Insurance by BogdanovaThe Encyclopedia of Insurance and Risk Science

RESEARCH DOSSIER

Property Insurance Market Nonrenewal and Geographic Retreat

Abstract. Property insurance market nonrenewal and geographic retreat describe a consequential reallocation of underwriting capacity across places, hazards, buildings, and policyholders. This monograph examines the contractual, actuarial, financial, regulatory, and socioeconomic processes through which existing property policies cease to be renewed, insurers restrict new business, admitted-market capacity declines, and residual or alternative markets absorb displaced exposure. The analysis develops a rigorous vocabulary of observation, distinguishes supply restriction from unaffordable offers and unmet demand, reconstructs the economics of catastrophe accumulation and reinsurance, and compares evidence from the United States, Australia, the United Kingdom, and continental Europe. Primary evidence includes U.S. federal and state market data, insurer-of-last-resort administration, insurance supervision, statutory protections, and official program accounts; the research literature supplies carefully qualified identification strategies and findings on housing, mortgage credit, and migration. Explicit measurement equations, illustrative calculations, data-quality diagnostics, and competing explanations prevent a rise in premiums or one insurer's withdrawal from being conflated with aggregate geographic abandonment. The central finding is that retreat is a multidimensional change in the availability, continuity, price, coverage, and institutional location of protection. Its policy significance lies in the distribution of catastrophe risk among insurers, households, mortgage lenders, governments, and communities, including the persistence of exposure after insurance arrangements change. The contribution is a documentary and analytical synthesis of established research, not a newly instituted proprietary theory or empirical estimate.

Keywords: Property insurance; nonrenewal; insurer withdrawal; geographic retreat; homeowners insurance; catastrophe risk; insurance availability; residual markets; FAIR Plan; reinsurance; protection gap; wildfire; flood; housing; mortgage delinquency; financial regulation; climate adaptation; causal inference.

Introduction

The continuity of property insurance is a spatial and intertemporal allocation problem. A homeowner or enterprise can encounter reduced protection because an incumbent carrier declines renewal, a carrier refuses to originate new policies, deductibles rise, an essential peril is excluded, the available premium exceeds household resources, or only a residual-market contract remains. Each outcome affects economic security differently. The risks insured against continue to exist when an insurer changes its portfolio. The financial burden moves among property owners, insurers, lenders, reinsurers, public pools, and ultimately communities.

The scale of the underlying issue is visible in the U.S. Treasury Federal Insurance Office's 2025 analysis of 2018–2022 homeowners policies. Using a ZIP-code-level dataset assembled with insurance regulators, it reported an average nonrenewal rate approximately 80% higher in the highest expected-loss ZIP-code quintile than in the lowest, alongside substantially higher average premiums. These are cross-sectional associations within a defined dataset, as explained in the Treasury's homeowners insurance findings (Federal Insurance Office, 2025). The findings do not mean that every nonrenewal was caused by climate change or that every affected resident lacked replacement coverage.

The subsequent NAIC market dynamics study (Czajkowski and Harms, 2026) documents substantial increases in company-initiated nonrenewals between 2018 and 2024, while also finding that the homeowners market remains broadly operational at aggregate level. The contrast between national financial functioning and localized deterioration of access is the central empirical puzzle. Portfolio diversification allows the industry to remain solvent while particular neighborhoods face declining carrier participation or higher prices. Aggregate sectoral stability therefore has limited power to establish local contract continuity.

The international comparison reinforces this distinction. The ACCC Northern Australia inquiry (2020) identified affordability as the principal market problem in its study area while still documenting substantial cyclone, flood, reinsurance, and insurer-cost pressures. European supervisors instead emphasize the insured share of catastrophe damage and protection gaps, as shown in the EIOPA protection-gap dashboard (EIOPA, 2025). Neither an affordability finding nor a low catastrophe coverage ratio by itself measures the number of policies an incumbent insurer declines to renew.

The research questions are therefore precise. How should nonrenewal and spatial retreat be defined and measured? Through which claims, capital, catastrophe, reinsurance, regulatory, and behavioral channels do insurance portfolios change? What follows for policyholder access, residual markets, household mobility, housing prices, and public risk-bearing? Which findings support causal interpretation, and where do observational limitations demand a narrower claim?

The study uses source-critical documentary comparison, elementary actuarial identities, jurisdiction-specific legal analysis, and examination of identified empirical papers. It makes no claim to independent access to policyholder-level insurer records, confidential underwriting manuals, proprietary catastrophe model outputs, or new causal estimation. Its substantive object is property market participation and renewal continuity; general property claims adjustment, catastrophe model calibration, and broad theories of insurance demand are treated only insofar as they explain observed retreat.

Definitions, Historical Context, and Primary Source Foundations

Contractual and institutional vocabulary

Nonrenewal is an insurer's decision not to continue an expiring contract for a subsequent term, subject to the applicable contract and law. Company-initiated nonrenewal must be distinguished from nonpayment cancellation, policyholder cancellation, midterm insurer cancellation, expiration without a replacement request, and the insurer's offer to renew with materially different conditions. A policyholder may retain contractual coverage after a nominal renewal while experiencing a large deductible increase or meaningful peril exclusion. Conversely, a declined renewal may be followed by a replacement policy from another insurer at a comparable price. These are separate observable outcomes.

Geographic retreat means a sustained reduction in offered, written, or maintained insurance capacity for defined risks in a defined geographic territory. It can be extensive, where fewer insurers participate, or intensive, where existing insurers limit capacity, restrict classes of property, tighten underwriting rules, reduce insured amounts, or alter coverage. Withdrawal may be announced by a carrier, inferred from policy transactions, or observed in a decline in quoted products. These indicators have different evidentiary strengths.

