Jeremy Siegel's Silly P/E

Jeremy Siegel has an absolutely astonishing argument in the WSJ today, designed to prove that stocks are quite cheap really: he says that S&P "miscalculates" the earnings of the S&P 500, for the purpose of calculating the index's p/e ratio.

The way S&P does it, they take the aggregate market capitalization of the S&P 500 (the price) and divide it by the aggregate earnings of the S&P 500 (the earnings). Seems sensible enough to me. But Siegel has a better idea: why not ignore the companies which lose money, and just concentrate on the companies which make money? If you do that, then stocks look much less overpriced!

Siegel doesn't ignore the money-losing companies enirely. But, he says, "firms with huge losses generally have extremely low market values", and so he decides to minimize their impact in his calculations by weighting earnings according to market cap. Clever, eh? But the idea that this is the "accurate" way of totting up earnings is just silly.

Earnings don't change according to market capitalization. The p/e ratio is an interesting animal: the numerator changes from day to day and even from second to second, while the denominator changes only once a quarter. It's an indicator of how the market (the numerator) is reacting to reality (the denominator). But under Siegel's method, the denominator changes every second as well. And rather than dividing the market by reality, we end up dividing the market by itself. Which is much less useful.

Update: Paul Kedrosky weighs in:

At the very least Siegel should have recalculated the series over history to show what his new cap-weighted earnings multiple would look like. Trotting out a single adjusted figure with no context of what cheap now means in S&P 2.0 Siegel-world is meaningless.