A common presumption is that decentralization is prone to a potential pitfall owing to the greater vulnerability of local governments to capture by local elites. We investigate the determinants of relative capture of local and national governments theoretically, in the context of an extended version of the Baron-Grossman-Helpman model of electoral competition with lobbying by special interest groups. A number of factors do provide support to the traditional presumption, such as reduced cohesiveness of interest groups, higher levels of voter awareness, and greater electoral competition at the national level. A number of other factors may, however, create an opposite tendency for lower capture at the local level. These include less electoral uncertainty at the national level, and a higher value of campaign funds in national elections owing to their fungibility across different districts. Relative capture also depends on heterogeneity across districts with respect to levels of local inequality and poverty: accordingly decentralization will tend to increase capture in high inequality districts and lower it in low inequality districts. Power-sharing between parties at the national level, due either to coalition governments or proportional representation, limits the extent of national capture. We conclude that empirical research is necessary to investigate the extent and determinants of relative capture.