2.5% of NFIP Policies, 48% of Claim Dollars: The Repetitive-Loss Imbalance

Congress built the National Flood Insurance Program (NFIP) to do two things at once: keep flood coverage available and affordable for property owners who need it, and keep the federal balance sheet from eating every large flood. Those goals collide hardest on repetitive-loss (RL) properties — structures that file paid flood claims again and again. The March 2026 GAO testimony that updated FEMA’s own counts puts the collision in a single pair of percentages: unmitigated RL properties are about 2.5% of NFIP policies in force, while properties with two or more losses have drawn 48% of cumulative NFIP claim dollars.

That is the book question this piece answers. Not “where does it flood,” and not “what should premiums be,” but what share of the policy book versus the paid-loss book sits in the repeatedly damaged slice. The interactive dashboard above walks the policy-versus-claim split, the growth of the unmitigated RL stock, the decade-scale widening of the imbalance, FEMA’s mitigation method mix, and a state inventory of multiple-loss and severe repetitive-loss (SRL) properties.

What “repetitive loss” means in FEMA’s vocabulary

FEMA tracks more than one definition, and the distinctions matter for reading any table. NFIP repetitive loss generally means an insured structure with two or more claim payments of at least $1,000 in any rolling ten-year window. Flood Mitigation Assistance (FMA) repetitive loss uses a damage-to-value threshold (repairs averaging at least 25% of structure value on two occasions, with Increased Cost of Compliance coverage in force on the second event). Severe repetitive loss is the hardest-hit tier: four or more claims above $5,000 that cumulatively exceed $20,000, or two building-only claim payments that exceed the building’s fair market value, with timing rules that prevent double-counting near-simultaneous claims.

GAO’s headline 2.5% / 48% pair mixes those lenses carefully. The policy share refers to unmitigated RL properties as a fraction of policies in force. The claim-dollar share refers to cumulative payments to properties that have experienced two or more losses — a multi-loss history lens that is closely related to, but not identical with, the live RL policy count. Treat them as the best public book-level comparison FEMA and GAO have put in front of Congress, not as a perfect one-to-one mapping of today’s insured RL roster to every historical dollar.

The book split: 2.5% of policies, 48% of claim dollars

Source: GAO-26-109045 (March 2026), citing FEMA. Policies in force: 4.58 million as of December 2025. RL properties still insured by NFIP: 112,640 as of January 2026 (≈2.5% of that book). Claim-dollar share (48%) is cumulative through December 2021 for properties with two or more losses.

The arithmetic overweight is brutal. If a segment holds 2.5% of policies and 48% of claim dollars, it is absorbing roughly nineteen times as much paid loss as a proportional book share would imply. The rest of the program — nearly every other policy — still pays a majority of claim dollars in absolute terms (52%), but it does so from 97.5% of the policy base. Cross-subsidy language is often vague; this table is not.

Two caveats keep the comparison honest. First, multi-loss properties also tend to face higher full-risk premiums under Risk Rating 2.0, so they are not free riders in every sense — they are still a concentrated claim engine. Second, cumulative claim dollars through 2021 embed the Gulf Coast mega-events of the mid-2000s and later storms; the share is a stock of history, not a single-year loss ratio.

The unmitigated stock is still growing

Mitigation has not kept pace with new RL designations. GAO’s updated stock series shows unmitigated RL properties rising from about 131,000 in 2009 to 176,000 in 2018 and 227,000 in 2026 — roughly a 73% increase from the 2009 baseline to the 2018 print alone, and continued growth into 2026. Over the same long window FEMA mitigated only 12,972 repetitive-loss properties across its grant programs (1989–2025). Not every mitigated structure was NFIP-insured, and not every unmitigated RL structure still carries a policy, which is why the insured RL count (112,640 in January 2026) sits well below the broader unmitigated stock (227,000).

The gap between stock growth and mitigation throughput is the operational version of the fiscal problem. Every year the program can acquire, elevate, relocate, or floodproof only a thin slice of the repeatedly damaged inventory, while storms and mapping updates keep minting new RL designations.

How the imbalance widened since the older GAO framing

A decade and a half ago, GAO’s shorthand was already alarming but smaller: RL properties were about 1% of policies and 25–30% of claims (GAO-10-1063T). The 2021–2026 FEMA/GAO framing — 2.5% of policies, 48% of claim dollars — is not a perfect apples-to-apples replacement (definitions and windows differ), but the direction is unambiguous. The repeatedly damaged slice of the book has become both a larger policy footprint and a much larger claim-dollar footprint.

That widening matters for reauthorization politics. NFIP’s current authorization is set to expire on September 30, 2026. Debates about rate caps, means-tested assistance, and mitigation budgets all orbit the same fiscal fact: a small inventory drives a large share of paid losses, and the inventory is not shrinking.

