Where the tradition stands

Ninety years of writing produces a lot of claims. Most arguments about value investing are unproductive because they mix three different kinds of claim together. Sorting them apart is most of the work.

Column one: what is confirmed

That sorting a market by a price multiple separated subsequent returns, in more than one sample, by more than one author, with the cheaper group ahead. Basu found it on New York Stock Exchange industrials between 1957 and 1971. Lakonishok, Shleifer and Vishny found it across two exchanges between 1968 and 1990 on three different multiples. Fama and French found book-to-market carrying explanatory power for average returns in 1992.

These are different samples, different periods and different sorting variables, and the direction is the same in all of them. That is the strongest thing the tradition has, and it is worth stating exactly: a past association between a low multiple and a higher subsequent return, measured repeatedly.

Column two: what is contested

Why it happened. The risk reading says the cheap group was riskier in a way the measurements did not capture and the extra return was the wage for bearing it. The behavioural reading says investors extended recent results too far and the extra return was the correction.

Both readings fit the same numbers, which is why fifty years of argument has not closed it. The risk reading owes a name for the risk. The behavioural reading owes an account of why the mistake keeps being made by people who can read the papers describing it.

Also contested: whether the association held up through the long stretch after 2007, when cheap-by-book-value companies trailed expensive ones for longer than most earlier dry spells on record. Three separate papers published in 2021 examined that stretch and did not agree on its cause, which is a different lesson.

Column three: what is open

Whether a specific printed threshold selects anything in a specific market on a specific date. This is not a question the literature can answer, because it is a question about now.

The figure above is one instance, and it carries the day it was computed: on the 2026-09-13 build, Graham's ceiling of 15 times earnings sat between the 25th percentile and the median, which is not the rare line most readers assume it is.

Ask the same question now and the market answers for itself. Across the 2,625 (built Oct 21, 2026) US companies with a computed price to earnings, the median is 19.84x (built Oct 21, 2026) and the cheapest quarter ends at 12.48x (built Oct 21, 2026). Those three are read when this page is served and each prints the build it came from. Where 15 falls between them is the whole of the open question, and it is open because the line was printed once and the market was not.

The same exercise on a different criterion

His financial-condition test asks for a current ratio of at least 2.0. Across the 4,220 US companies with a computed current ratio on 2026-09-12, the median is 1.87 and the 75th percentile is 3.65. So a floor of 2.0 sits just above the middle of the market: it removes slightly more than half of it.

Set those two side by side and something useful falls out. In this market, on these dates, the price test and the balance sheet test are about equally demanding, which is not what most summaries of Graham imply.

The honest summary

The tradition has a real empirical record, an unresolved explanation, and a set of printed thresholds whose usefulness depends entirely on the market you are standing in. Anyone telling you it is proven, and anyone telling you it is dead, is collapsing three columns into one.

Check yourself