A flood-exposure penalty that keeps deepening for over a decade.
Sun–Abraham (2021) event-study estimator on a repeated-sales panel of 8,148,080 U.S. residential transactions (LightBox, 2000–2020). The dot at year −1 is the reference period. After exposure, sale prices per square foot fall steadily relative to non-flooded controls, reaching −18.9% by year 12. Pre-trends are flat (joint test p = 0.6).
Source: Sandoval-Olascoaga, Shu & Porter, Frontiers in Environmental Economics (2025), Figure 2 · LightBox property transactions 2000–2020 · Sun & Abraham (2021) event-study estimator · Property + year/month fixed effects · Tract-clustered standard errors · 95% confidence band shown · Pre-trends joint test p = 0.6 · n = 8,148,080 transactions
$189,000
Flood penalty on a $1M home 12 years later
8.1M
Property transactions analyzed across the contiguous U.S.
32.3%
Of county salary & wages expenditure at risk in worst-affected counties
Why it matters
Prior estimates of the flood-price discount sat at 5 to 10% and faded within a decade. This paper estimates 19% and still deepening at year 12.
Prior empirical work on flood-exposed property values reported discounts of 5 to 10% and a "forgetting" pattern in which prices recovered within a decade. Both came from geographically narrow case studies (Miami-Dade, NYC after Sandy, individual counties in North Carolina or Georgia) and from standard hedonic regressions.
By assembling a nationwide repeated-sales panel and using the Sun–Abraham (2021) event-study estimator, we recover an effect more than twice as deep and entirely persistent, once the model conditions on neighborhood population trajectory. Where the flood lands governs the size and permanence of the discount.
How we did it
Same property, before and after, across eight million sales.
The unit of observation is the transaction. The outcome is log(sale price per square foot). The treatment cohort is the year a property's census tract first experienced a flood between 2000 and 2020, drawn from the Lai et al. (2022) NLP-extracted event database and the NOAA Storm Events Database. Property-level fixed effects absorb all time-invariant home characteristics. Year and month fixed effects absorb national macro shocks and seasonality. Standard errors cluster at the census-tract level.
The repeated-sales design restricts analysis to properties that transact multiple times across flood exposure, strengthening identification beyond standard hedonic models. Event time runs from −7 to +12 years, with e = −1 as the reference. We estimate the model separately for properties in census tracts that gained versus lost population in the 2000 to 2010 census, isolating the demographic moderator. A second stage estimates state-by-state elasticities of property-tax revenue to sale prices, then projects county-level fiscal impacts using First Street Foundation 30-year flood inundation projections and SSP-aligned block-group population forecasts from Shu et al. (2023).
What we found
A 19% penalty after 12 years, which splits in two by neighborhood demographics.
Pooled. Three years after a flood, properties sell for 2.5% less than identical non-flooded properties (p = 0.0001). By year 12, the gap is 18.9% (p < 0.0001). On a million-dollar home, that is a $189,000 penalty. And it is still deepening.
Split by population trajectory. In growing tracts, prices dip to −17.8% at year 8, then partially recover to −8.7% by year 11. In shrinking tracts, prices fall to −4.8% by year 5 and continue down to −15% by year 11, with no rebound. The penalty is permanent only where housing demand is also retreating.
Even before the flood. Growing-population tracts saw prices rise +3.05% per year faster than shrinking ones (2001 to 2022, statistically significant every year except during the Financial Crisis). The gap widened to +5.8% by 2022. Climate exposure compounds a demographic divergence that was already present before the flood.
The fiscal echo. The nationwide elasticity of property-tax revenue to sale prices is 0.71 (p = 0.001). A 1% price decline implies a 0.71% revenue decline. Twelve years after a flood, projected impacts on total county taxes range from 2.7% to 24.5% (mean 12.1%). At the upper bound, that is 32.3% of total county salary and wages, or 21.83% of education expenditure. The same shrinking communities that face the deepest price penalty lose the most fiscal capacity to adapt.
In communities already facing demographic and economic headwinds, flood events may exacerbate existing vulnerabilities, further constraining their tax base and financial capacity to invest in recovery or long-term adaptation. From the paper.
Cite this paper
Sandoval-Olascoaga, S., Shu, E., & Porter, J. (2025). Flood risk and property value changes: understanding the impact of climate event exposure in the context of population change. Frontiers in Environmental Economics, 4, 1615802. https://doi.org/10.3389/frevc.2025.1615802