Saturday, February 27, 2010
Well, while reading Scott Patterson's new book, The Quants, I came across a discussion on the Gaussian copula, a model developed by David X. Li of the Canadian Imperial Bank of Commerce (CIBC) to price CDOs (collateralized debt obligations). The copula Li used was developed by the 19th century mathematician, Carl Fredrich Gauss, who based it on the normal or bell-shaped curve. In the words of the author, "The Gaussian copula was, in hindsight, a disaster." It was a simplistic model whose assumptions belied the true prices and risks that investors faced. As the author pointed out, CDOs loaded with subprime mortgages had tranches rated at the AAA level; Li's model caused those tranches as well as others to be mispriced.
And so, WWGD? I.e., what would Gauss do? Hopefully, not assume that all distributions fit nicely into a bell-shaped curve. Hopefully, he would have recognized that highly complex instruments can't be necessarily priced by overly simplistic models.
There are may contributors to the '08-'09 market meltdown, and one might believe that this model may have been one of them. Risk remains a very difficult area of our industry, but we tend to at least assume that prices are accurate; alas, they aren't always. No matter what risk model you employ, if the prices are bad, no need to go any further.