Loading reports … Fetching the current index of long-form research.
Reports
Loading reports … Fetching the current index of long-form research.
Climate risk reaches balance sheets through insurance before it arrives as a disaster.
Insurance is where climate risk reaches balance sheets first. The premium reprices the mortgage payment before any disaster, and where private coverage is withdrawn the risk moves to residual plans, lenders, and public budgets, so the map of withdrawal is an early-warning map of housing, credit, and municipal stress. That map measures insurability, which combines hazard, exposure, rebuilding cost, and legal regime, and the channel works whichever of them dominates.
The insurance line sat inside the escrow payment for decades without changing anyone’s decision. It now changes decisions, and it does so before the house has burned, flooded, or lost a roof.
Insurance turns a risky structure into collateral a lender will finance. It protects the homeowner and the lender, and for decades its cost was small and steady enough to sit unnoticed inside the escrow line. The renewal notice has become the first climate bill many households see, and it arrives every year whether or not the house has suffered a loss.
A March 2026 Dallas Fed Economics article by Shan Ge, Stephanie Johnson, and Nitzan Tzur-Ilan measures the shift. Homeowners insurance premiums rose about 70 percent nationally from 2019 to 2025. Insurance took 14 percent of the average homeowner's monthly mortgage payment, principal and interest included, in 2025, against 10 percent in 2013.1 A household that qualifies on principal and interest can still fail on taxes and insurance, and in the markets where the premium has risen fastest the escrow account is where climate risk enters the household budget.
Treasury's Federal Insurance Office found the same direction in the policy data that preceded the recent repricing: average premiums per policy rose 8.7 percent faster than inflation from 2018 to 2022.2 The price rise is therefore older and broader than one hard insurance cycle.
The Dallas Fed authors used ICE McDash mortgage-level data for 2015 to 2023, which records the insurance payment itself and covers roughly two-thirds of the U.S. mortgage market, and linked it to credit reports and relocation records. What they report separates measured estimates from a projection.
A $1,000 increase in the annual premium raises the probability that a household moves by 0.54 percentage point. Premium increases pushed roughly 31,000 mortgages into delinquency in 2022. A third figure, about 203,000 additional delinquencies a year from 2025 to 2055, is of a different kind: it is what the authors' model produces if premiums follow First Street Foundation's projected 29.4 percent rise by 2055, and this report treats it as a scenario to test against.3
The transmission runs in four steps. Hazard and rebuilding cost set the expected loss an insurer prices. The premium sets part of the housing payment. The housing payment sets delinquency and relocation risk. Delinquency and relocation, repeated across a county, set the demand for its housing. Financially secure households can switch carriers, raise deductibles, retrofit, or move. Constrained households absorb the premium, cut other spending, borrow, or fall behind, and over time the sorting can begin to follow insurability as well as preference.
Catastrophe risk used to mean the peak event. In 2025 the losses that kept insured totals above $100 billion came from floods, convective storms, and wildfires.
Munich Re reports that natural disasters caused about $224 billion of global losses in 2025, and insurers covered about $108 billion. Weather accounted for 92 percent of all losses and 97 percent of insured losses, and no hurricane struck the U.S. mainland for the first time in ten years. Floods, severe convective storms, and wildfires produced $166 billion of total losses and about $98 billion of insured losses, against 10-year averages of $136 billion and $60 billion.4 Swiss Re's independent count puts 2025 global insured catastrophe losses at $107 billion, with secondary perils a record 92 percent of them.5
The United States carried most of it. Munich Re puts North American losses at $133 billion, of which about $93 billion was insured. U.S. severe thunderstorms caused $56 billion of losses, $42 billion insured, against 10-year averages of $39 billion and $29 billion. The Los Angeles wildfires of January 2025 caused about $53 billion of losses and about $40 billion of insured losses, the costliest wildfire disaster on record.6
Secondary peril began as a catastrophe-modeling category and is now a balance-sheet one. Hail, convective storm, wildfire, and flood losses are frequent, spread across wide areas, and fall on high-value roofs, vehicles, and structures, so they keep a loss year elevated without a named storm making landfall. A primary insurer can carry a peak event through its reinsurance tower. It carries frequent losses closer to its own capital, and it reprices the retail book to match.
