infrastructure · environment · family: the rulebook was written for a world that no longer exists
the levee invited the neighborhood
Flood control infrastructure success encourages floodplain development, increasing total losses
Problem statement
The U.S. invested massively in structural flood control — levees, dams, channelization. The National Flood Insurance Program (NFIP, 1968) was designed to provide affordable insurance while mapping flood risk. But the "safe development paradox" (its most-studied form is the "levee effect") means flood protection creates a false sense of security that attracts development into flood-prone areas. A national assessment of development around the regulatory 100-year floodplain — the policy side of the paradox, not the levee side — found 89,080 km² of developed land — 24% of all developed land in the study system — within 250 meters of the 100-year floodplain as of 2019, and projects that approximately 22% of all anticipated development growth from 2020 to 2060 will occur within that 250 m zone (Sanchez et al. 2024). Empirical gauge-record analysis shows levees raise flood stages upstream of and within leveed reaches, with abrupt, statistically significant increases of up to ~2.3 m at some sites (Heine & Pinter 2012, Hydrological Processes). The NFIP is roughly $22.5 billion in debt to the U.S. Treasury. Levees don't eliminate flood risk — development behind them raises the potential for greater damage if a levee fails in a flood that exceeds its design (Sanchez et al. 2024).
Why this matters
U.S. flood losses averaged $32.1 billion annually as of 2020 and are projected to rise 26% to $40.6 billion by 2050, borne disproportionately by poorer communities (Wing et al. 2022). Repetitive loss properties have historically accounted for just 1% of NFIP-insured properties but about 25–30% of flood claims, costing the program more than $12.5 billion (Pew 2016). The NFIP, intended to reduce flood losses, may instead unintentionally incentivize development in flood-prone areas (Sanchez et al. 2024), and Congress has required FEMA to allow many policyholders to pay discounted premiums that do not fully reflect flood risk — contributing to $36.5 billion in borrowing from Treasury since 2005 (GAO-23-105977). The self-reinforcing feedback is particularly insidious: protection draws development, and the more that is built behind a levee, the greater the potential damage if it fails in a flood beyond its design (Sanchez et al. 2024).
What’s been tried and why it hasn’t worked
Risk Rating 2.0 (phased in from October 2021) substantially improved the alignment of premiums with property-level flood risk, but statutory caps on annual rate increases mean GAO estimates it would take until 2037 for 95% of current policies to reach full-risk premiums, with an unfunded premium shortfall of roughly $27 billion (GAO-23-105977). Buyout programs exist but are extremely slow — NRDC's review of nearly 30 years of FEMA data found a median of more than 5 years from flood to buyout completion — already struggle to meet existing need, and are voluntary. Only about 24% of severe repetitive loss properties have seen action to mitigate their flood risk (NRDC analysis of FEMA data). FEMA flood maps are outdated (only one-third of U.S. rivers have been modelled by FEMA, and only one-quarter of those models have been updated in the past five years), the mapped high-risk areas do not account for changing development patterns (EESI 2026), and FEMA models are not mandated to account for climate change (Wing et al. 2022) — the maps often underestimate actual flood risk (Sanchez et al. 2024).
What would unlock progress
Mandatory disclosure of flood risk at property transaction (shifting information asymmetry). Prospective zoning that prevents new development in high-risk floodplains rather than trying to relocate existing development. Actuarially sound flood insurance with income-based assistance (separating the insurance pricing signal from affordability). Streamlined buyout programs for repetitive-loss properties. Dynamic flood mapping that incorporates climate projections and upstream development.
Entry points for student teams
A team could build a spatial analysis tool that overlays development permit data with flood risk models to quantify the rate at which floodplain development is increasing exposure in a specific metro area, making the "safe development paradox" visible to local planners. Alternatively, a team could design an improved buyout program workflow using digital tools to accelerate the currently 5-plus-year median process. Urban planning, GIS/spatial analysis, and policy design skills apply.
