Panel data GMM - Number of groups falls

[Karun Advani]

Dear All,

This may be a trivial question, but I am unsure why this happens. I have an unbalanced panel dataset with 112 groups (countries) from the period 1984-2016. When running a two-step gmm procedure using xtabond2 with the command:

xtabond2 grlhcountglob grlminc, twostep small gmm(L.lminc) robust

The number of groups falls from 112 to 54. Both variables used in the estimation are growth variables calculated by using log-differences. I was wondering why does the number of groups fall when running the two-step GMM procedure. Also, is there a way to find out the groups that have been excluded from the estimation?

Thank you in advance.

[Sebastian Kripfganz]

list if e(sample)

lists all observations after xtabond2 that have been used in the estimation. You can, of course, also replace list by summarize or any other command that might be of help.

[Karun Advani]

Thank you very much for your help Sebastian.