Am I correct with the two step system GMM xtabond2 command

[Badiah Eljahimi]

Dear Statalists,

I am doing two-step system GMM xtabond2

the dependent variable is rgdpg

endogenous variables are rgdpg ihs_inigdppc fdxstwo

exegnouse varaibles are ihs_inf ihs_gfcf ihs_gov ihs_trd ihs_lbor

the command that I used:

xtabond2 rgdpg rgdpg_lag1 ihs_inigdppc fdxstwo ihs_inf ihs_gfcf ihs_gov ihs_trd ihs_lbor y*, gmm(rgdpg ihs_inigdppc fdxstwo , lag(2 5) collapse eq(diff)) iv(ihs_inf ihs_gfcf ihs_gov ihs_trd ihs_lbor, eq(diff)) gmm(rgdpg ihs_inigdppc fdxstwo, lag(1 .) collapse eq(level)) twostep robust

I have read Roodman (2009). However, I still confused

1- Is this command correct?
2- Can I increase the number of lags in gmm(rgdpg ihs_inigdppc fdxstwo, lag(1 .) collapse eq(level))
3- I put the time variable y* at the beginning of the command but not in gmm () eq(level) is that ok!

I will be very grateful for any help

Thank you

Badiah

[Sebastian Kripfganz]

For the level model, usually just a single lag is used, i.e. lag(1 1) instead of lag(1 .), because the additional lags are redundant (if all possible lags for the first-differenced model were used). Using further lags is not wrong per se, but it is difficult to justify and might look as if you are fishing for results that best support your prior hypothesis.

Note that you must explicitly specify instruments for the time dummies, i.e. iv(y*, eq(level). It is not correct to leave them out.

The following presentation might also be useful:

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

[Badiah Eljahimi]

Sebastian Kripfganz Hi Sebastian, Thank you very much for your response. I will read the presentation slides surely it will be useful

Thank you
Badiah