A new NBER paper on why structural methods are better than IVs. They write that "different valid instruments answer different questions. The sign of the IV estimator can be different from that of the true causal effect."
But can anybody explain to me how do structural models help you get rid of the omitted variable bias? Is it just a matter of interpretation?
Subscribe to:
Post Comments (Atom)
2 comments:
1. Grande Urzua!! He is chilean and was my profesor in Universidad de Chile, now is working in frontier econometrics.
2. At this point, I trust everything Heckman says.
3. The omitted variable problem is for both, but in the impact evaluation estimations they mention, a big part of the problem is solved with diff-in-diff
Post a Comment