An insurance desert is a descriptive term for territory with severely restricted effective market access. Scientific use requires a declared threshold and geographic unit: for example, the percentage of eligible dwellings receiving an admitted-market quote of specified scope, or the percentage that retain a policy after an insurer nonrenewal. The existence of a FAIR Plan, surplus-lines option, or mortgage-lender-placed policy changes the institutional classification of access; it does not establish equivalence in coverage or cost.

Risk, hazard, and loss need separate treatment. A peril such as wildfire, severe convective storm, or flood interacts with exposed assets and their vulnerabilities to create physical damage. Economic loss includes effects beyond insured physical damage; covered loss depends on policy wording, limits, deductibles, exclusions, and insured interests; realized claim payment depends on adjustment and performance. The IPCC Sixth Assessment synthesis (IPCC, 2023) establishes climate-related risk and adaptation context, while attribution of a particular insurer decision requires portfolio, cost, and regulatory evidence.

Denominators and the object of measurement

Let N(g,t) be company-initiated nonrenewals in geographic cell g during period t, and P(g,t) the appropriately defined number of policies in force or eligible renewals. A descriptive nonrenewal frequency can be written

NRate(g,t) = 1,000 × N(g,t) / P(g,t).

The choice of P matters: a rate per thousand policies in force is not the same as the conditional probability of nonrenewal among policies offered for renewal. Neither is a measure of the percentage of households uninsured. Researchers must report the exact construction used by the source. The NAIC MCAS definitions describe market-conduct reporting items including company-initiated nonrenewals, company cancellations, and nonpayment cancellations (NAIC, 2026). Reporting practices and line-of-business scope constrain comparison across states and years.

A related insurer participation indicator is the number of distinct active underwriting companies writing relevant business in cell g in year t. A company can remain counted as participating even while the number of policies it writes falls sharply; one parent group may own several licensed underwriting entities. Active carriers, independent groups, offered quotations, policies issued, and total insured value therefore answer different questions.

For a renewal cohort, contract retention can be expressed as the share of expiring policyholders who obtain uninterrupted, substantively comparable protection, whether from their original carrier or another insurer. This outcome requires linked transactions and information on coverage quality. A publicly reported company nonrenewal rate alone cannot reconstruct it.

Historical development and changes in the observational record

Property insurance historically depended on territorial classification, construction, urban fire protection, actuarial pooling, and capital available to absorb correlated losses. Contemporary catastrophe exposure, replacement-cost inflation, and risk-sensitive underwriting have increased the importance of geocoded property characteristics. Changes in insurance technology and information can alter observed pricing and classification even when the physical hazard changes more gradually. A reported retreat thus may combine remeasurement of known risk, new physical exposure, claim experience, tighter financing, regulatory constraints, and strategy.

The modern data architecture includes state insurance department records, insurer financial statements, market conduct annual statements, residual-market administrative statistics, and nationally aggregated federal data. California's residential policy count series reports new, renewed, nonrenewed, and FAIR Plan policies by ZIP code, supporting spatial comparison while remaining sensitive to insurer thresholds, dwelling types, and reporting updates (California Department of Insurance, 2026). The 2025 California residential fact sheet documents the scope of statewide reporting, representing approximately 98% of the homeowners market by its stated measure and period (California Department of Insurance, 2025).

At national scale, the federal FIO data call includes over 246 million policy-year observations between 2018 and 2022 in insurer-submitted material aggregated by ZIP code, covering more than 330 insurers according to the Treasury release (2025). NAIC's separate 2018–2024 MCAS study has different reporting units and dates. Combining their quoted rates into a single trend without harmonizing definitions, insurers, products, and geography would create an artificial series.

Why a geography of insurance requires historical and institutional context

A hazard map identifies a location's exposure to a peril. An insurer's map of acceptable risks is the output of additional decisions about claims volatility, building condition, policy design, distribution, expenses, capital, reinsurance, and regulatory price permissions. The difference matters especially in sparsely populated regions, older housing stocks, peri-urban wildland interfaces, and coastal zones. A low count of voluntary policies can reflect underwriting withdrawal, premium constraints, low household demand, reliance on public coverage, and legacy settlement patterns at the same time.

Insurance protection also changes through public institutions. In the United Kingdom, Flood Re operates a defined reinsurance arrangement supporting household flood insurance, with a planned transition by 2039. France uses a statutory natural-catastrophe guarantee and public-backed reinsurance. Australia's cyclone reinsurance pool changes the cost of bearing specified peril exposure. These arrangements create distinct meanings of market participation and must be studied in their legal and fiscal contexts rather than ranked by a single national nonrenewal statistic.

Market Exit and Renewal Changes

Observable levels of contraction

Market exit occurs at several scales. An insurer can stop writing all new residential business in a state, cease writing only high-hazard ZIP codes, suspend new applications while honoring existing contracts, apply stricter risk-selection rules to renewals, nonrenew selected policy cohorts, or sell a renewal book to another carrier. These actions differ in timing, legal treatment, and immediate consumer effects. A national company's public exit announcement also does not establish that all exposures have been abandoned: policies may be transferred, renewed by affiliated companies, or placed in a residual market.

