Hi,

For my Master's thesis I would like to estimate a gravity model using three specifications:
Bil. trade = ...
Exp = ...
Imp = ...

What I am currently struggling with is how to structure my data. The main question I have is if I would have to structure my data at the product level (HS6) so that it looks as follows:

Country pair | Product code | Exports | Imports | Bil. Trade|

This would, inherently, mean that I will have a great deal of zero's in my dataset right? From what I've read, I believe the best way to overcome this issue is to use ppml.

I figure it would be easier to split exports and imports in seperate datasets perhaps. This way, I would have no issues with zero's I reckon, correct?

My last question concerns the method to be used. I've read many threads about gravity model estimation and it strikes me that the most common method used is ppml, but what fixed effects do I include?

Kind regards,

Tim van Ewijck