xtabond2 gmmstyle and ivstyle

[Yasutake Homma]

Hello statalist,

Hope you do well.
I am examining gender diversity and firm outcomes like tobinq by a dynamic panel system gmm.
So I would like to use xtabond2.

For example, Yi,t = bYi,t-1+bXi,t+bXi,t++X'b+i.year+ei,t

Yi,t=TobinQt
X1,t = women on board ratio1,t
X2,t= firm age
X'= other control variables

Sample stata code:
xtabond2 tobinq l.tobinq WOM firm_age i.Year, gmm(l.tobinq_w, lag(1 1) collapse) iv(firm_age i.Year) nodiffsargan twostep robust

My question is in gmm () and iv().
I input "l.tobinq" in gmm() and did "firm_age" and "i.Year" in iv(). However, I wonder which other control variables to go.
If I think the endogeniety, l.tobinq covers lagged all control variables so other variables go to iv() or in gmm() regardless of thinking the correlation of l.tobinq and other control variables?

My another question is about lag2.
If I want to add lag2 of Y,

xtabond2 tobinq l.tobinq l2.tobinq_w WOM firm_age i.Year, gmm(l.tobinq_w, lag(1 2) collapse) iv(firm_age i.Year) nodiffsargan twostep robust

I add l.tobinq like this code and allow it the lag2 by "lag(1 2)". May I ask you to check whether this is correct or not?

I highly appreciate if anybody could check the above.
Any comments and answers are appreciated.

Best regards;

[Sebastian Kripfganz]

I would always recommend to explicitly specify the eq() suboption to iv() (and gmm()). Otherwise, the command might not actually do what you want it to do. In general, be careful when you rely on default settings.

You do not seem to have specified enough instruments. WOM does not appear in your list of instruments. Right now, the model is underidentified, which should be apparent by the fact that some coefficient is omitted in the regression output.

With a second lag of the dependent variable, you can indeed simply use higher-order lags as instruments, as you have done.

More on dynamic panel data GMM estimation:

• https://www.statalist.org/forums/for...ent-conditions
• Kripfganz, S. (2019). Generalized method of moments estimation of linear dynamic panel data models. Proceedings of the 2019 London Stata Conference.