Bayesian Investor Blog

I just noticed some confused arguments between Jeremy Siegel and Standard and Poors over how to aggregate earnings for companies in a stock index to produce a meaningful report of what the companies in the index earned.
See here and here.

Siegel provides an example involving percent changes in Exxon-Mobil and Jones Apparel. But that has a weak resemblance to what S&P is doing. A more accurate analogy to what S&P is doing would use changes to market cap rather than percent changes. If Jones Apparel declined in market cap by $10 billion, it would hurt the index just as badly as a $10 billion decline in Exxon-Mobil’s market cap. Looked at that way, S&P’s approach looks sensible.

But since Jones Apparel has a market cap of less than $1 billion, the current bankruptcy laws make it far-fetched that Jones Apparel could lose more than $1 billion in market cap.

If you’re using earnings as a proxy for the health of the economy, S&P’s method doesn’t create a problem – the bankruptcy laws affect who loses money, but the money is still lost. But for an investor, Siegel has a point which is half right.

Siegel’s solution of weighting earnings by market cap may work well under any realistic conditions, but has no sensible theory behind it, and can fail badly under some far-fetched conditions. Imagine that Jones Apparel reports an unexpected one-time windfall of $1 trillion, which ought to raise the market cap of Jones Apparel by about $1 trillion. The way S&P computes S&P 500 earnings, an investor looking at S&P 500 earnings would see a strong hint that the value of the S&P 500 ought to rise by about $1 trillion. But under Siegel’s method, the initial effect on S&P 500 earnings would suggest a barely noticeable rise in the value of the S&P 500 of under $1 billion. Then at some point the Jones Apparel market cap would soar and the S&P 500 earnings would be recomputed with much different weights and investors would see a much different picture. So Siegel has proposed something which could result in a potentially large change in reported S&P 500 earnings without any change in the what shares someone who invests in the S&P 500 holds and without any changes in reported earnings.

Morningstar has a method (PDF) designed for evaluating portfolios that uses a harmonic weighted average and ignores companies with negative earnings. That has advantages, but the magnitude of losses provides some hints about how far a company is from profitability, so an ideal method should pay some attention to losses.

Siegel mentions comments by Shiller that suggest Shiller has better (but possibly impractical) ideas. I doubt Shiller’s analysis provides as much support for Siegel’s argument as Siegel claims.

Any sensible investor looks at a multi-year average of earnings along the lines suggested by Shiller, which minimizes the problems associated with faulty weighting of earnings.