Bayesian methods in applied econometrics

Christopher Simms, of recent Nobel fame, has a presentation [pdf] up on his website from 2007 that is a pretty good introduction to the difference between frequentist and Bayesian approaches to economics, and ultimately why a Bayesian approach is most prudent for economists.

This from his introduction sets up the basic premises of his presentation -

Introduction

  • The first part of the talk makes some unoriginal claims about the role of Bayesian thinking.
  • Despite being unoriginal, and obvious to me and to a minority of econometricians, they are unfamiliar, thought-provoking, or outrageous to quite a few econometricians.
  • A summer seminar talk seems a good place to restate claims that are both obvious and outrageous and that therefore (for one of these reasons or the other) are usually excluded from journal articles.
  • The latter part of the talk discusses some areas of econometric application where frequentist asymptotics seems particularly persistent and suggests how Bayesian approaches might become more practical and prevalent.
Although I'm not entirely comfortable with Bayesian statistics, this was an interesting presentation to read none the less. I plan on referring back to this in the future when I come across some decisions about methodology for research projects.