Gravity models are one of the preferred tools for evaluating the ex post effects of trade agreements. While highly effective at predicting trade flows, they face a major obstacle when it comes to assessing the effects of trade agreements: the scarcity of data from signatory countries. As a result, the effects estimated to date may be very far from reality, in both directions; either they have gone undetected or they have been overestimated. In these circumstances, using statistical methods suited to small samples is essential. By being more demanding regarding the significance of results, these methods help avoid overestimating the effects of trade agreements. However, given the scarcity of data, they cannot guarantee that a trade agreement that appears to have no effect does not, in fact, have one.