Wednesday, April 13, 2005

Unorthodox data-mining

OpinionJournal has a good piece today on UChicago professor Steven D. Levitt who looks to be an econometrician who has found interesting correlations using sophisticated data mining. One of his assertions--the legalization of abortion has contributed positively to the fall in crime rate over the last 20 years.

This is a contentious issue and I don't purport to know enough to make a judgment. Here is a rebuttal (via Amazon comments) to the abortion findings.

Data-mining can be extraordinarily edifying or risky if your methodology or assumptions are off even by a little bit. Levitt is at UChicago which is known for having a fair to middling economics department and is a winner of the John Bates Clark medal for best economist under 40 so his methodology is probably quite good. Then again Larry Summers was a winner but so was the execrable Paul Krugman.

Here is Levitt's homepage on which some of his papers are available. Freakonomics, in any case, looks to be a worthwhile read.

No comments: