Exposing disparities in flood adaptation for equitable future interventions in the USA

This paper investigates disparities in flood adaptation effectiveness across different US communities using a FEMA dataset of approximately 2.5 million flood insurance claims. The study employs CAUSAL.FLOW, a causal inference method based on deep generative models, to analyze the impact of flood adaptation interventions considering income, racial demographics, population, flood risk, education, and precipitation. While the program yields significant cost savings per household, these benefits are unevenly distributed, with low-income communities, particularly minority communities, experiencing sharply reduced savings compared to high-income, predominantly white communities. The findings highlight the need for future flood adaptation efforts to prioritize equitable support for all communities.

Lidia Cano Pecharroman, ChangHoon Hahn