help with xtendothresdpd, GMM dynamic panel threshold with endogenous regressors

[Victoria Kim]

I am using the code, xtendothresdpd, in Stata that performs estimations of a dynamic panel data with threshold effects and with endogenous regressors. From my understanding, with GMM, usually a balanced panel is used and maybe even strongly preferred. However, this code works with an unbalanced panel. I have several questions regarding this:

• How exactly does this code work with an unbalanced panel? How was the code created so that it works with an unbalanced panel?
• What is the reasoning/justification for allowing the GMM dynamic panel with threshold effects model to work with an unbalanced panel?
• What are the benefits of using an unbalanced panel as opposed to a balanced panel?

I would appreciate all the help I can get on this. Thank you!

[Sebastian Kripfganz]

I am not familiar with the specific command you are asking about, but I would just turn the questions around:
Why do you think a balanced panel is strongly preferred?

Stata (or Mata) has tools that support programming commands for panel data. Unless there are dependency structures across groups that need to be taken into account, this is fairly straightforward.

Allowing unbalanced data has the benefit that you do not need to discard data to make a panel balanced. More data usually means more efficient estimates.