Highlights :

 Abstract :
Influential empirical work by Rauch and Trindade (REStat, 2002) finds that Chinese ethnic networks of the magnitude observed in Southeast Asia increase bilateral trade by at least 60%. We argue that this estimate is upward biased due to omitted variable bias. Moreover, it is partly related to a preference effect rather than to enforcement and/or the availability of information. Applying a theory-based gravity model to ethnicity data for 1980 and 1990, and focusing on pure network effects, we find that the Chinese network leads to a more modest amount of trade creation of about 15%. Using new data on bilateral stocks of migrants from the World Bank for the year of 2000, we extend the analysis to all potential ethnic networks. We find, i.a., evidence for a Polish, a Turkish, a Mexican, or a Pakistani network. While confirming the existence of a Chinese network, its trade creating potential is dwarfed by other ethnic networks. The large heterogeneity in the trade-creating potential of different networks is, among other things, explained by the share of high-skilled immigrants, the degree of ethnic fragmentation, and GDP per capita.

 Keywords : Gravity model | International Trade | Network Effects | International Migration

 JEL : F12, F22