Tuesday, April 21, 2009

Tales from the Meltdown 16: "Ten out of Three is a Pretty Good Record, Actually"

Daniel Kahneman, one of the godfathers of behavioral finance, is interviewed by Guy Rolnik of Haaretz on the fallibility of models, intuition and experts in the face of our own psychology:

"Psychology today differentiates between two methods of thinking: There is the intuitive method, and there is the rational one. The intuitive method is characterized by rapid learning, and it concludes very quickly that what has happened the last three times will happen forever, again and again."

Why is it that we believe that if it has happened three times, it will happen again?

"I once told a story about this: We once traveled from New York to Boston on a Sunday night, and we saw a car on fire on the side of the road. A week later, again on a Sunday night, we were traveling and again saw a car on fire in the same place. The fact is, we were less surprised the second time than the first because we had learned a rule: Cars burn at this spot.

"We find this everywhere - the speed at which people create rules, norms and expectations, even when they know it's ridiculous. This is the intuitive method at work. It remains true that whenever I travel, I always look for burning cars at that spot."

Over the last 40 years you have demonstrated that we never employ the statistical method of thinking, just the heuristic one - rules of thumb. Some economists today say the models were indeed based on statistics, but the problem is that they were based on statistics according to which the market goes up, rather than on long-term statistics.

"When it comes to finance, people link risk to volatility, but in reality, there is no connection between the two at all. There is a connection, but not when we're talking about huge risks. Greenspan and others believed that the global system - by virtue of its being global - was ipso facto more stable. Then it turned out that while it might have been more stable, it was also more extreme.

"In the last half year, the models simply didn't work. So the question arises: Why do people use models? I liken what is happening now to a system that forecasts the weather, and does so very well. People know when to take an umbrella when they leave the house, or when it will snow. Except what? The system can't predict hurricanes. Do we use the system anyway, or throw it out? It turns out they'll use it."

Okay, so they use it. But why don't they buy hurricane insurance?

"The question is, how much will the hurricane insurance cost? Since you can't predict these events, you would have to take out insurance against many things. If they had listened to all the warnings and tried to prevent these things, the economy would look a lot different than it does now. So an interesting question arises: After this crisis, will we arrive at something like that? It's hard for me to believe."

The financial world's models are built on the assumption that investors are rational. You have shown that not only are they not rational, they even deviate from what is rational or statistical, in predictable, systematic ways. Can we say that whoever recognized and accepted these deviations could have seen this crisis coming?

"It was possible to foresee, and some people did. There were quite a few smart people with a lot of experience who said bubbles are being created and they have to be allowed to burst by themselves. But it turns out that this bubble did not have to be allowed to burst by itself. I have a colleague at Princeton who says there were exactly five people who foresaw this crisis, and this does not include [Fed Chairman] Ben Bernanke. One of them is Prof. Robert Shiller, who also predicted the previous bubble. The problem is there were other economists who predicted this crisis, like Nouriel Roubini, but he also predicted some crises that never came to be."

He was one of those who predicted 10 crises out of three.

"Ten out of three is a pretty good record, relatively. But I conclude from the fact that only five people predicted the current crisis that it was impossible to predict it. In hindsight, it all seems obvious: Everyone seemed to be blind, only these five appeared to be smart. But there were a lot of smart people who looked at the situation and knew all the facts, and they did not predict the crisis."

(via Mark Thoma)

No comments: