Monday, November 30, 2009

Confirmation bias

Confirmation bias is the tendency to unjustifiably assign greater weight to evidence that confirms your existing beliefs than to evidence that contradicts your existing beliefs. A particularly blatant form occurs when weight is given to confirming evidence that is in fact false. An example being the hoax which claimed states which voted for Bush in 2000 (or 2004) ranked low in average IQ which took in some liberals predisposed to believe Bush voters were stupid. Taleb in his book "The Black Swan" similarly defines (p. 308) confirmation error:

You look for instances that confirm your beliefs, your construction (or model) -- and find them.

Taleb claims this error causes people to underweight the possibility of unforeseen events, Black Swans. This is plausible, still it is a bit amusing to see Taleb apparently make the same error when he claims (p. 221):

Consider the following sobering statistic. Of the five hundred largest U.S. companies in 1957, only seventy-four were still part of that select group, the Standard and Poor's 500, forty years later. Only a few had disappeared in mergers; the rest either shrank or went bust.

The 74 out of 500 claim apparently comes from a book, Creative Destruction, by Foster and Kaplan. This work was criticized by Siegel and Schwartz who report 94 survivors as of year end 2003. And Standard and Poors reported 86 survivors after 50 years. So the 74 number appears to be low. Perhaps more seriously the claim that only a few had disappeared in mergers is wrong. Mergers are the most common reason for removal from the index and another reason (other than shrinking or going bust) is being taken private. Companies do go bust but it is not nearly as common as Taleb is implying.

No comments:

Post a Comment