Time varying importer fixed effects

[Christine Everaerts]

Hi all,

I was wondering whether someone could help me how to start with an OLS fixed regression with time varying importer fixed effects?
I currently have this code, but not sure whether it's right:

egen country_pair = group(Destination Origin)
xtreg ln_trade ln_dist ln_pop ln_gdp ln_tariff ln_Tech ln_NonTech rta comlang_off comrelig contig i.country_pair, fe

Thank you in advance for helping me out!

[Clyde Schechter]

I assume by "importer" you are referring to the variable that in your data is called Destination.

It is mathematically impossible in a model with country_pair fixed effects to estimate Destination effects. This is because even if these effects vary over time, they do not vary within country_pair. As you have included country_pair fixed effects, the Destination effects are colinear with those, so Destination effects cannot be separately estimated.

So the next question is, what are you actually interested in doing? If your intent is to simply adjust ("control", though you never actually control anything in observational data) for the effects of Destination on outcome, there is no need to do anything: because the Destination is a fixed attribute of the country pair, the country_pair fixed effect automatically adjusts that without requiring you to do anything more. If, however, you really need to estimate the separate effects of each Destination to answer your research question(s), then you cannot include country_pair fixed effects in your model. You might consider instead doing something like:

reg ln_trade ln_dist ln_pop ln_gdp ln_tariff ln_Tech ln_NonTech rta comlang_off comrelig contig i.Origin i.Destination

if that makes sense in terms of your research goals.