Quantifying flood risks for Dutch residential mortgages

Everyone living in the Netherlands at the time cannot forget the images of the dramatic and tragic flooding in Limburg province in the summer of 2021. Fortunately, no one died when the Geul, Roer and Geleenbeek rivers broke their banks and wreaked havoc. However, there was a huge amount of material damage. At the time, the Netherlands-based consultancy firm Zanders, was already focussing on climate risk management for financial institutions.

In response to the 2021 flood event, Zanders looked into quantifying the potential impact of flood risk on lenders’ housing mortgage portfolios, after being asked to investigate by a leading Dutch retail bank with an extensive residential mortgage portfolio. We discuss this investigation with Sjoerd Blijlevens, a Senior Manager at Zanders.

Could you outline your case study of the physical climate risk of a residential mortgage portfolio?

We did this study in cooperation with one of the larger Dutch retail banks which was interested in getting an initial assessment of how big the risk of climate change could be from a credit risk perspective on their substantial residential mortgage portfolio throughout the Netherlands.

What was your method?

We concentrated on looking into the materiality of two types of physical risk: wooden foundation pillar rot (paalrot) and flood risk. The study initially concentrated on flood risk. In order to determine the potential impact of flood risk, we first collected all the relevant information on the properties in the residential mortgage portfolio. One key component was the geographical location of the properties. We needed to know where each home was located, on a more granular level than postal codes. We also collected financial information related to the mortgages such as the loan amount and the market value of the home to determine the Loan-to-Value (LtV) ratio for each individual property. Finally, we collected information on flood risks, using the Climate Impact Atlas. This is a public source where you can find information on the probability of floods occurring in a certain area and what the potential flood depth could be for that location. This information is available for different climate scenarios.

What happened next?

At that point we had information on the properties’ underlying the residential mortgage portfolio, and information from a climate perspective in terms of flooding. The next step was to translate this flood risk into a potential financial impact for this bank. We reviewed a number of different approaches at that point. In the end, we opted to use damage functions. This method relates flood depth with property characteristics, such as what type of home it is, in order to estimate the potential impact to correct for, or offset, the damages.

Were those damage functions graded into flood levels, say 50 centimetres, then a meter, then 2 metres, and so on?

That is effectively how it works. Damage functions are depicted on two axes: on the horizontal axis you have the flood depth in meters, and on the vertical axis you have what is known as the damage factor, which runs from zero to one. A damage factor of one corresponds to €1,000 of damage per square meter. So, if your apartment is on the 3rd floor of a high rise, then there will be no direct damage to your property unless the flood level increases beyond four meters or so. However, if you have a family home, then a one-meter flood depth will already have financial consequences.

What were the advantages of this method?

This method allowed us to determine the link between potential floods and the financial cost for the owners of those homes. We then related this to IFRS 9 calculations (IFRS 9 is an accounting standard used to classify and measure assets and liabilities, also requiring expected credit loss calculations – ed), which included assumptions about the housing prices development over the next few years. We also included scenarios in which housing prices decreased, which would imply elevated risks.

What was the primary aim of your research?

The main motivation for this study was for the bank to get an idea of how material flood risk could be. To see if they might have to dive into this risk type in a lot more detail.

What were your key findings?
That the impact of flood risk on the expected credit loss that we calculated was more limited than we had anticipated beforehand.

Did that shock you?
We were certainly surprised. And even though there are certainly possible improvement points that you can make going forward, it is also important to realize that we were exploring expected credit losses, which is quite different than unexpected losses. Things could turn out quite a lot worse if you have a major flood. Obviously, many properties would be affected at the same time in that case. I think it is also prudent to be aware of what the damages could occur in an unexpected flood loss scenario.

Was this a positive conclusion?

For the bank it was a positive finding, and therefore for the homeowners too. Not all areas in the Netherlands are equally exposed to floods: the eastern and southern parts of the country are a bit higher, so there the flood risks primarily exist close to the main rivers. Flood risks are more pronounced in the western part of the Netherlands. Another factor that came into play was the strong increase in housing prices over the past couple of years that has resulted in relatively low LtV ratios. This means that you need to experience quite a bit of damage for the bank to incur a loss.

And banks obviously want to avoid losses?
Indeed. Problems for the bank potentially arise if damages cause property values to drop below the outstanding loan amount. If the homeowner goes into default at that time, the bank may end up with a loss.

All of us living in the Netherlands have seen the TV images of the extreme floods in Limburg in July 2021, how do those type of events fit into the analysis which you undertook in terms of climate risk?
Those events were an example of what we were attempting to quantify during our investigation. It is interesting to see that in the aftermath of the Valkenburg flooding that banks incurred hardly any losses. Much of the damage was covered by insurance companies, and also the government stepped in. So, it turned out not to be a substantial credit risk for banks in that specific case.

Surely housing prices were affected?

We made an analysis of how the Limburg flooding impacted housing prices in the region. We found that even though there was a short-term drop in housing prices soon after the flooding, prices quickly bounced back in line with the national trend.

So, the market “forgot” about the flood?
Quite soon afterwards, it seems. The main risk would be that if at some point consumers as a group start to recognise that in certain areas floods will occur much more frequently than we are used to, and that at some point that gets incorporated in housing prices. That was clearly not the case here.

What was for you the most interesting part of this case study?

The fact that we actually managed to start quantifying flood risk. In the area of climate risk, quantification is still a major challenge. This is partly due to data issues (lack of data and overall data quality) and also because a process needs to be established for translating a certain probability of a climate risk event into the financial metrics and risk measures with which a bank is familiar. That was an interesting process and I think we made progress towards the quantification of these types of risks.

What do you personally hope to gain from attending the Sustainable Finance and Climate Risk Event?

This whole climate risk field is strongly in development, and it will be interesting to be in touch with others working in the same field to see how they are progressing and to share experiences. To a certain extent we are all trying to invent the new wheel at the same time, and working together and sharing experiences will help to speed up the process. I hope we can find ways to improve comparability and consistency across the market.