Links 8/23/09


  1. rd

    On the "Flaw of Averages":

    I keep going back to one of my touchstones which is an excellent article by William Bernstein in 2001:

    In this, he looks at the mathematical Monte Carlo models and then compares them to historical game-changing events that occur around the world and points out that they occur on a much more frequent basis than you could ever get out of a statistical data set from several years of relatively orderly markets.

    As a result, the Monte Carlo methods are a good predictor of about 80% of the potential outcomes but are essentially irrelevant for the last 20%. After all, there is no statistical analysis from a financial market data set that will predict the outbreak of a world war with the potential invasion and destruction of your country 30-years out.

    As an engineer, we tend to use "characteristic values" instead of average values in design. These are typically anywhere from 1 to 3 standard deviations below the mean because we know that it is the weakest elements that dictate succes or failure. On top of that, we look for connections between design elements to prevent progressive collapse so that the failure of one small component does not cascade into the failure of many.

    It is approaches like this that mean that structures like the I-35 bridge in Minnesota can stand for 50 years under heavy use even when there is a fundamental design error embedded in the structure. The financial equivalent of that bridge would not have stood for 3-years the way Wall Street and US banks structure their finances.

  2. Anonymous

    This is one of the most successful blogs on NC for about a year or so. About agenda, I can clearly see LB-GS issue from the past blogs. This blog finally helped me to make up my mind on certain things. Thanks everybody. Goodbye.

Comments are closed.