Abstract There are two principal theories of why countries or regions trade: comparative advantage and increasing returns to scale. Yet there is virtually no empirical work that assesses the relative importance of these two theories in accounting for production structure and trade. We use a framework that nests an increasing returns model of economic geography featuring `home market effects' with that of Heckscher–Ohlin. We employ these trade models to account for the structure of regional production in Japan. We find support for the existence of economic geography effects in eight of nineteen manufacturing sectors, including such important ones as transportation equipment, iron and steel, electrical machinery, and chemicals. Moreover, we find that these effects are economically very significant. The latter contrasts with the results of Davis and Weinstein (1996, Does economic geography matter for interactional specialiazation?, NEBR# 5706), which found scant economic significance of economic geography for the structure of OECD production. We conclude that while economic geography may explain little about the international structure of production, it is very important for understanding the regional structure of production.