How climate risk could reshape Australia’s mortgage market

Climate risk is already changing the cost of owning some Australian homes. For many homeowners, the first sign will be on the insurance bill.

Premiums have risen in areas exposed to flooding, cyclones and bushfires, and some households are cutting back cover or going without it. That creates a problem for mortgage lenders as well as borrowers: most home loans require insurance, and a property that becomes difficult or expensive to insure may also become harder to value, refinance or use as security.

A home loan can run for 25 or 30 years, while home insurance is generally renewed each year. The lender is making a long-term decision about a property whose risk may be reassessed much sooner. That difference in time scale means the first signs may be felt by households well before they appear in banks’ portfolios. A premium that becomes unaffordable, a policy that is harder to renew or cover that no longer provides adequate protection, can affect a borrower’s finances, as well as the value of the property securing the loan.

“A property becomes hard to insure well before it becomes hard to sell or finance, so insurance is the transmission channel into everything else: property values, borrower income, collateral and, ultimately, bank credit risk,” said Associate Professor Kristle Romero Cortes, School of Banking and Finance at UNSW Business School.

Insurance is where the pressure starts

In parts of Australia, pressure is already showing up in growing insurance affordability stress and a rise in uninsured homes, although effects on property values and lending remain harder to trace.

The Reserve Bank of Australia has noted that unaffordable premiums in northern Australia are contributing to a rise in the number of uninsured homes. The Actuaries Institute estimates that 15% of Australian households – about 1.61 million – faced home insurance affordability stress in the year to March 2024, up from 10% in 2022. Of households with mortgages, 5% were affected, representing about $57 billion in outstanding loan balances.

The figures do not allow the increase in insurance affordability stress to be attributed solely to climate change. They do, however, show how a protection gap can develop before a mortgage itself becomes difficult to maintain. “The honest starting point is that the evidence base is still thin,” A/Prof. Romero Cortes said. “What we have mostly documents exposure and financial stress, not yet a clean causal picture of climate risk moving prices or lending decisions at scale.”

Learn more: How climate risk increases home insurance affordability stress

The question for borrowers and lenders is whether a property could become effectively unfinanceable. “Most mortgages require insurance; if premiums become unaffordable or insurers withdraw, the loan’s security is compromised, which can cut off new lending and refinancing and trap owners who can’t sell or refinance,” A/Prof. Romero Cortes said.

In its March 2026 Insurance Climate Vulnerability Assessment, the Australian Prudential Regulation Authority (APRA) estimated that around one in seven households in freestanding homes are currently uninsured, rising to about one in four by 2050 under both stress scenarios it tested. Weather losses were the main driver under higher physical risk, while construction-cost inflation played a larger role under higher transition risk. The scenarios were not forecasts.

Insurance need not disappear for this to affect the mortgage market. Rising premiums, narrower cover, and more uninsured homes could leave households with less room in their budgets and make lenders more cautious about the properties they are willing to lend against.

The risk is difficult to measure

Climate-risk modelling considers the interaction between the hazard, what is exposed to it and how vulnerable that asset is to damage. 

“In principle, what you would like to have is this overlapping between the physical hazard, the value of your property, and then you can calculate the annual average loss,” said UNSW Business School’s Dr Matteo Malavasi, a Senior Lecturer in the School of Risk and Actuarial Studies.

A flood may affect an entire building, but a ground-floor apartment is more vulnerable to damage than one on the sixth floor. The flood is the hazard; the property and its value are the exposure, while the apartment’s location helps determine its vulnerability.

Assessments can also take account of responses to climate risk: mitigation, through reducing emissions, and adaptation, through measures that protect properties from damage. The challenge grows when those risks are assessed across an insurer’s or bank’s portfolio. “You need to have a reliable hazard model,” Dr Malavasi said. “You also need to have a reliable estimate for your exposure.”

Insurers generally have detailed information about the properties they cover, including construction materials, building codes and location. The greater uncertainty often lies in the hazard model. Over a 30- or 50-year period, a property may change little, while climate risks compound and substantially alter the level of risk attached to it.

Models, therefore, need to account for changing conditions rather than simply repeat the past.

