PPML estimation with bilateral trade agreement dummies.

[Madhuri Kumari]

Hello everyone!

I am applying PPML estimation to analyze the impact of two trade agreements (both between two countries only, India-Japan and India-South Korea) on trade. My dependent variable is export values and independent variables are usual gravity variables and trade agreement dummies. For this, I am using a large dataset of 192 countries as reporters and partners as well. I have included India-Japan and India-South Korea trade dummies as independent variables along with exporter and importer diversion dummies. So, there are 6 dummy variables in total. I am using full set of fixed effects- Exporter-time, Importer-time and Exporter-Importer FEs. The results seem to be ok with Exporter-time and Importer-time FEs but something is definitely going wrong after including Exporter-Importer FE. Almost results turn insignificant.
Is there something wrong with my model or dataset? I am unable to figure out this issue.

[Joao Santos Silva]

Dear Madhuri Kumari,

I do not think this is strange at all; the pair fixed effects will be highly collinear with your variables of interest unless they have many observations equal to 1 and also many observations equal to 0. If that is not the case, your sample is simply not rich enough to identify the effect of those trade agreements with precision.

Best wishes,

Joao

[Madhuri Kumari]

Thank you so much sir for the reply. This would help me to explain my results.

Regards,
Madhuri