IAFE 2010 Annual Conference - Talk: Physics envy may be hazardous to your wealth

Physics envy may be hazardous to your wealth...YES, but how?
http://web.mit.edu/alo/www/Papers/physics8.pdf



This talk was quite interesting, delivered with quiet passion and confidence that was not obnoxious, or as one would say in Chinese 自信但不骄傲.

He proposed a thought provoking taxonomy for classifying various levels of risk and uncertainty

Taxonomy of Risk & Uncertainty:
Level 1: Complete Certainty i.e no randomness, as in physics, all parameters are deterministic;
Level 2: Risk without Uncertainty, i.e randomness but coming fully random variables whose law and parameters are completely identified;
Level 3: Fully Reducible Uncertainty;
With sufficient data, using existing statistical estimation methods, we can get to level 2.
Level 4: Partially Reducible Uncertainty
Parameters of random variables estimated as in Level 3may vary unpredictably
Level 5: Irreducible Uncertainty: Total uncertainty as to the nature and random variables driving observed data that can be improved with even more data.

Technical criticism:
At the very least,Level 2 & 3 should be merged. There is really no real life situation, except in controlled experiments, where the law of the random variable is written on the wall. You always have to estimate it. 4& 5 could be more tightly put together, so that really there are only 3 levels: No randomness, randomness with a precisely defined law and and randomness with a law that is not fully known.
Each of the last two levels creates the need for potentially redundant hedges.

(From memory 8 days later) One example he uses in the talk but that is not in the paper online and that caught my attention is the example of the Drawing of an urn containing let's say 100 balls colored black or white.

In the first example we know there are 50 white balls and 50 black balls.The question much one would pay to take part in a game where if one guesses correctly the color of the ball taken out of the urn, one receives let's say $100.

The conclusion here appears to be that one should pay $50.

In the second case we do not know how many black or red balls there are in the urn. Using iterative reasoning that boils down to what is known in information theory as a maximum entropy inference, it also appears one should pay $50.

He notes there is a psychological bias for red.

This leads me to my practical criticism of the paper

Practical Criticism:

This was an analysis of uncertainty from the viewpoint of a statistician/econometrician rather than from the viewpoint of a Risk Manager that is still prey to the mindset of a physicist in its analysis.

As I explained in my article in the Investment Professional last summer "BICs, the PPIP and the fallacies of expectations based risk management", the correct answer of what one should be willing to pay in the example he gives really is: "it all depends..."

Curiously, he refers admiringly in his online paper to Taleb's Black Swans without really addressing issues of hedging robustness that are the risk management dual correspondent of uncertainty issues.

References:

Smoke and Mirrors - BICS, the PPIP, and the Fallacies of Expectations-Based Risk Management

http://www.iafe.org/html/06182010.php

http://web.mit.edu/alo/www/Papers/physics8.pdf

http://www.newyorker.com/online/blogs/currents/2010/06/new-yorker-currents-nassim-taleb-on-risk-and-robustness.html