A scientifically useful record therefore follows a sequence: incumbent insurer identity; policy and hazard class; insured address and building characteristics; expiration date; notice and stated grounds; renewal alternative; quotations and coverage comparability; subsequent insurance status; and timing relative to disasters or regulatory changes. Where complete transaction linking is impossible, the report must state which segment of this sequence has been observed.

A geographic transition can be described by counts of (i) participating admitted insurers, (ii) insurer-group offers, (iii) admitted policies, (iv) written insured value, (v) insurer-issued nonrenewals, (vi) residual-market placements, (vii) surplus-lines placements, and (viii) properties with no effective coverage. Reporting all eight is usually difficult; selecting just one has inferential consequences.

Evidence from U.S. market-conduct data

The NAIC's 2018–2024 market study (Czajkowski and Harms, 2026) provides a direct measure of insurer-initiated nonrenewal activity. Its 2024 regional values were 11.7 per thousand policies in force in the Northeast, 14.2 in the Midwest, 22.0 in the Southeast, and 25.1 in the Western region. Expressed as percentages, these are 1.17%, 1.42%, 2.20%, and 2.51% respectively, with the explicit caution that a regional average masks substantial within-region heterogeneity. They are not percentages of homes becoming uninsured.

Across 2018–2024 the authors report nonrenewal-rate increases ranging from approximately 96% to 216% by region. A large percentage growth rate can coexist with a modest absolute policy fraction when the initial level was low. The study's regional unit does not identify any particular county or ZIP code as abandoned. Its broader finding that the homeowners insurance industry remained operationally strong also does not rebut local withdrawal: the aggregate industry and individual high-risk territories are different units of analysis.

The FIO ZIP-code evidence (2025), covering an earlier period, separately finds a strong expected-loss gradient. Highest-risk quintile ZIP codes paid average premiums of $2,321 during 2018–2022, about 82% above the lowest-risk quintile, and experienced nonrenewal rates approximately 80% higher. The expected-loss measure is a modeled building-hazard proxy and cannot substitute for insurer-specific underwriting files. Exposure quality, property value, prior claims, building age, local regulation, and household income can also vary systematically across ZIP codes.

The datasets together support a carefully limited conclusion: significant geographical and temporal differences in pricing and company nonrenewal are documented, and deterioration in some places is compatible with an otherwise functioning national insurance sector. These data do not identify the individual insurer's objective function, prove that every nonrenewed policy became uninsured, or measure the effect of a specific regulation without an identification design.

California as a study of differentiated retreat

The California Department of Insurance's wildfire and insurance data give an unusually useful spatial record of new policies, renewals, nonrenewals, and residual-market growth. The administrative distinction between policies issued by the admitted voluntary market and the FAIR Plan is vital. The FAIR Plan provides basic fire coverage when qualifying households cannot obtain it through ordinary market channels; differences in other perils and supplemental coverage remain economically important.

Nonrenewal protections are legally bounded. The Department's one-year wildfire moratorium guidance explains that qualifying residential properties within designated ZIP codes receive protection against wildfire-risk-based cancellations and nonrenewals for one year from the governor's emergency declaration, with separate provisions for total losses (California Department of Insurance, 2026). It does not establish permanent compulsory renewal or universal protection for every cause and every property.

The legal distinction between a notice period, a renewal right, and an emergency moratorium prevents a common analytical error. A notice rule regulates timing and information; a moratorium can temporarily constrain a specific insurer action; market-entry incentives seek future supply. None by itself determines the long-run risk price or cost of reinsurance.

California's Sustainable Insurance Strategy explanation (2025) connects permissible forward-looking catastrophe modeling and reinsurance cost treatment to insurers' commitments to write in distressed areas under a framework tied to an 85% market-share benchmark. That percentage is a specified regulatory market-participation requirement, not a statement that 85% of distressed homes are insured. Its effectiveness needs evidence on actual new writings, rate approvals, coverage breadth, FAIR Plan transfers, and realized post-disaster losses.

A Dallas Fed working paper on insurer-of-last-resort interaction (Taylor, Turland, and Weill, 2025) uses geographic border variation and difference-in-differences analysis of earlier California moratoria. The authors find short-run reductions in company-initiated nonrenewals and cancellations, with effects dissipating after the one-year protection elapsed, and no discernible effect on participation in the state insurer of last resort in the studied setting. This is strong evidence for the limited duration of one specific legal intervention; it is not a universal estimate for all disasters or future regulatory designs.

Florida and the problem of interpreting residual-market contraction

Florida provides a useful counterexample to treating every residual-market trend as permanent deterioration. The Florida OIR July 2026 report reports a substantially smaller Citizens Property Insurance portfolio than at its earlier peak, improved aggregate underwriting measures for domestic property insurers in 2025, and new admitted-company approvals. These supervisory facts support evidence of market re-entry and portfolio transfers in that period, while their association with reform packages does not alone establish a uniquely identified causal effect of each provision.

The Florida Surplus Lines Service Office's June 2026 market comparison describes the changing mix of admitted, Citizens, and surplus-lines residential business. A Citizens policy leaving the residual portfolio may be privately assumed, canceled, otherwise exited, or reclassified; depopulation counts must be joined to renewal outcomes before being equated with improved household protection. Likewise, the rise of surplus lines may maintain access for properties excluded from the admitted market while changing contractual and regulatory safeguards.

A proper evaluation follows cohorts transferred from Citizens to private carriers and asks how long they remain insured, at what total premium and deductible, with which covered perils, and with what future nonrenewal rate. Sectoral profitability and a declining residual-policy count are necessary evidence about the market but are insufficient measures of distributional outcomes.

