instruments in GMM

[honza]

Hi,
I am interested in dynamic panel data model and i have some problems with it. I don‘t know how the number of instruments is capped in GMM. Generally, the number of instruments produced will be quadratic in T. So in a model with one lag of the dependent variable, k strictly exogenous variables and p = T − 2 periods from which to from moment equations, there are k + p * (p + 1)/2 moment conditions. In my model there are 17 periods, 9 covariates, so 129 instruments. However, In my Eviews output there is instrument rank equals to 34 (this is number of cross section in my model). I have read that number of instruments can‘t be higher than number of cross-section. Still, I don’t know how the number of instruments is capped (I would like to know some citation or the “capped technique“, eventually).
Thanks,
Honza

[EViews Glenn]

EViews drops singular instruments, provided that identification is maintained.

[honza]

Thanks for your response,
so, can i claim that a number of instruments is capped in order to keep the number of instruments lower than the number of cross-section (because of weak Hansen test, biased estimation, etc.)? Is it ok? Is there some literature on this?
Thanks,
Honza

[EViews Glenn]

No. It's because the instruments that are dropped are redundant. If the resulting model is overidentified, one could still drop instruments in accordance with the weak instruments literature.

An analogue would be if you specified X and X*3 as separate instruments. We would drop the X*3 instrument as it is redundant.

[honza]

Thanks, i understand.
So, can i at least claim that redundancy of some instruments isn't so bad, because it (in eviews) capped some instruments?
(Because (for instance) the J test has problem with too many instruments and it is greatly weakened by instrument proliferation.)
Anyway, once again, thanks a lot and excuse my bad English. ;)

[EViews Glenn]

It's neither bad nor good. It's just a case of redundancy.