S&P Does the Earnings CorrectlyIn his "The S&P Gets Its Earnings Wrong" (op-ed, Feb. 25), Jeremy J. Siegel claims that Standard & Poor's systematically understates the earnings of the S&P 500. In his view, the recent losses of the financial companies in the S&P 500 should be discounted because of their diminished weights in the index.

His argument, however, fails the simple tests of both logic and index mathematics. A dollar earned or lost is the same, irrespective of whether it is earned or lost by a big index constituent or a smaller one.

Prof. Siegel's example of Exxon-Mobil illustrates why S&P's method of calculating earnings works. If large Exxon-Mobil earned $10 billion and small Jones Apparel lost $10 billion, index investors collectively -- and individually -- would bear a proportionate share of both Exxon's earnings and Jones's loss, despite the fact that the value of Exxon-Mobil's shares in the index portfolio is about 1,381 times the value of the Jones's shares.

To use an analogy, we could hypothetically view the S&P 500 as a single company with 500 divisions, with each division having earnings and an implicit market value. The smallest of these divisions could have an outsized loss that wipes out the combined earnings of the entire company. Claiming that these losses should be ignored or minimized because they came from a less valuable division is flawed.

Prof. Siegel's approach -- applying the weights based on market values to the results based on a company's earnings -- effectively mixes apples and oranges.

David M. Blitzer

Managing Director, Chairman of the Index Committee

Standard & Poor's

New York

I still think that Siegel is correct, and that Blitzer (and S&P) are missing the point. S&P is correct that the total earnings of the index is the sum of the earnings of the individual companies. But when computing a PE ratio, this approach is flawed if you want a PE ratio that can be comparable to an individual stock's PE. This is because the S&P 500 is NOT a multi division company in which the earnings of a bad division can wipe out the earnings of the rest of the company. These are separate individual companies.

When a stock with a tiny market value posts a massive loss, this loss will have a disproportional effect on the overall PE ratio of the index. If you want the PE of the index to provide some indication of the overall valuation of the market, you will have too high a PE using S&P's method.

I blogged on this a couple of days back. But my basic argument (and I think Siegel's) is that if you have a stock with a tiny value and a huge loss, it's impact on the PE should be trivial because if you buy the index today you'll only be buying putting a tiny amount of your money in that stock. By not value weighting the earnings (or losses) you are assuming that that stock makes up a much larger chunk of the index, when it doesn't.

Or put another way. You have two stocks in the index. One has a PE of 10, and is massive. The other has a negative PE and earnings equal to minus the earnings of the big firm. Is the PE of the index 10 or close to infinity? If you bought the big stock, you'd buy it off a PE of 10, and if you bought the tiny, big loss stock you'd buy it as basically an equity option.

So while S&P is mathematically correct, if they want a PE ratio that is actually economically meaningful, their approach is flawed.