Over the past decade, natural disasters in the US such as hurricanes, floods, and wildfires have caused over $1 trillion in damages to housing and infrastructure. The federal government has responded with over $200 billion in post-disaster assistance primarily through rebuilding subsidies for affected homeowners and grants to rebuild infrastructure. Yet significant questions remain about whether this largely place-based approach effectively serves affected households and communities. Strict screening processes can result in years-long delays or denials of assistance, renters receive minimal direct support, and by providing in-kind rebuilding aid tied to specific properties, these programs may prevent households from migrating even when such moves could provide a valuable form of insurance during recovery.
To evaluate the distributional impacts of natural disasters and the effectiveness of recovery policies, I analyze household decisions following Hurricane Maria’s devastation of Puerto Rico in 2017. Puerto Rico provides a unique setting for studying disaster recovery. As an unincorporated U.S. territory whose residents are U.S. citizens, affected households face meaningful migration decisions: they can relocate to the mainland without immigration barriers while qualifying for the same federal disaster assistance programs, yet face significant financial and social costs of moving. I combine high-frequency flight data and an original household survey with a dynamic structural model of migration, housing, and infrastructure investment to quantify how current post-disaster policies affect household welfare and to study alternatives.
Flight data reveal that 7% of Puerto Rico’s population migrated to the mainland US following Hurricane Maria, with approximately 40% of those who left returning within one year and 60% remaining permanently. Renters were significantly more likely to migrate permanently, while homeowners were more likely either to remain in Puerto Rico throughout the recovery period or to migrate temporarily before returning. Features of the rebuilding assistance application process likely contributed to this differential behavior. Survey evidence suggests that binding financial constraints prevented many additional households from migrating, even temporarily.
The model results reveal substantial variation in policy effectiveness. The model results reveal substantial variation in policy effectiveness. Infrastructure investment generates total household value above its fiscal cost, producing broad-based benefits for both renters and homeowners while supporting local economic recovery. In contrast, in-kind homeowner rebuilding subsidies generate a value to eligible homeowners well below cost. This low valuation reflects several features of current program design. Eligibility for rebuilding assistance requires homeowners to repair their specific damaged property, effectively conditioning aid on remaining in place during the application and approval process. Because these benefits are non-transferable and in-kind, they provide substantially less value than their fiscal cost, particularly for liquidity-constrained or distressed homeowners.
The model estimates suggest that modest reforms to disaster assistance could substantially improve outcomes without increasing fiscal costs. Small migration subsidies could allow households to temporarily relocate during the worst of the disaster’s aftermath while maintaining the option to return once conditions improve. Additionally, replacing or complementing in-kind rebuilding assistance with flexible cash transfers would permit households to allocate resources according to their specific circumstances. Flexibility improves outcomes without added cost.