A large body of empirical research finds that a pair of regions within a country tends to trade 10 to 20 times as much as an otherwise identical pair of regions across countries. In the context of the standard trade models, the large "border effect" is problematic, because it is consistent only with high elasticities of substitution between goods and/or high unobserved national border barriers. The author proposes a resolution to this puzzle based on vertical specialization, which occurs when regions or countries specialize only in particular stages of a good's production sequence. The author develops a Ricardian model of intra-national and international trade, and shows how endogenous vertical specialization magnifies the effects of border barriers such as tariffs. He calibrates the model to match relative wages, trade shares, and vertical specialization for the U.S. and Canada. The model implies a much smaller border barrier and border effect than previous estimates.