Sunday, March 07, 2010

Another Book for the Pile: Scott Patterson's The Quants

One of the problems with quantitative analysis in finance is that a lot of it relies on calculus. Calculus is a beautiful thing, but it also involves small changes--and when I say small, I mean infinitesimal.

This makes for a good approximation when analysts are dealing with movements of a few basis points. It becomes a big problem, however, when the economy goes through major structural shifts. The point came home to me some years ago, when I was trying to estimate the impact of changes in the tax code on house prices. I was using regression coefficients (which are essentially first derivatives), and got changes in prices that seemed way too big. The problem is that the relationship between house prices and taxes is non-linear, so simulations involving large changes in tax policy cannot be approximated with a linearization.

1 comment:

blueberry said...

Now that's something mathematical. We should know the ins and outs thanks for your explanation.