Developing a Spatial Regression Model Framework for Insured Flood Losses in Houston

Abstract

Events such as Hurricane Harvey in 2017, which struck Houston resulted in billions of dollars in reported flood insurance claims through the National Flood Insurance Program (NFIP). Currently, there is limited research investigating how hydrologic and socioeconomic drivers influence the location and magnitude of NFIP claims. Here, we present a statistical modeling framework of NFIP claims and claim amounts at the census tract level for 13 flood events affecting Houston from 2010 to 2019. We determine a relationship between insured losses and local hydrologic and socioeconomic variables and account for spatial dependency via eigenvector spatial filtering. We observed that communities with high policy densities within high-risk flood zones incurred the most insured losses, as would be expected, while census tracts with predominantly White, non-Hispanic, and Hispanic populations are linked to higher claim amounts. We additionally found that communities with a lower median income and larger elderly population are linked to more flood losses. Our modeling framework allows the evaluation of flood loss projections due to climate change and changes in policy density.