Our analysis of the Foolish Four over the last 50 years is likely to please no one. The results are mixed, but show some weak support for the proposition that the high yield/low price formula we've used does tend to select stocks that outperform. However, the outperformance is at a much lower level than we found in our original study and is not high enough to justify our continued endorsement.
A number of people have posted or emailed me asking what we now recommend instead of the Foolish Four. As is always the case, we recommend that you study various investing options and decide what is right for your own situation. For some, that may mean continuing the Foolish Four as the value portion of a diversified retirement portfolio, as I will be doing. (If you think that just because I am staying with the Foolish Four for another year you should, you definitely need to stick with index investing.)
For others, especially those who were attracted to the strategy because it was simple, we know of no simple alternative other than an index fund or index shares. The idea that one could wildly outperform the market in just 15 minutes a year turned out to be just too good to be true. The good news is that index investing requires even less effort than the Foolish Four. The bad news is that you will never beat the market. But matching the market over a long time period is a pretty good way to build wealth.
For Foolish Four investors who were attracted to the mechanical nature of the Foolish Four and its high potential returns and who are not bothered by the prospect of high volatility and risk, our Workshop strategies may be of interest. The Foolish Four Portfolio will become just a small part of the new Workshop Portfolio when WP debuts next year. Keep in mind that the Workshop area is highly experimental, and it is quite possible that strategies which right now look very promising will prove to be no longer-lasting than the Foolish Four. Any mechanical strategy is vulnerable to the dangers of misinterpreted data, changing market conditions, and over-popularization. The Workshop is not for those who are simply looking for quick "stock picks."
Now for some numbers. There is a huge problem with these results. They aren't going to make the critics happy, and I'm not happy about them, either, but not for the reason you might think. What we found was some support, weak though it was, for the Foolish Four method of picking stocks. At this point, it would be a lot easier to just say, yeah, the critics were right, the strategy was never valid. But the numbers don't support that.
I would love to fake it and say that the returns in the '70s and '80s were simply a result of a random statistical fluke, because I know that the Fool will be accused of trying to justify our mistakes when I explain what I think we really found. Besides, it's pointless to argue. No one is suggesting (any more) that the Foolish Four is likely to strongly outperform the market in the future. Anyway, here goes:
We started with a database of the 500 largest U.S. stocks from 1950 on. The data was purchased from the Center for Research in Security Pricing run by the University of Chicago's Graduate School of Business. They maintain the most accurate and extensive stock database we know of.
The one thing that most people agreed would be the acid test of our Foolish Four formula's ability to actually select stocks that outperformed was to test it on data from the "pre-discovery" period. That would be data from before the original study. The original study for the Foolish Four (current version) was conducted on data from 1961 through 1995.
The argument was also made that if the strategy worked for Dow stocks, it should work for non-Dow stocks as well, as long as they were more or less similar companies. Our database didn't identify Dow stocks, but we did eliminate utility and transportation stocks (which have their own, separate Dow index) when we ordered it.
So starting with U.S. companies that were not utilities or transportation companies, we sorted the database by market cap and used the largest 30 stocks as our "universe." (Why 30? Because the Dow has 30 stocks and we wanted to keep as many factors constant possible.) We call this universe of stocks the Top 30. The database had price and dividend data from the last day of each month allowing us to construct 12 portfolios per year, each held for 12 months.
When we look at the pre-discovery period, portfolios started from December 31, 1949, through November 30, 1960, we find that the Foolish Four portfolios outperformed the Top 30 stocks by 2.24%, slightly higher than the average outperformance across all 50 years. This is why I can't say, as much as I would like to (I don't want to prolong this debate -- really, I don't), that the results we saw in our original study were purely a product of "datamining." If all we had discovered was a statistical anomaly, we shouldn't see the same kind of returns in a completely different time period.
But when we look at the post-discovery period (1996-1999), we see that the Foolish Four stocks underperformed the Top 30 by an average of 5.25% per year. That is significant. It's a huge deficit. But it's only four years so it is less conclusive than the pre-discovery period.
When we look at our database during the original sample period (1961-1995) we see the Foolish Four outperforming by 2.40%. As noted Monday, over the entire time period the outperformance was 1.74%.
But all of those returns after 1960 are "contaminated" by the presence of Dow stocks. In other words, if the original high returns were the result of a random association, the stocks that caused that random association would still be in the data base. So we created two subsets of data.
One subset excluded Dow stocks, and one excluded Dow stocks and also limited the Top 30 stocks to no more than three companies in the same industry (the BSP method). I will be talking more about these tomorrow as well the returns of pure high-yield strategies such as the Dogs of the Dow, but the short answer is we didn't find anything very interesting or conclusive. We did see that when we took the Dow stocks out in the '50s, the outperformance of the Foolish Four dropped quite a bit.
While reasonable people probably could (and will!) interpret this data in many, many ways, my take is that the high yield/low price formula was mildly successful in identifying stocks that outperformed the universe of stocks from which they were picked PRIOR to 1996. But the outperformance was not nearly as high as we originally thought when looking at a shorter time period, and it was not high enough to engender any faith in future performance.
The one thing that really stands out in the data is how poorly the strategy has fared in the last several years. We don't know why, but this is at least consistent with the theory that dividends have become less important to investors and therefore high-yielding companies have been less attractive. It is also consistent with the theory that the strategy never worked and the recent returns are a "reversion to the mean," although the pre-discovery data would contradict that. It could even be consistent with the idea that the strategy is just having a very bad period due to the market's emphasis on growth stocks.
Essentially, such speculations are moot since we are no longer promoting the strategy. I have no plans to spend much time on further research although I believe that Bob Price, without whom none of this would have been possible, has a few ideas he wants to try. Either that or he's planning a Mediterranean vacation -- he keeps talking about Monte Carlo. (A Monte Carlo simulation is a way to test the statistical validity of the strategy.)
The table below contains the returns discussed above. Tomorrow we will discuss results from the other data subsets we developed. Within a few days I will make the database available to anyone who wants to double-check our results or perform tests on the statistical validity of the differences in the returns.
Sigh. These results are just not very satisfying to either side. But they're all we got.
Fool on and prosper!
Results of Foolish Four Research Project
1950-1999* Top 30 Fool 4 Diff.
All-months average: 13.46% 15.20% 1.74%
Standard deviation: 16.30% 18.27%
(*For reference the S&P 500 for this
approximate time period was around 13%.
1999 is a partial year.)
All-months average: 17.34% 19.71% 2.36%
Standard deviation: 16.00% 22.45%
All-months average: 6.84% 7.80% 0.96%
Standard deviation: 13.51% 14.10%
All-months average: 7.69% 11.37% 3.68%
Standard deviation: 16.85% 17.68%
All-months average: 15.78% 19.08% 3.30%
Standard deviation: 17.06% 18.19%
All-months average: 20.62% 18.72% -1.90%
Standard deviation: 13.28% 14.19%
*January through May portfolios only for 1999
All-months average: 17.65% 19.89% 2.24%
Standard deviation: 15.44% 21.60%
All-months average: 10.27% 12.67% 2.40%
Standard deviation: 15.75% 16.83%
All-months average: 28.92% 23.66% -5.27%
Standard deviation: 11.55% 14.81%