What does a fire settlement that does not go as planned actually cost? Part 3

Part 3: A worked example

In part 1, we looked at what a major fire does to the collateral, and in part 2, at what the documentation duty in the capital requirements regulation can cost the bank in capital. Here we make it concrete with a numerical example. Note: all figures below are stylized, but hopefully they can serve as a starting point if you want to swap in your own assumptions, or your bank's.

Assumptions

Assume a Krokstadelva-type row house has a market value of 4.5 million kroner before the fire, with the house accounting for 3.3 million and the land 1.2 million. We assume that rebuilding to today's building code (TEK) costs 3.9 million, more than the building was worth. That is not unusual outside the largest cities.

Picture two borrowers in the same row. Anne has paid down some of her loan and owes 3 million, a loan-to-value of 67 percent. Bjørn bought last year and owes 3.8 million, a loan-to-value of 85 percent.

At realization, we assume the bank applies a 15 percent forced-sale discount and 3 percent sales costs. An insurance payout to the mortgage holder, by contrast, is realized krone for krone. Full-value compensation is paid out as rebuilding cost only if the house is actually rebuilt, normally within five years. If it is not rebuilt, the market value is compensated instead. Here we assume that value is 2.8 million.

Two scenarios and a baseline

Realizable amount is market value less a 15 percent forced-sale discount, then 3 percent sales costs. Interest and costs during the period of default are not included.

Scenario A confirms what we wrote in part 1: a settlement that works makes the collateral better. Bjørn's loss given default (LGD) actually falls to zero, because the building is new and worth more than the old one. The bank has had its collateral upgraded, paid for by the insurance industry.

Scenario B is the opposite. The municipality refuses rebuilding, or sets conditions that cannot be met. The customer is paid the market value instead of the rebuilding cost, and the land is not compensated, because the relocation cover in natural damage insurance law is triggered only by a natural disaster, and fire is not a natural disaster. Bjørn's loss rate goes from 2.4 to 22 percent. He has not missed a single payment along the way.

Notice how much the loan-to-value ratio matters. Anne, who had paid down her loan, comes through the same scenario almost unscathed. The difference between the two is not the risk in the property, but how much buffer was there when it happened.

The hidden cost

Let's explore what happens in these scenarios to a borrower who lacks contents insurance. This is a plausible scenario. The Krokstadelva fire produced several examples of households left without cover, and one in ten Norwegian households today lacks contents insurance, according to the Consumer Council of Norway (see part 1). Crowdfunding campaigns are running for households that lacked building insurance (see this one and this one), and for households that lacked contents insurance, in the wake of the Krokstadelva fire.

Anne has building insurance, and the building is fully insured for full value. That means temporary housing is covered at market rent for the whole rebuilding period, and the mortgage itself does not become any harder to service than it was before the fire (assuming she keeps her job, which we will assume has nothing to do with the fire). But Anne does not have contents insurance. Building insurance only covers what is fixed to the building itself (kitchen fittings, bathroom, flooring), not belongings: clothes, furniture, electronics, kitchenware, everything she brought with her when she moved in. According to Finance Norway's contents insurance calculator, an ordinary Norwegian household's belongings have a replacement value of between 1 and 2 million kroner. Anne has to cover that herself.

On top of that, she loses what an ordinary contents policy would otherwise have covered, for example help documenting and valuing everything that was lost, and other extra costs that arise after a fire.

For Anne, this means she has to use her own funds or take out a consumer loan to refurnish, while the mortgage keeps running unchanged alongside it. What should interest the bank is how much other, more expensive debt Anne takes on at the same time as she services the mortgage. The bank will likely offer a payment holiday if her ability to pay falters anyway, but that postpones the burden rather than removing it, since interest normally keeps accruing and gets added to the loan balance. And a payment holiday is not free for the bank either. Under current accounting standards (IFRS 9), forbearance granted under distress can reclassify the loan as "in default" or "credit-impaired", which means the bank has to set aside more money against potential losses. The bank therefore has a concrete incentive to keep track of whether the customer has adequate insurance before a fire happens, not reactively after the customer has called in desperation.

The loss rate is unchanged. The probability of default (PD) is not.

In scenario A, Anne still has to replace everything she lost, but here she has a timeline: the house will be finished on the same plot, say within a year, and a stripped-down, temporary existence, borrowed furniture, minimal kitchenware, only the essentials of clothing, is manageable when you know it is temporary and when it ends.

In scenario B, that timeline does not exist. The municipality has refused rebuilding on the plot. Anne gets a cash payout instead of a finished house, and she has to establish herself somewhere entirely new, permanently, not temporarily. The whole cost burden of furnishing a full home arrives at once. She has no chance to build up funds first.

Notice what the LGD figures in the table above do not capture: they only measure the security the bank has in the building and the land. Whether Anne gets a new, better house (scenario A) or a cash sum and no plot to build on (scenario B), the contents bill is equally real in both, but the room to postpone it is completely different.

Now imagine rebuilding does not take one year, but four. That is not a made-up number. In Nesbyen, after the storm named Hans, some 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 was a condition for full insurance settlement. If rebuilding takes a year, a household that lost all its belongings might manage temporarily: borrow the essentials and postpone full re-establishment until the house is finished. If rebuilding stretches over four years, that is unlikely to be a real option (and even less so in scenario B, where there is not even a date to wait for).

Expected loss

The calculation for the bank's expected loss is PD times LGD times exposure. Here we have to assume something about the probability of default. It is worth being clear that this is an assumption, not an estimate from any model. We set PD at 0.4 percent under normal circumstances, and 2.0 percent for a household that has had to take out a consumer loan to replace lost belongings, on top of an unchanged mortgage.

Bjørn under normal circumstances: 0.4 percent × 2.4 percent × 3.8 million gives 365 kroner a year in expected loss.

Bjørn in scenario B, with no contents insurance: 2.0 percent × 22 percent × 3.8 million gives 16 720 kroner a year.

Forty-six times higher, on the same customer, in the same home, with the same payment history.

And then the portfolio

Assume 80 of the 113 homes in Krokstadelva were mortgaged, with an average loan of 2.5 million. That is around 200 million kroner in exposure on a single slope, likely spread across two or three banks.

If everything goes by the book, the event costs the banks almost nothing. If a quarter end up in scenario B, we are talking on the order of 10 million in losses spread across a handful of institutions.

That will not bring down any bank. That is not the point either.

The concentration risk is unknown

The point is that the whole range, from zero to 22 percent loss rate, is decided by three things no Norwegian bank monitors today.

The first is whether the insurance is sufficient, and whether it still exists. We looked at that in part 2: the duty to monitor this is already written into the capital requirements regulation (CRR), article 208(5), and there is no register that makes it possible to comply with it.

The second is what the municipality decides. Whether rebuilding happens, on what conditions, and how long it takes. That is decided in a planning resolution the bank is not a party to, is not notified about, and cannot appeal.

And the third, which does not appear in the table at all: 200 million kroner sat on a single slope without anyone knowing it in advance. The concentration was there long before the fire. It just was not measured, because no one measures property concentration in the unit a fire actually spreads through.

The numbers above are stylized. The concentration risk is not.

This is part 3 of three. Part 1 covers what a fire settlement does to the collateral. Part 2 covers the documentation duty in the capital requirements regulation.