Regulatory choices and renewal rights

Regulation changes the sequence and incentives of market exit. Advance notice enables search and financing adjustments; appeal and information rights facilitate correction of inaccurate property characteristics; temporary catastrophe moratoria offer immediate continuity; filing rules shape the adjustment of rates to loss costs; and public programs allocate extraordinary layers of losses. Their consequences depend on enforceability and capacity. When regulatory rates remain below the capital-consistent price of correlated risk for prolonged periods, a privately financed insurer may restrict writings or nonrenew in lieu of expanding a loss-making portfolio. Where allowed rates reflect updated risk and mitigation is credible, insurer entry may improve, with affordability implications for households.

This is a conditional economic proposition, not a claim that a single regulation caused any observed withdrawal. To identify causes, legal reforms must be located in time alongside hazard exposure, prior insured losses, reinsurance renewals, capital flows, construction inflation, firm-specific strategies, and the evolution of residual-market eligibility.

Capital and Reinsurance Costs

A renewal decision as a prospective risk commitment

At renewal, the insurer undertakes a new period of contingent obligations. Its decision depends on expected insured loss, the distribution of extreme aggregate losses, expenses, acquisition and servicing costs, the cost of reinsurance, capital requirements, permitted rates, risk mitigation, and the opportunity cost of scarce underwriting capacity.

For a portfolio j in period t, an elementary prospective adequacy comparison is

Expected premium revenue >= Expected claims and adjustment expense + operating expense + reinsurance expense + required risk-capital return,

with all terms evaluated on compatible timing, policy scope, and accounting bases. This is a useful decomposition of costs, not a binding regulatory pricing equation: taxes, investment income, reinsurance recoveries, policyholder dividends, operational constraints, and accounting methods may alter the complete calculation.

If the available premium is constrained while a property's expected loss or capital requirement increases, the insurer may reduce new writings or its share of accumulated exposure. Even if a policy remains individually profitable on an expected-loss basis, concentrated catastrophe risk can make another policy unattractive at the portfolio margin. In that case, geography matters because insured losses across nearby buildings rise together after a severe event.

Correlation, accumulation, and reinsurance attachment

For individual insured losses X_i, aggregate loss S = sum_i X_i, and portfolio variance Var(S) = sum_i Var(X_i) + 2 sum_{i<j} Cov(X_i,X_j). Correlation generated by common wind, wildfire, hail, or flood events increases aggregate tail exposure. The size and location of the portfolio, not merely the average expected loss per building, affect capital needs.

Reinsurance can reduce retained tail exposure through catastrophe excess-of-loss treaties, quota shares, or other structures. A stylized excess-of-loss layer with attachment A and limit L pays min(max(S − A,0),L), subject to contract terms. The cedent's retained cost includes premium paid for the layer, uninsured portions, reinstatement or exhaustion terms, and residual capital. A rise in modeled exposure or reinsurance price can change the margin on accepting another property even when the property's own loss record is benign.

A transparent hypothetical illustrates the distinction. Consider an insurer writing 10,000 otherwise comparable buildings with expected annual covered loss of $1,100 per building. A 20% increase in repair costs alone raises this component to $1,320, before any change in catastrophe frequency. If reinsurance expense increases from $250 to $400 per building, the arithmetic increase in these two annual costs is $370. That figure is neither a predicted insurance premium increase nor an observed market estimate; it excludes expenses, taxes, capital return, profit, rate approvals, mitigation and different coverage limits. It demonstrates why an underwriting decision can change without a new disaster at each individual property.

Catastrophe losses, repair inflation, and risk perception

The Swiss Re 2025 catastrophe analysis (Swiss Re Institute, 2026) reports $107 billion in global insured natural-catastrophe losses and approximately $220 billion in economic losses across its event classification. The series highlights major California wildfire and severe convective storm losses. These global totals establish material capital demand and the presence of an insurance gap; they cannot be mapped directly onto a single state's renewal probabilities. Differences in global catastrophe-loss methodologies and reporting conventions also make cross-publisher insured shares unsuitable for mechanical comparison.

The OECD 2026 market trends characterize catastrophe and climate risks among the most frequently identified risks in national supervisors' responses, while reporting broadly positive nonlife underwriting performance at aggregate level (OECD, 2026). This coexistence reinforces the need to distinguish systemic loss-bearing capacity from territorial risk appetite. Strong reported profits in one period do not eliminate insurer restrictions in locations where the prospective capital cost of additional concentrated risk is high.

A forward-looking view of physical risk is consistent with the IPCC synthesis assessment (IPCC, 2023). Climate change can affect hazard distributions, but insured losses also rise because of wealth accumulation in hazard-prone places, construction and labor costs, changes in deductibles, asset vulnerability, exposure concentration, and contract design. A causal argument about nonrenewal must isolate these interacting drivers rather than assign the entire phenomenon to one broad environmental label.

Prices, asset values, and the reinsurance capital market

The October 2026 revision of Keys and Mulder's mortgage escrow research constructs a large U.S. premium dataset from mortgage escrow transactions covering 2014–2024. The authors report stronger pass-through of disaster risk into premiums, interpret changes in the pricing gradient in light of global capital-market repricing, and estimate economically meaningful effects on home-price growth in highly exposed ZIP codes. This study supports an indirect channel from insurance-market financial conditions to asset prices. Its inferred insurance-premium observations and identification assumptions remain distinct from insurer nonrenewal microdata.

