It’s silly to blame risk-management calculations for sinking financial firms.

One cannot read very far into the literature on the current financial crisis without bumping up against Warren Buffett's somber warning, "Beware of geeks bearing formulas." This little brickbat of Omaha wisdom hits me like a hard punch in the gut. You see, I was once a geek. Yes, I was a management consultant, and for a time I made a living advising large banks on risk management.

It is now widely understood that over the past decade risk management at financial institutions around the world failed on an epic scale, and that this failure helped precipitate a recession that will cost millions of people their jobs and their homes. The humbling question is: How could this have happened? Weren't we—the banks and their advisers—supposed to be too smart to screw up in this way? The commentators who regularly cite Buffett's maxim have a theory. The problem, they say, is precisely that we thought we were too smart.

It does look pretty suspicious. I remember well the fishy smell that emanated from many of those financial models that I delivered to clients. When confronted with exotic financial instruments that are not regularly traded, for example, my fellow quants and I fabricated "virtual" markets, stashing heaps of dubious assumptions underneath our well-trimmed spreadsheets. All too often we ignored whole categories of risk—just because analyzing them wasn't part of our assignment or because they didn't fit neatly into our equations. We downplayed "liquidity risk"—roughly speaking, the unpleasant possibility that in the midst of a market panic when everyone is rushing for the exit, you may not be able to get out the door when it suits you. Neither were we too keen on "systemic risk"—the annoying possibility that everything, including all those pretty insurance policies that you thought were protecting your positions, might go up in smoke all at once. I never could shake the impression that the precise statistical predictions we offered to clients—often taken to decimal extremes—were like so much fine perfume sprinkled on a messy pile of guesswork.

At the bottom of every financial model there is in fact a stubborn lie—the pretense that financial markets operate in the manner of a physical process, subject to the iron laws of statistics, like atoms bouncing around in a thermodynamic equilibrium. This beguiling analogy makes it too easy for geeks like me to lose sight of a timeless truth: If atoms could talk to one another, then the laws of thermodynamics would get broken every day by clouds of stampeding gases.

Even more clearly than I remember the shortcomings of my precious models, however, I remember the reactions those mathematical confections provoked among those for whose use they were intended. On one of my first assignments, I informed—with much trepidation—the senior manager of a large financial institution that the mixed portfolio of securities and bizarre derivatives he oversaw incurred a potential risk over the coming year equal to $1,432,678,723, give or take a few pennies. I expected he might poke a few holes in the questionable reasoning behind our model. I was astonished instead to hear him reply, rather nonchalantly: "There is no risk. No way will the markets go down this year."

After many further meetings with him and his more-forthcoming colleagues, I understood that he was really just throwing down a marker. He wanted a good starting position in the impending negotiations. The less risky we deemed his assets, after all, the more of them he could buy and hence the more money he could make for the bank—and for himself. The assessment of risk that ultimately emerged from our surprisingly pliable models, in this as in many other cases, was not a "fact" resulting from a process of discovery but an agreement haggled out among various intensely interested parties (among whom one would have to include us, the consultants, since it was always in our interest to secure additional business from our clients).