Small N, Small T Panel for GMM-Sys

[Carolina Coletta]

Hi,

I have a dynamic panel with N = 22 (companies) and T = 15 (years) (the 22 companies are the population)

I understand that GMM requires a "small T, large N" panel, but I don't know if my panel is appropriate.

Is N = 22 to small for a GMM estimation, even considering that it's the population?

Thank you.

[Sebastian Kripfganz]

You cannot expect too much from such a small data set. You might have to make strong assumptions about the exogeneity of your variables (meaning that you cannot really allow for many predetermined or endogenous variables), in particular if you want to include more than just a few independent variables. The simpler your model specification, the better. Of course, you will then have to be cautious about the interpretation of your results.

You might want to ask yourself, if a dynamic model is really necessary. Maybe a simple static fixed-effects model can do a good job.

[Carolina Coletta]

Thank you!