Note to self to think harder about instrumental variables when I have time, if not Medicaid
July 25, 2011 § Leave a comment
As proof, instrumental variable approaches are basically useless, because even attacks from people who don’t really get that correlation isn’t causation are difficult to dismiss.
I’m willing to say IV estimates, when assumptions are close to met, are better than naive regressions, but how much better? I need to look at whatever work has been done on the robustness of IV estimates.
(edit: I don’t think the Tragedy of the Commons is the biggest issue, but it also deserves further consideration)