Good morning Statalist community,

I am studying the determinants of asset trade flows across 40ish countries, from 1997 to 2015, using the gravity equation. The data looks as follows:
Code:
input str22 country_o str7 iso3_o str44 country_d str3 iso3_d int year float(ln_asset_flows ln_gdp_d ln_distw) byte comlang_off float(eu common_leg) byte(contig comcur) double(asset_flows distw) byte(eu_o eu_d) long(d_country o_country)
"Australia"  "AUS" "Argentina" "ARG" 1997 17.596739   26.5947  9.39637 0 0 0 0 0 43869925.7721299 12044.573995607949 0 0 1  2
"Australia"  "AUS" "Argentina" "ARG" 2001         .  26.50859  9.39637 0 0 0 0 0                0 12044.573995607949 0 0 1  2
"Australia"  "AUS" "Argentina" "ARG" 2006 17.047169  26.30107  9.39637 0 0 0 0 0         25321600 12044.573995607949 0 0 1  2
"Australia"  "AUS" "Argentina" "ARG" 2007 17.378374 26.521635  9.39637 0 0 0 0 0 35264000.0000001 12044.573995607949 0 0 1  2
"Austria"    "AUT" "Argentina" "ARG" 1997         .   26.5947 9.371706 0 0 0 0 0                0 11751.146581945419 1 0 1  3
"Austria"    "AUT" "Argentina" "ARG" 2001 11.386568  26.50859 9.371706 0 0 0 0 0 88129.9999999998 11751.146581945419 1 0 1  3
"Austria"    "AUT" "Argentina" "ARG" 2002         . 25.540377 9.371706 0 0 0 0 0                0 11751.146581945419 1 0 1  3
"Austria"    "AUT" "Argentina" "ARG" 2003         .  25.77944 9.371706 0 0 0 0 0                0 11751.146581945419 1 0 1  3
"Austria"    "AUT" "Argentina" "ARG" 2004 14.817685 25.934366 9.371706 0 0 0 0 0          2724200 11751.146581945419 1 0 1  3
"Austria"    "AUT" "Argentina" "ARG" 2005 14.673918 26.130024 9.371706 0 0 0 0 0          2359400 11751.146581945419 1 0 1  3
"Austria"    "AUT" "Argentina" "ARG" 2006 15.700305  26.30107 9.371706 0 0 0 0 0          6585000 11751.146581945419 1 0 1  3
"Austria"    "AUT" "Argentina" "ARG" 2007  15.99396 26.521635 9.371706 0 0 0 0 0 8832600.00000001 11751.146581945419 1 0 1
As suggested by numerous authors in the literature, I have decided to include importer-year and exporter-year fixed effects to account for multilateral resistances. I do so by typing (I have Stata 14):
Code:
reg ln_asset_flows ln_distw comlang_off eu common_leg contig i.o_country#i.year i.d_country#i.year
where ln_distw is the log of distance, comlang_off is a dummy for whether the two countries share the same official language, common_leg is a dummy for whether they share the same legal system, eu is a dummy equal to 1 if both countries are in the European Union, o_country is the exporter country and d_country is the importer country.

The code runs just fine and produces the output, however it gives me some warnings that I am not sure how to interpret, such as
Code:
note: 9.o_country#1997b.year identifies no observations in the sample
and
Code:
note: 54.d_country#2003.year omitted because of collinearity
.

For quickness sake I tried to use reghdfe as follows:
Code:
reghdfe ln_asset_flows ln_distw comlang_off eu common_leg contig, absorb(o_country#year d_country#year)
and I get no error whatsoever. 4 singletons observations get dropped and MWFE estimators converge in 7 iterations, which I am not sure represent a problem.

My questions are the following:

1) Are the messages I get using reg a problem for the interpretation of my results (I am not interested in the FEs)?

2) If they are, browsing through other discussions I saw that it is sometimes suggested to change specification, but in this case I don't see how I could do that, given this is what is suggested by the literature. Would you have any suggestion?

3) Does the fact that reghdfe does not give any error message mean I can just ignore the above stated issues?

Thank you in advance to the community for your help!

Best,

Matteo