<FOOLISH FOUR PORTFOLIO>

Fool's Gold?
Part 1 of many

by Ann Coleman (TMF AnnC)

Reston, VA (May 10, 1999) -- A recent publication from the Halls of Academia raises a number of interesting points about the Foolish Four model and Dow Investing in general and calls into question the viability of the Foolish Four as an ongoing investment strategy.

I'll bet that got your attention.

Titled "Mining Fool's Gold," the article by Grant McQueen and Steven Thorley in the March/April issue of the Financial Analysts Journal is quite extensive and raises a number of points worthy of discussion. I believe that going through the article point by point will prove to be a worthwhile exercise, from the standpoint of learning about both the pitfalls of designing your own investing strategies and the pitfalls of criticizing something you don't understand.

In brief, the authors have used the Foolish Four as an example of data mining, a dangerous practice that has entrapped many an unwary researcher and lead to great expectations and inevitable letdowns. Their explanation of the practice is a good one:

Data mining is the practice of finding forecasting models by searching through databases for correlations, patterns, or trading rules. After searching through enough variables or rules, say 100 -- a researcher will find, simply by chance, about 5 that are statistically significant at the 95 percent confidence level. Data mining becomes a particular problem when the research proclaims the final pattern significant without providing the number unsuccessful mining attempts. Discussions of data mining are not new, nevertheless we believe that the recent explosion of data availability and the low cost of computing power make our investigation a valuable and timely reminder.

This is an excellent point and one which, if you have read this space for very long, you will have seen raised before. The problem is that the study takes the Foolish Four as its example of data mining without ever bothering to find out just how the strategy was conceived and/or how it has been tested since. Just like truth is the best defense against claims of libel, the defense for strategies "discovered" by data mining is that they have a valid rationale. That's why no one takes the correlation between the winner of the Super Bowl and the performance of the stock market seriously, but the fact that stocks that are growing their earnings per share rapidly do better than those that are not is quite easily accepted.

After describing what's wrong with data mining, McQueen and Thorley go on to say about the Foolish Four: "As researchers we are unable to fully analyze the significance of a strategy that worked in the past without knowing how many unsuccessful strategies were explored including unsuccessful strategies tried by the Gardners themselves."

A simple phone call would have provided the answer at least to the part about the Foolish Four. I still have the original study in my files, and today I will share it with you.

The Foolish Four was born on the Motley Fool's Beating the Dow message board (since renamed Dow Investing/Foolish Four) on AOL in October of 1994. The board was started to promote discussion of Michael O'Higgins's book Beating the Dow, and it quickly became one of the most active message boards in the forum.

Beating the Dow was an inspiring and exasperating book. The returns were wonderful, the idea simple and easy to understand. But it was extremely frustrating to those of us discussing it, because no way to independently verify the results was provided. Mr. O'Higgins provided average yearly returns but no individual stock returns.

A brief recap may be in order here. In Beating the Dow, Michael O'Higgins and co-author John Downes proposed three strategies to beat the market. All were based on the idea that the Dow stocks that paid higher dividend yields tended to outperform the Dow as a whole over the long run. This actually was not a new idea -- the phenomenon had been noted informally for many years, and it was later studied in depth by Jim O'Shaughnessey. In fact, you can see yearly returns for a High Yield 10 strategy (a.k.a. the Dogs of the Dow) that go back to 1929 by clicking here.

O'Higgins also claimed that if you simply bought the five lowest-priced stocks on that High Yield 10 list, you could improve your returns further. He claimed that, if you wanted to live dangerously, a single stock strategy that invested in the second-lowest priced stock of the High Yield 10 (the PPP or Penultimate Profit Prospect -- hey, we didn't name it!) had shown amazing returns.

Those of us who had read the book were frustrated by the lack of data to independently verify his claims. In describing the PPP, O'Higgins hinted that the lowest-priced stock might not be a good one. This suggested that the strategy might have room for improvement IF ONLY WE HAD SOME DATA. Grrrrrr!

