This paper proposes a new approach for creating country indices and applies it to the construction of an index ranking countries according to their current level of competitiveness in international goods exchange. Existing indices largely assign equal or in some other way arbitrary weights to their single indicators regardless of the information contained. By applying the so far neglected method of canonical correlation analysis, we obtain index weights which are themselves transparent outcomes of the data structure. In doing this, we consider competitiveness as a multidimensional concept which can be proxied by export volumes as well as inflows of capital and labour. Weights for these index components are derived based on their correlation with a set of development indicators. In this way, components are assigned larger weights which reflect to a larger degree the state of institutional quality and to a smaller degree exogenous first-nature advantage. As a consequence, the weight of capital inflows turns out to be driven to zero.