The global protection gap moved the other way in 2025, and the reason supports the thesis. Swiss Re reports that about 49 percent of 2025 economic losses were insured, the highest share in its records, and Munich Re puts the uninsured share near 50 percent against a 10-year average near 60 percent. The share rose because the year's losses fell heavily in the United States, where insurance penetration is high. Swiss Re estimates that 80 to 90 percent of catastrophe losses in emerging economies typically go uninsured.7
A falling global gap and a rising U.S. premium are two readings of one fact. The insured share is high where an insurance market still writes the risk, and that market charges for the privilege. The question this report follows is what happens at the margin where it stops writing.
Primary insurers buy catastrophe protection from reinsurers and capital-market investors, so the household premium is the last link in a wholesale chain. At the January 2026 renewals Guy Carpenter's U.S. property-catastrophe rate-on-line index fell 12 percent, while attachment points held where the hard market of recent years had set them.8 Cheaper reinsurance above an unchanged attachment eases the cost of the peak layer and leaves the frequent losses with the primary carrier, which is the layer secondary perils hit.
Swiss Re attributes more than 80 percent of the long-run rise in weather-related insured losses, 1970 to 2025, to exposure growth: more people, higher asset values, and higher reconstruction costs in exposed places.9 Hazard intensification adds to that, most clearly for North American wildfire. For U.S. premiums the split has been measured. Benjamin Keys's NBER Reporter summary of his work with Philip Mulder on mortgage escrow records reports that average nominal premiums rose 33 percent from 2020 to 2023, that higher construction costs explain slightly over half of that rise, and that disaster risk explains nearly 20 percent; a one-standard-deviation increase in disaster risk carried a $425 higher premium in 2023, up from $250 in 2018.10 Climate is therefore the minority input to the national premium and the growing one in exposed places. The split narrows what this report claims. Its subject is the insurance channel, through which every driver of the premium reaches the household on the same renewal notice, and climate is the driver whose share is rising where exposure is concentrated.
A coverage map needs a definition of the line it draws. This report draws it where the private market will no longer write a property at a price its mortgage payment can carry.
A property is uninsurable in this report's sense when its only available coverage comes from a residual plan or a public program, or when the private premium exceeds what the household's payment can carry. Three public readings track that line: insurer nonrenewal rates, the size and growth of residual plans, and the share of the payment the premium takes. A rising price shows a market still willing to write the risk. Nonrenewals and residual-plan growth show carriers declining to write it.
Treasury's Federal Insurance Office analyzed more than 246 million policies written by more than 330 insurers from 2018 through 2022. Homeowners in the 20 percent of ZIP Codes with the highest expected annual losses paid $2,321 a year on average, 82 percent more than those in the lowest-risk fifth. Their nonrenewal rates were about 80 percent higher, and their claims averaged about $24,000 against about $19,000.11
Treasury's data end in 2022, before most of the repricing in Part I. The most recent multi-state nonrenewal tabulation public at the June 2026 cutoff comes from the Senate Budget Committee's majority staff, which collected nonrenewal data for 2018 to 2023 from 23 insurers holding about 65 percent of the national homeowners market. Florida's statewide nonrenewal rate rose from 0.79 percent to 2.99 percent over those six years, and Louisiana's from 0.49 to 1.80 percent.12 At county level the rise is sharper, and it is concentrated where wildfire and coastal exposure meet development.
Two cautions travel with these data. They are a majority-staff analysis of carrier submissions, and the National Association of Insurance Commissioners warned that the submissions may contain inconsistencies. Industry economists replied that Florida and California drive the national picture. The county pattern answers the second point in part: Barnstable County, Massachusetts, and Jackson and Harrison Counties, Mississippi, sit beside six California counties and one in Florida.
California's FAIR Plan was built as an insurer of last resort. In the year to September 2025 it wrote 268,379 new policies, an average of 22,365 a month, 26 percent more than in the prior year. By September 2025 it held 645,987 policies, $696 billion of exposure, and $1.93 billion of written premium, up 169, 317, and 335 percent from September 2021.13 Its count kept rising after that: the Assembly Insurance Committee's background paper for its January 2026 oversight hearing puts the plan at 668,609 policies and $724 billion of exposure in December 2025. Both counts combine dwelling and commercial policies.14
The plan's growth is the private market's withdrawal recorded in reverse. State Farm General stopped accepting new applications for all business and personal property and casualty lines in California on 27 May 2023, citing construction costs, catastrophe exposure, and the reinsurance market.15 Households that cannot find a standard policy move to the FAIR Plan, and the plan then holds risks the private market has already screened out. The Los Angeles fires showed what that concentration costs: about $40 billion of insured losses in a single urban-edge event, a loss the size of a large hurricane.
