Benjamin Graham's Last Will: 10 useful rules for stock selection
What may be less known to you, though, is some quantitative work that Graham did towards just before his death (known as his "Last Will") with the help of a aeronautical engineeer James Rea, where he first identified the 10 best-performing stock selection criteria and then apparently distilled them into the 3 most important criteria.
It seems that, upon reading an article that Graham had written for Barron’s—“Renaissance of Value”— Rea forwarded some of his quantitative/screening research to Graham. This led to a three-year working relationship before Graham died, which culminated in two articles, one by Rea and one by Blustein (for Forbes) where they set out the findings of the research based on 50 years of back-testing as to the most effective value screening criteria for the US market.
Benjamin Graham's 10 Rules for Stock Selection
Here's the list that Graham came up with. The idea behind the rules is that the first five measure "reward" (by pinpointing a low price in relation to key operating results like earnings) and the second five "risk" (by measuring financial soundness and stability of earnings).
1. An earnings-to-price yield at least twice the AAA bond rate
2. P/E ratio less than 40% of the highest P/E ratio the stock had over the past 5 years
3. Dividend yield of at least 2/3 the AAA bond yield
4. Stock price below 2/3 of tangible book value per share
5. Stock price below 2/3 of Net Current Asset Value
6. Total debt less than book value
7. Current ratio great than 2
8. Total debt less than 2 times Net Current Asset Value
9. Earnings growth of prior 10 years at least at a 7% annual compound rate
10. Stability of growth of earnings in that no more than 2 declines of 5% or more in year end earnings in the prior ten years are permissible.
Unfortunately, the issue with these criteria is that, if all 10 are used, the criteria are just too onerous and are unlikely to result in a meaningful number of picks, especially with changing market conditions and business practices over time. The question which Graham and Rea explored is whether certain criteria can be preferred over others?
The Magic 3
The caveat here is that we've unfortunately not yet managed to get hold of the Graham-Blustein article, so this is based on secondary material. However, this source indicates that Graham found that the earning yield and the dividend yield criteria (i.e. the criteria numbered 1 and 3) to be by far the most important performance criteria, while also finding that criteria 1 in combination with criteria 6 would perform almost as well as all 10. Blustein (who wrote up the work for the Forbes article) apparently suggested that criteria 1, 3 and 6 were the most profitable (similar results have been found on the Johannesburg stock exchange). This ties to another source which indicates that
"Graham stated in a lecture at UCLA that if an investor just used earnings yield, dividend yield, and debt to tangible equity, they would get results double the DJIA".
Does it work?
The ten rules apparently produced market-beating returns for five of the six decades that Ben tested it on. Subsequent testing by Henry R. Oppenheimer from 1974 to 1981 found that:
"By using Graham's criteria (1) and (6) to select securities from the combined NYSE-AMEX universe, an investor could have achieved a mean annual return of 38 per cent! Use of criteria (3) and (6) and (1), (3) and (6) would have resulted in mean annual returns of 26 per cent and 29 per cent, respectively".
Before anyone gets too excited, though, it's worth noting that, after Graham died, Rea used the formula in a mutual fund known as the American Diversified Global Value Fund. Run by Rea’s son, it seems that it didn’t do well at all but we don't have any details on why - it may have been a management issue, rather than the rules themself. Interestingly,Old School Value has done some fascinating back-testing more recently for the US market which suggest that it does still work. Interestingly, though, he found that a combination of criteria 1, 2, 6 & 7 worked best - and didn't find criteria 3 (yield) to be additive.