Collinearity in Gravity Equations

[Adam Potocny]

Hello,

I am trying to estimate a GE, but I am running into an issue I can't wrap my head around. I am using importer and exporter time-varying FEs (to control for GDP, multilateral resistance, ...), and country pair time-invarying FEs (to control for distance, shared language, ...).

The problem is that when I generate RTA dummies (for my RTA of interest), the importer and exporter time-varying FEs perfectly explain two of the RTA dummies (RTA_importer and RTA_exporter, which measure whether an importer/exporter is part of the RTA (so only after its creation year)), and collinearity makes them drop from the ppml estimation. I however do need therse coefficient for interpretation. How can I solve this? I am using the ppmlhdfe package.

Thank you!

[Joao Santos Silva]

Dear Adam Potocny,

The RTA dummies will always be collinear with the fixed effects. So, either you use just a RTA dummy for the pair, or will have to drop the pair fixed effects.

Best wishes,

Joao

[Adam Potocny]

Dear Joao Santos Silva,

The intent of using three RTA dummies was to identify Vinerian trade effect. However, I am guessing that keeping the fixed effects is more important. Thank you!

Kind regards,

Adam P.