Florida’s insurance story is told as a comeback and California’s as a crisis. Both answer the same question, which is who holds the risk when private carriers retreat.
Florida's Citizens Property Insurance Corporation peaked at 1.42 million policies in October 2023. More than 546,000 policies moved from Citizens to private insurers in 2025, and by 31 May 2026 the count stood at 293,772, about 79 percent below the peak. Citizens' board recommended its first average personal-lines rate cut since 2015 in December 2025, and the 2026 multiperil rates fell an average 8.8 percent.16
Florida's reversal is the strongest objection to reading the coverage map as a climate map. Florida's 2022 and 2023 reforms cut the litigation costs that had driven its losses, and private capital returned within two years. The map does move with legal regime, and the thesis accepts that. The channel is unchanged: Florida's risk moved back to private carriers through the same door it had left by, and the depopulated book now faces its first hurricane seasons in private hands. Carriers rated by Demotech, often small and single-state, have taken many of the Citizens policies, and Sastry, Sen, and Tenekedjieva find that such insurers failed far more often than their ratings implied and that their policyholders were more likely to default on mortgages after a disaster.17 Whether the depopulated book holds is therefore a test still to come.
California supplies the regulatory reading of the same evidence. Carriers have long argued that the state's prior-approval rate regime kept forward-looking catastrophe models out of homeowners rates, so that rates lagged the wildfire risk and carriers withdrew instead of repricing. The December 2024 regulation that admits catastrophe models into rate filings is the state's answer to that argument.18 On that reading the map measures regulation as much as hazard, and the thesis takes the point. Regulation decides whether rising expected loss shows up as price, as it did in Florida after 2023, or as withdrawal, as it did in California before 2025. In both states the cost reached a balance sheet. Legal regime decides how much of expected loss private carriers will carry at a given time, and expected loss decides where the market settles over a decade.
The National Flood Insurance Program exists because private carriers would not write flood risk at a scale that kept mortgages available in flood-prone places. It had 4.58 million policies in force in December 2025, and 112,640 of them, about 2.5 percent, covered repetitive-loss properties in January 2026.19 Its debt to the Treasury is $22.525 billion against a statutory borrowing limit of $30.425 billion, leaving $7.9 billion of room.20
A correlated flood year that used the remaining $7.9 billion would turn into a congressional appropriation, and the program's authorization has been extended 35 times in the short term since the end of fiscal 2017.21 The flood template shows the end state of a withdrawal. The risk stays financeable because a public balance sheet carries it, and the price signal that would slow development in exposed places arrives weaker than the loss.
The losses a private carrier sheds stay in the economy. They move to whoever holds the property, the loan, the tax base, or the backstop.
When private insurers cut exposure, households carry more risk through higher premiums and thinner coverage, lenders carry more collateral risk, local governments carry more tax-base risk, and public programs carry more disaster risk. Insurance is the plumbing that lets property risk become credit, and where it narrows, credit narrows with it. The European Central Bank and EIOPA estimated in 2023 that only about a quarter of climate-related catastrophe losses in the European Union were insured, which shows the end state of a gap that has settled.22
The clocks in Exhibit 4 run at different speeds, and that is what makes the reallocation dangerous. Premiums reset at each renewal and reinsurance at each January, property assessments adjust over years, and public budgets usually react after the disaster. The first absorber to show strain is the household, and the last is the tax base. The link from premium to property value has been measured: the same NBER Reporter summary reports that the reinsurance price shock alone reduced 2023 home values by an average of $8,400, and in Florida's sea-level-exposed markets sales volumes began to diverge in 2013 while sellers cut list prices only slowly and unsold inventory built up.23 Volume moves first, then price, then the assessment roll. The FAIR Plan's 2025 recovery assessment shows the fiscal stage already reaching policyholders who never bought a FAIR policy: the Insurance Commissioner approved a $1 billion assessment on the plan's member insurers to keep paying Los Angeles fire claims, and allowed insurers to recover part of it from their own policyholders through a temporary supplemental fee.24
Exhibit 4. Insurance retreat reallocates climate loss without reducing it, and each absorber has its own clock
Who carries what private insurers shed · Power Curve synthesis
| What it absorbs | Where it surfaces | Expected lag | |
|---|---|---|---|
| Households | Higher premiums, larger deductibles, narrower coverage. | Delinquency, relocation, credit-card balances. | One renewal |
| Lenders | Collateral whose insurability is deteriorating. | Delinquency and refinancing risk in high-premium ZIP Codes. | One to three years |
| Local governments | Assessment values, transaction volumes, development economics. | Property-tax base, then ratings and borrowing costs. | Three to ten years |
| Residual markets | Risks the private market has already screened out. | Exposure growth, then assessments on remaining policyholders. | One to three years |
| Public programs | The tail the private market will not price. | Program debt, disaster appropriations, reauthorization fights. | After the event |
Source: Power Curve analysis of the Treasury, Dallas Fed, Senate Budget Committee, California FAIR Plan, Citizens, GAO, CRS, and ECB/EIOPA evidence cited in this report. The lags are reasoned estimates from the renewal, assessment, and budget cycles and have not been measured.
