Does the predictive power of variable moving average rules vanish over time and can we explain such tendencies?

Recent studies offer striking evidence that, in stock markets, the predictive power of fundamental variables and seasonal effects tends to diminish over time. In this article, we analyse whether this also holds for the popular variable moving average (VMA) rules of Brock et al. (1992). While previous research on this issue has strongly concentrated on US and emerging stock market indices, we fill a research gap by focussing on a wide range of developed market indices and individual stocks. Using a trend regression approach for a dataset covering 1972 to 2015, we find that most analysed trading rule specifications show negative trend coefficients. These results, robust in a variety of settings, indicate that VMA rule signals have steadily lost their ability to forecast future price movements accurately. Analysing several rationales for this outcome, we find that negative trends in the autocorrelation of stock returns are a highly promising explanation for the poorer performance of VMA rules because they are designed to capture autocorrelation.