Dear Stata forum:

I am trying to estimate a dynamic panel in which, beyond the lagged dependent variable (Y), I have another endogenous regressor (X), but I don't want to use gmmstyle instruments for it, only ivstyle instruments. The reason is that lagged values of X will still be suspitious to be endogenous, so I prefer to rely only on external purely exogenous instruments.

Can anybody tell me if this is the correct specification to do my estimate?

xtabond2 Y L.Y X Z, gmm(L.Y, lag(2 .) ) iv(Z instrument1 instrument2), twostep robust

Many thanks

Juan