Likelihood inference and the role of initial conditions for the dynamic panel data model

Lancaster (2002) proposes an estimator for the dynamic panel data model with homoskedastic errors and zero initial conditions. In this paper, we show is invariant to orthogonal transformations, but inefficient because it ignores additional information available in data. The condition trivially satisfied by subtracting observations from We that differencing out further erodes efficiency compared drawing inference conditional on first observations. Finally, compare method standard random effects approaches unobserved Standard implicitly rely normal approximations, which may not be reliable when very skewed some mass at values. For example, firms naturally depend period firm enters a new state. It seems unreasonable, then, assume process determining known or stationary. can instead make structural parameters conditioning