First difference or forward demeaning: Implications for the method of moments estimators

In this paper, we consider the method of moment estimation for dynamic panel models based on either forward demeaning (FOD) or first difference (FD) transformations to eliminate the individual-specific effects, using either all lags or one lag as instruments. We show that the Arellano.Bond-type generalized method of moment (GMM) based on FD is asymptotically biased of order c. using all lags or one lag as instruments where c=TN<. as N,T... For GMM based on FOD, it is asymptotically biased of order c. when using all lags, but it is asymptotically unbiased when using only fixed number of lags as instruments. We also discuss these findings in light of the simple IV estimator. Monte Carlo simulations confirm our findings in this paper.