Flooding remains one of the costliest natural disasters in the United States but impacts vary greatly across space. Key policy questions arise regarding potential heterogeneity in Marginal Willingness to Pay (MWTP) to avoid flood risk. In this paper, we estimate household preferences to avoid such risks in the presence of heterogeneous sorting using house sales across Florida in 2010. We use a discrete choice, residential sorting model to estimate heterogeneous preferences for flood risk. Our identication strategy relies on a boundary discontinuity design (BDD) to control for potentially confounding, yet unobserved, factors that are correlated with flood risk. In addition, we account for differential pricing of National Flood Insurance Program premiums including underlying flood risk and subsidies, which, if ignored, may bias estimates of willingness to pay. Understanding heterogeneity in preferences for such risks is key to assessing the extent to which vulnerable sub-populations systematically sort into higher risk areas. In addition to providing new valuation estimates for flood risk, our results have implications for the distributional impacts of US natural disaster policies and its potential role in exacerbating inequitable flood risk exposure in the US.