Chapter 10 — When Climate Enters the Contract
Climate enters homes through water, fire and cracks in walls. For many people, it will first appear in another form. A map will change colour. A premium will rise. Coverage will shrink. A bank will request further evidence, or a municipality will prohibit occupation. Risk will alter the present before it destroys the building.
At Soulac-sur-Mer, on the Atlantic coast of south-western France, the apartment block known as Le Signal did not first disappear beneath the water. It began by losing its use.
The seventy-eight apartments were built in the second half of the 1960s, when the building stood well back from the shore. Several decades later, coastal erosion had brought the edge of the dune within a few metres of the property [1, 2].
The building was still standing. Its balconies, walls and windows remained visible. Storms and erosion had reduced the strip of land separating it from the ocean. In January 2014, when only about nine metres remained between the building and the edge of the dune, the municipal authorities ordered an evacuation and permanently prohibited occupation [2, 3].
The loss fitted badly into the categories already available. This was not a house swept away in a single night, a sudden flood or a fire with a clear before and after. Erosion advanced gradually. It was too foreseeable to resemble an accidental event, yet its effects had become too certain for residents to remain.
In 2018, France’s highest administrative court confirmed that the erosion of the dune did not allow the owners to receive compensation from the national fund for major natural hazards, commonly known as the Barnier Fund [1]. The owners had to leave while remaining tied to their properties. They continued to face condominium charges, insurance costs and, in some cases, mortgage repayments. Some also had to pay for another home [4].
A special solution was eventually created. Legislation allowed compensation for the loss of use, and the first settlement agreements were signed in 2021 [5]. By the time demolition began in February 2023, the building had long since disappeared from the ordinary lives of its former residents. The concrete remained, but the homes did not [6].
Le Signal shows that social ruin can precede physical destruction. A property may lose its use, value or access to insurance while the walls are still standing. Climate then acts through law, credit and contract.
This reclassification is not confined to the coast. After several catastrophic floods in Canada, the Intact Centre on Climate Adaptation found lower final sale prices, longer selling times and fewer properties placed on the market [7]. Physical danger can therefore affect the liquidity of housing, meaning the ability to sell it within a reasonable time without accepting a steep discount.
Behind the premium lies a politics of territory. A house depends on a watershed, coastline, soil, road network, evacuation route and history of building permits. Construction standards, levees, drains, emergency services and public guarantees reduce some risks while shifting others [8]. An insurer never prices a place left in a state of nature. It prices a territory already protected, developed, neglected or exposed through collective decisions.
An address carries this history without explaining it. The wildfires that struck the Landes de Gascogne in 2022 affected a cultivated forest shaped since the nineteenth century around maritime pine. The area had also experienced intensified commercial use and the expansion of suburban housing. The exposure of homes came from the meeting of forestry, land policy, urban development and a changing climate [9–11].
In the autumn of 2022, a few weeks after the major fires in south-western France, a public meeting was held in Belin-Béliet. During the summer, fire had crossed the Landes de Gascogne, forced thousands of people to evacuate and burned tens of thousands of hectares. The town had not been destroyed, but everyone knew it might have suffered much more [11].
Residents, elected officials and forest owners tried to understand what had happened. A forester pointed to the housing developments built along the edge of the woodland. In his view, the problem came from homes placed too close to the forest. The nearer residents lived to the pines, the greater the chance of ignition and damage.
A resident replied that the pines had also been planted too close to the houses. She refused to let residents be treated as the sole authors of a danger from which they would be the first to suffer. The forest around their homes was not an untouched natural environment. It had been cultivated, harvested and replanted around one dominant species [9].
Another participant observed that the existing neighbourhoods could not simply be removed. The forest, however, could be managed differently. Why should the discussion concern only the houses, rather than the continuity of plantations, the diversity of species, firebreaks and the maintenance of forest roads?
They were speaking about the same place, but telling different histories. Some saw residents moving into contact with an activity that was already there. Others saw forestry and planning choices that had made the territory more vulnerable. Between these positions stood the municipalities that issued permits, households searching for affordable housing, landowners and businesses dependent on the forest [10, 11].
