The purpose of this paper is to exploit possible profitability through stock-price moving averages (MA), by using eleven MA at different frequencies as well as the Fama-MacBeth regression with 120 months as a rolling window to derive out-of-sample forecasted returns to individual stocks. Applying principal component analysis to optimize the series of MA helps improve the out-of-sample forecasts on which we form the out-of-sample forecasted return strategies (OSFRS). Evidence shows that, first, individually-employed MA do not well explain the cross-section of stock returns, but their joint predictive power does exist. It follows that the regression estimates could be interfered by certain biases (e.g., multicollinearity). Second, using many MA may not necessarily improve the forecasting power, implying that parts of employed MA are unimportant and generate noisy forecasts. Third, among comparable investment strategies based on hybrid information of MA, OSFRS with the principal components of MA outperforms all others.