Availability adds what price cannot show. A mortgage lender requires hazard insurance, so a property that cannot be insured at all cannot carry a conventional mortgage, whatever its price. A nonrenewal is recorded at the renewal date, months or years before the same stress can appear in a delinquency, an assessed value, or a rating action, and that ordering follows from the cycles that produce each reading. No public series yet measures the lag county by county, which is why this report calls the withdrawal map an early-warning map and gives the lags in Exhibit 4 as estimates. A nonrenewal also reflects past losses, and that does not make it a lagging indicator for the absorbers, because what matters for a lender or a city is whether the nonrenewal arrives before the delinquency and the assessment, and the renewal, assessment, and rating cycles put it first. That ordering is a design argument and has not been measured. The closest measured pairing runs through insurer failure: households insured by fragile carriers that later failed were more likely to default on their mortgages after a disaster. A price increase is visible on the renewal notice in the year it happens. A withdrawal compounds: an insurer that stops writing new business in a county lets its existing book run off through nonrenewals, the residual plan grows, and private capacity shrinks for years before a municipal analyst sees it in the tax roll. Monitoring nonrenewals and residual-plan growth is how a lender or a city sees the stress while it can still be managed.
Losses are rising, and the harder test is whether risk can be reduced faster than its price rises. Premiums reprice at each renewal while building stock and land use change over decades, and mitigation that would lower a property's expected loss can take years of permits, contractors, standards, and capital. The best outcome is insurable resilience: properties and communities that attract private capital because the underlying risk has been reduced, priced, and disclosed. Coverage that protects growing exposure without reducing risk only enlarges the problem the next renewal inherits.
That is the test the watchlist applies county by county. Where nonrenewals fall back while premiums stay high, the risk is being priced and carried, and the household pays for it in the open. Where nonrenewals keep rising and the residual plan keeps growing, the risk is being shed, and the absorbers in Exhibit 4 inherit it on their own clocks.
Each decision below names who acts, the reading that should prompt action, and what acting costs. The order runs from the household outward to the public balance sheet.
Mortgage underwriters. Treat the premium as a collateral variable. Underwrite to the payment including insurance at the next renewal, with a stress for a further rise, in any county where the 2023 nonrenewal rate exceeds the state rate. The cost is tighter qualification in exposed places, which is the risk showing up on time.
Bank and housing-finance supervisors. Add nonrenewal rates, residual-plan share, and premium share of payment to collateral reviews in high-risk ZIP Codes. Rising residual-plan dependence in a lender's footprint is the reading that should prompt a closer look, because it moves before delinquency does.
Municipal finance officers and rating analysts. Read residual-plan growth in a jurisdiction as a forward indicator of the property-tax base. Where FAIR Plan or Citizens share of local policies rises for two consecutive years, test the revenue forecast against slower transactions and flatter assessments before they appear.
State insurance regulators. Florida shows that legal and rating reform can return private capital within two years, and California shows that a residual plan can quadruple its exposure in four. Reform buys time and capacity. It holds only where expected loss is priced, so pair any rate relief with mitigation commitments that lower the loss being priced.
Adaptation funders. Fund the projects that make a property insurable at a durable price, and use that as the test of a project. Insurable resilience is a higher bar than coverage access, and it is the one a private carrier will lend against.
