Warren Buffett · The Superinvestors of Graham-and-Doddsville
He was not the first speaker. On 17 May 1984, Columbia Business School marked the fiftieth anniversary of Security Analysis with a seminar it had arranged as a fight: Michael Jensen of the University of Rochester — one of the hardest heads in the efficient-market camp — went first. His argument was that if a large crowd flips coins, a few will run long winning streaks by luck alone; the fund managers who beat the market year after year are those few. Not skill. Luck.
Then Buffett got up, picked Jensen's metaphor off the floor, and ran it further than Jensen had. He imagined 225 million Americans betting a dollar each on a coin flip every morning, losers dropping out, winners letting it all ride. After ten flips, about 220,000 are left. After twenty, 215 remain, each having turned a dollar into a little over a million. They start writing books called "How I Turned a Dollar into a Million in Twenty Days Working Thirty Seconds a Morning," and jetting around the country to seminars on efficient coin-flipping, taunting the skeptics: "If it can't be done, why are there 215 of us?"
Then he made his opponent's case for him: if 225 million orangutans had done the same thing, you would get the same 215 smug winners. The laugh came. Only then did the knife: but if you found that 40 of those winners came from one particular zoo in Omaha, you would go ask the zookeeper what he feeds them and what books they read. Omaha is where Buffett lives. He had put himself in the cage, then invited the room to go question the keeper.
The rest is nine tables. Walter Schloss, Tweedy Browne, his own wound-up partnership, Bill Ruane's Sequoia Fund, Charlie Munger, Rick Guerin, Stan Perlmeter, plus two pension funds he had steered toward value managers. Nine audited long-run records, all far ahead of the S&P 500, with almost no overlap in holdings — not one man calling the flip and fifty people echoing him. The one thing they share is an intellectual village: all trace back to Ben Graham, all look at two variables only, price and value. A concentration like that, he said, cannot be explained by chance.
What the talk left out is sitting inside the tables he printed himself. He writes that "this group has assumed far less risk than average." Yet Munger's partnership fell 31.9% in 1973 and another 31.5% in 1974 — both years much worse than the Dow. Guerin's Pacific Partners fell 42.1% and then 34.4%. By Mohnish Pabrai's account of what Buffett told him over a 2007 charity lunch (a paraphrase, not a transcript), Guerin was on margin, got the calls, and had to sell his Berkshire stock back to Buffett at around $40 a share. One of the nine superinvestors was nearly erased in those two years — and the word leverage never appears in the talk.
So how did it hold up? Partly. The most serious outside test came from AQR's Frazzini, Kabiller and Pedersen: Berkshire's Sharpe ratio for 1976–2011 was 0.76, the highest of any stock or mutual fund with more than thirty years of history. Not luck. But they also showed that once you control for two factors — betting-against-beta and quality-minus-junk — the alpha of Berkshire's public stock portfolio falls to 0.3% a year, statistically indistinguishable from zero; and that Buffett ran about 1.6-to-1 leverage, much of it insurance float costing only 2.2% a year, more than three points below T-bills. His edge was real. It was also translatable into a recipe you can now buy for a few basis points. Meanwhile Fama and French's 2020 re-examination found the monthly value premium for big value stocks falling from 0.36% (t = 2.91) in 1963–1991 to "a puny 0.05%" (t = 0.24) in 1991–2019.
The sharpest correction came from Buffett. In 1984 he wrote that "market prices are frequently nonsensical." In 2017, after winning a ten-year bet that an index fund would beat five funds-of-funds, he summed it up differently: "Though markets are generally rational, they occasionally do crazy things." And in his 2013 letter he had already published his instruction to the trustee of the cash left to his wife: "Put 10% of the cash in short-term government bonds and 90% in a very low-cost S&P 500 index fund. (I suggest Vanguard's.)" The man who called the efficient-market crowd the Flat Earth Society prescribed, for his own household, buying the whole market.
He picked up his opponent's weapon, wiped it off, and added a troop of orangutans.
