Monday, January 31, 2011

PE10 Weaknesses Continued

The PE10 is a useful ratio that aids the investor in determining the relative price level of the aggregate stock market. But as discussed on this site last week, the metric is far from perfect, as the arbitrary use of a 10-year period can bump up or push down the measurement for normalized earnings, thereby biasing the ratio. Another major weakness of the PE10 as tool for making historical comparisons has to do with a change in how corporations have returned money to shareholders.

It used to be that dividends were the main method by which corporations paid shareholders. But for various reasons, buybacks appear to have become the preferred choice for corporations looking to return cash to shareholders.

As it pertains to current year earnings, whether corporations doled out returns in dividends or buybacks in the past makes little difference. But because of the way Standard and Poor's records the earnings of the index (using the sum of bottom-up earnings per share), this can have a major effect on past years' earnings.

For example, if the index earned $1 in EPS in 2001, and since then companies have bought back 50% of their outstanding shares, the real "earnings power" represented by that $1 is actually $2.

This concept is perhaps easier to grasp with respect to individual companies. Consider GameStop's quarterly earnings, which show just over $50 million in after-tax earnings in the third quarter of both this year and last year. While the absolute earnings are roughly similar for both years, GameStop has seen more than 15% growth in year-over-year diluted earnings per share as a result of share buybacks. Which number do you think is more relevant, GameStop's EPS pre-buyback, or the fact that it earned just over $50 million in both years? I would argue that the latter is more relevant, and that the former can bias the earnings estimate lower.

Considering the fervor with which companies bought back shares this past decade, this difference can have a significant impact on the estimate of the index's normalized earnings level. Unfortunately, it's difficult to quickly adjust the formula to correct for this problem. The level of share buybacks is inconsistent from year to year, and shares have been repurchased at various prices, ranging from extremely high to extremely low.

Related to this problem is fact that the size of the capital stock has changed. As companies have re-invested the earnings that they have not paid out in dividends or used to buy back shares, they now have larger capital bases from which to earn profits.

Fortunately, there are other methods that can be used that are in keeping with the spirit of the PE10, without suffering from these drawbacks. For example, one could average the index's ROE over the last ten years, and multiply that number by the index's current book value in order to arrive at an estimate for normalized earnings. My rough calculations with imperfect data suggest a PE10 of around 19, and a PE8 (see here for why a PE8 might be more relevant right now than a PE10) of around 18.

Of course, this method still suffers several drawbacks. The biggest problem with using ROE, for example, is that it doesn't correct for different levels of leverage. Higher debt can result in higher ROE, but that doesn't mean these higher levels are sustainable.

Fortunately for value investors, we don't have to spend a lot of time valuing the market. Knowing that it is approximately fully-valued or somewhat overvalued is good to know in general, but our money is made with individual stocks. Figuring out how to make money on the index is a lot harder than it is with individual stocks, since with individual stocks you only have to swing at the most attractive pitches.

2 comments:

Unknown said...

I have thought a bit about this as well.

The ROE method makes a lot of sense. Are you aware of a source for long-term market wide ROE data? It would be interesting to play with those numbers.

Here is a link to a blog post exploring the PE10's relationship and possible flaws vis-a-vis inflation. In my opinion, both it and several of the comments it attracted are worth a read to anyone interested in this topic.

http://greenbackd.com/2010/08/27/graham-shiller-pe10-calculated-using-shadowstats-cpi/

Saj Karsan said...

Nice post and comments, Parker.

No I don't know a source for ROE. Also, I think using historical ROIC might be even better, since it corrects for past debt levels that may have skewed ROE in either direction. But no, I don't have an ROIC source either.