An insurance map can represent this proximity with a colour or score. It can estimate the chance that a house will be reached by fire. It cannot decide how much of the exposure belongs to the homeowner, forester, municipality or economic organisation of the region. By charging the address, the premium risks giving an individual answer to a question the meeting left open. Are the houses too close to the forest, or is the forest too close to the houses?
Geographers use the term riskscape to describe the meeting between a material hazard and the ways a territory is inhabited, governed and modelled [12]. The fire remains real. The disagreement concerns the causes retained, the responsibilities assigned and the scale at which action should occur.
A climate premium can therefore be accurate within a model and still leave a household without a solution. To judge it, we must ask what the household can still do. Can it pay, borrow, complete the work, challenge the information, sell or leave? A price becomes a boundary when it signals exposure without opening any realistic way to reduce it.
From Map to Contract
Insurance has long dealt with storms, floods, earthquakes and fires. It knows that some events strike many contracts at once. It builds reserves, holds capital and buys reinsurance. Climate change is not confronting an industry that had never encountered catastrophe.
What is changing is the combination of several trends. Some hazards are becoming more frequent or intense. Dangerous seasons last longer. Reconstruction costs are rising. Buildings, infrastructure and insured values continue to concentrate in exposed areas. Historical series then provide a weaker guide to future losses [13].
Three ideas help separate these changes. A hazard is the possible physical event, such as a flood, drought or fire. Exposure refers to the people, buildings and activities that may be affected. Vulnerability describes how severely they will be damaged and how well they can resist. The same rainfall can produce very different losses depending on sealed ground, the state of drains, basement construction and the resources households have to protect their property.
Climate changes the hazard. Urbanisation and planning change exposure. Standards, investment and unequal resources change vulnerability. Attributing every increase in losses to the natural event alone therefore misses part of the problem. Research on avalanches, for example, shows that exposure changes over long periods as building expands and also varies with seasonal movements of people [14].
Several Models for the Same Territory
Insurers can no longer extend past frequencies without adjustment. They use catastrophe models to simulate possible events, identify the buildings affected, estimate physical damage and convert it into financial loss. These models are especially important when claims are rare, records are short and portfolios are concentrated.
They do not produce a single future. In urban flood modelling, several approaches may explain past observations reasonably well and still disagree about which streets will flood next. A static map, a surface runoff model, a representation of the drainage system and a model connecting drains to the street do not describe exactly the same mechanisms [15].
This is sometimes called equifinality. Different models may fit observed events similarly while giving different results for rare events. For an insurer, these differences can alter expected loss, the capital required, the premium and the decision to cover an address [15].
Each projection also assumes a climate pathway, a condition of infrastructure, an adaptation policy and an evolution of exposed values. A premium therefore contains a view about collective decisions that have not yet been made. A municipality that maintains its networks, restricts new construction and finances protective work does not create the same future as one that postpones those choices.
Two forms of uncertainty are useful here. One is aleatory. No one knows which house will burn or which street will flood in the next event. The other is epistemic. It comes from incomplete knowledge, missing data or decisions not yet taken. It concerns the state of soils, the actual resistance of buildings, the effectiveness of levees and the maintenance of forests [16].
Some epistemic uncertainty can be reduced through inspection, research, maintenance or public action. Converting it immediately into an individual premium may make a household pay for ignorance or delay that is not entirely its own. A better model can reduce uncertainty. It cannot decide which increase is affordable, which place should be protected or how the cost of leaving should be shared.
Ortwin Renn and Andreas Klinke distinguish the complexity of causal mechanisms, scientific uncertainty and ambiguity about values [17]. The first two call for more knowledge and comparison between models. The third concerns what society chooses to protect and whom it asks to pay. More data do not resolve that disagreement.
When a Map Changes Function
A map may represent a past event, a modelled hazard, exposed property, vulnerability or possible loss. It may inform residents, guide emergency services, prohibit construction, open access to assistance or alter an insurance price [18]. These uses are not equivalent.
The prologue showed that a map does not set its own consequence. Here, the change of function matters. Moving from a hazard map to a pricing map is not the automatic continuation of a calculation. An institution decides to use the result to allocate a charge. Moving from the same map to non-renewal adds another decision. The institution no longer wishes to carry the risk.
A map selects a phenomenon, period, resolution, scenario and threshold. It may describe its chosen object well while saying nothing about debt, tenancy, the cost of work, the availability of contractors or the resources of a municipality. Like any map, it offers a way of seeing a territory and leaves some things outside the frame [19].
