Main content start

After disasters, wealthier neighborhoods rebuild stronger while lower-income areas fall further behind

Faculty Researcher

Extreme weather events such as hurricanes and floods can devastate communities. In a paper recently published in Nature, Jackelyn Hwang and colleagues tracked rebuilding in the aftermath of disaster across the United States. The study’s key innovation is the way it measures disaster recovery. Instead of relying on broad surveys or low‑resolution satellite proxies, the authors build a detailed, building‑by‑building dataset using historical Google Street View images of buildings identified with FEMA damage records. They assemble data for 106,782 damaged buildings (over 616,000 images) across 2,195 census tracts in 16 states and 12 US extreme weather events from 2007 to 2023. They then use the GPT-4.0 vision model to classify each building’s visible recovery path—whether it became an empty lot, was rebuilt to a similar structure, was rebuilt with improvements, or showed no visible change—so they can track how neighborhood “built environments” change over time at scale.

Using this approach, they find large and consistent inequalities in recovery. Lower‑income neighborhoods are much more likely to lose buildings entirely and less likely to rebuild with upgrades, such as elevated structures. Among buildings with visible change, 37.1% in lower‑income areas became empty lots, compared with 22.2% in middle‑income areas and 7.3% in higher‑income areas. Meanwhile, 81.8% of visibly changed buildings in higher‑income neighborhoods were rebuilt with improvements, versus 33.2% in lower‑income neighborhoods. Recovery also varies by region: areas affected by Hurricane Sandy show fewer empty lots and more upgrades than parts of the Gulf Coast affected by events such as Hurricane Harvey and the 2016 Louisiana floods. 

The authors connect these patterns to resources. Higher homeowner and flood insurance coverage among FEMA assistance recipients is linked to more rebuilding with improvements and fewer empty lots. They conclude that recovery can widen pre‑existing neighborhood gaps and that financial assistance structures may need rethinking to support more equitable resilience.