I am trying to estimate a model using first difference GMM estimator with the xtabond2 command and I was trying to understand whether my logic was consistent.

I have a panel of 130 countries and 18 years. My model looks something like this

enter image description here

Where z and q are two interacted variables and X are the other controls.

Given the potential endogeneity of the autoregressive parameter and the z and q variables I am estimating this via Roodman (2009) xtabond2 command, but I am not sure if I am instrumenting properly my interacted variable.

the code looks something like this. Please do note that in the STATA sintax I also add year fixed effects.

xtabond2 y l.y z##q x i.Year, ///
gmm(y l.y z#q q z x, lag(2 .) collapse) iv(i.Year) noleveleq small noconstant robust

What I am not sure about is how I should instrument my interacted term. Am I correct to think that I should also instrument - gmm() - the interacted (z#q) term? the idea is that z#q might be endogenous to y, hence I think I need separate instrument for it.

Moreover when I use the # and ## operator in xtabond2 STATA issues an error message stating:

factor variables may not contain noninteger values

So I am not sure how to go about this problem: do I need to create a separate variable with the interaction and include that separately or there is another way of going about this?

I thank you in advance for your reply



migrated from stackoverflow.com Oct 10 at 13:08

This question came from our site for professional and enthusiast programmers.

Your Answer

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

Browse other questions tagged or ask your own question.