The standard neoclassical model cannot explain persistent migration flows and lack of cross-country convergence when capital and labor are mobile. Here we present a model where both phenomena may take place. Assuming increasing returns and perfect capital mobility, the driving forces behind labor migration are the size and the composition of the workforce. The size of the workforce produces divergence: a larger workforce in one country implies higher wages, the wage gap induces in-migration and this makes the wage gap even larger. In this case migration is only beneficial for the labor importing country. The evolution in the composition of the workforce due to migration may have opposite effects. In one case in-migration keeps the wage of skilled workers high enough to induce people to qualify for skilled jobs. This improves the composition of labor in the receiving country only and it makes divergence a likely outcome. In another case labor migration keeps the relative wage of unskilled workers in the rich country high enough to induce all migrants to take unskilled jobs and reduce the skilled-to-unskilled ratio in that country. Then, migration is beneficial to the country which was first exporting labor.