Hello, I am using a unilateral gravity model to see the determinants of trade in ICT services. I used a RE estimation. Both GDP coefficients and WTO dummy are +ve and significant. However the coefficient of the GATS_commitments in the ICT had an insignificant effect on exports.

Since I have a selection bias because some countries in the dataset are not WTO members and thus don't have GATS commitments, I ran the heckman selection command but the results became weird. The coefficient of the GATS_commitments turned negative significant while the coefficient of the GDP of rest of the world became also negative.

My heckman command was : heckman lnexports GATS_total lnGDP lnrestgdp lnindwnetpercent1 lncpiscore lnrealEffectiveExchange2010 lnsecondarySchoolEnrolment evercolonized r1 r2 r3 r4 r5 r6 r7 , select (WTO_membership=Totaltimetotradedays Timeregpropoerty Procedurescontract) twostep first

ln exports are exports of ICT services, r are different regions

Here I used variables from the ease of doing business that have a significant effect on being a WTO member.

Thank you so much