These omissions do not make the map false. They only mean that it cannot decide alone who should remain insured, at what price and for how long. Physical danger constrains the available choices. It does not choose the colour, premium or right attached to a zone.
The same information can therefore support prevention, pricing or withdrawal. This shift is especially important when insurers, banks and public authorities rely on the same data [20]. An exposed zone may open access to a survey and a grant. It may also make credit more expensive and insurance harder to find. The quality of the model does not remove the need to examine the decision that gives it force.
Three uses of price follow. A blind price keeps contributions relatively broad while concealing an important difference in exposure. It may preserve access in the near term, but it can also support new construction or subsidise valuable property. A signalling price makes danger visible and is accompanied by time, assistance, credit or protective work. A boundary price raises the premium, deductible or conditions without opening a realistic route to adaptation.
The distinction does not depend on the amount alone. The same price may alert a household with savings and quietly force out a household that can neither finance work nor find another offer. A price has a preventive role only when an institution organises what comes next.
Four Architectures of Climate Insurance
No country chooses between a wholly free market and solidarity without prices. Climate insurance combines pricing, pooling, reinsurance, legal duties and public guarantees. The main differences concern how quickly exposure becomes an individual charge and which institution takes over when the ordinary contract reaches its limit.
France and National Pooling
France’s natural catastrophe regime, commonly known as CatNat, rests on national pooling. When an event receives official recognition as a natural catastrophe and the property policy covers the relevant damage, compensation is supported by public reinsurance and a state guarantee.
The CatNat surcharge does not vary directly with the exposure of each home.1 For the hazards covered, the address does not determine the whole contribution. The rule expresses a national solidarity towards catastrophes whose causes and effects often extend beyond the owner.
The first report from the French Observatory of Natural Risk Insurability gives a partial view of what this architecture still achieves. Using data from forty-one insurers representing more than 90 per cent of the detached home market, it classified 97.7 per cent of municipalities in metropolitan France as showing no tension in insurer presence. A further 1.6 per cent showed slight tension and less than 1 per cent moderate tension. None was classified as facing severe tension in the 2022 data [21].
The measure does not capture prices, the quality of coverage or the ease of obtaining a new contract. Several insurers may remain present in a municipality while particular homes receive no affordable offer. The indicator provides a useful baseline for observing withdrawal. It does not measure insurability on its own.
Address level data can also serve opposing aims. They can reveal whether selection is concentrated in exposed areas. They can help each company steer its portfolio towards less costly locations. The Langreney report warned of gradual demutualisation, as each insurer improves its own position while the companies that remain inherit a more exposed portfolio [22].
The increase in the CatNat surcharge on 1 January 2025 reflects the financial pressure on the regime [23]. It does not solve coastal retreat, disputes over drought related subsidence or delays in adaptation. France’s Court of Accounts has pointed to weaknesses in anticipatory policy, including the management of the coastline [24, 25].
National pooling protects against immediate fragmentation. It can also create the impression that compensation will be enough. Coastal erosion requires decisions about habitation. Some places will be protected and others left. The timetable and compensation must also be decided. Insurance can finance a loss, but it cannot choose the future of the coast by itself.
The United Kingdom and Flood Re
Flood Re creates a different compromise in the United Kingdom. Insurers can transfer part of the flood risk of eligible homes to a common pool. The household retains a policy sold in the ordinary market, while an industry wide levy helps finance highly exposed properties.
The scheme prevents flood risk from being converted immediately into a fully individual price. It is designed as a temporary arrangement. Its purpose is not only to subsidise current cover, but to buy time in which homes and territories can become less vulnerable before the scheme is due to end [26, 27].
The practical question is what happens during the time bought by pooling. Without work, planning rules and finance for adaptation, the promised transition becomes a delayed increase in price.
The United States and Residual Markets
The United States has a more fragmented architecture. The National Flood Insurance Program made flood insurance available in places where the private market was insufficient. It also supported development and property values while accumulating debt and implicit subsidies. Reforms intended to bring premiums closer to risk exposed the conflict between a more accurate signal and the ability of households to pay, adapt or leave [28, 29].
