The fire that wasn't in the model: Part 1

What Krokstadelva does to the collateral

When a row house in Krokstadelva caught fire on Friday 17 July, it was an entirely ordinary fire. A few hours later, 113 homes were gone and over 400 people evacuated, making it the largest residential fire in peacetime Norway since Hemnesberget in 1923. What made the difference was not the house fire itself, but the conditions around it. Dry vegetation, strong wind, high temperature.

Climate change is producing that combination more and more often, and it is barely present in Norwegian credit assessments.

Summer 2026, in short

While Krokstadelva was being put out, southern Europe was burning. Spain has passed 244 000 hectares across around 400 fires so far this year, six times the average of the last 20 years, and Spanish authorities describe the fire in Ávila as the country's largest in recent history. In Gironde, France, around 220 000 people had to be evacuated, and French authorities describe the fire as one of the largest since the Second World War. There have been major fires in several other countries, for example Croatia. Europe is warming more than twice as fast as the global average.

It is tempting to dismiss Krokstadelva as one unlucky fire in a housing estate, and the fires in France, Spain, or Croatia as one season of Mediterranean wildfires. But they are part of a pattern we are likely to see more and more often. One spark becomes a catastrophe when the fuel is dry enough, the wind strong enough, and the distance to vegetation short enough.

Fire is not flood, it is worse

Norwegian banks have one comfort when thinking about physical climate risk: the natural disaster scheme. Flood, landslide, storm, and storm surge are covered broadly, and insurers share the bill through the Norwegian Natural Perils Pool. It is a good scheme, but fire is not included.

A wildfire or a fire that spreads from one building to the next is an entirely ordinary fire claim under standard building insurance. The natural disaster premium is set by law and is the same for everyone regardless of location. Fire insurance has no such protection. It can be repriced, given stricter terms, and in principle an insurer can refuse to renew it. Norway has built a solidarity-based shield against flood and landslide, and skipped over a peril that could grow quickly with drier summers. That is a paradox, given that the Directorate for Cultural Heritage points out that few countries have as high a share of timber housing as Norway. So when 113 homes on the same slope were lost at the same time, a handful of insurers were left holding fire insurance payouts, and a handful of banks were left holding burned-down collateral.

Insurance veteran Sverre Bjerkeli estimated early on that Krokstadelva could cost over one billion kroner. Fremtind has registered 170 claims, Gjensidige 76, If around 70, Tryg close to 60, Storebrand 38, and JBF 31 claims, with estimated losses of up to 150 million. That is not a dramatic amount for the Norwegian financial sector, but it is a significant amount for one small area, and for the individual buildings in question. Assume that 80 of the 113 homes were mortgaged, with an average loan of 2.5 million kroner. That is around 200 million kroner in collateral on a single slope, likely concentrated in two or three banks. The risk does not disappear, it moves from one pocket to another.

The seven gaps insurance does not close

The assumption that "insurance covers it" is not directly wrong, but it is imprecise on at least seven points that all land on the bank directly, and that make fire risk, like climate risk, credit risk.

From the bank's side, a full settlement after a fire can be good news for the collateral. A building from 1975, with old wiring, the possibility of unknown defects, and an unclear maintenance backlog, gets rebuilt to today's building code, and becomes more fire-safe: fire compartmentalization, new electrical systems, balanced ventilation, and more energy efficient. The loan's remaining term is now shorter than the building's lifespan, and in many places construction cost will exceed the local market value.

But is this the whole picture, and is it only the settlement for a totally destroyed building that matters to the bank? The assumption has at least seven gaps, ones the bank currently has limited insight into, since it does not monitor the status of the customer's insurance (more on that in part 2).

Underinsurance. Full-value insurance is meant to cover rebuilding an equivalent building. But only what is legally required, and what the borrower should not already have fixed themselves before the fire. Any changes the borrower wants in the rebuilt structure beyond this must be covered by the borrower themselves. If they extend it or raise the standard without disclosing it, a reduction in payout quickly follows. Commercial buildings and holiday homes often have a fixed insured sum, so construction cost inflation hits directly. Older buildings with a maintenance backlog do not always get full value at all.

Relocation and land value. If a business customer's premises need to be relocated, the customer must have taken out insurance that specifically covers this. There is no automatic guarantee that either a new building or the land is covered. If a policyholder's home is hit by fire and a building ban or other conditions that are impossible to meet follow, the customer gets the building itself compensated. But not the drop in the value of the land. This is where fire differs from natural disaster damage. If a home is hit by flood or landslide and the municipality refuses rebuilding due to the risk of further natural damage, the insurer is also obliged to compensate the land's market value from before the damage. That obligation is triggered by a natural disaster, and fire is not a natural disaster. The land does not go to zero because of this. It still has road, water, and drainage access, it has a location, and it can have a use other than housing. But according to a 2022 analysis by Samfunnsøkonomisk analyse for the Ministry of Local Government and Regional Development, land value makes up between 20 and over 40 percent of a home's sale value, depending on centrality. The gap between what a buildable plot is worth and what a restricted plot is worth can easily be a million kroner or more. No one covers that gap. It stays with the borrower, who is still servicing a loan granted against the whole property, and therefore with the bank. In Nesbyen, after the storm named Hans, flood victims are still living outside their homes three years after the event, because a municipal building and subdivision ban pending a flood barrier assessment is a condition for full insurance settlement, and the municipality's decision has been delayed. The municipality also says that building applications where safety measures following Hans have not been carried out will be rejected until those measures are in place. What will the municipality of Drammen do about safety requirements and any conditions in Krokstadelva? How long will it take before a building and subdivision ban is finalized, and how long will it last?

