Opening the floodgates? Modelling spillovers from flood insurance protection gaps to UK mortgages – Bank Underground

Will Banks and Kemal Erçevik

When extreme weather hits, households typically turn to insurers to cushion the financial blow. But rising temperatures and greater exposure in high-risk areas could test the insurance sector’s capacity to absorb such losses. As the Financial Policy Committee has highlighted, climate change could create insurance protection gaps, leaving households vulnerable and shifting risks across the financial system. We have built a model to estimate potential protection gaps, finding that – under conservative assumptions – the share of UK mortgagors uninsured could increase from 5% today to around 7%–10% in 2050, or up to 16% following a severe flood event. While this would have substantial welfare implications, our model suggests the aggregate impact on lenders would be small compared to previous financial crises.

How might insurance protection gaps arise?

In the UK, most mortgagors have combined buildings and contents insurance covering flooding. That’s in part due to Flood Re, which subsidises insurance for high flood-risk households.

But that could change as temperatures rise. Under a very conservative emissions pathway scenario – representative concentration pathway 8.5 – floods are likely to become significantly more frequent and severe.

Alongside the planned end of Flood Re in 2039, this could push up insurance costs and reduce coverage. Imperfect information or limited appetite to underwrite the highest risks could lead insurers to withdraw from certain areas – as seen in California following recent wildfires.

Lower insurance coverage could lead risks to shift across the economy and financial system. House prices could fall if potential buyers start to factor in higher insurance costs. Banks could face higher losses on mortgages as collateral values fall, or if borrowers become more likely to default due to higher premia or uninsured flood damages. We have built a model which captures these dynamics in a stylised way (Figure A).

Figure A: A stylised figure of our insurance protection gaps model

First, we combine a six-digit postcode-level hazard exposure vulnerability model based on UK flood risk estimates (covering river, surface water, and coastal flooding) with an insurance pricing model and borrower-level data on mortgages and incomes. Then, we estimate expected flood losses and home insurance premia. We plug this into our existing scenario analysis toolkits to translate higher premia into house price and credit risk impacts.

As with all climate and insurance models, the results depend on many judgement calls. We use a stylised set of hazard models and a severe long-term climate scenario which is subject to deep uncertainty. While the model is well-calibrated to the UK financial system, the insurance premium, house price, and stress-testing models we use are illustrative, not definitive. So, the estimates here should be treated as indicative.

We use the model to answer three key questions.

1) How big might insurance protection gaps become?

In our modelled 2025 scenario, 95% of mortgagors have building and contents insurance and we assign each household an insurance premium, averaging around £430 per year (aqua bar in Chart 1). By 2050, we assume Flood Re ends and flood risk rises. We conservatively assume that insurers no longer cover previously subsidised properties, and that mortgagors stop paying for insurance if the premium increases above a threshold share of their income.

This leads the protection gap to as much as double, from 5% today to 6.8-10.2% in 2050 – not far from insurers’ 2021 estimates. This could leave as many as 910,000 mortgagors without flood insurance (orange bar).

That 6.8%–10.2% gap assumes insurers have perfect information about increases in flood risk. But the uncertainty and complexity in modelling risks at the property level mean information is typically imperfect, and therefore extreme events often lead to updated perceptions of hazard risk and higher premia. To capture this, we simulate an illustrative 1-in-100 flood year under RCP8.5 and assume that premia increase rapidly for those houses flooded. In that scenario, the protection gap increases substantially to around 1.39 million mortgagors (15.7%) in 2050 (purple bar).

Chart 1: Estimates of the mortgagor insurance protection gap

Note: Under scenario RCP8.5. Sample of 6.85 million mortgages has been upscaled to 8.8 million to reflect the whole UK mortgage market. Mortgage stock as of end-2024. Flood Re assumed to end in 2039. Estimates are subject to substantial uncertainty. ‘Uninsured’ defined as any properties for which insurance is estimated to be unaffordable (ie above a certain percentage of gross income, calibrated for each income decile using ONS Living Costs and Food Survey data) or unavailable. Aqua bar shows the approximate current coverage gap. Orange bar shows range of gap estimates under higher climate risk (RCP8.5 2050), no Flood Re and various risk reflective pricing models. Purple bar reflects houses uninsured after a 1-in-100 flood year leads to high markups in insurance premia for those houses flooded.

Sources: FCA Product Sales Data, Mitiga, RiskLayer, ONS Living Costs and Food Survey and Bank staff calculations. Full sources are available upon request.

Some assumptions we make are conservative – for example, that households’ willingness to pay for insurance is fixed over time, even as flood risk increases. But in other ways, we could be understating the risks. The hazard model we use is highly uncertain, and does not capture non-linearities or tipping points in the earth system, and we only capture the impact of flooding, not other hazards like windstorms or subsidence.

Either way, increases in flood risk under climate change could leave an increasing number of mortgagors uninsured against the negative impacts of extreme weather events. Notably, mortgagors without adequate insurance may find it harder to remortgage, particularly if lenders view uninsured properties as higher risk – further amplifying financial vulnerability in affected areas.

2) What could that mean for house prices?

