xtabond2 system-GMM lag instrument query

[Isaac vk]

Dear All,

I am conducting an analysis on a dynamic panel dataset using system-GMM.

I have estimated the model, as below, which gives me good results, however, I have included two instruments for log_lagGDPPC: log_lagGDPPC & L.log_lagGDPPC.
I am a bit unsure as to whether including these two instruments is valid, and would greatly appreciate it if anyone could help me understand why using these two instruments together provides significant results yet if either is removed the significance might fall away so dramatically. Is there a way to combine the two lags in one variable?

Equation:

xtabond2 logGDPPC log_lagGDPPC CELLSUB CAPFORMp GOVCONp POPG tm*, gmm(log_lagGDPPC L.log_lagGDPPC L2.CELLSUB L2.CAPFORMp, collapse) iv(GOVCONp POPG tm*, equation(level)) nodiffsargan twostep robust orthogonal


Many thanks in advance.

[Sebastian Kripfganz]

Statistical significance does not say anything about the validity of the instruments. You would not normally include L.log_lagGDPPC in addition to log_lagGDPPC in the gmm() option. For the transformed model, lags of the instruments are added automatically. For the level model, it is unusal to include additional lags. However, this should not normally have a big effect on the results.

More on GMM estimation of dynamic panel models:

• Kripfganz, S. (2019). Generalized method of moments estimation of linear dynamic panel data models. Proceedings of the 2019 London Stata Conference.