My question is why N>T is required when working with dynamic panel estimations based on system GMM, such as xtdpdsys at stata. Is that based on the potential lost of information due to orthogonal deviations?


  • $\begingroup$ Perhaps some parameters are allowed to vary by time period. $\endgroup$ Jan 27 at 16:35

Asymptotics are done as N goes to infinity and not T. If T is larger than N, this assumption is odd. True for most of what is commonly called panel data estimators.

  • $\begingroup$ Your point is good (+1), but I am not sure that this is the true reason (why several software do not compute it numerically). I have the feeling that it is related to the rank of the weighting matrix $W$. $\endgroup$
    – Bertrand
    Jan 29 at 12:45
  • $\begingroup$ This looks related but I'm not sure if its helpful. papers.ssrn.com/sol3/papers.cfm?abstract_id=1412035 $\endgroup$
    – mark leeds
    Jan 30 at 17:25

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.