Dear all,
I have a long panel N = 20 and T = 35 and try to estimate a dynamic model, say:
y_it = a + b*y_it-1 + c*X_it + u_i + e_it
Because of lagged dependent but N < T and T > 30, is it safe to go with the conventional FE model (xtreg ... , fe robust)?
If I apply Anderson and Hsiao (1981) FD estimator, all variables must be differenced.
What if I don't want to do so by estimating the within-FE model but instrument the lagged dependent y_it-1 with y_it-2 or y_it-3? Is it correct to mitigate potential bias due to lagged dependent?
Many thanks any reply,
I have a long panel N = 20 and T = 35 and try to estimate a dynamic model, say:
y_it = a + b*y_it-1 + c*X_it + u_i + e_it
Because of lagged dependent but N < T and T > 30, is it safe to go with the conventional FE model (xtreg ... , fe robust)?
If I apply Anderson and Hsiao (1981) FD estimator, all variables must be differenced.
What if I don't want to do so by estimating the within-FE model but instrument the lagged dependent y_it-1 with y_it-2 or y_it-3? Is it correct to mitigate potential bias due to lagged dependent?
Many thanks any reply,