Genes
Structural cousins: stuck the same way, in other fields
Sources
Sanchez, Georgina M., et al. (2024), "The safe development paradox of the United States regulatory floodplain," PLOS ONE 19(12): e0311718, Wing, Oliver E. J., et al. (2022), "Inequitable patterns of US flood risk in the Anthropocene," Nature Climate Change, U.S. GAO (2023), "Flood Insurance: FEMA's New Rate-Setting Methodology Improves Actuarial Soundness but Highlights Need for Broader Program Reform," GAO-23-105977, The Pew Charitable Trusts (Oct 2016), "Repeatedly Flooded Properties Cost Billions," Accessed 2026-08-20. go to source 1 ↗ go to source 2 ↗ go to source 3 ↗ go to source 4 ↗
Working notes
Notes made while writing and checking this brief, shown as written. They use working shorthand.
This is a "problems of success" case in the "protection-encourages-exposure" sub-type (closely related to the wildfire suppression case). The mechanism is: successful protection (levees, insurance) reduces perceived risk, encouraging development that increases actual risk. When protection fails (levee breach, 500-year flood), losses are catastrophic because of the development the protection enabled. Structurally parallel to environment-wildfire-suppression-fuel-accumulation. The NFIP's $22.5 billion debt makes the fiscal cost of the paradox concrete.
Reconciliation 2026-08-20: The Source line misattributed the anchor paper: "The safe development paradox of the United States regulatory floodplain" (PLOS ONE 19(12): e0311718, Dec 2024) is by Sanchez et al. (NC State), not Wing et al. — the paper exists and does contain the 24% / 89,080 km² / 22%-of-2020–2060-growth figures (all confirmed against the article). The ~$32B annual-loss figure belongs to Wing et al. 2022, Nature Climate Change ($32.1B in 2020, +26% to $40.6B by 2050) — now cited separately. The levee claim ("increase inundation extent by 25% of protected area, depths up to 2 m") appears in neither paper and could not be sourced; replaced with the verified empirical finding from Heine & Pinter 2012 (Hydrological Processes 26(21), doi:10.1002/hyp.8261 — stage increases up to ~2.3 m upstream of/within leveed reaches). The repetitive-loss statistic ("SRL = 2.5% of policies, ~50% of claims") did not match Pew: corrected to Pew's 1% of insured properties / 25–30% of claims / >$12.5B (Pew infographic, Oct 2016). NFIP debt $22.5B confirmed current (EESI, May 2026, https://www.eesi.org/articles/view/the-national-flood-insurance-program-is-perpetually-underwater-are-there-bipartisan-solutions); $36.5B borrowed since 2005 confirmed in GAO-23-105977. The Risk Rating 2.0 sentence ("Congress has repeatedly delayed implementation") was wrong — RR2.0 is in effect; replaced with GAO's 2037/95%-of-policies and ~$27B shortfall findings. Buyout 5+ years and ~24%-of-SRLPs-mitigated re-anchored to NRDC ("Going Under," Weber & Moore 2019, https://www.nrdc.org/resources/going-under-long-wait-times-post-flood-buyouts-leave-homeowners-underwater; "Losing (More) Ground," https://www.nrdc.org/bio/anna-weber/losing-more-ground-updated-severe-repetitive-loss-data). Two figures could not be sourced anywhere and were removed: "uninsured flood losses $10–17B annually" and "$2.3B federal mitigation 2014–2018." The loose ASFPM org citation was dropped. All URLs on the Source line verified live 2026-08-20.