Insurers usually make decisions over a single year, while climate change unfolds over decades. “The problem is relevant because mortgages have 30 years’ duration, and 30 years is still too short to see the impact of any material effect of climate change – but that’s where the hazards and the risks start to accumulate and compound,” Dr Malavasi said. “The level of complexity increases, and as complexity increases, so does uncertainty.”

"Even if we were to, say, achieve the Paris Agreement target or any sort of net-zero commitment, we’re still going to have roughly 20 to 30 years of potential damages to come because of the impacts of climate change"

MATTEO MALAVASI

A mortgage brings those timescales together: the lender makes a long-term decision while the insurance market reassesses its exposure each year.

“Climate projections are on a much more macro scale,” Dr Malavasi said. “When you go from the big, macro scale of the global climate and how it affects the atmosphere to the super-micro level of the asset, it becomes tricky – tricky to calculate something that is meaningful and that reflects the information you’d like it to reflect.”

Projections are not always available at the scale needed to assess an individual property. If that mismatch is not addressed, models can produce detailed numbers that do not accurately reflect the risk to the asset. The granularity of climate projections has improved in recent years, Dr Malavasi said, but there is still work to do before they can be reliably connected to hazards and exposures at the property level.

Responses to climate risk are also part of the picture. Mitigation aims to reduce emissions and future hazards, while adaptation aims to reduce damage to exposed properties through measures such as flood protection or changes to building standards. A household, council or government may invest in adaptation, but it is difficult to know how effective those measures will remain over the life of a mortgage.

“When you do adaptation, you don’t look, for instance, at protecting your house or your property for the next year; you do it on a much longer horizon,” Dr Malavasi said. “Ideally, you want to protect it for the entire duration of your mortgage, or even longer.”

Subscribe to BusinessThink for the latest research, analysis and insights from UNSW Business School

Even if emissions targets are met, the effects of past emissions will continue to influence physical risks. “Even if we were to, say, achieve the Paris Agreement target or any sort of net-zero commitment, we’re still going to have roughly 20 to 30 years of potential damages to come because of the impacts of climate change,” Dr Malavasi said. “They don’t stop the day we reach net zero.”

A model can be accurate and still unfair

That difficulty is not just about the quality of climate data – it also raises questions about what insurers are measuring when they use more detailed data to price risk.

Different prices can be justifiable when they reflect expected risk and comply with anti-discrimination law. A variable that predicts losses, however, is not necessarily a legitimate measure of physical risk.

“A potential rating factor being predictive does not, by itself, make it fair or lawful to use,” said UNSW Business School Associate Professor Fei Huang, also in the School of Risk and Actuarial Studies.

As insurers use more granular data and artificial intelligence, it can become harder to see why a variable is predicting claims. A model may show that two variables move together without explaining why. That matters because, as A/Prof. Huang has argued, an AI model does not need to use race directly to produce discriminatory results; information that appears neutral can still reflect protected characteristics.

In a recent Actuaries Institute article, A/Prof. Huang contended that catastrophe models need to be assessed for transparency, fairness and stability, not just aggregate accuracy. An AI model estimating flood or cyclone damage might appear accurate overall while relying on variables that act as proxies for a household’s financial circumstances or demographic characteristics.

She gave the example of payment frequency, which may reflect a household’s financial position more than its physical exposure to a disaster.

“The result could be higher premiums for households in lower-income areas even where their underlying flood risk is unremarkable,” A/Prof. Huang said. “This is particularly concerning because communities facing genuine climate risks may also be among those least able to absorb unaffordable premiums. The key question is therefore not simply whether a model is powerful and predictive, but whether the data is relevant to the risk being priced, whether the outcomes are fair across communities, and whether the insurer can explain and justify them,” she added.

That makes it important for insurers to distinguish genuine differences in physical exposure from the characteristics of the people who live in a particular place.

The issue also intersects with human-rights law. The Australian Human Rights Commission identifies banking and insurance as services covered by the Racial Discrimination Act and notes that apparently neutral practices can create indirect discrimination.

