returns for each stock in the ALSI covering normality and random walk theory for the duration of the two sub-periods and entire examination period. The results of these tests would provide some insight into the level of market efficiency of the JSE and to what extent this efficiency is cyclical. The results for the Jarque-Bera test and Q-Q plots are in agreement, with both tests presenting a strong case for non-normally distributed returns. By contrast, the results of the random walk tests are rather mixed. The results of the Ljung-Box and runs tests suggest that very few stocks in the sample have returns that are randomly generated while the results from the three different variance ratio tests convey quite the opposite in that all stocks in the sample have non-randomly generated returns. Mixed findings for this section are not an unexpected result given that this is the case in the literature as well. The results in support of the Adaptive Markets Hypothesis are present, albeit feint. A larger number of stock returns switch between being normally distributed and nonnormally distributed during the two sub-periods, however, with respect to predictability, as few as 30 stocks possessed returns that were found to switch between a state of predictability and nonpredictability.