By RiskThinking Team
Summary: Climate adaptation can be valued as a put option. The upfront capital expenditure is the premium, and the payoff is a permanent reduction in tail losses. Counting the capital released by that reduction shows adaptation is worth far more than average-based analysis suggests.
Plot every climate model experiment in the world and you get around 2,500 lines. They clump. Many researchers expect a two or three degree world, so that is where the papers pile up.
Dr Ron Dembo, founder and CEO of Riskthinking.AI and lifetime Fellow of the Fields Institute for Research in Mathematical Sciences, argues that the clumping is the problem. In a recent talk he set out why the tail of the distribution is where climate risk lives, and why adaptation is best valued as a financial option.
Climate is a radical risk problem. It cannot be forecast, and past data is a poor guide to what comes next. Even if the world settles at two or three degrees of warming, five degree events can still occur.
The evidence is already here. In a single year, Riskthinking.AI measured 837 events that had never been recorded before. It uses these black swans to back test its models. Hong Kong was recently hit by typhoon winds of around 300 km/h, against buildings designed for 200 to 220 km/h. Jamaica saw a hurricane with double the wind speed ever measured there.
The response is to treat every reasonable, science-based scenario as part of an ensemble. Weight them by likelihood, but never drop one. When an event is rare but catastrophic, impact matters more than probability.
Most climate risk analysis averages. Yet loss distributions are long tailed and skewed to the right. A single point in the tail can be ten times the average.
Ron separates three reference points. The baseline reflects building standards that may be 20 or 30 years old. The 95th percentile is what most risk teams look at, through measures such as Value at Risk (VaR) and Conditional Value at Risk (CVaR). Then there is the true tail. Use only one or two pathways and the tail events disappear from view. As Ron puts it, the middle is what everyone already believes and works on. The tail is what does the damage.
Geography adds another layer. A US company in the S&P 500 may hold its most exposed physical assets in Peru. Physical risk follows location, which is why it demands geospatial data in a way conventional finance does not.
Riskthinking.AI's analysis of Marriott shows what this looks like in practice. Under extreme climate, the company could lose 24% of the value of its physical assets over five years. Broken down hotel by hotel, the portfolio sorts into groups: act now, act within months, act within a year or two, and set aside for the moment. That is the adaptation question in three parts. Where, what and when.
Ron's central argument is that adaptation behaves like a put option. Money is spent now to avoid an uncertain loss later. The upfront capital expenditure (capex) is the option premium.
Ron calls this framework climate capital option theory. It values adaptation as an option, counting both the losses it removes and the capital it frees up.
The insight is in what most analysis leaves out. Insurance lasts a year. Adaptation is effectively permanent, often for 20 years or more. Removing part of the tail permanently reduces the capital an organisation must hold against it. Banks work this way, and every startup founder knows liquidity is king.
So the true value of adaptation is the risk eliminated, minus the adaptation cost, plus that capital benefit. Counted properly, the option is worth far more than it first appears.
Three further lessons follow from option theory:
Take a manufacturing facility exposed to a correlated chain of hazards. A typhoon brings flooding and runoff, made worse by drought beforehand. The figures are illustrative, based on a real case. The adaptation is rooftop hardening, at a capex of 2.2 million.
| Measure | Without adaptation | With adaptation |
| Average loss | 2.8 million | 0.9 million |
| CVaR (average of the tail) | 42 million | 8 million |
| Probability of a large loss | 12% | 2% |
An average-based view sees a 2.8 million problem. The tail view shows a 42 million one, and a loss on that scale did occur in the real case in Indonesia. Removing that part of the tail is worth 8.5 million in capital benefit alone, which delivers a strong return on the 2.2 million spent.
For insurers, the option value quantifies tail reduction. For banks, it lowers balance sheet volatility: stronger homes mean a steadier mortgage book. For regulators, it offers an auditable, comparable number.
It also bears on insurability, which is eroding. In 17 US states, insuring a house is already difficult. Without insurance, funding becomes harder and the cost of capital rises, with effects that ripple through the economy. Adaptation is one of the few levers that pushes back.
Climate risk has no agreed standard, so Riskthinking.AI chose to offer one. Through the Climate Risk Commons, it has open sourced its full stochastic modelling pipeline, free for non-commercial use. It is the same engine that calculates physical climate risk on the Bloomberg Terminal.
The aim is parity. A researcher on a laptop in Mozambique can use the exact same tools as an analyst at Barclays in London.
AI extends this further. An analyst in Toronto can send work to a colleague in Chicago, who loads the same skill in Claude and gets the identical model, data and environment. Collaboration of that kind was not possible before.
Averaging tells a comforting story. The tail tells the true one.
Adapting to Our Future Climate, The Fields Institute for Research in Mathematical Sciences. You can watch the full presentation here.
Image credit: Phosphor Tailings Pond, Near Lakeland, Florida, USA, 2012. © Edward Burtynsky, courtesy Nicholas Metivier Gallery, Toronto