Well, in a former life I was a newspaper reporter, and that experience taught me to get on the phone and dig. My contribution to the Foolish Four strategy was not to propose a strategy that beat Beating the Dow but to dig up some numbers and test the proposals that were pouring in from the board. I had no luck contacting Mr. O'Higgins, but John Downes, the co-author of BTD, who was running a Beating the Dow newsletter by then, faxed me the actual BTD stocks and their annual returns back to 1973.

To get a handle on something O'Higgins had hinted at -- a possible dichotomy between the lowest-priced stock and second-lowest priced stock (the PPP), I created a spreadsheet and started analyzing the average returns of all stocks that were in the number 1 position each year, vs. all stocks in the number 2 position each year, vs. number 3, etc. Several interesting things showed up -- it turned out that the lowest-priced stock really was a dog. It had an average return of around 8%, way below the market. The second stock was indeed wonderful, returning close to 30%. Numbers 3 and 4 were lower than # 5, and the returns for the 5 highest priced stocks (numbers 6-10 of the High Yield 10 group) were, when averaged, no better than the Dow as a whole.

Was this data mining? Actually, I suspect that data miners would be insulted if I said it was.

Strangely enough, we realized that this was not a definitive study -- even without the benefit of academic advice. But it did answer the questions that we had been asking about Beating the Dow. Still, it was obvious that the average return of all stocks that just happened to fall into a certain spot on the roster each year could not possibly be significant. Without some logic to explain why, it was obvious that the higher average return of all stocks ranked 5 vs. all stocks ranked 4 was simply an artifact of the data. But what about the first two stocks, with average returns of 8% vs. 30%, when all the others were averaging between 11% and 20%?

We are only talking about 21 data points, but a difference that great suggests that something is going on. (I didn't know much about statistics back then, but common sense doesn't require a Ph.D.) O'Higgins's comment that the lowest-priced stock is often a company in financial trouble seems to provide the explanation. Rather than being beaten down but poised for a rebound, perhaps in too many cases that lowest-priced stock was being beaten down for good reason. As for the surprising performance of the second lowest priced stock, well, I don't know that any of us would have bet our life savings on that continuing, but it was certainly worth exploring further.

Once it became obvious that the first stock was a real loser, I tested 4 portfolios that varied only in one factor -- the number of stocks included. (We didn't have the data to test any other factors.) They all beat Beating the Dow, which was all we were concerned about at that time. The two- and three-stock portfolios were felt to be rather too risky, although you can certainly make that argument about a four- or five-stock portfolio, too. At any rate, a general consensus developed that a four- or five-stock portfolio that dropped the first stock and optionally doubled up on the second-lowest priced stock was probably a good strategy to try if one wanted to do better than the Beating the Dow 5.

This was the genesis of the Foolish Four. It can't be called classic data mining -- there wasn't enough data to mine! There were no prices, no dividends, just a list of ten stocks, supposedly the ten highest-yielding Dow stocks at the beginning of the year, listed in price order from lowest to highest.

Tomorrow -- Did we let it go at that?

Fool on and prosper!

Today's Stock Lists | 1999 Dow Returns

05/10/99 Close

Stock Change Last -------------------- CAT -1 7/16 61.75 JPM + 13/16 135.69 MMM +1 5/16 94.19 IP -1 56.13

Day Month Year History FOOL-4 -0.58% 1.53% 30.91% 32.86% DJIA -0.22% 2.02% 20.27% 19.80% S&P 500 -0.35% 0.38% 9.35% 9.62% NASDAQ +0.91% -0.65% 15.22% 16.80% Rec'd # Security In At Now Change 12/24/98 24 Caterpillar 43.08 61.75 43.34% 12/24/98 22 Int'l Paper 43.55 56.13 28.87% 12/24/98 9 JP Morgan 105.51 135.69 28.60% 12/24/98 14 3M 73.57 94.19 28.02% Rec'd # Security In At Value Change 12/24/98 24 Caterpillar 1034.00 1482.00 $448.00 12/24/98 14 3M 1030.00 1318.63 $288.63 12/24/98 22 Int'l Paper 958.12 1234.75 $276.63 12/24/98 9 JP Morgan 949.62 1221.19 $271.57 Dividends Received $29.45 Cash $28.26 TOTAL $5314.27 </FOOLISH FOUR PORTFOLIO>