“…The general FE model is defined as follows [68]: lnEX i,j,t = 0 + 1 lnCBEC i,t * BR i,t + 2 lnGDP i,t + 3 lnGDP j,t + 4 lnDIST i,j + 5 COMLANG i,j + 6 CONTI i,j + 7 FTA i,j,t + i,j,t lnEX i,j,t = 0 + 1 lnCBEC i,t * lnDIST i,j + 2 lnGDP i,t + 3 lnGDP j,t + 4 lnDIST i,j + 5 COMLANG i,j + 6 CONTI i,j + 7 FTA i,j,t + i,j,t lnEX i,j,t = 0 + 1 lnCBEC i,t * t + 2 lnGDP i,t + 3 lnGDP j,t + 4 lnDIST i,j + 5 COMLANG i,j + 6 CONTI i,j + 7 FTA i,j,t + i,j,t Some studies use FE estimation to deal with the possible heterogeneity and endogeneity issues in a gravity model [69,70]. In this way, all time-invariant factors that are specific to a country pair as well as other unobservable characteristics (such as political, or economic characteristics) of the country pairs can be captured by the country-pair dummies [71][72][73]. However, a major shortcoming of the FE estimation is that any time-invariant variables in a gravity model, such as distance, will be absorbed into country pair FE, not allowing the estimation of coefficients on these variables.…”