Reinsurers and insurance-linked investors. The 2026 softening added capacity above unchanged attachments. If attachments fall while secondary-peril losses stay near the 2025 level, the wholesale market is pricing the frequent layer below its loss experience, and retail availability will reverse again when it corrects.
Housing and economic-development officials. Risk-based pricing is economically coherent and falls hardest on lower-income households, which face the same signal with fewer options. Means-tested mitigation support keeps the price signal intact while widening the set of households that can act on it.
The thesis is that insurance is where climate risk reaches balance sheets first, and that the map of coverage withdrawal is an early-warning map of housing, credit, and municipal stress. It breaks on any of four readings, each observable in public data. It breaks if residual plans shrink across several states for reasons beyond one-off legal change, with California FAIR Plan policies flat or falling through an active wildfire season. It breaks if private carriers re-enter high-nonrenewal counties at adequate rates and hold those books through a major loss year, as Florida's depopulated book now must. It breaks if premium growth stops reaching mortgage performance: a Dallas Fed or similar reading that finds no delinquency effect from premium increases after 2022 would remove the first link of the chain. And it breaks if adaptation and code changes lower expected losses fast enough that nonrenewal rates in the counties of Exhibit 2 fall back toward their 2018 levels. The watchlist that follows carries each of those readings.
| Variable | What a move would mean | Reading | Falsifying trigger | Horizon |
|---|---|---|---|---|
| California FAIR Plan policies in force | A rising count confirms that private retreat continues in the state with the largest residual plan | 645,987 (Sep 2025), +39% year on year | Flat or falling for two quarters through an active fire season | Next FAIR Plan policy statement |
| Florida Citizens policy count | A count held low through a major storm shows Florida’s reversal is durable and limits the thesis to markets without reform | 293,772 (31 May 2026) | Stays below 400,000 through a landfalling major hurricane | Q1 2027 |
| Nonrenewal rates in exposed counties | Falling rates would mean carriers are writing the risk again at a price the market accepts | Florida 2.99% statewide; Lake County, CA 7.56% (2023) | Florida back near its 2018 rate of 0.79% | Next federal or NAIC data call |
| U.S. property-catastrophe reinsurance | Lower attachments would pass wholesale relief to the frequent-loss layer primaries now retain | Rate on line −12% at 1 Jan 2026; attachments unchanged | Attachments lowered at two consecutive January renewals | 1 Jan 2027 |
| Insurance share of the mortgage payment | A falling share would ease the payment squeeze in Part I | 14% (2025) | Back below 12% nationally | Next Dallas Fed or ICE reading |
| NFIP borrowing room and authorization | Long-term reform would show the public backstop can price and mitigate risk as well as absorb it | $7.9bn of room; short-term extension to 11 Dec 2026 | Long-term reauthorization with repetitive-loss mitigation at scale | 11 Dec 2026 |
Each falsifying trigger is set in the direction of the reading that would weaken the thesis. Readings as published by June 2026, except the NFIP row.
Source: California FAIR Plan Association; Citizens Property Insurance Corporation; U.S. Senate Committee on the Budget; Guy Carpenter, as reported by Artemis and Reinsurance News; Federal Reserve Bank of Dallas; Congressional Research Service. Data through 31 May 2026.
The rule that follows from the six readings is simple to apply. Where a county's nonrenewal rate runs above its state's and its residual-plan share is rising, a lender should underwrite to the premium expected at the next renewal, a rating analyst should test the tax base against slower transactions, and a regulator should ask whether rates are lagging the risk. Where both readings fall back, the risk is being carried at a price, and the map has stopped warning.
This report was published on 12 October 2026 on data through June 2026. The readings below are as of that cutoff, except the NFIP authorization date, which is updated to the extension in force at publication.
