Warming temperatures, changing weather patterns, and rising sea levels can interrupt operations, damage assets, and disrupt supply chains — all of which can hit a company’s bottom line.
To minimize their potential losses, publicly traded companies turn to private vendors that create models to predict such disasters.
These companies and their investors may not know that the projections derived from different climate models have different validation methods. They also may not know what happens when these models’ outputs consequently don’t match or look vastly different.
Clifford Rossi, the director of the Smith Enterprise Risk Consortium at Maryland University, adds that investors shouldn’t assume that a model assessing a company’s climate-risk exposure has been independently vetted or is accurate.
These ‘unknowns’ were apparent in a recent study by the nonprofit CarbonPlan that compared climate model vendors and found discrepancies, as reported by Bloomberg.
Oriana Chegwidden, a research scientist at CarbonPlan, told Business Insider that the study requested flood and fire risk estimates for post office locations across New York and California from nine analytics companies but got responses from only two. The small sample found a broad agreement about risk levels regionally but not locally or between neighborhoods and houses.
Related stories
Simply put, the numbers in a report can look different depending on which company made the assessment. And if you’re using one vendor, you’re not getting the full story, said Oriana.
The transition to a lower-carbon economy will also be costly for companies as they adjust to meet lower emissions standards. Jin Oh, a senior director at Climate at Moody’s, says the costs from damages and transitioning will trickle into company cash flows, indirectly or directly. The open-ended question is: How much and when will costs hit? The answer will vary depending on the sector and where a firm’s operations and suppliers are.
The 2-layer climate model problem
CarbonPlan’s study uncovers an issue far deeper than a few discrepancies.
Climate models have a two-layer problem. The first starts with the scientific community. Scientists worldwide use a set of agreed-upon inputs called the shared socioeconomic pathways, which are based on past trends, to project future scenarios. Inputs include greenhouse gas emissions, energy use, and population growth. The output is then freely available to the public.
However, the range of outcomes diverges significantly. The relationship between different elements such as land, water, and air, and how they interact with one another when shifts happen is contested. At the extreme end of these outcomes, there’s the “hot model problem,” which is too hot to make any sense because it runs much hotter than other models, said Timothy Canty, a professor of atmospheric and oceanic science.
Below is a chart from the Intergovernmental Panel on Climate Change’s Summary for Policymakers that demonstrates the five variations of models. The projection lines within the shaded red areas are within the hot model problem.
IPCC report Summary for Policymakers
Climate vendors use these public models for their black-box models, but it’s not clear which version, Canty said. They then add proprietary inputs and downscale their models to a property level. Some models aren’t as detailed about things such as land elevation, which is where the second layer of problems comes in.
“There’s concern that it’s getting over-interpreted,” Canty said. “The worst-case scenario, a vendor tells a company, ‘you need to take all of these actions’ because of some climate risk threat that never happens.”
Consequently, companies that end up spending billions on something that was never an issue won’t listen to any climate modelers again. And you end up with the cry-wolf syndrome, Canty said.
From model to valuations
“The models are not ready to make hard-money decisions off of; there’s too much uncertainty around them,” said Rossi, who has been sifting through vendor climate models.
Investors need to realize that there are enough of these issues to call into question the accuracy of any of these readings, Rossi said. If these models are used to project the impact climate risks pose to a firm’s long-term cash flows or asset valuations in 10, 15, or 20 years: the projections may not be as bad, but it also may be worse, he added.
There are several scenarios where companies may get caught in the crossfire of this dilemma, Rossi said. Fossil-fuel extractors could get lower valuations; shipping companies could miss projected sea levels at certain ports; insurance companies underwriting property may underestimate or overestimate risk; and banks assessing their portfolios for risk exposure on assets could miscalculate.
It’s an even bigger issue when firms like banks use the outputs from climate models as part of an integrated assessment model that projects damage estimates. Depending on the temperature projections over the next 20 to 30 years, there will be an estimated loss in GDP and other economic activities, which drives future profitability estimates and capital requirements, Rossi noted.
“Well, guess what? The damage models that are used to form that relationship are some of the most simplistic and most prone to error that can be,” Rossi said. He also pointed to real estate, which is another big sector where these models matter. Companies like Zillow use things such as flood risk ratings that may not be accurate or precise. But if a buyer uses the data to negotiate a price down, these could end up in future legal action, he noted.
Canty said vendors do hire climate scientists or have them on their boards of advisors to ensure that the models are being used appropriately. But the company is still trying to sell something, and there’s always a concern that it is straying. Right now, there aren’t enough industry standards or transparency to know these things.