Crash: On the Likelihood of Outlier Events

A fascinating article on "Risk Management" in The New York Times has got me thinking.

The article traces the origins of a standard risk management models in the financial sector, known as Value-at-Risk (VaR), and why they failed to predict possibilities that caused the current economic crisis, such as bank under-capitalization and massive derivative losses due to sales of "insurance" on mortgage-backed securities (also known as credit-default swaps).

Would you fly on an airline that had a 99% track record of the flight arriving safely? What if only 1% of their flights crashed? The airline would be out of business. After all, 1% is just one out of a hundred, which, in the grand scheme of things, is not that rare at all, especially when there are millions of potential opportunities to hit that 1%.

So, what is an acceptable risk when we book a seat on an US carrier? Apparently it is somewhere around .000223% (or roughly 2 out of one million).

What seems apparent is that psychology has everything to do with how we perceive risk. As the author of the Times article sums it up: "People tend not to be able to anticipate a future they have never personally experienced."

People who live on the Gulf coast have experienced a lot of hurricanes. So, mostly, they likely feel like they know how to handle them. Until an outlier like Katrina arrives.

Millions use bridges or rely on levees every day to live in and move about their home region without much thought to their failure, despite decades of warnings about aging US infrastructure from the Army Corps of Engineers, until the Minnesota bridge collapsed and the New Orleans levees broke.

To a traveler who only takes two flights each year, a .000223% chance of crashing seems very likely to not occur, although every time the flight hits turbulence that traveler is probably thinking about the possibility of crashing. Even more so, the airline employees who schedule hundreds of thousands (or even millions) of flights each year know that an accident is a virtual inevitability-- though even they weren't fully prepared for an outlier like 9/11/2001.

Any mother who is told there is 1% chance her unborn child will have a serious genetic disease probably thinks that is way too much risk. A doctor who sees thousands of patients a year knows she will eventually have a patient who suffers from the rare disorder and is going to take a 1% risk seriously.

But apparently a 1% or so chance of a series of events that at some time in the future could cause serious economic carnage was routinely ignored as the financial industry made billions for years, assuming that housing prices only move one way: up. Granted, there would have been a high price for any corporate officer who warned of the possibility of huge losses in credit-default swaps, a concept hedge fund manager Jeremy Grantham labels "career risk":

It’s what I call the Goldman Sachs Effect: Goldman increased its leverage and its profit margins shot into the stratosphere. Eager to keep up, other banks, with less talent and energy than Goldman, copied them with ultimately disastrous consequences. And woe betide the CEO who missed the game and looked like an old fuddy-duddy. The Board would simply kick him out, in the name of protecting the stockholders’ future profits.

Everybody else was doing it. And the VaR models continually affirmed little risk. In fact, significantly less risk than most other assets. Since, as the Times article lays out, most of the models only measured very short term risk and the long term models usually only drew on data from the previous two years, the more housing prices rose, the less likely the risk seemed that they would fall. In the end, the VaR models proved fairly useless in predicting the systemic collapse that occurred before it was too late.

Thus one insurance company, which is in the business that mastered the actuary table, managed to sell over $400 billion worth of un-hedged credit-default swaps (plus g-d knows what else) and has cost the US taxpayer over $150 billion. So far.

Let's not mention all the other financial institutions that have to date cost over $300 billion. Or the fact that none of the financial sectors' shareholders, employees, and managers will have to pay back the billions they pocketed over the past five years.