No. of Recommendations: 2

Harmy,

Very good questions.

One of the reasons I have posted this on the board here is in the hope that some experienced people can provide input and help to make the rules more concrete.

There is a very good case for choosing either the 1st year earnings or the 2nd year earnings. I've always struggled with the idea of using past earnings when you are investing for the future. Surely the future matters more. The trouble with using the average is that a very good past year, maybe with a few extraordinary items could bolster the figure even for a company whose future earnings are in decline. The same applies to using the maximum of the three figures. On the plus side, if the maximum is in the future, say the second year, you could pick up the stock at very low price now.

Using 2nd year earnings could lead to longer holding times, but fewer transactions, since many of them will not change in the second year. You could be holding some duds for a long time before they either pay off or drop out of the list. CLI (at the top of the list last year and this year could be an example). Two year forecasts are also likely to be less reliable, so you would need to watch them regularly for revisions. This latter point is probably what leads to the greater variability. Very close monitoring of these revisions could actually lead to more transactions. The market quite probably looks this far ahead, though, since it has the best correlation.

Using 1st year earnings is probably the safest and best. The numbers are fairly reliable , except in a roaring bull market. Surprisingly, in practice, the results are almost identical to using two year forecasts, except for the actual order of the rankings. In other words, you usually get the same stocks, just ranked in a different order. If you are producing the portfolio at this time of the year, just as annual reports and new earnings estimates are being published, the 1st year earnings makes sense. At other times of the year, the second year earnings may be more useful. I haven't done enough testing yet to see if this strategy works at any time other than between August and October.

Using the maximum seemed to me to be more logical. I think it would be better to limit it to the maximum of the two forecasts, ignoring the past earnings. That way you cover a stock that is just turning around and whose earnings are growing rapidly for the foreseeable future, while getting it at this year's price, instead of next year's. Unfortunately, in practice, it is not as good as the previous two, from what back testing I have done so far. Possibly it combines the disadvantages of both methods. Possibly it loses more 'quick gainers' – those that rise in the coming year—than it gains in the following year. Either of the first two methods would have picked up the best undervalued stocks in one of the two years anyway.

For those with a statistical bent, here are some interesting numbers, showing the correlation between the metric and the actual growth over one year.

Metric Correlation

Exp 2Yr Growth 65.9%

Exp 1Yr Growth 58.8%

Best Exp Growth 43.1%

Exp Growth (at 1-8-2001) 35.0%

Yield 35.0%

P/Book ratio 8.6%

Return on Capital 7.8%

Return on Equity 4.5%

Future Exp Growth (1-8-2002) -10.5%

Market Cap -11.7%

Yield -14.4%

Old P/E -14.7%

P/E Growth ratio -20.2%

Old Price -24.2%

'Exp 2Yr Growth' was 15 times the two year forecast divided by the price at August 2001.

'Exp 1Yr Growth' was 15 times the two year forecast divided by the price at August 2001.

'Best Exp Growth' was the result using the best of those two.

'Exp Growth' was 15 times the actual earnings, divided by the price at August 2001.

'Future Exp Growth' is the same, but using August 2002 price. It is there for comparison.

For those without a statistical bent, here is an explanation of what you can take from that list. Take P/E Growth ratio. It is negative, meaning that stocks that had higher values of P/E Growth ratio coincided with lower returns. It also means that in 20% of cases, it was significant. In 80% of cases, it was no better than picking the stocks out of a hat. As any gambler knows, that's not a bad edge all the same. Correlation using the 15 rule are quite extraordinary. Statisticians however, prefer to see the correlation at greater than 90% for statistical confidence. Investors will have to settle for being less rigid..

About ninemsn Finder. I use cut and paste or export. I have tow PCs, with different versions of Excel. One will not accept the export very well. Copy and pate is a bit tedious – there are about seven pages, but it is workable.

The following should be compulsory reading before considering this strategy:

http://www.fool.com/portfolios/RuleMaker/2002/rulemaker020807.htm

http://www.fool.com/DripPort/2002/dripport020815.htm

Chris