[Joao Santos Silva]

Dear Francesco Rossi,

In that context, units are unlikely to be independent and therefore clustering is indeed a poor solution. I guess you will have to code that (and maybe make it available in SSC?).

Best wishes,

Joao

[Francesco Rossi]

Dear Joao Santos Silva,

I'll give it a try, and if I succeed, I'll make it available on SSC. However, I'm not very comfortable writing complicated programs in Stata on my own—I prefer R.

How much effort do you think it would take to implement your estimator in R?

Thanks again for your suggestions!

Best,
Francesco

[Joao Santos Silva]

Dear Francesco Rossi,

The estimator is very easy to implement, so it would be easy to do in R. If you do it and make it available, please let me know and I'll put a link in my page.

Best wishes,

Joao

[Muritala Ogunsiji]

Dear Joao Santos Silva,

I must confess that I have benefited a lot from your posts, which have motivated me to join the forum. Thanks so much.

Please, is the xtqreg appropriate for a panel model with an endogenous covariate and an interaction term where the term is interacted with the regressor of interest? Your paper is only explicit about the estimator's appropriateness for cross-sectional data with endogenous covariates. If the estimator is inappropriate in this case, could you please recommend an appropriate one? I have unbalanced panel data for 82 countries over 24 years.

[Joao Santos Silva]

Dear Muritala Ogunsiji,

I am glad you have found my posts useful. As for your question, that command does not deal with endogeneity, so it won't work. Because you have T=24, you can to estimate using ivqreg2, also described in the paper, and include the fixed effects as dummies. This is not ideal, but maybe you can try and see if you get sensible results?

Best wishes,

Joao

[Muritala Ogunsiji]

Dear Joao Santos Silva

Thanks for your swift response. I will explore the suggestion.

Regards,
Muritala