Most institutional climate scenario analysis starts with a choice between two extremes: a 1.5°C world of rapid decarbonisation, or a 4°C hot-house. Regulators ask for both, and both fill stress tests and disclosure reports. What gets lost is that these two anchors sit at opposite ends of the distribution and almost nowhere near its centre. Each works as a stress test, and neither was ever built as a forecast.

What climate scenario does the world actually track?

Most investors already sense the tails are unlikely. The harder problem is what replaces them, because picking a scenario off a chart feels arbitrary and building one in-house invites accusations of bias. So the tails get used by default, even by the people who find them implausible, because nothing else is on the table.

This research closes that gap with a probability rather than a pick. We let the observed emissions record run forward many thousands of ways and measure which standard scenarios each trajectory most resembles. Every scenario across the IPCC and NGFS families receives a weight, calibrated on decades of measured emissions instead of assumptions about future policy or technology. None are discarded.

Where the probability actually lands

The mass concentrates hard on the middle of the distribution, on a world of partial and inconsistent climate policy. A single scenario, NGFS Fragmented World, carries more than 70% of the weight on its own, which exceeds every other scenario in the library combined.

Blending the library by probability produces one central pathway, and it implies 2.3°C of warming above pre-industrial by 2100. Both the 1.5°C Paris target and the 4°C hot-house sit outside the likely range, at close to zero probability. The full distribution, including where the remaining weight sits and what each bucket means, is set out in the paper.

What changes in a transition risk number

The effect on risk measurement is larger than most investors expect, and it is the part of this work with the most immediate portfolio application.

Run a sample of major listed companies through a transition risk model under a 1.5°C-aligned pathway and every one of them scores a uniform maximum, the signal that would flag each as effectively uninvestable. Run the same companies under a fragmented-policy world and the sample pulls apart, with utilities and airlines falling into the low single digits while genuinely exposed businesses stay pinned at the top. A saturated signal cannot tell you which holdings carry more risk than which, and answering that question is the whole point of running the analysis.

The paper names the companies and publishes their scores under both scenarios.

Reporting and investing are separate jobs

None of this argues against the scenarios regulators prescribe. Mandatory disclosure is a stress test, and it deliberately asks about worlds nobody expects to occur. For compliance, use the prescribed pathways exactly as specified; introducing your own view of what is probable invites the perception of bias into what should be a clean disclosure.

Investment decisions run on the opposite logic. Materially reporting a risk and materially managing one call for different inputs, and 2.3°C is uncomfortable under either. It implies serious and costly disruption, and it puts the 1.5°C target out of reach on current evidence. The argument here is about calibration rather than complacency.

Inside the paper

• The full probability distribution across the standard scenario library, with the weight on each of the nineteen scenarios
• The probability-weighted emissions pathway to 2050, and the warming distribution behind the 2.3°C central estimate
• Transition risk scores for ten named listed companies under both a 1.5°C-aligned pathway and a fragmented-policy world, at risk-year 2040
• The methodology: data sources, the cumulative-carbon basis, scenario library composition, and the limits of what the probabilities can be read to mean

Download the full white paper below.