Recent experience with disasters and terrorist attacks in the US indicates that state and local governments rely on the federal sector for support after disasters occur. But these same governments are responsible for investing in infrastructure designed to reduce vulnerability to natural and man-made hazards. This division of responsibilities - state governments providing protection from disasters and federal government providing insurance against their occurrence - leads to the tension that is at the heart of our analysis. We explore these tensions building on the model of Persson and Tabellini (1996). We show that when the federal government is committed to full insurance against disasters, states will have incentives to underinvest in costly protective measures. We then show that when the central government cannot verify state investment choices, the optimal insurance system would be designed to reward states that succeed in avoiding disasters and punish those that do not, thereby giving states an incentive to increase investment in protective infrastructure. However, this raises the question of whether the central government can credibly commit to such a scheme, and we find in a simple political model that it cannot. In our political model, the central government will decrease transfers ex-post if a state provides protective infrastructure that increases its expected uncertain income, generating a soft-budget constraint for states. This provides an additional incentive for states to underinvest in protective infrastructure. We discuss these results in light of disaster policy in the US.