The financial mechanism has at least two steps. The price or availability of insurance changes the recurring cost and security of property ownership; capitalization then affects demand and the discounted value of housing services. The extent of price transmission depends on mortgage requirements, household wealth, insurance alternatives, expected catastrophe losses, local housing supply, and migration opportunities. These dependencies limit the transferability of any one paper's estimated dollar effect to other regions or dates.

Mitigation and risk engineering

Risk mitigation can restore a property's insurability when it materially changes expected damage or claims tail exposure. Examples include roof strengthening, ember-resistant construction, defensible space, flood barriers, elevation, drainage, electrical safety, and maintenance. The economic effect depends on the peril, verified engineering effectiveness, insurer recognition of mitigation, local hazard interdependence, and the time required to retrofit a building.

At the neighborhood scale, mitigation has collective features: one building's defensible space may matter less when surrounding structures are combustible; drainage improvements may shift rather than eliminate floodwater; and infrastructure resilience depends on coordinated maintenance. This creates an evidence requirement for any insurance incentive scheme: measure physical loss reduction and contract renewal outcomes, rather than assume that a subsidy or certification alone changes risk.

Insurance availability is therefore determined at the intersection of physical loss prevention, capital allocation, and credible contractual recognition of risk reduction. Reinsurance support can help absorb rare losses, while engineering adaptation can reduce the underlying losses that need to be financed. The instruments act on different terms in the underwriting equation.

Policyholder Access and Migration

Effective access is more than the existence of a policy

Property owners need protection that is available when sought, affordable under the household's budget, adequate relative to exposure, and durable across renewal periods. These dimensions can diverge. A quote with a very high deductible may satisfy a mortgage lender's nominal insurance condition while leaving the owner unable to finance a large repair. A residual-market fire policy may preserve coverage for one peril while requiring separate protection for other hazards. A timely replacement policy may retain nominal limits but exclude a newly material exposure.

For a household h, let available annual premium be p_h, disposable resources after essential expenditure be y_h, deductible be d_h, and expected uninsured repair liability be u_h. The observed p_h/y_h ratio measures one aspect of payment capacity. Full financial protection also requires the household to finance d_h and u_h after a disaster; mortgage escrow, income volatility, and savings affect those amounts. Any empirical affordability threshold needs a definition of income, household composition, policy limits, and insured peril.

The OECD catastrophic-risk protection report (OECD, 2026) distinguishes insurance availability, affordability, take-up, and adequacy in its comparison of public policy instruments. These are not synonyms. The EIOPA European dashboard instead uses hazard-specific modeled risk and historical insured/economic damage to examine protection gaps across 30 European countries. A country-level gap can reflect nonpurchase, contract exclusions, low insured limits, public arrangements, and uninsurable exposure; it does not estimate territorial nonrenewal without additional observations.

Insurance pressure, credit, and residential mobility

Property insurance often interacts with mortgage covenants. A decline in coverage can trigger a search for replacement, insurance monitoring by servicers, higher escrow requirements, or lender-placed coverage where contract and law permit. Insurance premium inflation can reduce available cash for principal, interest, taxes, and maintenance. Risk-based property values may also affect credit and collateral assessment.

The revised Dallas Fed household relocation study (Ge, Johnson, and Tzur-Ilan, 2026) uses household-level microdata and instrumental-variable strategies to report that larger insurance premium increases are associated with greater relocation toward lower-risk and lower-cost areas and increased mortgage delinquency; estimated relocation responses are stronger among less financially constrained households, while delinquency pressures are concentrated among more constrained households. The result clarifies why observed migration and remaining neighborhood exposure can move in different directions. Affluent residents may be better positioned to relocate; financially constrained residents may absorb higher premiums, greater underinsurance, or payment arrears.

Population movement cannot be inferred directly from insurer nonrenewal counts. A household may buy replacement insurance, sell, rent, reduce coverage, or remain uninsured. To study migration, individual transaction records must be linked to address transitions, mortgage and income variables, and the timing of premiums and hazard events. Local housing trends, employment, school access, taxes, and regional economic shocks can confound the association.

A distributional analysis must also address renters, condominium associations, small businesses, and households without mortgages. Renters' access to contents and liability cover is different from a homeowner's building policy. Condominium master insurance affects association fees and loan eligibility. Small-business interruption exposure extends beyond the building's repair cost. One homeowners-policy dataset cannot substantiate population-wide market retreat for all property interests.

United Kingdom: catastrophe protection through reinsurance support

The United Kingdom's Flood Re demonstrates a different institutional approach. Rather than directly underwriting every high-risk property, the scheme supplies eligible flood reinsurance under defined legal and operational conditions. The Flood Re 2025–2026 annual accounts report that more than 742,000 properties had benefited cumulatively from policies ceded since inception; as of March 2026 the ultimate policy count was estimated at approximately 353,000. The same report describes insurance quotation availability for high-risk properties within its monitored sample and the scheme's financing and reserves (Flood Re, 2026).

These data show the operation of a market-support institution, with the important qualification that cumulative properties assisted, annual in-force cessions, households at flood risk, and households protected against every peril are separate denominators. The scheme's 2039 transition horizon also means that current accessibility must be evaluated together with flood resilience, future risk-reflective pricing, and the distribution of support. A stable reinsurance pool can moderate acute market contraction while leaving long-term exposure reduction as a separate policy task.

Northern Australia: high cost with different availability conditions

The ACCC Northern Australia final report (2020) concluded that affordability rather than outright availability was the central supply-side concern in the region it examined. This result is essential counterevidence against treating every high-hazard market as a geographic insurance desert. The inquiry identifies cyclone, flood, claims cost, reinsurance, competition, commissions, and other pricing components as important cost drivers.

