A large and growing literature links high levels of ethnic diversity to low levels of public goods provision. Yet while the empirical connection between ethnic heterogeneity and the underprovision of public goods is widely accepted, there is little consensus on the specific mechanisms through which this relationship operates. To gain analytic leverage on the question of why ethnicity matters, we identify three families of mechanisms – what we term preference, technology, and strategy mechanisms. Our empirical strategy is to identify and run a series of experimental games that permit us to examine these mechanisms in isolation and then to compare the importance of ethnicity in each. Results from experimental games conducted with a random sample of 300 subjects in Kampala’s slums reveal that successful collective action among homogenous ethnic communities in urban Uganda is attributable to the existence of norms and institutions that facilitate the sanctioning of non-contributors. We find no evidence for a commonality of tastes within ethnic groups, for greater degrees of altruism toward co-ethnics, or for an impact of shared ethnicity on the productivity of teams.