Wednesday, May 24, 2006

The short version of Beta...

...being the worst way to measure "risk".

Beta is the standard deviation of returns. Standard deviation requires a distribution to be meaningful. A normal distribution is used. Capital market probability distributions are no-where near "normal" (i.e., mean, mode, median are equal). The crash of 87 was more than 13 standard deviations away from the mean of a normal distribution. The probability of a NINE standard deviation move is .0000000000000000000001049. A 13 standard deviation is three times that size, and I don't want to type in any more zeros. Add to that the fact that volatility is volatile and you have to be VERY careful about making any conclusion or decision based on "beta".

Speaking of volatility, there are academic bubbles right along with market bubbles. The two often develop in parallel. The market goes up for awhile, and the risk managers all start high-fiveing each other while counting their bonuses from their superior performance. They give lectures at Universities to MBA students and give the impression that they have built an iron-clad way to shield themselves (but never their clients!) from any adverse exposure.

It will be interesting to see who are the losers here. I don't anticipate the largest players being hurt...they are too close to the information flow (Goldman Saches fired its technical analysts some time ago...why employ people who analyze price movements ex post when you know the biggest trades ex ante???). I expect some long/short and global macro hedge funds to suffer, along with all the fools who came late to the party.

No comments: