The surface water trap: risk for banks and insurers - Part 2

Part 2: The documentation requirement and surface water

In part 1 we saw that surface water is rarely one thing in legal terms. It may be covered as natural perils damage if it comes from a “wild stream” (villbekk). If it is surface runoff or sewer backflow, it may be only partly covered and subject to a reduced payout. In either case it is priced more freely than the statutory natural perils premium. Flooding and surface water are often caused by the same storm but are insured differently, and the flood zone maps banks consult only show flooding. Here in part 2 we look at how that uncertainty can affect the capital requirement when the bank cannot say precisely whether a property held as collateral is actually adequately insured against this particular risk.

The requirement to know whether collateral is adequately insured already exists

The Capital Requirements Regulation (CRR, Article 208(5)) states that the property taken as collateral must be adequately insured against the risk of damage, and that the bank must have procedures in place to monitor that the insurance is adequate. Some banks have permission from the supervisor to calculate their capital requirements with internal models. These are often called “IRB banks”, and they include the Nordic commercial banks, DNB, Sparebanken Norge, the regional banks in the SpareBank 1 Alliance, BN Bank and Sparebanken Møre. IRB banks are exempt from this requirement for exposures older than 2025. Other Norwegian banks use the standardised approach for their risk weights (“standardised banks”), and they must meet the insurance requirement for all exposures.

The EBA’s guidelines on environmental scenario analysis (EBA/GL/2025/04) point in the same direction and appear to be less tolerant of the IRB banks. They ask banks to take account of existing and future gaps in insurance coverage. They also describe reduced coverage in climate-exposed areas as a mechanism that can drag down values and creditworthiness (paragraphs 26 and 57; the guidelines apply to Norwegian banks from 1 January 2027).

Both sets of rules assume that the bank can actually see the risk the insurance is meant to cover. For surface water, it usually cannot. Most banks consult the flood zone maps from NVE, the Norwegian Water Resources and Energy Directorate, and those maps do not show surface water. They can also be fairly outdated and coarse for flooding.

For surface water, a check at loan approval is not enough to meet the insurance requirement. A property that was adequately insured at approval can have its cover reduced after a damage event without the bank finding out, and without anything having changed with the building or the ground itself. Physical changes can also occur during the life of the loan that increase the risk, and these vary between buildings and locations. If green areas on or near the plot are built over, water can suddenly run into the building during a downpour instead of soaking into the ground as it did before. That risk grows as climate change makes downpours more intense, larger in volume or both, especially for buildings that are poorly located within a catchment.

…but do banks actually know the answer?

As mentioned in the article on fire risk, Norway has no central register of current building insurance policies, nor of which policies have been reduced after previous claims. A bank can hardly look up an address and find out whether surface water cover is still in place today. Nor can it see whether the customer received only half the payout last time because a required protective measure had not been carried out. The insurers know this. They also know which customers have seen their premiums rise rapidly because of high surface water risk, and which customers pay high premiums for the same reason. As far as I know, however, insurers have no obligation or routine to tell the bank about any of this, or to inform the bank if a policy is cancelled.

Customers can also spend a long time in limbo before their insurance is settled. After Storm Hans in 2023, 149 homeowners in Nesbyen received notice of a possible ban on building and subdivision. The event involved both flooding and surface water, and flooding was probably the main problem, but it is still a relevant example. Nine months later, in May 2024, 117 of the homeowners had been cleared and 12 had received a permanent ban. Twenty properties remained in what the municipality itself called a “yellow zone”, neither cleared nor finally resolved. Only in September 2026, more than three years after the event, was the flood protection project approved, so these properties will probably now get an answer. The example shows how long a case can stay open when several parties need to agree. An ordinary flooded basement is unlikely to make the national news. The case will probably stay marked as “insured” in the credit file, because no one has any reason to check again.

The Nesbyen case also shows that “who decides” is not always a purely technical insurance question. In autumn 2023, the Norwegian Natural Perils Pool and the municipality disagreed about what required a building application, and several insurers asked their customers to hold off on applying. The longer the application was delayed, the longer both a refusal and the right to full compensation were delayed. Neither problem is unique to river flooding. The same uncertainty, on a smaller scale and without TV cameras, can arise around a perfectly ordinary surface water claim. Is the damage covered, who decides, and how long can the answer wait before anyone really cares?

