Decomposing the predictive performance of the moving average trading rule of technical analysis: the contribution of linear and non linear dependencies in stock returns

On several occasions technical analysis rules have been shown to have predictive power. The main purpose of this work is to decompose the predictive power of the moving average trading rule and isolate the portion that could be attributed to the possible exploitation of linear and non linear dependencies in stock returns. Data for the General Index of the Athens Stock Exchange are filtered using linear filters so that the resulting simulated “returns” exhibit no serial correlation. Applying moving average trading rules to both the original and the simulated indices and using a statistical testing procedure that takes into account the sensitivity of the performance of the trading rule as a function of moving average length, it is found that the predictive power of the trading rule is clearly weakened when applied to the simulated index indicating that a substantial part of the rule’s predictive power is due to the exploitation of linear dependencies in stock returns. It is also found that the contribution of linear dependencies in stock returns to the performance of the trading rule is increased for shorter moving average lengths.