We estimate the effect of RTAs on bilateral exports by means of a gravity model analyzing its sensitivity to different specifications and methods. RTAs generate a sizable positive effect. However, shifting to country-pair and time-varying fixed effects systematically reduces coefficients. Nevertheless, the RTA effect is consistent across methods and specifications. The RTA effect attributable to particular trade agreements displays high variability. While most RTAs increase trade, others present non-significant or negative results. We apply robustness checks to individual RTA estimates by presenting PPML time-invariant fixed effects and next to these, country-pair and time-varying fixed effects estimates. Thus, 38.2% of RTAs are positive and significant in both specifications. RTAs trade creation effects tend to prevail over trade diversion effects.