By
First, let's define the problem. The Foolish Four, at least our official portfolios and probably many of those started at other times (see Starting Here, Starting Now), is not beating the market as consistently as it did in the past, and when it has beaten the market the margin is much smaller than we expected from the past history of the strategy. Yep, that's a problem.
Those who see this as proof that the strategy is flawed (let's call them the Loyal Opposition) have been very vocal of late. To describe the LO position broadly, they believe that the high returns we found in the original backtest were the result of a statistical anomaly, not a repeatable phenomenon. As proof of this they say that the strategy has not "worked" since it was discovered. (They say other stuff too, but that's the most important point.)
This is the classic definition of something that is due to chance rather than underlying principles. Looking at the historical record, you can see an association between certain characteristics and high returns, but the returns evaporate when you try to use those characteristics to pick high-returning stocks going forward. The association was really only due to chance.
I believe that there is a different reason why we are seeing mediocre performance from the strategy. I believe that it is due to the fact that the Dow began including companies that paid insignificant dividends starting in 1991 and adding more each time the Dow was revised. This has reduced the "pool" of high-yielding companies from which the Foolish Four is picked. If this is the case, then simply enlarging that pool by including all blue-chip companies, not just those listed on the Dow, may very well restore the strategy to health. (Although a third explanation, that value investing in old line, blue-chip stocks is not going to be very competitive in a prolonged bull market like we are seeing, may still be blunting the returns. If that's the case we will need an economic shake-up to get things back on track, which is a good argument for simply adopting a long-term perspective. See yesterday's column for a brief discussion of this topic.)
One can make excellent arguments for either side. But I don't see any definitive arguments being made because all arguments, including mine, are simply different interpretations of the same data. What we need is more data. The Beating the S&P strategy gives us more data, but it only goes back 13 years, not long enough to really, definitively settle anything. And even BSP did badly last year, although its long-term average for the past 13 years is stellar.
The significance of the BSP data is that the same high yield/low price strategy appears to work on non-Dow stocks. If that is the case, and 13 years is not enough to prove it to my satisfaction, then we have an "out of sample" test that is strong support for the idea that low price/high yield characteristics really are associated with higher-than-average returns. In other words, the Foolish Four is not the result of a statistical anomaly, but a real phenomenon.
That still doesn't prove that the association will hold up going forward. Nothing will ever prove that. But it gives us a much stronger basis for following the strategy.
So we need more data. And I'm working on that.
In fact today I want to ask those of you who are statistically literate and interested in this topic to suggest a methodology that we could follow that would prove or disprove to your satisfaction the validity of the Foolish Four, specifically the RP formula OR the high yield/low price method, as a stock picking strategy. Don't worry about trying to prove the unprovable, please. Future returns are never guaranteed. Just tell me what it would take to prove that the past returns were not a fluke.
Since I'm asking, I'd better specify that we already know that the strategy did not work well in the '60s so anything that says that it has to have worked "forever" (forever being defined as more than 30 years ago) is not acceptable. I'm not interested in suggestions that set up impossible standards. As far as I'm concerned, anything that has worked for the past 30 years is good enough, and if market conditions were different 40 years ago, that's not going to worry me.
What I am looking for are constructive suggestions such as: Divide your sample companies into two groups. The strategy should outperform the group as a whole over X years at Y level of confidence.
My aim is to develop a protocol that will satisfy all of us somewhat (and probably no one completely) and then to dig up that data and test it. Then we will have something to discuss. In the meantime, I'm going back to writing about more fun topics, like earnings estimates and that perennial favorite, stock splits!
Fool on and prosper!