Buttonwood: Back to the Shiller p/e

ONE key reason why I have been pessimistic about the outlook for the US stockmarket is based on the use of the Shiller price-earnings ratio. Ben Graham, the doyen of securities analysis, devised a version of this measure, but it has become associated with Robert Shiller, the Yale professor who can claim credit for calling both the dotcom and housing bubbles in his book Irrational Exuberance. He maintains the data on his own website.

The Shiller version tries to eliminate the effect of the economic cycle on valuations; without it, stock markets look expensive when earnings collapse in recessions and look cheap when earnings are high in booms. So it averages earnings over 10 years, and adjust them for inflation; at the moment, the p/e is 21.5, well above the historical mean.

So what? The noted quant, Cliff Asness, head of fund management group AQR, published a third quarter commentary in which he looked at future equity returns when the Shiller p/e was at current levels. The average 10 year real return was just 0.9%. Indeed, if you rank years by Shiller p/es, then you get an almost perfect relationship between valuations and future returns; real returns when the Shiller p/e was at its lowest ranged between 10.3 and 10.4%.

Now Mr Asness points out that there is a danger of small-sample bias and that the standard deviation of results is quite high. Nor is the Shiller p/e much use as a market timing signal; it was already high in the late 1990s and got even higher by the spring of 2000.