Michael Jensen had spoken first, and his point was unmistakable: the managers who beat the market year after year are no different from people who have flipped twenty heads in a row. Buffett took the microphone and did not dispute the metaphor. He inflated it to absurdity. Two hundred and twenty-five million Americans, out at sunrise, a dollar a head, losers dropping out, winners letting the whole pot ride. Ten days later, 220,000 remain, each up a little over a thousand dollars. Twenty days later, 215 remain, each having turned one dollar into a million.
He then described those 215 people: first modestly confiding their coin-flipping technique to attractive strangers at cocktail parties, then writing books, then "jetting around the country attending seminars on efficient coin-flipping and tackling skeptical professors."
But then some business school professor will probably be rude enough to bring up the fact that if 225 million orangutans had engaged in a similar exercise, the results would be much the same—215 egotistical orangutans with 20 straight winning flips.— Warren Buffett, The Superinvestors of Graham-and-Doddsville, 1984
He had finished his opponent's argument for him, and finished it better. Only then came the turn:
if (c) you found that 40 came from a particular zoo in Omaha, you would be pretty sure you were on to something. So you would probably go out and ask the zookeeper about what he's feeding them, whether they had special exercises, what books they read, and who knows what else.— Warren Buffett, The Superinvestors of Graham-and-Doddsville, 1984
Omaha is his own city. He locked himself in the cage and sent the room off to interrogate the keeper. The keeper's name was Ben Graham.
In 1984 Warren Buffett was 53 and chairman of Berkshire Hathaway, but not yet the symbol he became. He had never written an academic article in his life. This is the only one — and he was dragged onto a stage to produce it.
His relationship to the occasion was awkwardly close. He had taken Graham's course at Columbia in 1951 and got an A+. He then offered to work for Graham for free and was turned down. He tells this himself, and it is the best laugh in the piece:
I offered to go to work at Graham-Newman for nothing after I took Ben Graham's class, but he turned me down as overvalued. He took this value stuff very seriously!— Warren Buffett, The Superinvestors of Graham-and-Doddsville, 1984
After much pestering he was finally hired. From 1954 to 1956 he was one of four staff at the "peasant" level at Graham-Newman. In 1957 he opened his own partnership; in 1969 he closed it on purpose.
Why that year. By 1984 the efficient-market hypothesis had won the academy. The capital asset pricing model and beta were in the textbooks, "risk" had been defined as price volatility, and the Graham-and-Dodd habit of estimating what a business is worth and waiting for a discount was disappearing from the syllabus — as Buffett notes at the end, "the academic world, if anything, has actually backed away from the teaching of value investing over the last 30 years." Choosing the fiftieth anniversary of Security Analysis to stage this had a faint air of a memorial service, and the school set it up as a contest: Jensen for the academy, Buffett for the method being pronounced dead. Jensen went first. AQR's later paper records the scene's standing in the field: efficient-market academics held that Buffett's success might simply be luck, the happy winner of a coin-flipping contest, "as articulated by Michael Jensen at a famous 1984 conference at Columbia Business School."
In the room sat Columbia's faculty and students, and a good number of people still making money with the method. Buffett's job was not to convert Jensen. It was to keep the village on the map.
He opens on the textbooks. Is the Graham and Dodd approach out of date? "Many of the professors who write textbooks today say yes." He quotes one of them anonymously: "If prices fully reflect available information, this sort of investment adeptness is ruled out." Then comes the most economical counterpunch in the piece, two words:
Well, maybe.— Warren Buffett, The Superinvestors of Graham-and-Doddsville, 1984
What he does next is unusual for an argument of this kind: he builds his own obstacle first. The whole thing turns, he says, on the fact that these winners "were all well known to me and pre-identified as superior investors, the most recent identification occurring over fifteen years ago." Then he holds the knife to his own throat: "Absent this condition—that is, if I had just recently searched among thousands of records to select a few names for you this morning—I would advise you to stop reading right here." He adds that all the records are audited, and that he knows many of the people who invested with these managers: the checks they actually received over the years match the stated numbers.