Collateral value moves independently of the settlement. Even with a full settlement, land values in a burned housing area can fall, and they fall for all the neighbors too, including those who had no damage at all. We see the same pattern where properties become hard or impossible to insure: market value accounts for the risk long before anyone has actually made a claim. As this is being written, several homes in and around the fire-affected area in Krokstadelva are listed for sale. We do not know how many of them were listed before 17 July, how many of the sellers have already bought something else, whether they will actually manage to sell, and if so, whether they get the price they had counted on. For a seller who has already bought a second home, the answer to that last question is not academic. It decides whether the bridge loan runs for three months or eighteen. And a bridge loan is the bank's exposure, granted against an assumed sale price, in a market that changed within a few hours on a Friday in July. The neighbors who escaped without damage have no insurance claim. They have a property in an area that will, for the foreseeable future, look like a building site, in a location any prospective buyer can now google. Collateral value can move for the whole neighborhood. Insurance only covers those who lost something.

Time horizon. A loan for a home or a commercial building generally runs for at least 25 years, often longer, even though it is typically refinanced along the way. An insurance policy is renewed every year, can be repriced, and can in principle be declined. In southern Europe and parts of the United States, that mechanism has already kicked in, and the insurance gap is growing faster than the losses. In Norway this has not been an issue, because Norway has not been a country with climate-driven major fires in residential areas. Now Norway is a country with climate-driven major fires in residential areas.

Cost. Fire risk is not pooled between insurers, and can therefore be priced freely. That makes it the one physical climate hazard where a customer can, in principle, be priced out or refused cover altogether. We look more closely at pricing, and at what it does to the bank's capital requirements, in part 2.

Missing cover. Building insurance is close to universal in Norway, and for good reason: the bank requires it at loan approval, and natural disaster cover follows automatically with fire insurance. Note the asymmetry here: fire insurance is what triggers natural disaster cover, but fire itself is not one of the perils the natural disaster scheme covers. And "close to universal" is not the same as "all". Just under 20 households in Krokstadelva were fully or partly uninsured. A couple said they believed home insurance came included in the package they bought with the property. The Consumer Council of Norway estimates that around one in ten households lacks contents insurance. The loan does not disappear even if the collateral does. Can every household that has to buy back all their belongings from scratch still service the mortgage?

Claim reduction. If a building has been hit by the same type of damage multiple times, and/or the owner has received an order from their insurer to make repairs but has not done so, the insurance payout can be reduced if a fire occurs. This has likely been more common in connection with flood, landslide, and storm surge so far. But as a result of increased focus on fire exposure, and on what increases or reduces actual risk, insurers may well become more eager to have customers clarify whether they have ignition sources (for example an arborvitae hedge, old electrical wiring, a non-compliant EV charger, a missing stove guard, electrical machinery, e-bikes, e-scooters, or solar panel installations) near the building. Then a reduction in payout quickly follows if the burned site shows that the information the customer gave was incorrect.

What this means for credit work

How exposed to fire a mortgaged property is depends on four things: the vegetation, terrain, and soil around the building; the weather, drought, and wind at the time; emergency response and water access nearby; and the building itself, meaning roofing material, year built, and safety measures. A bank issuing a 25-year loan cares little about acute conditions like tomorrow's wind forecast, but a great deal about the other three. Two neighboring homes in Stenseth can have very different fire risk: one might have a shorter distance to the forest edge, a steeper slope, a narrower access road, and drier vegetation. That is why Telescope's climate risk exposure score weights historical insurance claims data, so the difference between two neighboring properties can be quantified, not just described. Telescope covers fire risk today, and works with partners to break it down the same way we already do for our existing surface water flood interfaces.

Norwegian banks are still learning how climate risk fits into the credit process, and in practice the work so far has focused on flood, surface water, and landslide. That is where the data exists, and that is where EBA's expectations have bitten first.

Fire, by contrast, has been treated as an insurance matter. Krokstadelva is a reminder that this treatment rests on assumptions: that fire insurance is priced the same for every building, that the settlement always comes, that it is always sufficient, and that it always makes the collateral whole again.

The practical question is not whether the bank should start refusing loans near the forest edge. It is whether the bank knows where in its portfolio the forest edge actually is. Concentration is the variable that separates a manageable single loss from a real problem: how many mortgaged properties (or insurance claims, for that matter) sit within the same fire spread zone, with the same access road, in the same type of vegetation? These are questions we at Telescope work on every day, and can offer banks and insurers solutions to help manage.

And then comes the hard question

Say the bank actually manages to answer where in its portfolio the forest edge is. Can it document that the mortgaged properties are insured at all, and that the insurance is sufficient?

The regulation assumes the answer is yes, and has done so for a long time. In practice, the bank checks this at loan approval, and after that it is unclear whether much more happens. We look at that more closely in part 2, which covers the documentation duty in the capital requirements regulation and what it can cost. In part 3, we calculate what a fire settlement that does not go as planned actually does to the loss rate on a single loan.