THE CHOIR AT BEAUVAIS

In the thirteenth century, Beauvais set its builders an almost reckless target: a Gothic choir taller than any yet built, raised in a town without the population or the money to obviously justify it. The vaults climbed to 47.5 metres above the floor, higher than anything that had come before, pushing the pointed arch and the flying buttress to the edge of what either technique had ever been trusted to carry. Part of that vaulting came down in 1284, not long after workers had finished it.

The rebuilding that followed added extra piers and narrowed the spans between them, trading a visible measure of the original’s ambition for a quieter measure of margin, and the choir has stood ever since, through wars, revolutions and the ordinary erosion of eight intervening centuries. Ambition came back anyway, three hundred years on: a soaring crossing tower, finished in 1569, went up while the nave meant to brace it still existed only on paper. That tower stood for four years. It fell on Ascension Day, 1573, with the bracing nave still unbuilt. A later attempt to finally add the nave began in 1600 and stopped after a single arch, leaving the ambition on record without ever quite finishing it. Beauvais Cathedral still stands today, nearly eight centuries after construction began, missing the one part a complete cathedral is ordinarily supposed to have.

Socrates claimed only one advantage over the famously clever men of Athens, and it was a strange one to claim as an advantage at all. Those Athenians were confident they possessed knowledge they simply did not have. Socrates, by his own account, understood only the size of his own ignorance, and refused to build anything on top of that gap by pretending otherwise. By any ordinary measure of the craft, the masters of Beauvais ranked among the most capable builders of their age. Their brilliance was the problem, in exactly the sense Socrates meant it: skill at the calculation makes it easy to mistake the calculation for the ground it is standing on.

Benjamin Graham put a name to that same Socratic caution in the final pages of his book The Intelligent Investor, and ranked it, without qualification, the single most important concept the whole discipline owns: the margin of safety. It means buying meaningfully below one’s own estimate of what something is worth, so that room remains for the estimate itself to be wrong. The word doing the real work in that sentence is estimate. Earnings can disappoint. The discount rate can move. A competitor can arrive that nobody in the room had modelled. The analyst, not least, can simply be mistaken about his own reasoning, in ways invisible to him at the moment he is reasoning.

Personal history explains some of Graham’s own insistence on this point. His fund lost close to seventy per cent of its value in the 1929 crash, which he later blamed squarely on a failure of humility: a safety-first reputation that had, in practice, drifted into speculating on earnings no sober observer should have expected to arrive. The margin of safety, as he wrote it up afterward, reads like a personal debt being worked off in print, one principle at a time.

A margin of safety works as an advance confession, written directly into the price paid or the structure built: an admission that the person doing the estimating might simply be wrong, settled at the one point in the whole process where no meeting, no committee and no later argument can talk it back out again, well before any actual disaster has had the chance to arrive and prove the point. A courtier studying the broad foundation beneath a merely adequate tower sees stone doing nothing useful. An engineer studying the same foundation sees a structure built to survive its own architect having got something wrong.

Engineers outside cathedral-building have formalised the same instinct under its own name. Ordinary structural design commonly calls for members built to carry one and a half to twice the maximum load anyone expects them to bear in practice, and considerably more than that for anything, a lift cable, a pressure vessel, whose failure would kill the people standing nearest it. Nobody expects the maximum load every day. The factor exists for the day somebody mis-estimated what the maximum load would be. A passenger lift’s main cables are commonly rated to hold several times the car’s maximum load, precisely so that the one ride that turns into an emergency never depends on an estimate someone got only approximately right.

Elegance, left to its own preferences, always wants the margin gone, because the margin looks exactly like doubt wearing a disguise of extra stone or unspent capital. In good years an unused margin reads as inefficiency, and the neighbouring tower that skipped it looks bolder for as long as the weather holds. Capital kept in reserve earns a visible, immediate cost every quarter, charged against an invisible, merely probable benefit that may never need to be collected at all.

This is the specific difficulty with defending a margin of safety in front of anyone who rewards nerve: the entire case for it rests on a future failure that, if the margin works, never actually happens. An investor who holds cash through a rising market looks timid for exactly as long as the market keeps rising. A fund that insists on buying only well below its own valuation watches faster-moving competitors report better returns for years at a stretch, with no applause available anywhere for a margin that is merely sitting there, unused, doing the only job it has ever had to do.

Graham’s own investing method made the logic almost comically literal. He sought out companies trading for less than the value of their cash and inventory alone, after every liability was subtracted, so that even a thoroughly mediocre business, treated as scrap rather than as an enterprise with a future, would still return the purchase price. Critics, not unreasonably, called it a cigar-butt approach: picking up a single free puff from a stub someone else had already thrown away and smoked past the point of pleasure. Graham’s own point was narrower and less romantic than either the praise or the criticism suggests. He was making sure that being wrong about the business, which he fully expected might happen, still left him approximately whole, regardless of whether anyone ever called the business itself wonderful.

Committees are where this argument is actually lost, far more often than in any individual’s own head. Anyone who has sat through a budget review has felt first-hand just how difficult unused stone, or uncommitted capital, is to defend to colleagues who can see a specific, named, numerically attractive opportunity that the caution is currently declining to fund. The margin has no champion in the room except an argument about a failure that has not happened yet and, with any luck, never will.

Skipping the margin at scale has its own well-documented ending. Lehman Brothers ran at roughly twenty four times leverage in 2003 and had pushed that figure to about thirty one times by 2007, meaning a decline of only three or four per cent in the value of its assets was enough to erase its entire equity cushion. For years the extra leverage simply produced extra returns, exactly the way skipping a margin always does, right up until September 2008, when a decline nobody had priced in arrived in a single, calamitous week, and there was no cushion left to absorb it.

Decades usually settle the argument more convincingly than any committee meeting does: whichever structure is still standing once the weather the margin was quietly built for has actually arrived makes its own case, without needing to speak.

Beauvais makes the final comparison almost too neat to need stating. The choir that was rebuilt wider than strictly beautiful, piers added at the cost of a little of the original vision, is the part of the cathedral still open for a service today. The tower that was raised fast, elegant, and ahead of the structure meant to support it lasted four years before Ascension Day finished the argument for good.

An analyst is being quietly teased this quarter for holding too much cash and buying too little, while a colleague two desks away posts better numbers on names bought at full price and better still. Nobody in the room is being irrational. One of them is simply pricing in a mistake that has not happened yet, a precaution that costs something every ordinary year and earns its entire keep in exactly one: the single year it is the only reason anyone in the room is still solvent. Socrates would probably have recognised the colleague with the better numbers as the more impressive man that quarter, and the quieter one as the better bet across every quarter after it.