Problems with a Two step System GMM model

[zaineb belkhiria]

Hello everyone.
I am using a two step system GMM model. You will find attached the code I used. I applied the durbin-wu hausman test and I detected my endogenous variables. I put gmm( endogenous variables) and iv( exogenous variables).
code:
xtabond2 MATURITY l.MATURITY LEVERAGE GD BI BoardSize DUALITY FIRMSIZE ROA TANGIBLEASSETS GROWTHOPPORTNITIES LIQUIDITYRATIO if REGION==0 ,gmm(l.MATURITY BI BoardSize FIRMSIZE TANGIBLEASSETS , lag(2 11) collapse) iv (LEVERAGE GD DUALITY ROA GROWTHOPPORTNITIES LIQUIDITYRATIO) robust twostep

Is the applied code correct ?
is the delayed dependent variable must be significant ?
when I applied the code on several lag(a b), I found that among my variables only two are significant. AR(2) can only be verified by 1%.
I don't know what the problem is and how I can solve it?
Please do not hesitate to give me your ideas and your suggestions.
Best regards.

[Sebastian Kripfganz]

Note that the iv() instruments are not just assumed to be exogenous with respect to the idiosyncratic error component but are also assumed to be uncorrelated with the unobserved firm-specific effects. This is a strong (random-effects style) assumption, which is often hard to justify. With xtabond2, I always recommend to explicitly specify the eq() suboption to avoid unintended consequences from the default settings.

As an alternative, my own xtdpdgmm package comes in its latest version with the xtdpdgmmfe command, which allows you to simply specify exogenous and endogenous variables, and the command will choose the right instruments for you; see https://www.statalist.org/forums/for...84#post1675484