California places insurance prices within a more visible regulatory procedure. Since Proposition 103, rates must be open to review so that they are not excessive, inadequate or unfairly discriminatory [30]. This review does not remove increasing losses. Tight price constraints may encourage insurers to reduce supply. Greater pricing freedom may preserve supply while making coverage unaffordable [31].
The California FAIR Plan then acts as a residual market, meaning a last resort arrangement for homes that cannot readily obtain ordinary cover. It prevents complete exclusion, but concentrates difficult risks and often provides narrower protection. Evidence from mortgage lending suggests that reliance on the plan has spread beyond the best known wildfire areas and entered the ordinary financing of housing [32].
The risk has not disappeared. It has moved to another institution. A residual market can leave a door open while separating ordinary policyholders from those dependent on a more expensive or fragile contract. The insurance crisis then becomes a credit crisis before every house has suffered a loss.
Regional Pools and Public Capacity
For a government, catastrophe creates an immediate need for money. Emergency shelter, water, roads, imports and public services must be financed before international aid, tax revenue or borrowing can arrive.
Several regions have therefore created catastrophe pools. The Caribbean Catastrophe Risk Insurance Facility, African Risk Capacity and comparable arrangements pool part of the risk across countries and buy reinsurance internationally [33]. They are not designed to reimburse every loss. They release funds quickly so that governments can act during the first weeks.
These covers are often parametric. Payment depends on an index, wind intensity, rainfall or a modelled loss rather than a detailed assessment of every damaged property. Speed comes with basis risk, meaning the risk that the index and the actual loss do not match [34]. A population may suffer severe damage without triggering payment, or receive an amount far from the loss experienced.
Malawi faced this problem after the drought of 2015 and 2016. The African Risk Capacity model initially underestimated the number of people affected because an assumption about the maturity of maize did not fit the varieties being grown. After dispute, correction and recalculation, a payment of $8.1 million was made in 2017 [33].
The conflict had moved into parameters, the crop chosen as a reference, the threshold and access to expertise. Parametric insurance shows that rapid payment also depends on a prior representation of the world. The model can provide resources for action. It must still be open to evidence from the people and institutions whose exposure it describes.
Climate Injustice Has an Address
Climate effects unfold across territories that were already unequal. In the United States, home insurance now connects physical danger more directly to credit prices and property values. The Congressional Budget Office describes the central tension. Rising losses and uncertainty push premiums upwards, while strict price regulation may reduce the supply of insurance [35]. A household may therefore face either a price it cannot afford or the absence of an offer.
This mechanism operates in places shaped by segregation, differences in income and unequal infrastructure. Climate does not create those divisions. It carries them into maps, property values and terms of cover.
Research on climate gentrification in Miami-Dade County shows how elevation and anticipated risk can affect residential change [36]. Areas on higher ground or with better protection become more desirable. Households able to pay, wait or renovate have more choices. Others remain in homes whose value is becoming less secure while the cost of insurance rises.
Insurance can then turn a map of hazards into a map of solvency. The history of insurance redlining is relevant here. Refusals were once justified through congestion, crime, building condition and supposed neighbourhood quality. Territorial reasons presented as technical extended racial hierarchy and restricted access to credit [37].
Climate maps now rest on better established physical processes. They still enter territories produced by that history. The truth of the hazard does not make the distribution of its consequences just. An address may be objectively exposed even when the resident had little part in creating the exposure.
Property value is not a pure measure of danger either. It reflects the quality of the home, services, scarcity of land, coastal access, available credit and expectations of public protection. In Carteret County, on the coast of North Carolina, some highly exposed properties retained high values because access to the ocean and the view outweighed the expected discount for flood risk [38].
Uniform assistance for coastal property may therefore support a household with no realistic way to move, or preserve an expensive second home. A fair policy must read the hazard map together with income, debt, occupancy and the ability to leave. Equal exposure does not create equal social vulnerability.
That difference also appears in mobility. An owner with savings may complete work or accept a loss on sale. A heavily indebted household may become trapped in property that is difficult to sell. A tenant may face displacement and higher rent without receiving compensation for the building. Territorial classification has different effects depending on a person’s position in the housing system.
A territorial bad risk is not always abandoned at once. It may remain insured at a price that removes the resources needed for work, through a segment offering less choice and more uncertainty. Exclusion then takes the form of degraded inclusion.
