From Hazard to Capital: The Gap Climate Risk Still Has to Close
I’ve been in a number of conversations recently where the science is genuinely impressive - detailed hazard maps, multiple climate scenarios, clean outputs, the kind of work that would have felt out of reach even a few years ago.
And then, almost inevitably, someone asks a different kind of question. Not about the model. Not about the assumptions. Something much simpler. ‘What does this actually mean for us?’ Or more bluntly, from a credit lead on one call: ‘Are you telling me to change anything because of this?’ That’s usually where I think things get a bit uncomfortable. Not because the work is wrong. In many cases, it’s excellent. But because the answer isn’t obvious, and everyone in the room knows it.
The gap isn’t data
We’ve made real progress on hazard. Flood, wildfire, heat, wind, we can now get to a level of granularity that simply wasn’t available before. In some cases, down to individual assets, with forward-looking scenarios layered on top. If the question is ‘where is the risk?’, the industry is in a much better place.
But that’s not the question most financial institutions are trying to answer. They’re trying to decide whether to lend, how to price, how to allocate capital, whether something is material to a portfolio. And somewhere between those two worlds, things start to break down. People often describe the chain as:
Hazard → Damage → Loss → Financial Impact → Decision
Which is directionally right. But in practice, most of what’s available today only really gets you through the first part of that. And the rest is where it becomes less neat, and much more dependent on judgement.
Where it actually gets difficult
One of the more honest conversations I had recently was with a credit team looking at coastal exposure. They had hazard data, good data. They had some view on potential damage, less certain, but workable.Where things got harder was everything after that. How does that translate into expected loss over the life of a loan? How does that flow through to pricing? At what point does it become material enough to change a decision?
Those aren’t purely modelling questions. They’re judgement, governance, and comparability questions. And they don’t get resolved just by adding more data, or even better data. You can have something that is technically robust and still not have something you can take to a risk committee with confidence. As one CRO put it to me, quite dryly: ‘A perfect model that nobody acts on is just an expensive hobby.’
A structural mismatch
Part of the issue is that we still, implicitly, expect the market to adopt climate risk analytics in the same way insurance did. It won’t. Insurers built their businesses around probabilistic loss models. The outputs feed directly into underwriting and capital frameworks. Banks and asset managers are coming at this from a different place. What I see instead is much more modular adoption. Teams pull in pieces, hazard layers, scenario outputs, scores, and try to integrate them into existing credit models, valuation frameworks, and portfolio tools. That’s entirely rational. But it also means there isn’t a clean, end-to-end translation. And without that, the burden of interpretation still sits with the user.
What people are actually trying to do
In most of these conversations, the sticking point isn’t understanding risk. It’s using it. ‘I can’t take this to committee in this form.’ I’ve heard that, or something very close to it, more times than I can count. Not because the science is wrong. But because the output doesn’t quite connect to the decisions people are accountable for, or the way those decisions are made. That connection matters more than anything else. Not just insight, but how that insight fits into existing workflows, how it interacts with internal models, how it holds up under model risk management. This is also where newer approaches, including AI-driven analytics, are starting to surface more clearly. There’s an expectation that more dynamic, forward-looking signals will help close the gap.
But in practice, that introduces a different tension. The more sophisticated, and less transparent, the model becomes, the harder it is to defend. So the question shifts slightly. Not just ‘is this better?’ But ‘can I stand behind this in front of a regulator, or a board?’ Right now, that balance isn’t settled.
The external pressure is building
What’s changed over the last 12 to 18 months is the tone. The Network for Greening the Financial System is framing physical risk in terms of financial stability and capital impact. The Bank for International Settlements has made the point that physical risk is now systemic, but that consistent approaches to incorporating it into credit frameworks are still missing. And supervisors are becoming more direct. Recent European Central Bank commentary suggests that climate risk is still applied unevenly, and only partially embedded into capital planning. So the direction is clear. This is moving from understanding risk to demonstrating how it affects decisions.
The part we don’t talk about enough
From the outside, it can look like the market is moving quickly. More data, more platforms, more partnerships. But when you get into the detail, there’s still a lack of consistency. I saw a team recently comparing outputs from multiple providers for the same portfolio. All credible. All well-constructed. All slightly different. The question wasn’t which one was ‘right’. It was: ‘Which one fits into how we actually make decisions?’ That’s a harder problem. Because it’s not just about accuracy. It’s about fit, trust, and ultimately accountability.
Where this leaves us
This isn’t just a modelling challenge, and it’s not just a data challenge. It’s a question of how climate risk fits into financial decision-making in practice. The next phase of this market won’t be defined by who has the most detailed hazard layer. It will be defined by who can connect that hazard to capital, in a way that aligns with how decisions are actually made, and how they’re governed. Because ultimately, that’s the test. Not whether we understand climate risk. But whether we can use it. And in quite a few of the conversations I’ve been part of recently, we’re not quite there yet.