Analysis · The Graham lens

Graham's 22.5 Rule, Measured

Measured 29 July 2026 · 844 companies across 30 indices

Benjamin Graham gave defensive investors two price limits, and then, in the very next sentence, permission to break either one. That second sentence is the part most explanations leave out. We ran it across every company SageRanks tracks to see what it actually changes.

The sentence most explanations skip

Graham's price tests for the defensive investor, set out in chapter 14 of The Intelligent Investor, are quoted often enough to have become clichés: pay no more than 15 times earnings, and no more than 1.5 times book value.

Quoted alone, they read as a checklist. Graham did not write them that way. He added immediately that a low earnings multiple can justify a correspondingly higher asset multiple, and offered a rule of thumb: the two figures multiplied together should not exceed 22.5.

Notice what that constant is. 15 × 1.5 = 22.5. The product limit is not a third rule sitting on top of the other two. It is the same budget, spent freely. A company at 10 times earnings is allowed 2.25 times book. A company at 30 times earnings is allowed 0.75. The total price of a unit of earnings and a unit of assets stays exactly where Graham set it; only the mix changes.

That one sentence turns a rigid checklist into a trade, and it is the difference between failing a company on price and failing it on formalities.

What it does to 844 companies

771companies reporting both a P/E and a price-to-book
103pass both limits on their own
184pass once offsetting is allowed
81rescued by that single sentence

The rule is not a rounding error. It adds 81 companies to a list of 103, which is roughly 79 per cent more names, and it does so without loosening Graham's standard by a single unit of value. Every one of those 81 companies still costs 22.5 or less for a unit of earnings and a unit of book value combined. They simply pay for it in a mix the checklist does not expect.

Of the 844 companies we track, 771 report both figures. The rest are missing one or the other, most often a price-to-book, and we exclude them rather than let a company pass a test it was never measured against.

The two shapes of a rescued company

CompanyMarketP/EP/BProduct
SemapaPSI 2033.80.6622.3
Mapletree Industrial TrustSTI 3027.20.7921.5
FreseniusDAX 4016.71.2721.2
ING GroupAEX 2513.21.6321.5
Hikma PharmaceuticalsFTSE 10012.01.8422.0
FortescueASX 2011.01.9821.7
Barrick GoldTSX 609.82.1921.5
IAGFTSE 1007.03.0621.4

Two patterns run through the list, and they are mirror images.

The first is asset-cheap and earnings-expensive. Semapa trades at nearly 34 times earnings, far past Graham's limit, but at two thirds of its book value. That shape usually means a cyclical business near the bottom of its earnings cycle, or a holding company whose reported profits understate what it owns. Graham, whose entire method started at the balance sheet, would have wanted to look.

The second is earnings-cheap and asset-expensive. IAG trades at 7 times earnings and more than 3 times book. Airlines lease their fleets and carry thin equity, so book value flatters nothing; judged on assets alone the company looks expensive, and judged on earnings it is one of the cheapest names in the index.

Both shapes are invisible to the checklist. Both are precisely what Graham's sentence was written to catch.

What the rule does not fix

Offsetting widens the net. It does not change what kind of company falls into it.

SectorCompanies passing
Financial services53
Consumer cyclical22
Real estate22
Industrials20
Basic materials12
Energy12
Utilities11
Communication services10
Consumer defensive8
Healthcare5
Technology5

Of the 184 companies that pass, 53 are in financial services: banks, insurers and asset managers, close to one in three. Real estate adds 22 more. Technology and healthcare contribute five each, fewer than one in thirty between them.

That is not a data problem and it is not a fault in the screen. Book value measures what a company owns minus what it owes, and that figure means something very different for a bank, whose balance sheet is the business, than for a software company whose main assets are a codebase and the people who wrote it. Graham's asset test was built in an age of factories and railways. Pointed at 2026, it quietly becomes a screen for balance-sheet-heavy industries.

We publish that breakdown rather than smoothing it away, because a value screen returning banks and property is telling you something true about the method, not about the market. If you want to compare a bank with other banks instead of with software companies, the sector rankings do exactly that.

Two criteria out of seven

Something else is worth saying plainly, because most screens built in Graham's name do not say it. The price tests are the last two items on a list of seven. The other five are:

  1. Adequate size.
  2. A sufficiently strong financial condition, which Graham put at a current ratio of at least 2 to 1.
  3. Earnings stability: positive earnings in each of the past ten years.
  4. An uninterrupted dividend record going back twenty years.
  5. Earnings growth of at least a third per share over ten years, measured on three-year averages.

We test the sixth and the seventh. We do not test the first five, because our data source gives us a company as it stands today and not as it stood in 2006. Four of the five need a decade or two of history, and a screen that quietly drops them is not running Graham's method: it is running the two cheapest parts of it.

That matters for how much weight the result deserves. A company can pass the 22.5 rule comfortably while having lost money in four of the last ten years and never paid a dividend, and Graham would have rejected it at criterion three without ever reaching the price test.

Where our version is looser than his

One honest difference. Graham specified 15 times the average earnings of the past three years, precisely so that a single unusually good or bad year could not decide the verdict. SageRanks uses trailing twelve-month earnings, because that is what our data source reports consistently across 844 companies in 30 markets.

For a stable business the two are close. For a cyclical at the top or bottom of its cycle they are not, and this is the one place where our version of the rule is more permissive than Graham's. A miner at peak earnings will look cheaper here than his arithmetic would have allowed. Reading the sector mix above with that in mind is worth more than any single position in a ranking. The full weighting, and every other deviation we know about, is on the methodology page.

How to read the result

Passing the 22.5 rule is not a verdict. A screen answers one narrow question, which is whether a company is priced like a defensive investor's stock. It says nothing about whether the business is any good, whether the earnings will persist, or what the balance sheet is hiding. Graham used his limits to decide what to research, never what to buy, and a rule that rescues 81 companies from a technicality is only useful if you then go and read about them.

Sources

Benjamin Graham, The Intelligent Investor, chapter 14, "Stock Selection for the Defensive Investor." The seven criteria quoted above, including criterion 6 (price no more than 15 times the average earnings of the past three years), criterion 7 (price no more than 1.5 times the book value last reported) and the rule of thumb that the two multiplied together should not exceed 22.5.

SageRanks measurement, 29 July 2026. 844 companies across 30 indices, of which 771 reported both a trailing P/E and a price-to-book. Underlying figures from Yahoo Finance, collected by our own pipeline. Every number in this article comes from that single snapshot, and all of them move with the market. The weightings, thresholds and known deviations are documented on the methodology page.

Nothing else is cited here, deliberately. Where this article makes a claim about Graham it points at the chapter you can check it in, and where it makes a claim about the market it points at a measurement you can rerun.

See the 184 companies that pass Graham's screen today →
Graham Buffett Lynch Fisher Greenblatt
Not investment advice. SageRanks is an analysis and education tool. Nothing here is a recommendation to buy or sell any security. Figures were measured on 29 July 2026 and move with the market. Do your own research. That is rather the point.