The Hazards of Wheat Germ?
by Ethan Haskel (Cormend@aol.com)
Baltimore, MD (January 20, 1999) -- While reviewing the 1998 Official Beating the S&P (BSP) Portfolio returns recently, we noted that the lowest-priced stock of the highest ten yielders of the BSP 30, namely Chrysler, was the gold medal-winning stock for the year. Chrysler, now merged into DaimlerChysler (NYSE: DCX) returned nearly 73%. The Official BSP Portfolio had not included this lowest priced stock, a decision that cost it about 13 percentage points in return last year. These observations lead us to reexamine why we drop the lowest-priced stock in choosing our Official BSP Portfolio and opt instead to invest in the second through sixth lowest-priced stocks.
Flash back to 1995. Mel Gibson's Braveheart takes the Academy Award for best picture. The National Academy of Recording Arts and Sciences had just swallowed Alanis Morissette's Jagged Little Pill, awarding her the Grammy for album of the year. Remember Newt Gingrich? The "Master of the House" was named Time Magazine's "Man of the Year." Finally, the backtesting for Beating the S&P was winding down.
From the first eight years of backtesting (1987-1994), the initial results seemed clear-cut. The lowest-priced stock, called BSP 1, had the lowest return of all the five lowest-priced stocks. It was the worst performer fully 63% of the time, or in five of eight years. This BSP 1 stock had a negative return on average, losing about 4 percent each year, compared to the stocks in the other four positions, which had an average gain of about 12 percent. Meanwhile, Foolish work by our own Ann Coleman (known then to us old-timers as "Ann of the Many Numbers"), indicated that the lowest-priced stock of the Beating the Dow 5 was also, on average, a significant underperformer.
The decision seemed fairly clear cut at the time. With the information at hand, we'd drop the lowest-priced stock and invest in the next lowest-priced four or five stocks. Was this a Foolish decision?
Let's look at the numbers. The following are the average returns for the past 12 years for the BSP 1 through BSP 6 stocks:
BSP 1 +14.3% BSP 2 +14.5% BSP 3 +25.5% BSP 4 +19.6% BSP 5 +19.3% BSP 6 +28.3% BSP 1-5 +18.7% BSP 2-5 +19.7% BSP 2-6 +21.4% BSP 3-6 +23.2%
Please remember that these returns represent the simple arithmetic mean for each of the 12 years and not compound annual growth rates.
A few comments:
-- The BSP 1 stock still has the lowest return rate compared with the other stocks, although its return appears indistinguishable from the second lowest-priced stock.
-- From 1995 to 1998 the trend has reversed, namely that BSP 1 has returned an average of about 50% compared to a 35% average for the next four lowest-priced stocks.
-- The standard deviations for the returns of the individual stock positions, though not shown here for brevity's sake, are very high, indicating marked variation of the returns for each position over the past 12 years. Thus, many of the perceived differences in returns may not be actual differences but rather statistical flukes. More on this in another column.
-- We may well be observing the phenomenon of "regression toward the mean." I can't emphasize enough the importance of this principle when analyzing stock screen results and, especially, their variations. Simply put, if one measures an outcome for a group that seems much different from the average outcome for the population in general, on the next test the outcome for the same group will tend to look more like the outcome for general population.
For instance, after eight years of backtesting, we found the BSP 1 stock to return much less than the other stocks in the BSP. Regression toward the mean would predict that on retesting, that stock would fare better on average relative to the other stocks than it had done previously. That is, over the next eight years the BSP 1 stock would tend to gravitate more toward the average return of the BSP stocks as a whole.
So should we still be dropping the lowest-priced stock in the Beating the S&P Portfolio? For the time being, I'll say "yes." This opinion is based on the observation that BSP 1 remains among the worst performers, especially in the number of times it truly tanks. With 12 years of backtesting, the BSP 1 stock was the worst performer of the BSP 5 stocks fully half the time, a statistically significant finding. As important is the fact that there seems to be something about the lowest-priced stock that may well indicate general underperformance -- much longer-term data from the Dow dividend strategies bear this out. Even the RP variation of Beating the Dow, which doesn't strictly rank stocks on the basis of price, throws out the #1 stock based on many years of backtested data.
We should still keep ourselves open to the possibility that we're wrong, that the observations we're making in this screen or others may represent statistical flukes or merely "bad science."
Here's some favorite dialogue from Woody Allen's 1970s comedy Sleeper, about a cryogenically frozen health food store owner who wakes up to find himself many years into the future:
Scientist A: Has he [the store owner] asked for anything special?
Scientist B: Yes, why for breakfast... he requested something called "wheat germ, organic honey, and tiger's milk."
Scientist A: Oh, yes. Those were the charmed substances that some years ago were felt to contain life-preserving properties.
Scientist B: You mean there was no deep fat? No steak or cream pies or... hot fudge?
Scientist A: Those were thought to be unhealthy....
Without proper understanding of statistical sampling concepts, what we previously considered health food may turn out to be the junk diet of the future. I haven't really done justice to the principle of regression toward the mean in this short column, but here's a very nice expanded explanation. Keep this concept handy, though, when evaluating the information filtered through the Workshop. Let's properly assess today's high-flying screens so we may avoid tomorrow's crash landings.
Beating the S&P year to date returns (as of 01-19-99):
Schlumberger (NYSE: SLB) +7.3% Kimberly-Clark (NYSE: KMB) -6.2% Campbell Soup (NYSE: CPB) -19.1% Ford Motor Co. (NYSE: F) +9.0% BankAmerica (NYSE: BAC) +6.4% Beating the S&P -0.5% S&P 500 +1.9% Compound Annual Growth Rate from 1-2-87: Beating the S&P +20.6% S&P 500 +17.9% $10,000 invested on 1-2-87 now equals: Beating the S&P $94,900 S&P 500 $72,100