Reconciliation 2026-09-30: Round-6 read of the whole write-up against every source it names, by a first reader and then a second reader on the same day, who went over each repair and fetched every source again. Read in full: Sanchez et al. 2024 (PLOS ONE article text, https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0311718); Wing et al. 2022 (open-access article at https://www.nature.com/articles/s41558-021-01265-6); the Pew infographic PDF (https://www.pew.org/-/media/assets/2016/10/repeatedly_flooded_properties_cost_billions.pdf); the EESI article of May 19, 2026 (https://www.eesi.org/articles/view/the-national-flood-insurance-program-is-perpetually-underwater-are-there-bipartisan-solutions). GAO-23-105977 returned 403 from gao.gov to a script; the full report PDF was read from the Internet Archive copy of https://www.gao.gov/assets/gao-23-105977.pdf. Both NRDC pages returned 403 from nrdc.org; the Internet Archive copies of the "Going Under" landing page and full report PDF (Weber & Moore, September 12, 2019) and of the "Losing (More) Ground" post (January 9, 2024) were read in full. Heine & Pinter 2012 could not be read in full (Wiley blocks scripts; no open-access copy is listed by Crossref, OpenAlex or Semantic Scholar); its abstract and citation record were read from Crossref (doi:10.1002/hyp.8261: Hydrological Processes 26(21): 3225–3240). Changes: (1) "following the 1927 Mississippi flood, succeeding at protecting specific areas from routine flooding" and "Both succeeded at their stated objectives" are in none of the sources; removed. "Mapping flood risk" as part of the NFIP's design is kept: GAO lists "identifying flood risk" among the program's objectives and "flood insurance and floodplain mapping activities" as what it carries out, and Sanchez et al. say the NFIP "relies in part on FEMA flood maps". (2) "also called the 'levee effect'": Sanchez et al. say research on the paradox "has predominantly focused on the 'levee effect'" and their own study concerns the regulatory floodplain line ("policy effects"), not levees; the text now says the levee effect is the most-studied form, and the 24% / 89,080 km² / 22% figures are introduced as an assessment of development around the regulatory floodplain. "False sense of safety" now follows the source's "false senses of security". (3) "they transform frequent small losses into infrequent catastrophic ones" is not stated in any source; Sanchez et al. say development behind levees raises "the potential for greater damage if the levee fails during a flood event that exceeds its design"; the sentence now says that, keeping the source's "potential". (4) "instead subsidized floodplain development ... by offering premiums that do not fully reflect flood risk" was stronger than the sources and joined two claims neither source joins: Sanchez et al. say the NFIP "may unintentionally incentivize development in flood-prone areas"; GAO says Congress has required FEMA to allow many policyholders "discounted premiums that do not fully reflect their properties' flood risk", which "has contributed to" the $36.5 billion borrowed from Treasury since 2005; EESI adds that NFIP communities must adopt ordinances that minimize development in high-risk areas. The sentence now states the two claims separately at the sources' strength. (5) "political pressure demands maintaining and upgrading protection rather than retreat, locking in the exposure permanently" and "The structural lock-in makes retreat politically impossible once development exists" appear in no source; both removed, the first replaced by Sanchez et al.'s chain (protection draws development, which raises the potential damage if a levee fails in a flood beyond its design). (6) Buyouts "underfunded": the NRDC report says the programs "already struggle to meet existing need"; the text now says that. (7) "FEMA flood maps are outdated (many pre-date current development) ... systematically underestimates": the parenthetical is in no source; replaced with Wing et al.'s statement that only one-third of U.S. rivers have been modelled by FEMA and only one-quarter of those models updated in the past five years, EESI's statement that the mapped high-risk areas "do not account for changing development patterns", and Wing et al.'s statement that FEMA models are not mandated to account for climate change; "systematically" dropped because Sanchez et al. say the maps "often" underestimate risk. Confirmed against the sources and left as written: the Source-line citations for Sanchez et al. (PLOS ONE 19(12): e0311718, Dec 31, 2024), Wing et al. (Nature Climate Change, 2022), GAO-23-105977 (July 2023, title as given) and Pew (Oct 2016); 89,080 km², 24% of developed land in the study system (2,330 counties of the conterminous U.S.), within 250 m of the 100-year floodplain as of 2019, and about 22% of 2020–2060 growth in that zone; Heine & Pinter's abrupt, statistically significant stage increases up to 2.3 m upstream of and within leveed reaches (abstract); the $22.5 billion NFIP debt (EESI, May 2026); $32.1 billion in average annual losses in 2020, rising 26% (26.4%) to $40.6 billion by 2050, borne disproportionately by poorer communities (Wing et al.); 1% of NFIP-insured properties, 25–30% of claims, more than $12.5 billion (Pew, which calls them "repeatedly flooded properties"); $36.5 billion borrowed from Treasury since 2005, Risk Rating 2.0 from October 2021 ("substantially improves ratemaking"), 2037 for 95 percent of policies, and the $27 billion shortfall (GAO); the median of more than 5 years from nearly 30 years of FEMA data and "voluntary buyouts" (NRDC report), and 24% of severe repetitive loss properties mitigated (NRDC, 2024). Not checked: Heine & Pinter's body text; the "What Would Unlock Progress" section is the write-up's own proposals. No title change. Genome tags untouched.