Protection gaps could have a range of substantial social and economic impacts – particularly for households left uninsured. We focus on their financial impacts, asking if related house price falls could be large enough to disrupt the provision of vital services to UK households and businesses.

Building on previous Bank work, we model three channels through which house prices are discounted: higher expected insurance premia (in net present value terms); estimated unaffordability/unavailability of insurance, and expected flood damages for uninsured homes.

In aggregate, the house price falls in our model are small compared to those seen in previous financial crises, at around 1%–3% in our central case, or 3%–5% following an extreme weather event. This is much lower than the year-on-year fall of 15.6% seen after the global financial crisis, or the 28% assumed in the 2025 Bank Capital Stress Test scenario.

But national averages mask large impacts in some regions: Chart 2 shows that as many as 18% of mortgaged properties could see a fall bigger than 10% following a severe flood event, and almost 3% could experience falls over 30% under conservative assumptions. These estimates fit with estimates the Bank published previously, but this model tests further sensitivities and scenarios.

Chart 2: Distribution of falls in house prices under different scenarios

Note: Under scenario RCP8.5. House price falls reflect the net present value of expected increases in insurance premia, discounts for estimated insurance unaffordability or unavailability, and the value of flood damages for uninsured properties. Orange bars are consistent with the 6.8%–10.2% protection gap in Chart 1, purple bars with the 15.7% gap. Ranges reflect different discounting and insurance pricing assumptions.

Sources: FCA Product Sales Data, Mitiga, RiskLayer, ONS Living Costs and Food Survey and Bank staff calculations. Full sources are available upon request.

3) Could these effects lead to losses for banks?

To explore spillovers from house price falls to banks, we use a simple stress testing model. We calculate current and stressed loan to value and debt-servicing ratios to capture the impact of lower house prices and higher household expenditures on the probability of default and loss given default of UK mortgages.

This model captures the direct impacts of insurance premia and lower house prices on affordability, but does not capture the indirect impacts of flooding on consumption, leverage, output or inflation. So, our analysis is partial.

Intuitively, increases in mortgage impairment rates are small relative to large macro stresses to which UK banks are stress tested, even under severe assumptions (Chart 3). This effect is not even across the country though – Chart 3 shows the impacts could be nearly four times the average in higher flood-risk regions.

Chart 3: Impact of different scenarios on two key measures of bank credit risk

Note: Under scenario RCP8.5. Expected impairment rates reflect changes in loss given default and probability of default due to changes in loan to value and debt-service ability ratios due to higher flooding and lower insurance coverage. Aqua diamonds reflect baseline expected impairment rates with no flood impacts. Orange bars correspond to the range of house price falls in the orange bars in Chart 2. Purple diamonds consistent with the top of the range of the purple bars in Chart 2. Green line is a weighted average of the purple diamonds. Gold line is based on the published results of the 2025 Bank Capital Stress Test. The hazard model we use suggests high flood risk in the North East and North West of England, though the relative distribution of risks differs between hazard model providers.

Sources: FCA Product Sales Data, Mitiga, RiskLayer, ONS Living Costs and Food Survey and Bank staff calculations. Full sources are available upon request.

Our benign aggregate result reflects three well-established features of the UK mortgage market:

1. The large majority of UK mortgages on banks’ books have a loan to value ratio of below 70%, insulating mortgages from falls in house prices.
2. UK flood insurance coverage is high, and pricing is affordable relative to many other countries.
3. Flooding is a geographically bound risk driver, meaning large impacts even across large areas average out to smaller impacts at the national level.

That said if a severe flood event were to coincide with other stresses, it could amplify financial stability risks. Protection gaps could also pose risks to smaller lenders with portfolios concentrated in high-flood risk areas.

Conclusion and policy implications

Our model results suggest that UK mortgagor protection gaps are unlikely to threaten banking system solvency, but this is subject to many caveats. We only cover the existing stock of buildings, as we cannot capture the exposure or resilience of planned new builds. We do not yet capture increases in risk after 2050, or other key perils, assets (eg commercial real estate and non-mortgaged houses), or contagion channels (such as macro impacts).

Our results are relevant to financial regulators:

The results could also inform judgements by governments about the impacts of climate change, including costs to households from increased flood risk, the end of Flood Re, and the potential benefits of planning and adaptation measures to improve financial and physical resilience.

Flood protection gaps could substantially impact economic welfare. While our results suggest they may not threaten banking system resilience, they could have a range of negative consequences for affected households, as well as lenders exposed to high-risk areas, and wider economic growth. Given the relevance of protection gaps to many stakeholders, cross-industry collaboration will be needed to help prevent climate-related risks from spilling over to the wider financial system.

Will Banks and Kemal Erçevik work in the Bank’s Cross-cutting Strategy and Emerging Risks Division.

The authors are grateful to Howden Re (Rowan Douglas, Man Wai Cheung, Wes Hibbert, Tim Edwards, Dr Nidia Martinez and Naomi Price), RiskLayer (Professor James Daniell, Dr Bijan Khazai and Dr Andreas Schaefer) and Mitiga (Dr Alex Marti and Dr Foteini Baladima) who provided extensive insurance sector insights, flood hazard and other climate modelling for this work.

If you want to get in touch, please email us at [email protected] or leave a comment below.

Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.