Using a model that features biased heterogeneity, factor proportions, and labour market frictions, I obtain four results: (1) firm-level skill premium increases in the skill intensity of the industry; (2) firm-level skill premium may be higher (not lower) in the skill abundant country; (3) general equilibrium effects driven by factor proportions push the average skill premium upward in the skill intensive industry and in the skill abundant country; (4) the skill premia are more evenly distributed in the skill intensive industry and in the skill abundant country. These results run counter to the prediction of factor proportion models.