The Golden Rules?
[Editor's Note: Jim Stevens is on vacation this week. In his place we present a reprint from former Workshop writer-since-moved-to-the-Foolish-Four Ethan Haskel. It originally ran on November 4, 1998.]
BALTIMORE, MD (August 12, 1999) -- Some time ago, we discussed the importance of finding an optimal database from which to select stocks for a proposed screening strategy. Let's call such a potential new strategy the "Beating Everything Super Trust," or BEST for short. How do we choose those few golden nugget stocks for our BEST Portfolio out of a database that may contain hundreds or even thousands of companies?
We obviously need some rules. But what kind of rules work best? I've listed below six guidelines for writing the rules for any screening model. What follows might be considered the "rules for the rules."
1. The rules should be specific and objective. Give 10 people the rule book, and each should be able to come up with the same list of stocks. No hunches are allowed. No ambiguities are tolerated.
2. The rules should use relative rather than absolute criteria. For instance, let's examine a rule that states: "First, choose all companies from the database that have a market capitalization over 25 billion dollars." The "25 billion dollars" figure represents an absolute number. Such a rule might work this year, next year, and even five years from now. But what about 15 years hence, when half the stocks in the S&P 500 Index might have such a market value? Similarly, a rule that states: "Choose all stocks with dividend yields over 3%" might be useful today but much less so when interest rates are 10% and the average stock yields 6%.
A much better rule for choosing large company stocks for our BEST Portfolio might be: "First, choose the 100 largest companies from the database." The stocks chosen using values relative to other stocks hold up well for all time points in all market conditions. Whether it's 20 years ago or 20 years from now, you'll still have the stocks you want, without needing to change the rules midstream.
3. The simpler the rules, the better. Uncomplicated guidelines are easier to work with and inherently less prone to errors or manipulations. The medieval monk William of Ockham said it best about 700 years ago: non sunt multiplicanda entia practer necessitatem, or entities are not to be multiplied beyond necessity.
4. The rules should be based on sound investment principles. Without such a foundation, the possibility for data-mining errors increases dramatically. Data-mining traps are set when large numbers of variables are analyzed and applied nonsensically to obtain the best fit for a hypothesis. For instance, check out the essay "Cheating the Dow," in which I showed that certain Dow stocks that ended the year with a particular fraction in the closing price perform phenomenally. Obviously, the conclusion is spurious. Claims based on questionable assumptions should be regarded even more skeptically than usual.
Time-tested stock formulas often best form the basis for excellent screening rules. Such formulas may include relative strength for growth stocks, or the price to sales ratio (or dividend yield) for value stocks. Often, combining proven stock characteristics in novel combinations leads to fruitful new approaches.
5. The rules should select an appropriate number of stocks. This rule is self-explanatory. For instance, many of James O'Shaughnessy's screens are not appropriate for individual investors. They choose 50-stock portfolios, which are better suited for fund managers than individuals.
6. The rules should select portfolios that can be backtested. Without the ability to do backtesting, the usefulness of any screening strategy is diminished greatly. Old databases should be available that allow you to choose the stocks as if you actually had been investing at that time.
There likely are a number of other appropriate guidelines that I've overlooked. If you've got some others, let's discuss them on the message boards. Who knows, maybe you too can come up with your very own BEST Portfolio strategy!