The PE10 is increasingly becoming a common method for value investors to determine whether the broader market is cheap or expensive. Not only does this method have a logical appeal to it, but data suggests that the magnitude of the market's subsequent 10-year returns is related to its PE10 level.
But while there is a relationship between a market's PE10 level and its future returns, the following chart (courtesy of My Money Blog) illustrates that the relationship is rather approximate:
For example, in both 1902 and 1965, the PE10 stood around 23. But in the ten years following 1902, the market's real return was 50%, while in the ten years following 1965, the market shrank by almost that much!
Why is there such wide divergence between the market's relative price level and its long-term returns? I would argue that part of the reason has to do with the arbitrary 10-year time period inherent in the PE10 calculation. What you actually want in the denominator of the P/E calculation is a normalized earnings number. Sometimes the E10 (the average earnings of the last ten years) does give you a good approximation for that figure. But other times, such as today, it may not.
For example, in many E10 periods, one recession and one market expansion occurs. These periods may result in an E10 that closely approximates a normalized earnings number. Today, however, two recessionary periods (2001-2 and 2008-9) are contained in the last ten years. I would argue that this arbitrarily biases the PE10 upwards, since the denominator of the PE10 includes only one economic peak period but two economic troughs.
If you throw out the 2001-2 period in the PE10 calculation (so more like a PE8), you get a ratio of about 19, compared to the headline number of 21, a 10% difference. The data used to create the PE10 is made freely available by Robert Shiller, and you can play with it yourself by downloading it here.