With the premise nailed down, the coin-flipping contest begins. The passage is a funnel, tightening as it goes:
Both men agree that pure chance produces 215 winners. Buffett is not arguing about the count; he is arguing about origin. He reaches for cancer epidemiology: a rare cancer with 1,500 US cases a year, 400 of them in one little mining town in Montana. You would not call that coincidence; you would look at the water. "You would not necessarily know the causal factors, but you would know where to search."
Then — and this is the most skilful stretch in the piece — he starts dismantling his own argument on his opponent's behalf. He raises two conditions that would make a cluster meaningless. First: perhaps 100 of them were simply copying one terribly persuasive personality. He calls heads, a hundred followers call heads, and you have identified one case as a hundred. Second: suppose a strongly patriarchal society where every family of ten calls with the father. Then 215 winners are really 21.5 families, and the apparent hereditary factor is nothing at all. He puts the statistician's objections on the table before the statisticians can, and only then answers: the intellectual patriarch supplied the theory, but each student applied it his own way, and their holdings barely overlap. Schloss's biggest positions were "such stalwarts as Hudson Pulp & Paper and Jeddo Highland Coal and New York Trap Rock Company" — a deadpan joke, since nobody has heard of any of them. Tweedy Browne's names were obscurer still; Ruane worked with big companies; Munger concentrated. "These records do not reflect one guy calling the flip and fifty people yelling out the same thing after him."
The nine records follow: Walter Schloss (never went to college, took Graham's course at night, 28¼ years, partnership up 23,104.7% against 887.2% for the S&P); Tom Knapp's Tweedy, Browne; Buffett's own partnership, closed in 1969; Bill Ruane's Sequoia Fund; Charlie Munger; Rick Guerin's Pacific Partners (19 years, up 22,200% against 316.4%); Stan Perlmeter; and the Washington Post and FMC pension funds.
Scattered between them are asides that have each since been quoted on their own. On who can take the idea at all: "It's like an inoculation. If it doesn't grab a person right away, I find that you can talk to him for years and show him records, and it doesn't make any difference… It's instant recognition, or it is nothing." On scale: "Size is the anchor of performance." On the direction of risk and reward, he produces a revolver: offered $1 million to spin a six-shooter with one cartridge and pull the trigger once, he declines; offered $5 million to pull it twice, that would be a positive correlation between risk and reward. Value investing runs the other way — buying a dollar bill for 60 cents is riskier than buying it for 40 cents, and the expected reward is smaller. Cheapness is not the source of risk. Cheapness is the subtraction from it.
It lands on a bridge:
You don't try and buy businesses worth $83 million for $80 million. You leave yourself an enormous margin. When you build a bridge, you insist it can carry 30,000 pounds, but you only drive 10,000-pound trucks across it.— Warren Buffett, The Superinvestors of Graham-and-Doddsville, 1984
The last paragraph answers the question a businessman would ask: doesn't publishing the secret shrink your own edge? He replies that the secret has been out for fifty years and that in the thirty-five years he has practised it he has seen no trend toward value investing at all. "There seems to be some perverse human characteristic that likes to make easy things difficult." Then the sentence that has been printed on a thousand covers: "Ships will sail around the world but the Flat Earth Society will flourish."
The piece has been quoted for forty years, and almost all of the quoting is the orangutans and the Flat Earth Society. Hardly anyone reads the nine tables. The tables hold what the argument could not, and in places they contradict it.
First, the number the clients got is not the number people quote. Schloss's table prints two columns: the partnership compounded at 21.3% a year over 28¼ years; the limited partners — the people whose money it was — got 16.1%. The five-point gap is the general partner's compensation. Buffett prints both columns honestly and footnotes the reason; posterity quotes 21.3%. Guerin's table is starker: the overall partnership gained 22,200% over 19 years, or 32.9% annually, while limited partners gained 5,530.2%, or 23.6%. Munger's: 19.8% overall, 13.7% for limited partners. The legend of the superinvestors uses the first number in each pair. The investors lived the second.
