Economic damages from flooding are expected to grow, given the connection between hydrological systems and climate change. Yet, there are few studies analyzing flooding damages for the entirety of the contiguous U.S. that clearly measure the role of hazard, exposure, and vulnerability variables, and that are also suitable to perform spatially detailed predictions of future damages. We constructed a panel database for all U.S. counties between 1999-2018 encompassing: (1) property and human damages from fluvial and pluvial flooding, (2) river discharge, (3) the construction patterns of buildings, and (4) the incidence of flooding events to create proxy variables of flooding risk and its hazard, exposure, and vulnerability. Because damages are censored at zero, we perform an econometric regression using the method of trimmed least squares estimators described by Honoré (1992). The resulting estimates indicate the exposure variable as the main driver of flooding risk. We use these estimates to map the counties with highest predicted flooding risk for the next decade.