Thanks, I have used this before, but it doesn't work for panel regressions, i.e. a fixed effects regression with a sample of countries over a period of time.
It is likely the case that - cmp - (SSC; SJ) will allow you to get something like a panel-fixed effects 3SLS. - cmp - is exceedingly flexible, accommodates instrumental variable estimation, multiple equations, and panel data (though you'll have to specify the fixed effects yourself).
You might try posting your question to statalist.org as I believe that David Roodman (the creator of the package) follows statalist and at the very least you could get some good suggestions - though you may have to give more details (it's currently not clear that - cmp - would fit either - it is simply my guess that it would).
Not sure if you solved this, since it's old, but reg3 works. To have fixed-effects, you just use i.country (or i.year or i.id). And to approximate robust standard errors, you can use the bootstrap prefix.
Hi folks, on a slightly related question, is there a way to include a multilevel in reg3? One of my endogenous variables is the same for all of my individuals each time period. But, the dependent variable is related to that endogenous term. Thanks!