I have unbalanced Panel data set of 20 subject/group/Cross section ; 11 temporal or time period (annually) i.e. N=20; T=11. Since T is finite based on literature System GMM is more consistent and efficient.

I have a 3 dependent variable taking at a future time period i.e. Yit+1 Yit+3 and Yit+5; Two Interaction Variable i.e. Dummy*Xit^2 and Dummy *Xit; One Main Variable i.e. Xit and 7 Control variable X2-X8; Plus a time dummy and country dummy variable.

My issues is that

1. i dont know the right number of lag for the level variable to include in the difference equation as instruments and the number of Lag for difference variable to includes in the LEVEL Eqaution considering the fact that i intend to use the dummy time and cross section variable (Overidentification and Instrument proliferation)

2. Should i use all the right hand variable as instrument or what do you suggest