Monthly stock market returns are predictable when we refine the observation intervals of the variables used to predict these returns. Contrary to other predictability studies we find high out-of-sample adjusted R2s of up to 7% using economically important commodity returns. Shorter intervals reveal predictability consistent with near efficient markets based on price changes in industrial metals. More historical intervals expose predictability consistent with gradual information diffusion based on energy series. This predictability is robust to data mining adjustment, the inclusion of control (including economic) variables, and unrelated to time-varying risk. Inflation explains part of this predictability, but not all.

Keywords: observation interval, return predictability tests, market efficiency, gradual information diffusion, market timing, quantitative investment techniques, commodity

Continue reading here.