Computer model predicts the length of a household’s displacement in any US community after a disaster

One of the human impacts of natural hazards is household displacement. Destructive floods, wildfires, earthquakes and hurricanes often force people to leave their homes—some briefly, others for months or indefinitely.

Most disaster risk assessments, used by insurance companies, government agencies, development banks, and academic researchers to predict the potential future impacts of natural hazards, fail to account for hardships incurred by household displacement.

Instead, they focus on direct economic losses, a metric that “often highlights the wealthiest as the most at-risk,” says Nicole Paul, Ph.D. candidate at University College London. “But observations from past disaster events often show that poor and marginalized people have the greatest recovery needs.”

In a study published in Risk Analysis, Paul and her colleagues used recent, disaster-related data from the Household Pulse Survey (HPS) to train a computer model to predict the length of household displacement and return outcomes after a disaster.

The study is the first to use state-by-state data from the U.S. Census Bureau to quantify the contribution of different factors (including household size, tenure status, educational attainment, and income per household member) on household displacement and return.

According to the HPS data, 1.1% of American households were displaced due to disasters between December 2022 and July 2024. Hurricanes were the most common disaster type cited by displaced households, while other households reported floods, fires, tornados, and “other” hazard types.

Survey responses from 11,715 households that experienced disaster displacement were used by the researchers to fit predictive computer models for household displacement into three classes: emergency phase displacement (returned in less than one month), recovery phase displacement (returned after one month), and not returned (potentially permanent relocation).

Although most households returned relatively quickly, 20% were displaced for longer than a month and 14% had not returned by July 2024. The geographical locations of households revealed significant differences among states:

• Households in Louisiana and Florida were about 6.8 and 4.4 times more likely to be displaced, respectively, compared to the national average. Yet they returned home faster than households in many states.
• Displaced households in Minnesota, North Dakota, and Connecticut were likelier to take more than one month to return.
• Displaced households in Alaska, Hawaii, and Kansas were considerably less likely to return to their homes (by July 2024).

“The duration of displacement is key to understanding the human impact of a disaster,” says Paul. “Short-term evacuations can save lives and be minimally disruptive, while protracted displacement is associated with significant hardships for families.” Those challenges can include disruption of education, income and/or job loss, and various psychological effects.

To understand the impacts of future disasters, Paul adds, the computer model can combine estimates of physical damage with socioeconomic characteristics to predict the duration of household displacement within a community and therefore help inform risk mitigation strategies that reduce displacement risks for members of that community in future disasters.