Using external instruments in -xtabond2-

[Mihir Sharma]

Dear Statalisters,

I am working on Stata 15 with a panel data of about 400 observations (i=20, t=20). I am estimating a dynamic panel data, Arellano Bond (1991) first-difference GMM model using -xtabond2-.

xi: xtabond2 y l.y x1 x2 x3 i.year, gmm(y x1 x2 x3, lag(2 6) collapse) iv(i.year) robust noleveleq

In this specification, x1 is endogenous. I also have two other variables z1 and z2 that I want to use as instruments for x1. How can I do this? One option is to include z1 and z2 in the ivstyle-instruments as follows.

xi: xtabond2 y l.y x1 x2 x3 i.year, gmm(y x1 x2 x3, lag(2 6) collapse) iv(i.year z1 z2) robust noleveleq

Is this correct? Is there a way to formally test the validity of my instruments? I am not using a 2SLS estimator (using -ivreg2-) because of the dynamic nature of my specification (i.e. the presence of the lagged dependent variable l.y).

Many thanks,
Mihir