The Limits of Econometrics, and Roots of Modern Applied Micro

I just stumbled upon this article by David Freedman, published post-mortem.  A lot of this looks familiar to me: you can find pieces of it in David Freedman's book, Statistical Models, Theory and Practice, which I highly recommend.

Perhaps I've blogged about this before, but I rather suspect that David Freedman's critical writings on use and misuse of regression analysis formed the basis of so-called "applied micro," which grew out of Princeton University, and the work of Ashenfelter, Card, Krueger, Angrist and others.  An occasional citation will clue careful readers to this connection, particularly the teaching of natural experiments, and David Freedman's canonical example:  Snow on Cholera.  Some modern references to Snow refer to Freedman; many do not.  But I'm pretty sure it was in fact Freedman who dug this seminal work out of the dustbin of history and used it to inspire invigorated new empiricism that searches for natural experiments and rigorously tests maintained assumptions in regression models of observational data.

The one thing about Freedman that can be a little frustrating is his terseness.  There's a lot he doesn't say but only implies.  Still, he says enough to lay bear the nakedness of the heroic assumptions made in much applied econometrics.  The sins remain plentiful, even among his descendants who practice applied micro today.

Nevertheless, I think empiricism today is much better than it was twenty years ago.  I think David Freedman deserves much of the credit for that.  And I think he's still worth reading and re-reading, if we want to improve honesty in applied econometrics.