Dear all,

I am currently working on my master thesis and I would like some guidance, since the dataset includes some zero-trade flows, I figured that PPML would be the most applicable method. I want to estimate the effect of the Iranian migrant stock on trade flows using a gravity model. I have a panel spaning from 2010 to 2017 with 116 exporter countries and Iran as the importer. I wonder if it is possible with this set-up to use country-specific fixed effect, if so, is it most appropriate to use fixed-effects for both the exporters and the importer or just for the exporting countries?

One thing I tried was this:
Code:
egen exp_time = group(Exporter Year)
tab exp_time, generate (EXPORTER_TIME_FE)
egen imp_time = group(Importer Year)
tab imp_time, generate (IMPORTER_TIME_FE)
egen pair_id = group(Exporter Importer)
tab pair_id, generate (PAIR_FE)
ppml ExportstoIran PAIR_FE* lnGDP lndist lntotmig lnMS lnMStimeswithout lnpop BureaucracyQuality Democraticaccountability LawandOrder Corruption contig comlang_off
Is this a correct way of using the fixed effect in the case where I just have one importer or should I use the fixed effect for just the exporters? Or none?

Also, I tried both PPML and PPMLHDFE (without country-specific fixed effect) and they generated very similar results. The codes I've used was:
Code:
ppml ExportstoIran lnGDP lndist lntotmig lnMS lnMStimeswithout lnpop BureaucracyQuality Democraticaccountability LawandOrder Corruption contig comlang_off
and
Code:
ppmlhdfe ExportstoIran lnGDP lndist lntotmig lnMS lnMStimeswithout lnpop BureaucracyQuality Democraticaccountability LawandOrder Corruption contig comlang_off
If someone could provide me with the pros and cons of each of the approaches I would be very grateful.

Also, if I want to estimate the effect of the Iranian migrant stock just for 2016 and 2017 speficially, my method was to generate an interaction term which was the migration stock times a year dummy (with 1 for 2016 and 2017). Is this a solid way to estimate the effect for these particular years?

Thanks in advance.
Best regards,
Gabriel Bladh