The Graham number and its limits
A one-line formula for the most a defensive investor should pay. It's quick and it's everywhere, and it was built for a market of factories and railroads. How to use it, and when to ignore it.
Test 7, solved for the price
Graham never wrote down a "Graham number". The formula people use under his name is test 7 from lesson 3, rearranged. If the P/E can be at most 15 and the P/B at most 1.5, their product can be at most 22.5. Solve that for the price and you get a ceiling.
It's the most a defensive investor should pay. It was never meant as a fair value or a target price, though it's often quoted as one.
There's one quiet change from the book. Graham's test used average earnings over three years. Most websites, ours included, plug in the latest twelve months instead, which makes the number jumpier.
Back to Marlow
Marlow sits just under the ceiling, which matches lesson 3, where it passed test 7 through the 22.5 clause. There isn't much margin here. Graham's defensive investor could buy it, but he wouldn't be getting a bargain.
Two inputs, two ways to fool it
The formula only has two inputs, so anything that bends earnings or book value bends the answer.
Kestrel may well be expensive, but the formula can't tell you. Most of what Kestrel owns, the software it wrote and the customers it won, was paid for through the income statement and never reached book value. Buybacks make it worse. If book value is zero or negative, the square root has no answer at all.
On this year's earnings an $80 price looks 37% below the ceiling. On normal earnings it's about 9% above. Graham's three-year average would soften the peak a little. For a business this cyclical we'd average seven to ten years.
Graham on growth stocks
Graham didn't think every company should be valued on its assets. In chapter 11 of the 1973 edition he offered a rough rule for growth companies, built on normal earnings and the growth rate expected over the next seven to ten years.
A company with no growth deserves 8.5 times earnings. One expected to grow 7% a year gets 22.5 times. At 10% growth the multiple is 28.5, and at 15% it's 38.5. A few points of optimism move the answer a long way, and Graham knew it. That sensitivity is the reason he trusted assets and past earnings more than any growth forecast.
Where you'll see it here
The Graham number appears among the key metrics on each company's Summary tab, and again in the Valuation tab next to the seven tests. In the Valuation tab we leave it out of the blended fair value for REITs and for companies with a return on equity above 20%. In both cases book value says too little about what the business is worth.
Our rules of thumb for using it:
- Use it on steady, asset-heavy businesses with ten years of profits, such as distributors and manufacturers.
- Swap in a three-year average EPS, or a longer one for a cyclical.
- Read it as a ceiling. A price well below it is interesting. A price above it raises a question about intangibles, and that isn't automatically a no.
- Ignore it for asset-light companies with high returns on equity, and for anything with negative book value.
For those businesses, our ROIC lesson and the cash-flow models in DCF will tell you much more.
What still holds
Graham's formulas show their age. His habits don't. Have your own idea of value before you look at the price. Buy with room for error, prefer companies that can survive a bad year, and spread your bets.
Much of what Buffett and Greenblatt did later starts from there. Greenblatt's version, a ranking of good businesses at cheap prices, is in DL 364, the Magic Formula.
Takeaways
Sources for this lesson
- Benjamin Graham, The Intelligent Investor (1973), chapter 14 (test 7) and chapter 11, Security Analysis for the Lay Investor
- Benjamin Graham, The Intelligent Investor, revised edition with commentary by Jason Zweig (2003)