You can't blame the model.

No matter what happens with the rest of this Greatest Recession Since the Depression, let's stop blaming it on runaway quantum finance.

Math is not the problem. It gives answers. It provides precision (sometimes illusory). And, translated into computer coding, it runs endlessly and tirelessly.

But somebody had to make it up in the first place. And maintain it. And manage it.

That is where all the math wizardry on Wall Street can fall apart.

Too often, says Dan Borge, the man who quite literally wrote "The Book of Risk," an assumption gets built into a model because an assumption is needed for a model to work.

A low-level technician puts it into the model. The model gets handed off to some higher up. No one pays attention. Everyone's just trying to complete the model.

The model never gets challenged. Take the example of home prices. In 2005, 2006 and 2007, a lot of asset-backed securities were created on the assumption that ... home prices would go up. If home prices were to go down, the model might even show that something really unpleasant happens.

But decision-makers, not the model, said that was not going to happen. The assumption didn't get challenged. The result? An explosion in derivatives. Then an implosion in credit ... when home prices fall, nationwide, two quarters in a row. For the first time in half a century.

"I don't think people read between the lines when it comes to technology,'' says Michael Panzner, a member of the faculty at the New York Institute of Finance.

In effect, the algorithms that run Wall Street are not put through the same kind of rigorous examination as in other industries, where, say, lives are at risk. At Boeing, calculations get checked by four or five people. "That doesn't happen here,'' said Borge, principal designer of the first enterprise risk management system: the measure known as risk-adjusted return on capital, for Bankers Trust.

What this says is that a lot of risk management today is wasted on risk managers who get caught up in creating ever-more elegant models with ever-greater capacity to model the last three years, rather than the next three.

The search, at this point, is for the Black Swan, that "highly improbable event"-like a downturn in housing prices-that changes predictions, models and reality, radically.

The problem with Black Swans is they, almost by definition, can't be predicted. They are unanticipated events.

The problem is two-fold, says Bill Sharon, chief executive of Strategic Operational Risk Management Solutions in New York. There are also blue, green and yellow swans coming along. And you can spend so much time building lists of all possible events that could possibly go wrong, you don't spend enough time focusing on what can go right. Which will get you through the calamities.

Models should help decide what businesses you should be in, what tacks to take, what prices or fees to charge and what to invest in.

But if you get preoccupied with the apparatus of risk management, you won't see what's really going on.

Risk management, says Sharon, is a human exercise. The fact-finding that goes with getting the right inputs for the model from the right people in your organization. From the exercise of judgment on the output.

It's minds, not models, that matter most. In the case of a mortgage-driven securitization market, it wouldn't take the most sophisticated mind or model to pick up on the fact that the Case-Shiller Home Price Index had fallen two quarters in a row. For the first time. That's an anomaly worth noting. And adjusting for. But which only Goldman Sachs among big investment and proprietary trading houses seems to have noted.

The rest of the world should have eyes in the field looking for such anomalies, all the time, says Borge. And bringing these potentially reality-shaking facts to the attention of the top decision-makers in any risk-taking firm.

Now comes commercial real estate. This "seems to be the accident waiting to happen," says Panzner. This, he says, is a tsunami about to play out over the next two or three years, in a bit of a replay of the past two. "I don't think there's any way around that issue. Zero," he says.

This time, "we can see it coming,'' says Sharon. "But I don't think anyone knows what to do."

One thing's clear, he says, though. The answer is not to rely on an algorithmic tool "as if that's the answer."

The answer is to think about what can go wrong.

That is not getting put in a model.

This article can also be found at

Register or login for access to this item and much more

All Information Management content is archived after seven days.

Community members receive:
  • All recent and archived articles
  • Conference offers and updates
  • A full menu of enewsletter options
  • Web seminars, white papers, ebooks

Don't have an account? Register for Free Unlimited Access