We propose a decomposition of the border effect in international trade by controlling for differences in competition in local markets. An extension of the Hotelling (1929) model shows that the availability of local substitutes increases price dispersion and biases the estimation of the border effect. We test these predictions using detailed price database at the supermarket level for Uruguay. This stylized setting makes it possible to control for other potential explanations of the border effect (i.e., exchange rates, taxes, or transport costs). We find that for those goods without local competitors the border estimation increases substantially, while for those goods that do have local competitors the effect of border is negligible. As the literature suggests, results should be even larger for different countries than for different cities. The methodology developed in the paper allows a finer explanation for understanding the relevance of borders in price dispersion.