Nassim Taleb lecture

So down to London just for the Taleb lecture this evening at LSE. As I went past the New Academic Building just after 6:00, there was already a queue forming so I abandoned plans for food and joined up with the rest. The main lecture room was soon filled and an overflow room was arranged.

The lecture was originally going to have been aimed at various philosophical issues related to probability, but it was apparent that credit-crunch related events had rather swamped this plan. So we got 20 minutes on probability and the remainder of the lecture on more current events. Taleb's prediction that Fannie Mae and Freddie Mac had business plans susceptible to a "black swan" has clearly made him even more of a "sage" than he was before!

Taleb's main probability theme relates to social sciences and the impossibility of knowing the probability distribution that you are facing. This is especially important where the "impact" of the event is critical. He is referring to these as "fourth quadrant", his matrix being thin vs fat tails on one side, and simple vs complex payoffs on the other. The fourth quadrant is fat-tailed / complex payoff. The "real" fat tail remains invisible until it hits with a huge impact. Taleb had some stats related to higher moments of financial market distributions aimed at showing that very few events account for the majority of the fat tail.

His favortite model remains the turkeys being fed in the run up to Thanksgiving (or Christmas I suppose, depending on his audience). Just before the big event hits, turkey epistemology would have found it impossible not to conclude that the hypothesis "the farmer takes care of us" was true (via some form of probability and inductivism)

And once something is non-Gaussian, you don't know how it is non-Gaussian.

So with fat tails, there is no central limit theorem, no method for calculating variance, no applicable stress-testing as the past doesn't predict the future, no linear regression, all data is non-predictive, and this applies to all economic data. Things quickly look Gaussian in the middle but take an infinite time to reveal their non-Gaussian tails

And 99%+ of all social science is in the fourth quadrant - so economics has precisely zero precision in prediction yet "kills" by pretending that it does - "stop the doctors from killing more patients". Accept the humility of what we cannot know. That it is the best we have is no excuse.

He was especially angry at Bernache, of the US Federal Reserve, and especially his "new era of moderation" speech from a couple of years ago. But Taleb himself has no solutions to the current problems - "I can prevent the patient getting ill, but cannot cure the illness".

In answer to a question from an A-level student about studying economics, he replied "don't"! He briefly discussed his personal options strategies which turned out to be sell at the money, but long tails - something that in a past life I just didn't have the discipline to do.

And a final word - econometrics is a fraud. It has not delivered. GARCH is no better than simple and simple is no good. It is like trying to predict lottery winners by examining what clothes they were wearing when they bought their tickets!

I am now the owner of signed copies of both "Fooled by Randomness" and "Black Swan". An excellent evening