[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

[sedki zn]

Dear Professor Joao Santos Silva
I hope this msg finds you well.

I am currently working with panel data that exhibits strong cross-sectional dependence, and I plan to employ a quantile regression framework. I was considering extending the Common Correlated Effects (CCE) approach—by including cross-sectional averages of the dependent and independent variables—to control for unobserved common shocks within the quantile regression setting.

I would be very grateful to know your view on whether this approach is methodologically appropriate and consistent with the spirit of CCE when applied to quantile regression.

Thank you very much for your time !

Kind regards,
Sedki

[sedki zn]

Dear Professor Joao Santos Silva
I hope this msg finds you well.

I am currently working with panel data that exhibits strong cross-sectional dependence, and I plan to employ a quantile regression framework. I was considering extending the Common Correlated Effects (CCE) approach—by including cross-sectional averages of the dependent and independent variables—to control for unobserved common shocks within the quantile regression setting.

I would be very grateful to know your view on whether this approach is methodologically appropriate and consistent with the spirit of CCE when applied to quantile regression.

Thank you very much for your time !

Kind regards,
Sedki

Continued

More, I must admit that I am somewhat surprised to observe that a growing number of papers published in reputable journals document the presence of cross-sectional dependence and non-stationarity (often followed by cointegration tests), yet subsequently proceed to estimate MM-QR models without explicitly addressing these issues within the quantile regression framework itself.

I would greatly appreciate your perspective on how this should be interpreted from a methodological standpoint.

[Joao Santos Silva]

Dear sedki zn,

I am afraid that I have no experience with CCE, so I will not comment on that. On your second point, in the paper, we do not address non-stationarity or co-integration, so the authors you mention should justify what they are doing.

Best wishes,

Joao

[sedki zn]

Dear sedki zn,

I am afraid that I have no experience with CCE, so I will not comment on that. On your second point, in the paper, we do not address non-stationarity or co-integration, so the authors you mention should justify what they are doing.

Best wishes,

Joao

Thank you very much dear !
Actually not a single paper. I have checked more than 10 highly indexed papers and all of them confirm the non-stationarity then confirm the I(1).
Once they confirm data stability (long run relationship), they justify the use of MM-QR!

Kind regards