Suppose Finanstilsynet, the Norwegian FSA, were to require documentation of ongoing monitoring of the bank’s insurance risk related to surface water. Banks would then have, in practice, the same three options as for fire, with the same weaknesses:

• Annual self-declaration from the customer. This is cheap, but it has limited documentary value and does not capture reductions in cover whose implications the customer has not understood.

• Data-sharing partnerships with insurers. These make sense professionally, but they may require both privacy clearance and technical integration.

• The bank’s own cover against collateral losses. The EBA opens the door for banks to buy this. The cost then has to be priced into the margin, on a risk the bank today hardly measures precisely enough to price correctly.

What it costs not to be able to answer

If the collateral is not recognised as security for a loan because the bank cannot document adequate insurance against, for example, surface water, the risk weight on the loan can jump sharply. Our calculation is for a bank using the standardised approach. Under CRR3, such a bank splits the mortgage in two. The part within 55 per cent of the property’s value gets a risk weight of 20 per cent, and the rest gets 75 per cent. We take a loan of NOK 3 million secured on a home worth NOK 4 million, a loan-to-value of 75 per cent. All figures are stylised.

With recognised collateral, the risk-weighted exposure is NOK 1.04 million, an effective risk weight of just under 35 per cent. Without recognised collateral, we assume the whole loan gets 75 per cent, which gives NOK 2.25 million. The risk-weighted exposure grows by NOK 1.21 million, to about 2.2 times the original amount, and the CET1 capital requirement rises with it. With a stylised requirement of 15 per cent, the requirement goes from NOK 156 000 to NOK 337 500. That is NOK 181 500 more capital behind a single loan. If equity costs 7 percentage points more than debt, the extra capital costs roughly NOK 12 700 a year, or 0.42 per cent of the loan. The jump is larger for loans with a lower loan-to-value. At 50 per cent, the starting weight is 20 per cent, and the risk-weighted exposure rises from NOK 0.6 million to NOK 2.25 million, or 3.75 times.

Say the standardised bank has a hundred such loans in the same catchment. Together they total NOK 300 million, with a combined risk-weighted exposure of NOK 104 million. If 20 of the loans lose recognised collateral:

• the risk-weighted exposure increases by NOK 24.2 million (23 per cent)

• the CET1 capital requirement increases by NOK 3.6 million

• the capital cost increases by around NOK 250 000 a year

Banks with internal models (IRB) see the effect through their models’ loss given default and through the output floor, so the calculation is an illustration of the mechanism. What triggers the jump is not a weaker ability to pay or falling house prices. It is that the bank cannot produce a documented answer to a simple question about a house that is still standing.

As mentioned in the article on fire risk, a realistic outcome is hardly that an entire portfolio is disqualified overnight. More likely, the supervisor makes a remark in its assessment of capital needs (SREP), sets a deadline and adds a possible Pillar 2 add-on if the deadline is missed. Such an add-on could hit the entire portfolio, not just the individual loans in question.

The missing interface

A common approach to geographical concentration is to draw a radius around each building. That approach is poorly suited to surface water, because water does not follow a circle. It follows a catchment and a drainage network. For both the bank and the insurer, the relevant unit is therefore the catchment defined by a municipality’s surface water map, not the straight-line distance between two buildings. Banks and insurers both have an interest in the same buildings, and both are supervised. Even so, they have no ongoing channel between them today for sharing how the risk in an area is changing and how it is split between them over time.

Questions for the bank’s own portfolio

These are the questions I would ask myself if I worked in a bank today:

• How large a share of lending, and of collateral value, sits in areas of high surface water risk?

• Which of these buildings have water damage cover that holds up, what is their claims history, and have required protective measures actually been carried out?

• How many properties held as collateral are concentrated in the same catchment or low point, and what is the total exposure there?

• Is surface water included as a separate scenario in the bank’s ICAAP, distinct from river flooding, or is everything treated as a single flood category?

• If a surface water case is left unresolved between customer, municipality and insurer, how long can it take before anyone at the bank picks it up?

• What thresholds has the board set, and what happens when they are breached?

• Where in the credit decision is the figure used?

Conclusion

Surface water is usually less dramatic than a major fire or a river bursting its banks. Usually it is just a basement. That is exactly why it is easy to underestimate as a credit risk. Each event is small enough to be handled as an individual insurance claim, and never big enough to trigger a review of the whole portfolio. The banks that are well prepared know which properties share a catchment before the next cloudburst, not after.

This is part 2 of three. Part 1 covers what surface water does to collateral, and why the risk falls into several different coverage categories. Part 3 is a worked example that puts numbers on what repeated damage does to loss given default.