About the Data

Before providing more details about the data below, we address a common question: Can the map and data be used for local case studies? Yes, but with a few caveats:

1. The map only covers FEMA-funded buyout areas initiated between 2007-2017. Data identifying those areas from FEMA, via a FOIA request. (See below for more details.)

2. For buyout participants ("policy movers"), we include only those who were verified to be resident homeowners when they accepted their buyout. (I.e., We exclude landlords who took buyouts that required tenants to move. The reason is that we are interested in shifting focus away from managed retreat policy as being one simply of property acquisition to one that also involves residential relocation.)

3. Nonparticipants ("market movers") include both owners and renters who moved.

Issues of Confidentiality

To help maintain confidentiality, we locate all movers in each zone of retreat to the centroid of their encompassing census tract. Destinations are then randomly “blurred,” or adjusted, to within a few blocks of subsequent destinations.

FEMA Buyout Data

Production of the map began by identifying US homeowners who voluntarily sold their homes and moved through local programs funded by the Federal Emergency Management Agency’s (FEMA) Hazard Mitigation Grant Program (HMGP) between 2007 and 2017, specifically its property acquisitions, or managed retreat, program. Since 1993, officials have implemented the program in more than 500 cities and towns in every state. Records for participating property owners were recently released through a petition filed under the Freedom of Information Act.

Of the more than 40,000 records publicly released by FEMA in 2019, approximately 22,000 contain the name of the property owner. Of those owners, approximately 14,250 (65%) are identifiable as (noncommercial) occupants of the properties they sold. We focus only on owner occupants, or resident homeowners, because we are interested in residential retreat, not the sale and demolition of rental or vacant properties, as important as those dynamics are.

Specifying Zones of Retreat

We define zones of retreat by first identifying owner-occupant participants in FEMA’s Home Mitigation Grant Program (FEMA 2024) and then drawing a half-mile buffer around their acquired property. We use a radial buffer rather than administrative units (e.g., census tracts) to define our zones because boundaries for the latter may include bodies of water that flood into adjacent units. We use a half-mile distance because that length is short enough for nearby residents to know about buyouts in their zone but still long enough to account for residents who may not have received buyout offers but nonetheless considered moving in response to local flooding and implementation of managed retreat.

Identifying and Tracking Movers from Zones of Retreat

Within our designated zones of retreat, there are no available data indicating who was offered a buyout; there are only data indicating who accepted an offer. We refer to households who accepted a buyout offer as policy movers; we refer to all other movers, who can include renters as well as homeowners, as market movers.

To identify and track policy and market movers, we had to innovate because no publicly available data are sufficient nationwide. We turned to files produced by Data Axle, a private-sector vendor. Data Axle produces annual residential files that include the names and current addresses of more than 150 million U.S. adults. To update these files each year, the company uses more than a hundred sources, including real estate tax assessments, deed transfers, and other public records (voting registrations and utility connections) as well as change of address notifications, credit card billing statements, loyalty programs, and the like. Because the company sells the data widely to major corporations wishing to identify consumer locations for bill collection, targeted marketing, and other commercial communications, Data Axle is strongly incentivized to ensure that their data are accurate and up to date. Recent studies have assessed the validity of the data using the American Community Survey as a benchmark (Acolin et al. 2022; Ramiller et al. 2024). Their general conclusion is that the Data Axle files are indeed valid, with a tendency to under-represent younger adults (ages 18-24) and the unemployed. These studies also indicate that avoiding the use of Data Axle’s imputed sociodemographic characteristics and focusing strictly on tracking residential locations over time offers the greatest validity. This is the approach we take for the interactive map.

The annual Data Axle files start in 2006. Beginning there, we use the files to locate owner-occupant, buyout participants (policy movers) and then track where they moved next. We also use the data to rebuild the residential population living within a half-mile buffer of that participant during the year buyouts commenced in their community. We then follow those nearby neighbors over time to see if they moved. If multiple participant-centric buffers overlap in an area, we eliminate redundant observations so that each household in our dataset is observed only once. To identify and track policy movers we use the name and address provided in FEMA’s HMGP database; to identify and track nearby neighbors, or market movers, we use and follow the first adult listed at a given address in Data Axle’s household record.

For all movers and regardless of the year in which buyouts began in their zone of retreat, we extend the window of observation for relocation to the same time point: December 2022, five years after the last observed buyout in the publicly available FEMA data. We take this approach because buyouts can take years to finalize and because we are interested in the total, or cumulative, relocation of residents from respective zones of retreat following buyout implementation. To ensure all households in our study were present when buyouts began in their zone of retreat, we exclude all residents who arrived after buyouts commenced. This restriction has the added benefit of controlling for variation in local residential repairs and redevelopment following a recent disaster, thus maximizing comparative validity across zones of retreat that might otherwise experience very different recovery trajectories nationwide (Pais and Elliott 2008).

Measuring Distance Moved

To measure distance moved from origin to next address, we use a batch-processing variant of Google Maps to estimate the shortest driving distance (in miles) between origin and next observed address. We use driving distance rather than straight-line distance because our interest lies in residents’ continued attachment, or connectivity, to communities of origin, and driving distance offers a better proxy for that attachment than straight-line distance (Roberto 2018).

References

Acolin, A., Decter-Frain, A., & Hall, M. 2022. “Small-area Estimates from Consumer Trace Data.” Demographic Research 47: 843-882.

Benincasa, Robert. 2019. “Search the Thousands of Disaster Buyouts FEMA Didn’t Want You to See.” National Public Radio Online, March 5, 2019. https://www.npr.org/2019/03/05/696995788/search-the-thousands-of-disaster-buyouts-fema-didnt-want-you-to-see.

Federal Emergency Management Agency (FEMA). 2024. “Hazard Mitigation Grant Program.” Last modified September 9, 2024. https://www.fema.gov/hazard-mitigation-grant-program.

Pais, Jeremy and James R. Elliott. 2008. “Places as Recovery Machines: Vulnerability and Neighborhood Change after Major Hurricanes.” Social Forces 86(4): 1415-1453. https://doi.org/10.1353/sof.0.0047.

Ramiller, A., Song, H., Parker, J. R., & Chapple, N. 2024. “Residential Mobility and Big Data: Assessing the Validity of Consumer Reference Datasets.” Cityscape: A Journal of Policy Development and Research, 26(3): 209-232.

Roberto, Elizabeth. 2018. “The Spatial Proximity and Connectivity Method for Measuring and Analyzing Residential Segregation.” Sociological Methodology 48(1): 182-224.