Hello,

Can someone please help me understand two issues I have?

I am doing a gravity model with panel data on several countries. My goal is to estimate a treatment effect that is observed for one country. The main issue I am having is properly including exporter-time and importer-time fixed effects. Xtset is pair_id year.

My questions:

(1) If I put the treatment variables : i.treat##i.time BEFORE exporter-time fixed effects, I can estimate treatment. But if put them AFTER exporter-time fixed effects, results for treatment are completely omitted. I suspect that the latter is correct, since it seems that maybe Stata accounts for exporter-time fixed effects first then the treatments?

(2) I understand that my group of exporters are all either treated or not treated, and this is why they are dropping out of my fixed effects. However I really do not want to omit exporter fixed effects, is there any obvious work around?

thank you,
Donovan.