The ACCC 2025 insurance monitoring report finds that the cyclone reinsurance pool moderated premiums for certain medium- and high-cyclone-risk households and businesses, with stronger effects in higher-risk locations, while noting that premiums remained high for many consumers. Reinsurance support can therefore improve affordability even when a market was previously able to supply coverage, and measured price reductions do not necessarily imply equivalent reductions in physical hazard.

The case demonstrates the value of studying the policy instrument's financial incidence: insurer premium paid to the pool, savings passed through to insureds, eligibility, cross-subsidies, fiscal exposure, and mitigation incentives. A program that lowers quoted premiums without changing exposure can improve near-term coverage and still require sustained hazard reduction for long-term stability.

France and continental Europe: statutory pooling with climate pressures

France's CatNat system combines compulsory disaster-cover extensions of eligible damage insurance policies, a uniform surcharge, a declaration procedure recognizing eligible natural catastrophes, and a central reinsurance role for Caisse Centrale de Réassurance backed by the state. The French Court of Auditors' 2026 sustainability assessment finds increasing pressure from catastrophe experience, including drought-related soil shrinkage and swelling, and describes the challenge of preserving territorial pooling while controlling public financing (Cour des comptes, 2026). Its projections are scenarios, not certain forecasts of realized insured loss.

A statutory guarantee changes what insurers must include in an eligible contract; it does not ensure that every household purchases the underlying policy or that the insurer remains equally willing to offer the primary property contract. The distinction is a central limit of mandatory peril coverage as a response to geographic withdrawal. It can broaden the insured perils for policies that exist while still leaving the extensive margin of access to primary insurance and the fiscal allocation of extraordinary loss.

Across Europe the EIOPA protection-gap evidence should be read by hazard, country, model version, and historic period. A large estimated catastrophe gap may be caused by low take-up, exclusions, public disaster aid expectations, or weak property insurance infrastructure. Combining it with nonrenewal measures requires product-level and jurisdiction-level harmonization.

Empirical Verification, Uncertainty, and Competing Explanations

An auditable research design

The first research task is to define geography and population. ZIP codes, census tracts, counties, insurer underwriting territories, flood zones, and wildfire hazard boundaries are not interchangeable. ZIP code changes across time can produce artificial trends, and a company that writes a handful of policies can appear active in a territory even when effective quote availability is minimal. A property-level spatial crosswalk with documented boundary year improves continuity.

The second task is to define the policy population. Data should identify homeowners versus dwelling-fire, owner-occupied versus rental, admitted versus surplus-lines versus residual-market, personal versus commercial, and coverage for particular perils. A renewal cohort should record policies scheduled to expire and nonrenewed by the company, distinguishing midterm cancellations, nonpayment lapses, and policyholder-initiated switching. The NAIC market-conduct taxonomy supplies a starting point for U.S. definitions; other jurisdictions require their own contract and supervisory categories.

The third task is to link events. At minimum, record insurer identity, policy year, property location, hazard score and source edition, claims history if authorized, premium, deductible, limit, renewal offer, rejection, subsequent placement, and public program participation. Public reports often suppress small cells for confidentiality, which can systematically reduce visibility into sparsely populated high-risk locations.

The fourth task is to distinguish descriptive analysis from causal identification. Aggregate nonrenewals may rise after a fire because of hazard reassessment, prior losses, capital constraints, shifts in the insurer population, catastrophe model recalibration, rate regulation, and construction-cost inflation. A credible causal design should specify a source of variation, treated and comparison groups, timing, estimand, pretrends, interference, and threats from changing sample composition.

Observational indicators and worked arithmetic

Consider a hypothetical district with 10,000 policies in force at the start of a reporting year and 250 company-initiated nonrenewals. A rate stated per thousand opening policies is 25, or 2.5%. If 200 of the affected households subsequently obtain comparable new policies, the number of continuously uninsured households attributable to the nonrenewal cohort cannot exceed 50 at the specified observation date, assuming no other changes. If 70 transferred to a residual plan and 130 entered other admitted or surplus-lines coverage, the replacement market composition changed even though most households retained a policy. Without linked transactions, the proportion uncovered is unknown. The example is arithmetic, not actual district data.

Now suppose the count of insurers drops from 12 to 10 while total policies grow from 10,000 to 10,500. This is an extensive-margin reduction in carrier participation with an increase in aggregate policies. Conversely, the number of carriers can remain 12 while total insured value declines and average deductibles rise sharply. Either event may indicate deteriorating effective access. A simple carrier count or policy total is inadequate on its own.

A third hypothetical illustrates geographical aggregation. Two ZIP codes each begin with 1,000 policies and a nonrenewal rate of 1%. In the second year, the first rises to 5% while the second remains at 1%. The unweighted pooled rate is 3% only if their denominators remain equal. If the high-risk ZIP code shrinks to 100 covered policies while the other grows to 1,900, the in-force-weighted pooled rate is about 1.2% despite an extreme local retreat. Reporting the geographical distribution alongside weighted totals is indispensable.

For a simple insurance protection gap, the usual loss-share calculation is (economic disaster loss − insured disaster loss) / economic disaster loss, where denominators must cover the same events and valuation basis. It is invalid to compare insured payouts from one claims-peril set to economic losses from another without adjustment. This ratio is a damage-financing statistic, distinct from household insurance take-up, policy renewal, and access.

