Ao, thanks for the progress report.
That's disappointing to hear that the test isn't working like we'd hope. I think the most efficient thing to do would be to write up a summary of exactly what you've done with the 1996 tests first; depending on what's going on there, we might change course in a way that would render making the 2015 version of this test unnecessary. I'm not sure that you and I have the exact same thing in mind for the 1996 test, so I want to double check the exact simulations and specifications you ran. In particular, notice that in my write up for the binary choice estimator, I had shut down "direct" utility. That test assumes utility takes the form u(c-r), not f(c) + u(c-r). Does the test work better when that is imposed?
I should mention that it's totally fine for you to prioritize finishing up your finals for now. But when you get back to this, i think it would be best to start by writing up a summary of the 1996 results, and then for us to talk.
Thanks!