Hi All,

I'm running a PPML regression on bilateral export pairs, using the OECD TiVA dataset (2018 revision). I'm specifically focusing on service subsectors for my analysis, ie, the dependent varies by service sector. I have a theoretical question here: when we include GDPs of the exporter and importer as explanatory variables, is it always national output? For instance, if I'm using transportation services exports as my dependent, should the GDPs still be aggregate GDP of the countries, or should it be transportation services output? My instinct tells me it should be the former, since services exports feed as inputs into the production of many commodities and not just traded as final exports.

My second query relates to certain errors that I received while running PPML that I can't understand. I used ppml_panel_sg when I used the BaTIS dataset, since this is supposed to be faster and BaTIS is more comprehensive than TiVA. I included exporter-time and importer-time fixed effects to account for multilateral resistance. I followed the specifications suggested here (https://fmwww.bc.edu/RePEc/bocode/p/ppml_panel_sg.pdf). I get the following errors:

"lngdp_origin omitted due to collinearity issues over lhs>0
lngdp_destination omitted due to collinearity issues over lhs>0"

My predicted is also turning out weird - values keep repeating and many pairs of predicted are even negative. R-sq is very low, about 0.02.

What might be the reasons for this?

Regards,
Saunok