Quasi-experiments, spillovers, and model stability

Differences-in-differences, event studies, regression discontinuity near regulatory or hazard boundaries, and carefully designed instrumental variables can isolate components of behavior under assumptions. The California moratorium research uses a geographic and temporal design that informs the effect of temporary legal protection in its studied setting. Such designs require assessment of policyholders just across borders, shared hazard exposure, migration of insurers and households, concurrent regulations, and selection of risk levels.

Instrumental-variable studies of insurance premiums and mortgage outcomes can be informative while requiring a credible exclusion restriction: the instrument should affect housing and credit outcomes through the proposed insurance-cost channel rather than independently through disaster news, local credit tightening, or migration demand. The Ge, Johnson, and Tzur-Ilan study identifies its strategy and reports heterogeneity; the monograph does not reproduce its underlying confidential data or recalculate standard errors.

Model validity becomes particularly delicate when catastrophe hazard surfaces, building data, and insurance supply change together. A newly granular wildfire model may identify elevated loss potential in a place that has not physically changed; a roof-age database may newly classify buildings as high risk; or the actual hazard can intensify. The distinction is between change in estimated risk and change in the risk-generating process. A sensitivity audit should compare model vintages on the same exposure portfolio, then compare exposure vintages under the same model.

Data construction must also handle catastrophe tail variability. One calm year cannot establish a permanent reduction in insurer cost; one exceptional catastrophe does not automatically establish a permanent trend in frequency. Capital providers price the full prospective loss distribution, including severe correlated scenarios. A resilience intervention should be evaluated over enough time and physical events to detect meaningful changes, with uncertainty ranges explicit.

Competing explanations and negative evidence

Several explanations may produce similar observed withdrawal:

  • Physical hazard and vulnerability may increase expected insured losses.
  • Reconstruction inflation may raise severity independently of hazard frequency.
  • Higher reinsurance premiums or reduced catastrophe limit may raise retained risk costs.
  • Investment and capital constraints may increase the opportunity cost of concentrated underwriting.
  • Rate approval delays and contract restrictions may alter insurer margins and the feasible menu of products.
  • Distribution practices or insurer-group restructuring may alter carrier counts without reducing underlying household access.
  • Households may voluntarily lower coverage, switch insurer, or leave the market because of price or preferences.
  • Public residual-market eligibility and depopulation programs may move policies between institutional categories.
  • Better geocoding, inspection, or catastrophe modeling may change underwriting classification without immediate new damage.

These alternatives can be jointly true. A strong study reports what data discriminate among them, rather than assigning all movements to one factor. The NAIC report's methodological discussion acknowledges aggregation and market-conduct data limitations; California's own administrative reporting uses specified insurer and policy categories. Comparing these official designs is part of the evidence, not a nuisance to be erased.

Negative findings must remain visible. The ACCC Northern Australia inquiry found affordability to be the main issue in its defined region rather than a general inability to obtain insurance; Florida's 2026 market review reported re-entry and improved sector statistics in its defined period; the California moratorium evaluation found time-limited renewal effects. An international theory of universal and irreversible retreat would be contradicted by these distinctions.

Scientific limits and source audit

This study verifies source identity, dates, disclosed methodology, and specifically cited conclusions through accessible official publications, supervisory material, and research abstracts or summaries. It does not independently replicate the MCAS raw data, FIO ZIP-code microdata, restricted insurer policy records, or the econometric results of academic studies. Legal observations are expressly jurisdiction-specific and refer to identified official guidance or statutory sources as of the research cutoff. Where a source is an institutional press summary, the claim is bounded by that document's stated scope; a press release is not treated as a substitute for unseen contract-level evidence.

The reference set includes publisher reports, source institutions, research working-paper records, and a peer-reviewed 2026 journal article whose abstract addresses geographic discontinuity after the 2017–2018 California wildfire seasons. The Di Vincenzo California study (2026) identifies an altered relationship between wildfire risk and insurance outcomes after the major fire season, using ZIP-code evidence. Its finding offers independent support for spatially heterogeneous post-disaster changes, though its research design and outcome definitions must remain distinct from the NAIC's national regional nonrenewal series.

This evidence base supports qualitative mechanism tracing, direct reproduction of stated summary statistics, transparent arithmetic, and comparative institutional interpretation. It does not support newly computed cross-country renewal rates, claims of exact causal shares attributable to climate, or a single global estimate of geographic withdrawal.

Integrative Comparative Scientific Synthesis

A spatial reallocation of contingent financial obligations

The evidence supports a unified analytical interpretation: property insurance retreat is a change in the spatial allocation of contingent financial obligations. A building remains in a hazard territory whether the insurer renews, a public scheme reinsures the exposure, or the owner remains uninsured. The location and form of financial protection change, while physical exposure and potential public disaster costs may persist.

This interpretation follows from contract and balance-sheet mechanics. Private insurers can shift catastrophe accumulation by restricting underwriting or changing limits. Reinsurance transfers specified layers of that accumulation across capital providers. Public residual markets and statutory pools accept exposure under different eligibility and financing rules. Households that forgo protection assume more retained loss, while lenders may respond through insurance requirements and credit terms. The same locality may appear financially viable at national sector scale and vulnerable at household scale.

The comparative cases identify four materially different institutional transitions. California illustrates geographic renewal pressure alongside emergency continuity rules and an expanded residual market. Florida illustrates movement between residual and private portfolios, with recent re-entry in some segments. Northern Australia illustrates high premiums in a market where an official inquiry identified affordability as the dominant problem and reinsurance support altered price incidence. Flood Re and France CatNat illustrate risk-sharing institutions that support certain peril coverages while preserving obligations to finance future catastrophe costs. The distinctions are observed features of existing systems rather than a new doctrine.

