Insurers reprice every year; banks lend for decades. How climate risk passes from insurance into credit, the evidence already in the data, and what banks and insurers should do.
Introduction
A bank lending against a warehouse, a villa or a hotel asks for proof of insurance at origination. The loan runs for 15 or 25 years; the policy runs for one. At each renewal the insurer can raise the premium, add an exclusion or decline to renew. Most credit models assume none of that happens.
That mismatch is the subject of Bridging the Climate Risk Disconnect, published in October 2026 by Environmental Defense Fund (EDF) with D.A. Carlin & Company. Its argument applies directly to GCC banks running climate stress tests.
One risk, two horizons
Banks and insurers often carry the same asset: the insurer pays for damage, the bank holds it as collateral. They manage that shared exposure separately, on different horizons and with different data.
An insurer that withdraws from an exposed market protects its own book, but the risk stays with the borrower and with the bank holding the loan. A bank that then tightens credit there starves the area of the money needed to adapt, which is what would have kept the assets insurable. Each decision is rational alone. Together they make the problem worse.
The scale is large. Swiss Re Institute put global economic losses from natural catastrophes at USD 318 billion in 2024, of which USD 137 billion was insured. That left 57%, about USD 181 billion, uninsured in a single year.
The transmission is already in the data
- Home values. Keys and Mulder (NBER Working Paper 32579, revised October 2026) used more than 74 million premium records inferred from US mortgage escrow data. A repricing of catastrophe risk in global reinsurance markets cut 2024 home values by close to USD 44,000 on average in the most catastrophe-exposed tenth of zip codes, and by around 11% where risk was also rising.
- Delinquency. In research published by the Federal Reserve Bank of Dallas (Working Paper 2505, revised October 2026), a USD 1,000 rise in the annual premium increases the probability of a 60-day mortgage delinquency by 0.38 percentage points, about 22% above the average rate. The effect concentrates among borrowers with lower credit scores and higher loan-to-value or debt-to-income ratios.
- The UK mortgage book. Public First, for the UK Sustainable Investment and Finance Association (March 2026), estimates that around 430,000 mortgaged homes in England could become uninsurable or very expensive to insure by 2050. Flood Re, the scheme that keeps flood cover affordable for high-risk homes, ends in 2039, inside the term of a mortgage written today.
Bank models have not caught up. In a 2025 survey of 32 banks by UNEP FI and Global Credit Data, only 18% incorporated climate risk factors into internal ratings-based models, and 62% used their climate risk results for reporting because regulation required it.
Two blind spots in bank credit models
The report builds on two ideas from the University of Cambridge Institute for Sustainability Leadership (CISL), in its May 2026 report on Resilience-Adjusted Credit Risk.
The insurance blind spot. Credit models treat insurance as a condition met once, at origination, when it is a variable that can change at every renewal. As insurers withdraw from exposed markets, that assumption becomes a risk nobody is measuring.
The revenue paradox. A borrower who borrows to build flood defences becomes a safer credit, but the loan raises their leverage and a standard scorecard can mark them down for it. The avoided loss appears nowhere in the rating.
Behind both sits the horizon. Probability of default, loss given default and capital rules typically work on a one-year window, so a climate effect that builds over a decade barely registers.
The feedback loop and wrong-way risk
The two directions of transmission compound. An insurer reprices or withdraws. Property values fall and default risk rises. Banks tighten credit. Rebuilding and adaptation slow. The insurable pool shrinks, insurers face greater concentration, and they reprice again. The loop closes on the places that were already most exposed.
For a bank this is wrong-way risk: the insurance meant to protect the collateral retreats just as the collateral loses value and the borrower weakens. A model that treats insurance as fixed understates losses exactly when they arrive.
| Exposure | What changes when cover reprices or withdraws | Where it shows in credit |
|---|---|---|
| Residential mortgages | Premiums strain household budgets; insurance cost lowers the property’s value | Probability of default and loss given default both rise |
| Commercial real estate | Premiums compress net operating income; layered programmes and captives obscure true cover | Debt service capacity falls; collateral valuations lag |
| Corporate and supply chain lending | Business interruption from damage at a supplier, a customer or the borrower’s own site | Cash flow and continuity, rarely priced at origination |
What decides whether cover stays affordable
There is no agreed threshold at which an asset becomes uninsurable. The report points to eight factors: the type of peril, with severe and spatially correlated hazards hardest to pool; confidence in the pricing; whether the market is hard or soft; geographic concentration in the insurer’s book; exclusions that leave a policy in force without covering the peril that matters; insurer appetite, which can change at renewal for reasons unrelated to the asset; syndication, where one panel member’s exit can leave a layer unplaced; and client size, since large clients can use captives or self-insure.
