ivpoisson gmm command

[Aboubacar SAIDOU]

Dear all, dear Prof. Joao Santos Silva


I have read the paper "ENDOGENEITY IN COUNT DATA MODELS: AN APPLICATION TO DEMAND FOR HEALTH CARE", but I have a misunderstanding when it comes to using ivpoisson in gravity models.

Indeed, in the paper, I read that this approach is suitable for count data. However, this is not the case for trade data as in gravity models.

I want to know why this method is also suitable in the context of gravity models.

Best regards.

SAIDOU

[Joao Santos Silva]

Dear Aboubacar SAIDOU,

Nothing in the method prevents it from being used with data that are not counts. However, you have to be cautious when using that estimator for gravity equations because of the possible incidental parameter problem.

Best wishes,

Joao

[Aboubacar SAIDOU]

Thank you very much for this clarification Prof

Really appreciate it.

Best wishes,

SAIDOU

[Aboubacar SAIDOU]

Dar Prof, please i have an other question

How to compute R-square and informations criterias like AIC and BIC with ivpoisson?

Best regards.

[Joao Santos Silva]

Do not compute those; they are not meaningful in this context.

[Aboubacar SAIDOU]

Thank you Prof,


I'm comparing ivpoissoin estimator with ppml and ols estimators in order to choose the best one in my context. I imagine that in order to perform it one need criterias. Is there an other criteria that i can use ?


Best regards.

[Joao Santos Silva]

Dear Aboubacar SAIDOU,

The choice should be based on the suitability of the estimator to estimate the model of interest, not on goodness-of-fit. OLS is logs is unlikely to be valid, so that i probably out. The choice between PPML and IVPoisson depends on whether you need to use instruments and fixed effects.

Best wishes,

Joao

[Aboubacar SAIDOU]

Thank you very much prof.

Best wishes,

SAIDOU