xtabond2 command / equation () suboption)

[Zeinab Ghamloush]

Dear all,
I am using the xtabond2 command on stata to deal with the endogeneity problem in my model. After reading Roodman (2009) and Kripfganz, S. (2019), I still have some questions regarding the suboptions. Can you help me please?
I have N=133 firms and T=9 years. I want to examine the relationship between the debt maturity and the presence of women in the board of directors. My dependent variable is the debt maturity. My independent variables are the proprtion of women in the board, a set of control variables, time dummies and sector dummies. I am not sure that my control variables are strictly exogenous.
I am trying to run the xtabond2 command. I read that we should put first the dependent variable, the lagged dependent variable, then the list of the independent variables with time and firm dummies. Then after the comma, we identify the endogenous and predetermined instruments under the gmmstyle() option, after that, we identifiy the strictly exogenous instruments under the ivstyle () option.

For instance:

xtabond2 DM L.DM Prop_FD LEV AM MTB F_SIZE F_SIZEsq Volatility Z Abearnings Z B_SIZE B_IND DUAL BIG4 yr* i.sector, gmm (L.DM LEV AM MTB F_SIZE F_SIZEsq Volatility Z Abearnings Z B_SIZE B_IND DUAL BIG4 Profitability , lag (2 4) equation(??) collapse) iv(yr* i.sector, equation(??) ) twostep robust orthogonal small

My problem is that i am still confused regarding the suboptions. I know that the equation () suboption specifies which equation should use the instruments (i.e only differenced equation (eq(diff)), only level equation (eq(level)) or both (eq(both))). My questions are:
1) How to know which equation suboption should i choose? Shoud i put all the gmm and iv instruments once with eq (diff) and other with eq(level)?
2) If not, how to know which instruments should be put while using equation(diff) and which ones should be put while using equation (level)?
3) And regarding the number of lags, does it differ from differenced equation to level equation?

Sorry for my naive questions but this is my first time using system GMM and i am confused about the subptions.

Your answer will be highly apreciated.
Thanks in advance,
Zeinab

[Zeinab Ghamloush]

Any answers would be really appreciated! Thank you!

[Zeinab Ghamloush]

Sebastian Kripfganz if you can help this would be very appreciated! Many Thanks!

[Sebastian Kripfganz]

At this general level, I cannot give you a better answer than what you can find in the literature you have referenced. For a difference GMM estimator, you would choose eq(diff). That's your starting point. For a system GMM estimator, you would then add further instruments for eq(level) to that initial specification. I am afraid, applying this estimator requires understanding its very nature, which I cannot outline here in a few sentences, and therefore I need to refer back to those references. The following video might help: https://www.youtube.com/watch?v=5EnPiBMUYE4