Private Models, Public Decisions
Catastrophe models are often built by specialist firms. Insurers, reinsurers, investors and regulators use them to estimate future losses. Their expertise becomes public in effect when their results organise access to housing, credit or insurance.
Preparing Scenarios without Mistaking Them for the Future
A scenario does not state exactly what will happen. It tests what institutions might still be able to do if losses exceeded recent experience [39]. Its value depends on the decisions, budgets and responsibilities it activates.
In Japan, flood maps prepared for very heavy rainfall show expected inundation, water depth, shelters and evacuation routes. The model is connected to decisions about warning, movement and the continuity of services [40, 41].
Preparation does not guarantee that the worst event has been imagined. In 2011, the tsunami exceeded several local assumptions. At Rikuzentakata, thirty-five of the sixty-eight designated evacuation areas were inundated [42]. The lesson is not to abandon scenarios. It is to treat them as revisable tools rather than boundaries of the possible.
A report cannot replace the means to act. Before Hurricane Katrina, evacuation scenarios existed, but the resources needed to carry them out were not always available [43]. An institution can describe a catastrophe reasonably well and still fail to protect those most exposed.
From Scenario to Price
A catastrophe model links several operations. It creates possible events, places them in space, identifies the buildings affected, converts physical intensity into damage and then converts damage into financial loss. At each stage, uncertainty may be summarised or made less visible.
The final map can look more certain than the calculation behind it. A comparison in Los Angeles County found substantial differences between a private model of future flood risk and an open method developed by researchers. The two tools assigned similar vulnerability levels to only a minority of the properties studied. Other comparisons have also found disagreement between vendors [44].
The Los Angeles fires of 2025 revealed a related problem. Some models represented fire spreading from wildland into developed areas more effectively than fire moving from building to building through embers, wind, density and construction materials [45]. A home could appear relatively safe on a vegetation map and still be vulnerable to an urban conflagration.
Disagreement does not mean that all models are equally useful. It means that disagreement should travel with the decision. When a premium, refusal or planning rule depends on a model, the institution should disclose the main assumptions, uncertainty and credible alternatives.
Ulrich Beck uses the expression relations of definition for the rules and powers that determine what will count as risk, cause, evidence and damage [46]. In climate insurance, those relations distribute authority among researchers, model providers, insurers, public authorities, courts and residents.
Comparing predictive performance is therefore not enough. We must also ask who controls the data, chooses the scenario, fixes the threshold and can challenge the consequence. A private model becomes a territorial decision in practice when banks, insurers and public authorities adopt its categories without independent expertise.
Models translate local environments into distributions of loss, capital needs and reinsurance products [47]. This translation is necessary to finance insurance. It captures only what enters the estimate of possible loss. Attachment to place, local activity and the ability to leave must be brought back when the institution decides what follows.
The Last Payer
The insurer stands at the end of a long chain. A public authority allowed construction. A bank financed it. A developer chose materials. The municipality maintains networks. The owner completes or postpones work. When water or fire reaches the house, the claim file brings together the effects of these decisions.
The insurer is neither responsible for the whole climate problem nor a witness without power. It cannot remake a watershed or move a neighbourhood. It does possess a particular form of knowledge. Claims bring together damage that had appeared isolated and reveal weak construction, fragile networks and territorial concentrations.
This position has historically led insurers to support inspection and safety standards. It creates a duty to warn and cooperate. It does not confer government of the territory.
The company itself is divided. Underwriting considers which risks enter the portfolio. Actuaries measure commitments. Finance follows capital and investments. Reinsurers, regulators, rating agencies, banks and model providers add further constraints. No single actor has to decide that a municipality should be abandoned. Each protects its own balance sheet, and their combined choices may produce withdrawal without a clear author [48].
Insurers are also investors [49]. On the liability side, they promise to pay future losses. On the asset side, they invest the resources supporting that promise. Through commercial underwriting, they also decide which activities to cover. A coherent climate policy cannot examine home insurance premiums while ignoring these other decisions.
The insurer’s role remains limited, but it is real. It can publish aggregate information on losses and non-renewals, recognise effective adaptation, finance assessments and warn before damage makes action more expensive. When it benefits from public reinsurance, a state guarantee or the administration of a compulsory scheme, those advantages may justify duties of transparency, continuity and prevention.
