I have a panel of firm level data, and am trying to estimate firm level total factor productivity by GMM using lagged and differenced variables (value added, total asset, and labor) as instruments.
Would the following command, followed by predicting the residual, be appropriate?
gmm (lnvar - {b1}*lntar - {b2}*lnlab - {b0}), instruments(L.lntar L.lnlab L.lnvar L2.lntar L2lnlab D.lntar D.lnlab)
And in this case of lagged and differenced variables as instruments, how is it different from using xtabond?