Questions regarding the gravity model, control functions, and instrumental variables

[Karen Jyo]

Hello everyone,

I have some questions regarding the gravity model, control functions, and instrumental variables. In the context of the gravity model, PPML has become the most commonly used estimator.
However, when it comes to instrumental variables, existing commands like ivpoisson and ivppml do not seem to accommodate fixed effects, and most of the literature still relies on 2SLS for instrumental variable estimation.

With the publication of Chen and Roth (2024), however, using a log(y+1) transformation of trade value followed by 2SLS for addressing endogeneity no longer seems like a good choice.
I am now considering using PPML with a control function approach to handle endogeneity, but I am uncertain whether this is feasible.

I currently face two questions:

1. Can PPML be used in the first stage of the control function?
2. If so, should the first-stage PPML estimation be performed using Bootstrap?

I have seen discussions on similar topics before but have not found a definitive answer. The example codes provided also seem to follow the xtreg + xtpoisson .

Could anyone help clarify these questions? Thank you!

[Joao Santos Silva]

Dear Karen Jyo,

If you are using cross-sectional data, I suggest you have a look at the user-written command ivgravity.

Best wishes,

Joao

[Karen Jyo]

Dear Karen Jyo,

If you are using cross-sectional data, I suggest you have a look at the user-written command ivgravity.

Best wishes,

Joao

Thank you Professor Joao.
I use panel data from 100 countries, 100 HS 6 products, and 20 years.
And I will check the ivgravity. This is a new command for me.
But I still want to know if I can use PPML in the first and second stages of the control function approach cuz I am curious and learning more about econometrics.