Market functioning and social protection are separate testable propositions

The proposition that an insurance sector is solvent, profitable, or expanding in premium volume does not determine whether an individual neighborhood has competitive, adequate, and financially sustainable property coverage. Similarly, a high number of nonrenewals does not prove final loss of protection if policyholders obtain equivalent alternatives. The direction of welfare change depends on the replacement contract and its cost.

A complete local assessment therefore traces an expiring policy through notice, available alternatives, accepted coverage, premium and deductible, and eventual claim-paying capacity. If a policyholder moves from a broad admitted homeowners contract to fire-only residual coverage with a separate expensive difference-in-conditions policy, numerical retention may conceal a reduction in coverage quality. If an insurer nonrenews after improved geocoding but another insurer enters with mitigation-sensitive rates, incumbent withdrawal may coexist with effective competitive adaptation.

Distributional and geographic effects

Insurance can influence who remains in hazard-prone housing. Empirical evidence on premiums, housing prices, and mortgage delinquency suggests channels through which risk costs alter location decisions. Wealthier households may be able to absorb premiums or move; less liquid households may accept higher deductibles, defer maintenance, or face payment distress. These responses can amplify unequal exposure across neighborhoods even without differential physical hazard growth.

For research, distributional claims should be tested with household resources, age and construction of property, renters versus owners, mortgage status, prior claims, and available subsidies. A policy that expands market supply but raises premiums may help some households and exclude others. A subsidy that maintains affordability may transfer risk-bearing to a public balance sheet. The optimal choice depends on explicit fiscal, coverage, adaptation, and equity objectives; one objective cannot be inferred from a carrier participation count.

A practical agenda for verified monitoring

A robust supervisory monitoring program would publish consistent, anonymized geographically disaggregated indicators of company nonrenewals; new-business quotations and acceptance; cancellations by cause; admitted, residual, and surplus-lines placements; premium and deductible distributions; coverage limits and peril exclusions; and replacement continuity. It would publish definition changes, small-cell suppression methods, and hazard-data vintages. Confidentiality protections should be combined with meaningful geographic aggregation rather than eliminating access measures altogether.

A regulator evaluating a reform should publish the original intervention date, affected eligibility groups, rate and insurance-mix changes, benchmark exposure measures, and plausible counterfactual communities. Where records allow, link nonrenewal outcomes to mitigation and household displacement without exposing identifying records. Public reinsurance programs should additionally disclose contingent fiscal liability, attachment points, and whether savings reach the insured households targeted by the intervention.

The analytical priority is causal and contractual precision. Distinguish an underwriting cap from a claim of permanent market exit; a renewal notice from actual termination; a low catastrophe insured-loss share from nonrenewal; an insurer group's corporate restructuring from geographic capacity loss; and a statutory guarantee from the household's actual purchase of eligible primary insurance.

Interpretation of the evidence and original scholarly contribution

The original contribution of this monograph is an integrated, source-grounded comparison of nonrenewal measurement, hazard-related accumulation, reinsurance finance, regulatory continuity, residual-market transitions, and household and housing effects. It establishes an explicit audit path from reported nonrenewal through replacement coverage, warns against denominator and geographical aggregation errors, and demonstrates arithmetic examples that can be reproduced without private data. These are analytical syntheses and methodological clarifications from existing insurance economics and supervisory evidence. No new proprietary authorial theory, model family, doctrine, or empirical discovery is claimed.

The most important research consequence is that geographical withdrawal cannot be assessed conclusively from a single market indicator. The scientifically adequate object is the trajectory of effective risk protection for defined property populations, alongside the financing institutions and hazard conditions supporting that protection.

Conclusion

Property insurance nonrenewal and geographic retreat are measurable transformations in the continuity and allocation of risk protection. Company-initiated nonrenewals, underwriting restrictions, insurer participation, policy counts, premium increases, and residual-market transfers each record a different part of the process. Their separation is fundamental to sound inference.

The U.S. FIO evidence documents a substantial association between expected catastrophe losses and more expensive, less continuous homeowners coverage for 2018–2022. The NAIC's 2018–2024 study shows increases in company-initiated nonrenewal rates across regions, while its aggregate financial picture remains compatible with localized deterioration. California's regulatory and administrative records demonstrate short-term renewal protections and the importance of distinguishing the voluntary and residual markets. Florida provides contemporary evidence that some residual-market exposure can return to private insurers. Australia, Flood Re, and France show how public financing mechanisms alter premiums, availability, and catastrophe liability in different institutional systems.

An insurer's renewal decision reflects prospective expected loss, correlated tail risk, replacement-cost inflation, reinsurance terms, capital allocation, contractual coverage and regulatory conditions. The economic consequences extend to household liquidity, credit, asset values, location decisions, and the balance sheets of public risk-bearing schemes. The underlying physical hazard remains relevant even when financial institutions reallocate who pays for loss.

The documentary and scholarly record supports a qualified but substantial conclusion: territorial risk protection is becoming a central dimension of insurance-market performance, and its deterioration or recovery must be tested with geographically specific, contract-comparable, cohort-linked data. Public measures should be evaluated by actual continuity, coverage adequacy, net household cost, physical risk reduction, and fiscal sustainability. This methodological conclusion rests on existing research and documented institutions and requires no new theory. Independent policy-level, cross-jurisdiction causal estimation remains a legitimate future research task.

References