I’m often irked by economists’ love of applying crude statistical techniques in situations where institutions seem far more important. Don’t think I dislike statistics; the kind of hoops [Chris Lightfoot](http://www.ex-parrot.com/~chris/wwwitter/) was able to do are inspirational, almost magical. And I’m all in favour of using whatever techniqe provides the most useful result. But something feels wrong in fiddling with the data you have until you find a pattern that seems to make reasonable predictions, without even thinking about the underlying mechanisms.
Greg Mankiw’s [tips](http://gregmankiw.blogspot.com/2007/10/my-bet-on-nobel.html) for the Nobel prize in Economics trigger this worry. He predicts Eugene Fama, Robert Barro or Martin Feldstein for the prize- on the basis that they’re the most-cited economists who aren’t already Nobel laureates:
[A]s a purely predictive matter, highly cited economists usually get the prize eventually. In this old citation ranking, the top five most cited economists are all Nobelists. As of today, the prize has gone to more than half of the top 30 (and some of the others may win it in the future).
Mankiw admits this is a pretty crude measurement. It doesn’t say anything about the tendency to split Nobel prizes between economists with related work, and is probably more effective at predicting ‘people who will eventually win the prize’ than ‘people who will win the prize this year’.
But it works, so it’ll do. And we don’t need to worry about how the Nobel committee actually make their decision; it all comes down to a popularity contest.
Incidentally [Cosma Shalizi](http://www.cscs.umich.edu/~crshalizi/weblog/479.html) dipped his toes into this area a few months ago. Being Cosma, he skipped past the obvious debate, worried about the differing citation patterns in different areas, and then described a measure of journal popularity that sounds very similar to PageRank. He points to [eigenfactor](http://www.eigenfactor.org), a project which does just that – and with pretty pictures to go with it.