Second, "this group has assumed far less risk than average" is half-refuted by his own tables. The sentence reads: "It's very important to understand that this group has assumed far less risk than average; note their record in years when the general market was weak." It holds for Schloss (down 8.0% in 1973 against the S&P's 14.8%; down 6.2% in 1974 against 26.6%) and still better for Tweedy Browne (up 1.5% in 1974, when the S&P fell 38.1%). But:
Those figures are printed in his own Table 5 and Table 6, fewer than two pages from the sentence about low risk.
Third, what actually happened to Guerin in those two years is not in the talk at all. By Mohnish Pabrai's account — he won a charity lunch with Buffett in 2007 and later relayed what Buffett said, noting that he was giving the sense of it rather than a verbatim transcript — Guerin was carrying margin loans, got margin calls when the market collapsed in 1973–74, and had to sell his Berkshire stock back to Buffett at around $40 a share. The relay cannot be cited as a primary record, but the −42.1% and −34.4% are Buffett's own figures, and they fit it exactly. One of the nine superinvestors came close to being wiped out in those two years, and the reason — leverage — never appears in an article whose entire thesis is that only price and value matter.
Fourth, the method he describes is not the whole method he was using. The argument is that these people look at two variables and ignore everything else. But Buffett's own machine had a part he did not mention. Frazzini, Kabiller and Pedersen pulled apart Berkshire's balance sheets for 1976–2011 and estimated that Buffett ran about 1.6-to-1 leverage; that insurance float made up 36% of Berkshire's liabilities on average; and that this float cost him about 2.2% a year, more than three percentage points below the T-bill rate over the same period. A meaningful share of his return came from borrowing the cheapest money in the world — and in the 1984 story, where there is only price and value, that part is not there.
Fifth, the denominator never arrives. His defence is pre-identification: these people were not picked after the fact out of thousands of records. That stops the crudest form of selection bias, but not another question: how many students did Graham teach, and how many of them did badly with the method? The only place Buffett supplies a denominator is Graham-Newman, where three of four "peasants" were traceable — a sample of four. The other eight records come from his own partnership, the Sequoia Fund he set up to house his own former clients, his long-time partner Munger, Munger's friend Guerin, Perlmeter from the same office building in Omaha, and two pension funds he personally persuaded to hire value managers. Their holdings really were independent; he proves that carefully. The process that put them on the list was not independent at all. The sole criterion for inclusion was that Warren Buffett knew them and rated them.
PARTLY HELD UP · as of 2026-10
His narrow claim won: the nine were not lucky, and neither was he — the most rigorous outside test to date, designed specifically to route around his selection problem, confirms it. His wide claim lost. The academy was never persuaded to say the market was wrong; it turned value into a factor you can buy today for a few basis points. The value premium itself has run near zero for thirty years. And Buffett's own language, and the will he left for his wife, have both retreated to a position the 1984 Buffett would have mocked.
1992: the academy conceded the fact and kept the explanation. Fama and French showed with US data from 1963–1991 that high book-to-market "value" stocks beat growth stocks over the long run. That amounts to conceding the excess return Buffett was pointing at. Their interpretation, though, was that value stocks are riskier and the premium is compensation for bearing that risk — not mispricing. The efficient-market frame did not change a word; it gained a factor. Buffett wanted "the market is wrong." What he got was "our model needed another term."
Around 1995: the other academic response was no response. The article barely exists in the economics journals — one search of 23,000 economics papers turned up only twenty references to any publication by Buffett. William Sharpe in 1995 called him "a three-sigma event" (roughly one in 370), and Michael Lewis in 1989 called him a big winner produced by a random game. For this page I located only the bibliographic record of those two criticisms, not the originals, so I report the fact and do not reconstruct their arguments.
2013–2018: the most serious test, and it splits. In Buffett's Alpha, Frazzini, Kabiller and Pedersen say outright that the thing they are testing is the 1984 debate, and they name the weakness in Buffett's defence: "Ex post selecting successful investors who are informally classified to belong to Graham-and-Doddsville is subject to biases." So they changed tactics and took Berkshire apart instead. The result: a Sharpe ratio of 0.76 for 1976–2011, higher than any other stock or mutual fund with a history of more than thirty years; an average annual excess return of 19.0% against the market's 6.1%; a market beta of just 0.7. That half is entirely Buffett's. Jensen's luck hypothesis is dead.
The other half is less kind. Add two factors to the regression — BAB (long low-beta, short high-beta) and QMJ (long quality, short junk) — and the annualised alpha of Berkshire's public stock portfolio drops to 0.3%, indistinguishable from zero. Their conclusion:
Buffett's returns appear to be neither luck nor magic, but, rather, reward for the use of leverage combined with a focus on cheap, safe, quality stocks.— Frazzini, Kabiller & Pedersen, Buffett's Alpha, NBER WP 19681, 2013
It is the fullest verdict the debate has received, and an awkwardly shaped one: Buffett wins "not luck" and loses "not replicable." His method was written down as a recipe — by AQR, an asset manager that packages those very factors and sells them.
2015–2016: one of the nine stepped on the mine they all warn about. Sequoia was the one place Buffett recommended to his own partners when he wound up the partnership in 1969. At the end of 2015, Valeant Pharmaceuticals was its largest holding at 19.3% of assets (reporting earlier that year put the position above thirty per cent). Valeant fell roughly 80% in the first half of 2016; Sequoia was fully out by June, and co-manager Bob Goldfarb left in March. In the fund's own 2016 annual report to the SEC, the fund fell 6.90% that year while the S&P 500 gained 11.96%. The same filing carries another number: $10,000 invested at the fund's 1970 inception was still worth $3,612,432 at the end of 2016. The method won over forty-six years. The discipline failed in a single position.
2020: the excess return mostly vanished. Fama and French went back over the fifty-six years from 1963 to 2019. The monthly premium of big value stocks over the market was 0.36% (t = 2.91) in the first half — exactly the sample they had published on — and in the second half, 1991–2019, it fell to what they themselves call "a puny" 0.05% (t = 0.24). Small value fell from 0.58% to 0.33%. But they refused to pronounce it dead. Monthly premiums are too volatile: they state plainly that they "don't come close to rejecting the hypothesis that out-of-sample expected premiums are the same as in-sample expected premiums," while also being unable to rule out a wide range of much lower values. If Fama and French will not render a verdict, that is the most honest answer available to the prophecy that those who read their Graham and Dodd will continue to prosper: nobody knows whether the soil in that village is still fertile.
The sharpest correction is his own. In 1984 he wrote: "I'm convinced that there is much inefficiency in the market… In fact, market prices are frequently nonsensical." In late 2007 he bet Protégé Partners that an S&P 500 index fund would beat five funds-of-funds over ten years. The 2017 scorecard: the index fund up 125.8% (8.5% a year), the best of the five up 87.7%, the worst up 2.8%. Summing up that victory in the shareholder letter, his wording was:
Though markets are generally rational, they occasionally do crazy things.— Warren Buffett, Berkshire Hathaway 2017 shareholder letter
"Frequently nonsensical" had become "generally rational, occasionally crazy." And back in the 2013 letter he had already made part of his will public — for the cash left to his wife, the instruction to the trustee reads:
Put 10% of the cash in short-term government bonds and 90% in a very low-cost S&P 500 index fund. (I suggest Vanguard's.)— Warren Buffett, Berkshire Hathaway 2013 shareholder letter
He adds that he believes this will beat the results of "most investors — whether pension funds, institutions or individuals — who employ high-fee managers." In 1984 he said those who read their Graham and Dodd would continue to prosper. In 2013 the prescription he wrote for his own household was: don't read, just buy the whole market. This is not a contradiction — he always said most people could not do it, and the nine were his proof that a very few could. It is only that over forty years what got remembered was the orangutans and the Flat Earth Society, and not the condition he nailed down at the start and never moved: I knew these people in advance.