“Risk-based pricing removes the informal cross-subsidy that hid this problem,” A/Prof. Romero Cortes said. “It hits hardest those least able to move, retrofit, or self-insure the gap, likely deepening existing housing affordability and regional inequality pressures.”

More accurate risk pricing may help keep insurance markets viable, but passing the full cost to households in high-risk areas would leave those least able to respond with fewer options.

What banks can – and cannot – see

Australian regulators have begun testing climate risk in bank and insurance portfolios. These exercises can show where exposure is concentrated, but they are scenario tests rather than forecasts.

A/Prof. Romero Cortes pointed to APRA’s 2022 Banking Climate Vulnerability Assessment, which examined the five largest banks under two long-term climate scenarios and found that potential losses were concentrated in specific regions and industries, with mortgage-lending losses higher in northern Australia. The banks indicated that they could respond by adjusting their risk appetite, reducing exposure to higher-risk areas, or scaling back high-loan-to-value lending.

APRA’s 2026 Insurance Climate Vulnerability Assessment builds on that analysis by examining how falling home insurance coverage could affect the financial system. It identifies two routes into bank risk: households facing uninsured weather losses may struggle to maintain mortgage repayments, and damaged or devalued housing collateral increases a bank’s loss when a borrower defaults.

A/Prof. Romero Cortes said stress testing is useful because it shows where exposure is concentrated, but she also pointed to its limits, noting that APRA’s 2022 assessment did not yet flag systemic concern. “Its limits are its own design: long-run scenarios to 2050 carry wide uncertainty, results vary by each bank’s modelling choices, and it doesn’t capture second-round effects like insurance withdrawal feeding into credit losses – the exact dynamic already emerging in northern Australia,” she said.

"Annual insurance repricing is, in practice, a faster and more current risk signal than static flood or bushfire maps"

KRISTLE ROMERO CORTES

The Reserve Bank has described the direct effects on the financial system as limited so far, with risks concentrated in particular regions and sectors. For banks, the practical response over the long term is to reassess regularly rather than wait for certainty.

“No one can forecast that precisely, so periodic reassessment beats a single point-in-time judgment,” A/Prof. Romero Cortes said. “Annual insurance repricing is, in practice, a faster and more current risk signal than static flood or bushfire maps.”

Annual repricing can also pass the cost of uncertainty directly to households before the market decides how that risk should be shared.

Who should carry the risk?

Dr Malavasi pointed to the federal government’s Cyclone Reinsurance Pool as an example of how risk can be spread beyond the communities most exposed to it.

“The Cyclone Reinsurance Pool helps share cyclone risks across a much wider area – not only within specific locations where risks are more concentrated, which, coincidentally, are also the areas where perhaps more vulnerable communities live,” he said. “That should, in some sense, counterbalance that effect, supporting the risk-sharing mechanism at the core of the insurance industry.”

The pool illustrates one way to spread costs more widely, but it does not remove the need for adaptation or settle what should happen to homes that remain difficult to insure or refinance. The immediate challenge will be felt through ordinary decisions – insurance renewals, property valuations, loan applications or refinancing requests – that affect who can afford to remain in a place, borrow against it or sell.

Learn more: Can accounting fix global climate disclosure challenges?

A/Prof. Romero Cortes emphasised that no single policy response will be sufficient to address the challenge. 

“Standardised risk disclosure lets the market price risk predictably; adaptation investment is the one lever that reduces actual risk rather than reallocating it; reinsurance pooling can keep private insurance viable in the highest-risk areas,” she said. “And there needs to be a deliberate answer, transition support or a public backstop for what happens at the margin, rather than leaving the market to sort it out.”

Five recommendations for industry professionals

1. Lenders can review insurance status and renewal costs regularly when assessing property exposure and mortgage security.
2. Insurers can check whether model inputs reflect physical risk, document how pricing decisions work and review outcomes across communities.
3. Risk teams can compare hazard data with property-level exposure and vulnerability, and make uncertainty clear when projections do not support precise estimates.
4. Boards and executives can consider how insurance withdrawal or reduced cover could affect lending, refinancing and property values in exposed regions.
5. Policy teams can assess adaptation, risk pooling and transition support as separate responses to the insurance protection gap.