Return to Revisions

I think the use of a nonlinear estimation process (probit) in order to produce the first stage of your IV is not appropriate. IVs are an instance of a broader class control function approachs. For example, as with typical two-stage IV approaches, your SE will be off unless properly adjusted.