Robert Shiller and John Campbell initially published on P/E Ratio predicting later real returns in this 1988 paper. In 1996 Shiller followed up the research with this easily digestible paper on the topic. In summary he found that the random walk theory does not look right and that current valuations have a strong predictive ability for future long term returns. Over the short term noise prevents prediction, whereas over the long term, ten years, current valuation levels are predictive of future returns.

As I have said before the voting machine should be used for the short term, the weighing machine for the long term.