For a lender, these are early-warning indicators. Tracking premiums, deductibles, exclusions and appetite for material exposures shows insurability deteriorating before a loss does.
What each party should do
Banks: keep insurance and asset risk information current over the life of the loan; stress credit against higher premiums, narrower cover and withdrawal; and decide in advance what happens to pricing, collateral requirements, risk appetite and limits when insurability deteriorates.
Insurers: model second-order effects, concentrations and correlations; size the underwriting at risk under different scenarios; and reward measurable resilience through premiums, deductibles and multi-year structures.
Both together: test their methods on common hypothetical portfolios, exchange targeted information product by product, and pilot products that link adaptation finance with better insurance terms. It can work: Link Asset Management in Hong Kong secured an 11.7% cut in property insurance premiums for 2025 after demonstrating the effect of a HK$5 million flood resilience investment. The Global Commission on Adaptation (2019) put the benefit-cost ratio of adaptation at 2:1 to 10:1.
Regulators: supervise banks and insurers as one system, extend scenarios to contagion, recognise measurable resilience in capital frameworks, and plan for communities that lose access to insurance and credit.
Policy is moving unevenly. The PRA’s Supervisory Statement SS5/25, published on 3 December 2025, replaced SS3/19 in full and covers both banks and insurers. The US federal banking agencies withdrew their 2023 climate risk management principles on 16 October 2025. Voluntary alliances have proved fragile: Net-Zero Banking Alliance members voted to cease operations in October 2025. Shared definitions, data and methods will outlast pledges.
What it means for GCC banks
The CBUAE’s ICAAP guidance lists climate risk among the minimum risks that stress test scenarios should cover (paragraph 113). It expects each bank to consider three to five scenarios with graded severities (paragraph 119) and to run a reverse stress test at least once a year (paragraph 124). For other GCC regulators, see our comparison of GCC central bank climate rules.
Three steps follow:
- Make insurance a scenario variable. For the most exposed collateral, model higher premiums, narrower cover and withdrawal, and trace the effect on affordability, collateral values, probability of default and loss given default.
- Use the reverse stress test. Ask what combination of a physical event and an insurance market retreat would break the capital plan. That finds the concentrations a standard scenario misses.
- Start from the hazard data. National hazard figures are the base for asset-level analysis; see National Adaptation Plans in the GCC and our guide to climate stress testing and NGFS scenarios.
Insurance cover is a fact on the day a loan is signed. Over the life of the loan it is an assumption, and climate stress tests should treat it as one.
Sources: Environmental Defense Fund and D.A. Carlin & Company, Bridging the Climate Risk Disconnect, October 2026, and the primary studies it cites, each checked against the issuer’s own publication; PRA SS5/25; CBUAE ICAAP Guidance, section VII. All as held at 2 November 2026.
Frequently asked questions
How does insurance withdrawal affect bank credit risk?
Through both sides of the loan. A borrower facing a sharp premium increase is more likely to fall behind on payments, which raises the probability of default. The same higher insurance cost lowers the value of the property, which raises loss given default even if the borrower keeps paying. Where cover is withdrawn altogether, an uninsured loss hits the borrower's ability to repay and the bank's collateral at the same time.
What is the insurance blind spot?
It is a term from the University of Cambridge Institute for Sustainability Leadership's 2026 work on Resilience-Adjusted Credit Risk. It describes credit models that assume insurance cover stays in place for the life of a loan, when in exposed markets it is being repriced or withdrawn. The risk stays invisible until a loss exposes it.
What is wrong-way risk in climate lending?
Wrong-way risk arises when the protection a lender relies on weakens at the same time as the exposure it protects. In climate lending, insurers retreat from an exposed area just as property values fall and borrowers come under strain, so collateral value, creditworthiness and insurance cover all deteriorate together.
Do GCC regulators require banks to stress test climate risk?
In the UAE, yes. The CBUAE's ICAAP guidance lists climate risk among the minimum risks that stress test scenarios should cover, alongside credit, market, liquidity and other risks, and expects each bank to consider three to five scenarios and run a reverse stress test at least once a year. For the position of each GCC central bank, see our comparison of GCC central bank climate rules.