From Compensation to Adaptation
Insurance finances the loss that remains after reasonable precautions. It cannot make every accumulation of exposure sustainable for ever. With climate risk, insured loss is often formed long before a claim, through zoning, permits, standards, maintenance and work that was never completed.
The Sendai Framework distinguishes the creation of new risk, the reduction of existing vulnerability, preparedness and the financing of residual loss [50]. This sequence places insurance more accurately. It acts mainly when possible loss must be shared. It can also act earlier through data, advice, incentives and rules for rebuilding [51].
The floods around Valencia in October 2024 show the difference between paying and transforming. One year later, Spain’s public catastrophe compensation consortium had processed 98 per cent of claims for property damage and business interruption and paid about €3.84 billion [52]. The compensation system had mobilised substantial resources. Major flood protection works and lasting changes in planning remained slower [53].
An institution may be effective at settling an identified loss and much slower when it must finance a catastrophe that may never occur. Chapter 13 returns to this problem of the non-event. Here, it shows why the strength of an insurance regime cannot be measured only by the amount paid after disaster.
Many levers lie outside the policy. Flood damage depends on sealed surfaces, drainage, zoning and protective works. Wildfire damage depends on vegetation, electricity lines, materials, distance between buildings and evacuation routes. The risk of a home develops throughout its life, from planning and permitting to finance and maintenance [54].
When these decisions are postponed, their cost returns as a premium, deductible or non-renewal addressed to the household. Pricing then individualises a collective failure. This does not justify keeping prices artificially low. A price that conceals danger can support new construction and shift costs to future generations [55]. Bringing the price closer to risk without acting on its causes is not a prevention policy either.
Moving from sorting to adaptation changes the order of decisions. A map should not lead first to an increase and only later to a search for solutions. Where risk can be reduced, it should open an assessment, identify the actor able to act, estimate the cost and organise finance.
Insurance can contribute without governing the process alone. Municipalities control planning and some infrastructure. Banks finance property. Owners act on buildings. The state sets standards, redistributes resources and can carry losses that no private portfolio can retain. A risk signal becomes useful when these powers are connected.
Organising Retreat
In some places, remaining will become too dangerous or costly. Denying that possibility would prepare further disasters. Departure, however, cannot be left to non-renewal letters.
Managed retreat concerns timing, compensation, debt, tenants, owners, employment, public services and social ties [56]. It is not simply the movement of buildings away from danger. It requires decisions about who can leave, who may return and what the vacated land will become.
The contract can restrict this possibility after disaster. Compensation calculated for rebuilding in the same place may be insufficient to buy elsewhere, even when the household does not wish to remain [57]. Claims rules therefore govern geography as well as payment. Before loss, they should state how debt, replacement value and rehousing will be handled when rebuilding in place is no longer reasonable.
Retreat may also redistribute land. After the 2004 tsunami, safety zones prevented some fishing communities in Sri Lanka from rebuilding along the coast, while tourism projects received more favourable treatment [43]. A rule presented as protection from danger can displace residents and open valuable land to other uses.
A policy of retreat therefore needs public expertise, understandable information, contestable models, support for protective work, a transition timetable, relocation funding, tenant protection and rules for debt. Without such a procedure, climate is governed by an accumulation of private decisions. A premium rises, an insurer refuses, a bank withdraws finance and a municipality gradually loses its residents.
Private withdrawal creates a policy without a name. It does not announce that a territory is being abandoned. It says only that cover is unavailable, the deductible has risen or the market is under strain. Residents discover separately that the place where they live has already left the ordinary promise.
This leads to the next chapter. Calling a territory uninsurable may describe a real technical limit. Losses may be too correlated, too certain or too costly for a particular contract. The word may also hide the end of a chain involving models, reinsurance, delayed prevention, credit and public support. Before treating uninsurability as a state of the world, we should ask which institution can no longer carry the risk, with what coverage, for how long and after which opportunities for adaptation.
References
The CatNat surcharge is an additional compulsory contribution applied to French property insurance contracts. Since 1 January 2025, it represents 20 per cent of the part of the premium covering property damage in home and business policies. Its rate does not vary with the particular exposure of the insured property.↩︎