In this paper I discuss the nature of the political constraints that the World Bank faces in delivering basic services to the poor. The main problem arises because the Bank has to work through domestic governments which have political aims different from helping the poor. The conceptual approach attractive to economists and central to much recent thinking in the World Bank, particularly the 2004 World Development Report, is the notion of "politician proofing". Given that political incentives derail good policies, how can those policies be politician proofed? I argue that evidence and theory suggests that such an approach is ultimately futile, basically because we simply do not understand the relevant political incentives. I discuss alternative policy strategies and conclude that what is required is a much more fundamental assessment of what type of political equilibria deliver services to the poor. As I illustrate with the case of Botswana: once the political equilibrium is right, everything goes right and politician proofing is redundant.