Any fool can data mine and find spurious correlations. The real test is how they do out of sample. Robert Shiller became famous in 1996 with his “irrational exuberance” claim (or at least one degree of separation from famous, as it was when Greenspan repeated this claim that the public took notice.)
Shiller’s model looks at the ratio of stock prices to an average of inflation-adjusted earnings over the previous 10 years. When he made the call this ratio was near the mid-20s, well about the historical average of 16. So how’s this model done since?
Not well at all. I just saw the S&P hit 2001 as I wrote this. Back in 2010 his model was telling us that the S&P was 20% overvalued, when it was at 1070. Take a look at the ratio over time.
What do you see? I see a trend line that seems to have shifted upward from 16 in the mid-1990s, precisely when he made the irrational exuberance claim. The new average looks like about 25. And the current ratio is above 25. So the out of sample test was about as complete a failure as one can imagine. The model simply did not perform out of sample.
A few other points:
1. Shiller says that he is still buying stocks at extremely lofty values in 2014, even though in 2010 the S&P was already way overvalued at 1070. Huh?
2. Shiller was just awarded a Nobel Prize in economics. Not for creating a new model like the EMH, but rather for arguing that the EMH was wrong. But his alternative model has performed poorly in the 2 decades since it was first made famous. I guess the standard for experimental proof is a bit weaker in economics than in chemistry or physics.
PS. Of course I am not claiming Shiller is a “fool.” When I first saw his model I thought it was the most impressive anti-EMH model out there. The EMH needs to be challenged by the very best, and challenged frequently. But in the end I suspect the EMH will always come out on top, as other models fall by the wayside.
PPS. The Sumner model: 25 is the new normal for the PE10 ratio. Until it isn’t.