Difference GMM (xtabond) and System GMM(xtabond2) results variations Panel data

[Sehri Far]

Hi,
I have a sample of N=30 , T=36

Using fixed effect and difference GMM I'm getting the same results. These findings also support the theory.



The problem I have with Difference GMM (xtabond) is the large number of instruments.

Using xtabond2, I'm getting totally different results. I have tried different lags but nothing changes.

In the case of system GMM (Xtabond2), the p values for hansen j and hansen c tests are also close to 1.000.


Now I have two questions.

i) How can i reduce the number of instruments using xtabond (difference GMM).

ii) Why i have this much variation in results using Xtabond2 system GMM, is there any issue with the model ? How can i fix this variation in results?


I'll appreciate suggestions.

Thanks

[Sebastian Kripfganz]

You probably still have too many instruments in your system GMM estimation. I present some strategies for the reduction of the number of instruments in the following presentation:

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