The classic theory of fiscal federalism suggests that different people should have different governments. Yet, separate local governments with homogeneous constituents often end up doing poorly. This paper explains why and answers three questions: when regions are heterogeneous, what determines if power should be centralized or decentralized? How many levels of government should there be? How should state borders be drawn? We develop a model of political agency in which voters differ in their ability to monitor rent-seeking politicians. We find that rent extraction is a decreasing but convex function of the share of informed voters, because voter information improves monitoring but also reduces the appeal of holding office. As a consequence, information heterogeneity makes centralization appealing as a way of reducing rent extraction. Conversely, taste heterogeneity prompts decentralization as a way of matching local preferences. We also explain why the proliferation of government tiers harms efficiency. We find economies of scope in accountability: a single government in charge of many policies has better incentives than many special-purpose governments splitting its budget. Thus, a federal system is desirable only if information varies enough across regions. Our model implies that optimal borders should cluster by tastes but also ensure diversity of information. Quantitatively, our findings suggest excessive government fragmentation in the United States.