This study investigates the role of countries’ skills endowment for comparative advantage. It tests the theoretical model of Ohnsorge and Trefler (2007) who argue that it is the bundling of various skills at the worker level and their joint distribution that matter for trade specialisation. This departs from the literature assuming that workers are endowed with only one type of skills, generally measured by educational attainment. The model’s predictions are tested using information on cognitive skills from the Survey of Adult Skills (PIAAC) and Trade in Value Added (TiVA) data. Results show that workers' skills bundles and their distribution have larger effects on specialisation than countries’ endowment of capital per employee, or the relative endowment of workers possessing different levels of education. Furthermore, this study tests the model of Bombardini et al. (2012) and finds evidence that the within-country dispersion of skills significantly affects specialisation patterns.