We investigate the relationship between exporting, importing, and wage premia using a rich matched employer-employee data set. We improve on the previous literature (i) by using a new methodology to quantify the contribution of an extensive set of worker- and rm-level observable and  unobservable characteristics to the wage gap, and (ii) by controlling for the import as well as the export activity of the firm. These two innovations allow us to avoid large biases that characterized the previous literature. A robust result is that the hiring policy of exporters is quite different than the one of importers. While firm size and sales are, to different extents, important components of the wage gap both for exporters and importers, importers hire workers that are overwhelmingly more able than the average. Workers at exporting firms, on the contrary, are no different in terms of unobserved time-invariant characteristics.Our analysis provides a useful guidance for recent theories that aim at explaining participation both in export and import markets and at including non-neoclassical labor market features into trade models.