As noted above, Table 1 reports standard errors grouped only at the pair level. [39] examines in a recent paper the consequences of non-compliance with the interdependence of disturbances in several dimensions, with model models for the structural gravity of cross-sectional data and bilateral trade panels, if concluded. These authors conclude that ignoring multi-channel clusters leads to misleading conclusions about the relevance of preferential trade agreements of different types, since multi-tiered consolidation has a significant impact on standard errors in commercial cost variables. To address this problem, Table 2 presents the results of multi-level grouping that allow for a correlation in the notion of error in all possible cluster dimensions. The grouping clearly has no impact on point estimates. The results show that despite significant differences between the two types of standard errors (grouped standard errors are higher), all estimates of positives at the 1% meaning level remain statistically significant and all previous conclusions remain unchanged. The reference specifications contain only a forgery for all preferential (reciprocal) trade agreements. In order to examine the problems raised in this paper, we argue with the variable “dummy” PTA in three different ways by interacting with the mannequins, whether or not the exporter and/or importing countries are beneficiaries of non-reciprocal preferential trade agreements. First, we will divide this mannequin into two models depending on whether the exporter is a beneficiary country (PTAXben) or a developed country (PTAXdev). Second, from the importer`s point of view, we dedicate the PTA-Dummy into two parts (designated by PTAMben and PTAMdev). Finally, and most importantly, we divide the PTA model into four models taking into account the group to which all the trading partners of the pair belong: the export recipient and the developed importer (PTAXbenMdev), exporter and importer recipient (PTAXdevMben), exporter and importer (PTAXbenMben) and exporter and importer (PTAXdevMdev). For more clarification, we should consider the first of these decoys as an example. PTAXbenMdev is a model who is the unit if the country i is a non-reciprocal trade agreement and if the J country is a developed country and they share adherence to a reciprocal preferential trade agreement and, moreover, zero.

If we compare the estimated coefficient of this variable with the coefficient obtained for NRPTAXbenMdev, we can test whether reciprocal trade agreements or not are the best at encouraging exports from developing countries. Similarly, we can compare the export performance of industrialized countries participating in reciprocal (PTAXdevMben) and non-reciprocal (NRPTAXdevMben) trade agreements with recipient countries. With regard to the comparison between reciprocal agreements (EPA) and non-reciprocal agreements, these results show that the two models are positive in terms of exports from recipient countries to developed countries and that despite the estimate of the point for reciprocal agreements is greater than for non-reciprocal agreements, the difference between them is not statistically significant.