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
EDIT for more context:
What I am essentially try to do is measure the impact of economic sanctions on trade. Western countries have imposed sanctions on Russia and, as a retaliatory measure, the Russians did the same. Here, it is important to note that these sanctions do not involve the same productgroups or codes. For measuring purposes, I include 8 dummies. That is, I include 4 for the Western sanctions and 4 for the Russian sanctions to cover each distinct scenario.
For example:
Imposing country Sanctioned product
yes yes
yes no
no no
no yes
I have derived my data from COMTRADE at HS6 product level. In doing so, I have taken Russia as always being country i for simplicity purposes in terms of interpretation.
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
EDIT for more context:
What I am essentially try to do is measure the impact of economic sanctions on trade. Western countries have imposed sanctions on Russia and, as a retaliatory measure, the Russians did the same. Here, it is important to note that these sanctions do not involve the same productgroups or codes. For measuring purposes, I include 8 dummies. That is, I include 4 for the Western sanctions and 4 for the Russian sanctions to cover each distinct scenario.
For example:
Imposing country Sanctioned product
yes yes
yes no
no no
no yes
I have derived my data from COMTRADE at HS6 product level. In doing so, I have taken Russia as always being country i for simplicity purposes in terms of interpretation.