Mitigation exists — and mostly buys houses out

When FEMA does mitigate flood risk at the property level, it leans heavily on acquisition. From fiscal years 1989 through 2025, FEMA mitigated 95,762 properties across its hazard-mitigation assistance programs. Acquisitions accounted for 69,415 of them (≈72.5%). Elevation accounted for another 22,039 (≈23%). Relocation and floodproofing fill the remainder. About 77% of mitigated properties were funded through the Hazard Mitigation Grant Program (HMGP), with Flood Mitigation Assistance and the Building Resilient Infrastructure and Communities (BRIC) lineage covering most of the rest — noting that BRIC’s 2025 termination fight and subsequent court order sit outside the RL-share question but shape future capacity.

GAO’s earlier cost benchmarks (2008–2014 FEMA data) put average federal cost near $136,000 per acquisition and $107,000 per elevation. Acquisition eliminates residual flood risk by converting the parcel to open space in perpetuity; elevation reduces risk but leaves the structure in place. Both are expensive relative to annual premiums, and both face the process frictions GAO catalogs: multi-year timelines, local capacity limits, nonfederal cost shares, and voluntary-participation “checkerboarding.”

For the book question, the mitigation mix is a reminder that the federal answer to RL concentration has mostly been buyouts, not rate engineering alone — and buyouts have not scaled to the stock.

Where the multiple-loss inventory concentrates geographically

OpenFEMA’s NFIP Multiple Loss Properties inventory — and public compilations drawn from it — puts the national multiple-loss count near 237,000 properties, with roughly 35,000 classified as severe repetitive loss. The top of the state list is familiar Gulf and Atlantic flood geography: Texas, Louisiana, and Florida lead multiple-loss counts, followed by New York and New Jersey, then North Carolina, Pennsylvania, Virginia, Mississippi, and Alabama. SRL shares within those state inventories vary; New Jersey’s SRL fraction of its multiple-loss stock is notably higher than New York’s in the compiled ranks.

These state ranks are inventory counts, not policy-share percentages. A state can lead the multiple-loss list because it has a large insured coastal book, a deep historical claim file, or both. Still, for congressional district politics and FMA targeting, the concentration is the point: the national 2.5% / 48% imbalance is not evenly sprinkled across the map.

Risk Rating 2.0, surcharges, and the premium shortfall

In April 2023 FEMA implemented a pricing approach that more closely aligns premiums with risk. RL policies can face an additional surcharge after more than one loss on or after April 1, 2023 — GAO’s South Carolina single-family example shows a base premium of $1,507 picking up a $482 surcharge after a second claim and $964 after a third. That is a direct price signal aimed at the multi-loss book.

Price signals alone do not close the fiscal gap quickly. As of December 2022, GAO estimated that 66% of NFIP policies still carried discounted premiums, that the 2023 premium shortfall was about $2.7 billion, and that the path to 95% of policies at full-risk premiums stretched toward 2037 under statutory annual increase caps — with a multi-year shortfall in the mid-twenties of billions of dollars. Since 2005, FEMA has borrowed about $38.5 billion from Treasury to pay NFIP claims. Flooding remained the most expensive U.S. natural-disaster type in FEMA’s 2024 damage accounting (over $8 billion that year alone).

GAO’s policy recommendation set pairs mitigation targeting — aim scarce buyout and elevation dollars at properties driving the premium shortfall, which disproportionately include RL — with congressional options such as means-tested assistance instead of blunt rate caps. As of March 2026, FEMA had implemented four of nine GAO recommendations from the related body of work; Congress had not enacted the means-tested assistance redesign.

Caveats and reading rules

A few limits should travel with every share cited here. Definition drift: RL, FMA-RL, and SRL are not interchangeable, and the 48% claim-dollar figure is a multi-loss cumulative, not a live RL-policy loss ratio for calendar 2025. Vintage mismatch: policies in force (Dec 2025) and insured RL counts (Jan 2026) are newer than the December 2021 claim-dollar cumulative. Mitigation incompleteness: FEMA’s property-mitigation totals include non-RL structures; RL-specific mitigations (12,972) are a subset. State inventories compiled from OpenFEMA are labeled estimated in our dashboard where GAO does not publish a matching policy-share cut. Affordability: concentrating mitigation on the highest-risk, highest-shortfall properties can improve actuarial soundness and, over time, affordability for remaining owners — but acquisition removes households from neighborhoods, and elevation does not erase residual risk.

None of those caveats erase the headline. A thin slice of the NFIP book still sits on a thick slice of the paid-loss history. Until mitigation throughput or rate design changes that arithmetic, repetitive-loss properties will remain the clearest single window into NFIP’s fiscal exposure.

Sources

• [GAO-26-109045Flood Risk Mitigation (March 2026)](https://www.gao.gov/products/gao-26-109045)
• [GAO-23-105977FEMA rate-setting and NFIP fiscal exposure](https://www.gao.gov/products/gao-23-105977)
• [FEMA OpenFEMANFIP Multiple Loss Properties v1](https://www.fema.gov/openfema-data-page/nfip-multiple-loss-properties-v1)
• [GAO-10-1063Tearlier RL policy/claims framing](https://www.gao.gov/assets/gao-10-1063t.pdf)