Friday, January 16, 1998
The Daily Workshop
Report
by Robert Sheard
(TMF Sheard)
LEXINGTON, KY. (Jan. 16, 1998) -- If you've been reading my Workshop columns for any length of time, you're familiar with my newest screen, the Keystone model. I developed this screen in 1997 and then refined it to make the screening process simpler once I acquired some historical data with which to back-test the theory.
Originally I began with a list of the 100 largest American stocks (by market capitalization) and then narrowed that field to those that carried either a 1 or 2 ranking for timeliness in the Value Line Investment Survey. This approach yielded anywhere from 20 to 35 stocks, which I then sorted by 26-week total return to select those with the best relative strength.
To make the screen more regular once I began back-testing it, I abandoned the idea of a fixed field of 100 stocks and opted for a fixed number of Keystone stocks I wanted to work with -- 30. In order to effect the change in the screen, I now follow this process when screening for the Keystone stocks each week:
I start with all American stocks ranked 1 or 2 for timeliness. Then I rank them by market cap and select the 30 largest. The final screen is still the 26-week total return. This revised method produces a somewhat similar group of stocks each week, but because it forces us to include 30 stocks every time, there are occasions where stocks get included that have market caps that lie outside the top 100 stocks. (They're still all large caps, of course, but the field is expanded a bit on occasion.)
In my original back-testing of this approach I wasn't able to filter our foreign stocks automatically, so my group of 30 stocks often included three or four foreign companies, sometimes more. That means I wasn't replicating precisely the field of 30 stocks we had been striving for.
So I went back to the historical data this time and recreated the screens and counted down the market cap list until I was sure I had exactly 30 American stocks before I made the cut-off. Fortunately, this re-run of the historical data shows virtually no difference in the long-term returns since 1986, although some individual years varied, of course.
As this re-run of the numbers is more precisely the screen we're now calling the Keystone approach, however, we should be using these numbers in our comparisons rather than the ones I've been posting for the screen.
Year Key5 Key10 Key15 1986 22.0 23.8 29.9 1987 8.1 11.3 10.0 1988 8.5 6.9 11.7 1989 59.2 52.6 47.4 1990 -7.7 -2.9 -1.3 1991 71.8 60.0 51.4 1992 12.4 3.3 2.8 1993 36.3 15.1 8.0 1994 9.0 11.3 9.7 1995 43.8 45.5 44.1 1996 38.2 34.2 30.9 1997 56.6 55.6 49.4 1998 0.5 0.9 0.2 (through 1/7/98)
What I find most comforting is that there's a steady progression in higher returns as the model gets more concentrated from 30 to 5 stocks, suggesting that the final screen of 26-week total return is, in fact, a good choice.
Here are the compounded returns for various size groups of Keystone stocks. In 1989 I'm missing the returns for two stocks, Time and Squibb, which merged with other companies, of course, and disappeared from the database. Neither stock was among the top 15 for 1989, though, so the returns are only understated for the portfolios with 20 or more stocks that year. Adding those returns back in when I locate them will probably raise the 1989 returns for those larger groups by two or three percentage points.
5 stocks 27.7% 10 stocks 24.7% 15 stocks 23.0% 20 stocks 21.9% 25 stocks 20.3% 30 stocks 19.8%
I also find it extremely comforting that even selecting all 30 stocks over the past 12 years would have generated returns better than the S&P 500 Index (16.9%) and better than the more concentrated Beating the Dow 5- and 10-stock models (18.9% and 18.8% respectively). Next week, I'll post a file or include in one of my Workshop columns the list of 30 stocks for each year that generated these numbers. Fool on!
[Want to be the first Fool on your block to get a copy of Robert Sheard's forthcoming book? Click here to pre-order your copy of The Unemotional Investor.]