This paper presents evidence of spatial variation of ethnic political mobilization. The ethnic politics literature maintains that voters choose co-ethnic candidates because they expect to receive more benefits from co-ethnic officials. If the argument is true, however, a rational voter should take into account her ethnic self-identification in voting only if she expects ethnicity to influence the process of distribution of private and public goods. Given the characteristics of public goods (non-rivalry, non-excludability), a voter should only expect to receive extra public goods from a co-ethnic official if she lives in a neighborhood or village with many other co-ethnics since only such areas would be targeted with local public goods. This identification of sets of co-ethnics will critically depend on their degree of spatial ethnic homogeneity. If this argument is true, we should expect to see ethnic voting in areas of relative ethnic homogeneity, but not in areas of ethnic mixing. In this paper I exploit the highly contrasting levels of ethnic spatial homogeneity of rural vs. urban areas in Uganda to show that candidate choice in a recent presidential election conformed to the expected pattern.