California finalized two regulations in December 2024 that let insurers use catastrophe models in rate filings and price in the net cost of reinsurance, on condition that major carriers write in wildfire-distressed areas at no less than 85 percent of their statewide market share. The Department of Insurance completed its first model review in mid-2025, and the filings that follow will show whether carriers accept the bargain at scale.25 Florida enters the 2026 hurricane season with the smallest Citizens book in years, and whether private carriers hold it through a major storm is the open test of Part IV. The NFIP was authorized to 30 September 2026 at the cutoff and now runs on a short-term extension to 11 December 2026; long-term reauthorization remains stalled in Congress.26
Every figure traces to a named, dated source published before the June 2026 cutoff, and each was checked against that source before this edition went to press. Catastrophe losses are Munich Re's inflation-adjusted figures from its January 2026 release, so the 2025, 2024, and 10-year readings in Exhibit 1 and Table A1 share one basis; Swiss Re's independent count is cited beside them where it differs. Dallas Fed readings are as published: the 70 percent and 14 percent figures are national readings, the 0.54 point and 31,000 figures are model estimates, and the 203,000 figure is a projection conditional on First Street Foundation's premium path. Treasury readings are quintile averages across more than 246 million policies for 2018 to 2022, which predate most of the repricing and so understate the current spread. Nonrenewal rates are the Senate Budget Committee majority staff's tabulation of 23 insurers' submissions for 2018 to 2023, which the NAIC cautioned may contain inconsistencies. Residual-market figures are each program's own published counts over its own window, so the California and Florida windows differ.
This edition replaces three plates. The premium-index line and the mortgage-share bars of the previous edition carried intermediate years that no published source states, and the protection-gap series mixed vintages; each is replaced by the published readings alone. The previous edition's California FAIR Plan figures for March 2026 were not found at a primary source and are replaced by the plan's own September 2025 statement. The lags in Exhibit 4 are reasoned estimates. The first edition carried seven exhibits and this one carries four; the figures of those cut remain in the text and in Tables A1 and A2.
| Dataset | Source | Frequency | Coverage | Retrieved | Used in |
|---|---|---|---|---|---|
| Premiums, payment share, delinquency estimates | Federal Reserve Bank of Dallas; ICE McDash | One-off | 2013 to 2025 | 24 Mar 2026 | Part I |
| Global and U.S. catastrophe losses | Munich Re NatCatSERVICE | Annual | 2024, 2025, 10-year average | 13 Jan 2026 | Exhibit 1, Table A1 |
| Insured catastrophe losses, drivers | Swiss Re Institute sigma | Annual | 1970 to 2025 | 19 Mar 2026 | Part II |
| Premiums, nonrenewal, severity by ZIP quintile | Treasury Federal Insurance Office | One-off | 2018 to 2022 | 16 Jan 2025 | Part III |
| Nonrenewal rates by state and county | Senate Budget Committee majority staff | One-off | 2018 to 2023 | 18 Dec 2024 | Exhibit 2, Table A3 |
| FAIR Plan exposure, policies, premium | California FAIR Plan Association | Quarterly | Sep 2021 to Sep 2025 | Sep 2025 | Exhibit 3, Table A2 |
| Citizens policies in force | Citizens Property Insurance Corp. | Monthly | Oct 2023 to May 2026 | 31 May 2026 | Exhibit 3, Table A2 |
| NFIP debt, policies, repetitive loss | CRS; GAO | One-off | Dec 2025 to Mar 2026 | 25 Mar 2026 | Part IV |
| Property-catastrophe reinsurance pricing | Guy Carpenter, via Artemis | Renewal | 1 Jan 2026 | Jan 2026 | Part II, watchlist |
Every headline figure is the reading published before the June 2026 cutoff. Munich Re restates prior years in current dollars each January, so its 2024 total of $368 billion is the January 2026 restatement. Citizens publishes a count each month; the 31 May 2026 count is the last one inside the cutoff.
Insured share is insured losses divided by total losses, rounded. Munich Re rounds its insured figures as “about”.
| Period | Total losses | Insured losses | Insured share |
|---|---|---|---|
| 2025 | 224 | 108 | 48% |
| 2024 | 368 | 147 | 40% |
| 10-year average | 266 | 107 | 40% |
| Non-peak perils, 2025 | 166 | 98 | 59% |
| Non-peak perils, 10-year average | 136 | 60 | 44% |
| Tropical cyclones, 2025 | 37 | 6 | 16% |
| Tropical cyclones, 10-year average | 106 | 42 | 40% |
| U.S. severe thunderstorms, 2025 | 56 | 42 | 75% |
| U.S. severe thunderstorms, 10-year average | 39 | 29 | 74% |
Source: Munich Re, “Natural disaster figures 2025,” 13 January 2026. Data through December 2025. Behind Exhibit 1 and the protection-gap readings in Part II.
Hart, Brendan. 2026. “The Uninsurable Map.” The Power Curve, October. https://brendanhart.com/reports/uninsurable-map.
