Social predictors of flood exposure are critical but vary by location, Princeton study finds

A new study led by Princeton researchers reveals that commonly used indicators of social vulnerability - such as poverty, age, and race - do not predict flood exposure uniformly across Texas communities, raising critical questions about the reflexive use of these indicators in flood risk assessments and disaster planning.  

As climate change and development intensify flooding across the United States, disaster planners and policymakers are increasingly turning to social vulnerability indicators - such as poverty rates, racial demographics, or age - to assess which communities are most at risk. But a new study led by Princeton University researchers finds that these indicators may not be effective for every community. The connection between flood exposure and social factors varies significantly by location, which calls into question the “one-size-fits-all” approach of vulnerability measures in flood risk planning.

“These indicators are popular because they’re readily available, things like poverty, race, or age are included in most national datasets,” says STEP Ph.D. student and lead author Shelley Hoover. “However, social vulnerability isn’t a static concept. It's deeply shaped by local histories, unique community dynamics, and a multitude of other factors. Qualitative researchers have long warned us that using the same indicators to assess risk in every community overlooks how those factors actually function on the ground, an insight our spatial analysis reinforces.” 

Three Key Approaches

The research team pursued three key approaches to investigate if different social factors were consistently associated with flood exposure across communities, or if there was spatial variation. 

First, they conducted a literature review of U.S.-based studies to determine which social and economic indicators are most commonly associated with increased flood exposure. Second, conducted a multiscale geographically-weighted regression (MGWR) analysis of these relationships and examined how they vary across space, using Texas as a case study. Lastly, they compared these results with those from a traditional global statistical  approach, which assumes the same relationships hold everywhere.

The Results

The results suggest that relationships between social vulnerability and flood exposure are far more context-dependent than previously assumed. The literature review showed little agreement on how different social indicators relate to flood exposure across studies, suggesting that no single factor has a consistent impact nationwide. 

The spatial analysis in Texas reinforced the literature review findings, revealing that 14 out of 20 indicators had spatially varying relationships with flood exposure.  Age, for example, was associated with both elevated and reduced exposure depending on the region. The global model, by contrast, showed a uniform association with higher exposure statewide, masking this variation. This highlights a key limitation of traditional global regression methods: they miss localized relationships.  

The strength of MGWR analysis lies not only in its ability to detect local variation, but also in its capacity to identify the spatial scale at which each indicator operates. While some factors, like age, show highly localized associations with flood exposure, others demonstrate more regional or global consistency.

“This approach is powerful because it shows the spatial scale at which each social factor relates to flood exposure,” says Hoover. “Understanding the scale can provide insights into conditions that may push different groups into flood-prone areas. For example, we found poverty and flood exposure to operate at a very localized scale, associated with both higher and lower flood exposure depending on location. This makes sense since some flood-prone areas include desirable waterfront properties, while others reflect patterns of disinvestment." 

The researchers also highlight the limitations of relying on single indicators to understand vulnerability, as this may miss intersectional dynamics that overlap and present a more comprehensive picture of unique forms of risk. 

“This research is just a starting point, and we’re still making overgeneralizations that need to be acknowledged when applying this work to flood mitigation efforts,” explains Hoover. “For example, we found the indicator ‘Hispanic’ to be associated with higher exposure in nearly every region of Texas. But ‘Hispanic’ encompasses a wide range of communities with different histories, migration patterns, and relationships to land and infrastructure. These distinctions are difficult to capture using current indicators due to data limitations and statistical constraints, but recognizing what may be left out is essential. Without that, we risk overlooking critical dimensions of vulnerability, capacity, and resilience.” 

The Implications

With social vulnerability indicators playing an increasing role in flood risk reduction policy, the researchers emphasize avoiding a one-size-fits-all framework for indicator deployment.  MGWR offers a replicable methodology to identify which social indicators explain heightened exposure, and if the relationships with exposure are spatially consistent or are highly localized.  The approach is generalizable to vulnerability assessment for other climate hazards at a variety of geographic scales. 

Eric Tate, co-author and Professor of Public and International Affairs at Princeton University, notes the need for quantitative vulnerability modeling to use more place-specific approaches in assessing flood vulnerability.  

“Work in flood risk reduction has improved over the past decade by integrating social vulnerability models into policy and planning,” says Professor Tate.  “The MGWR methodology used in this study expands the analytical tool box for developing social vulnerability indicators that are tailored to specific hazards and places.”

The paper, “Spatial Heterogeneity in Social Vulnerability to Flood Exposure,” was published in Natural Hazards on May 15th, 2025.  The authors include Shelley Hoover (School of Public and International Affairs, Princeton University) and Eric Tate (School of Public and International Affairs, Princeton University).