Climate change modelling: Benchmarking now and to 2050
ILS managers have admitted having a challenge on their hands in finding the best methods to reflect climate change in their predictions and pricing, as insurer regulators and investors are increasingly asking questions on this topic. The latest vendor model studies on climate change have sparked debate among ILS firms about how best to incorporate this data into catastrophe models.
In March, there was pushback on modeler AIR after it used representative concentration pathway (RCP) 8.5 to predict a 20% increase in US hurricane damage by 2050.
Similarly, RMS last month announced its latest model release with the headline-grabbing prediction that insured average annual losses could grow by as much as 24% for North Atlantic hurricane by 2050, based on RCP 8.5.
The Intergovernmental Panel on Climate Change (IPCC) adopted the RCPs in 2014 as a method to describe how concentrations of greenhouse gases in the atmosphere will evolve in the future. They range from lower bound 2.6, which assumes all emissions stop now, through 4.5 and 6.0, to upper bound 8.5, which describes a future where no action Is taken to curb emissions.
Both modelling agencies emphasize that their ultimate climate change products will cover a range of scenarios, but the initial focus on the upper-end scenario begs the question of whether the ILS industry will move toward a common baseline framework for discussing the long-term impact – and if so what that baseline currency should be.
Some fear that the 8.5 is too extreme to be used in this way. Leadenhall Capital Partners said there is growing consensus that 8.5 is a less plausible scenario that’s unlikely to happen.
“I wouldn’t be surprised if we saw that upper bound come down a bit. We’d probably fall more into the bracket of 4.5 to 6.0, as a sensible stress-test range,” said Leadenhall Capital Partners chief underwriting officer Jillian Williams.
Fidelis Insurance commented that while the scientific consensus may be settling at about 3-4, this was “difficult to pin down. We are making short-term measurements based on something that will take a number of years before there is clarity,” said Fidelis Insurance group chief risk officer Charles Mathias.
At ILS firm Aeolus Capital, head of research Peter Dailey said he found the RCPs informative. Both worst-case and best-case scenarios are unlikely he noted.
“But they are instructive because they put a bound on physically how bad or how good the climate can get. It’s probably going to be somewhere in between, but it’s useful to know that range of possibility.”
The IPCC will soon replace the RCPs completely anyway, with new shared social economic pathways representing a nuanced picture of how economies will grow. The new approach will be based on the latest data from the Coupled Model Intercomparison Project (CMIP 6), an update to CMIP 5, on which the RCPs were based.
Paul Wilson, partner and head of non-life analytics at Securis Investment Partners, said: “The newer [scenarios] are far more relevant in my view. For climate change, ensuring thinking and data is latest is very important.”
The RMS climate change models will cover the range of all four RCP scenarios to allow clients to select the one that suits them best and remodel their portfolio out to the future. “We do not recommend [one over another]. One of the steps the industry will have to take is to appreciate uncertainty over RCPs,” said Joss Matthewman, senior director of climate change product management.
AIR Worldwide director of climate change research Peter Sousounis said that the RCP 8.5 scenario is not the only one that it will address, and that it has completed studies for clients on lower-range scenarios.
The agency expects to release a study later this year based on the Central Bank of England’s climate biennial (CBES) tests, which include scenarios for early and late as well as no climate policy actions – which Sousounis said have loss parallels across the RCP spectrum.
“In short, we understand clients have different interests and business needs and we will continue to offer solutions for different climate scenarios based on that interest.”
Most market participants accepted that it would be difficult to have one scenario act as a baseline industry currency.
One Bermudian reinsurer said a single view “means we’d all be wrong at the same time”, arguing that “diversity of views, within reasonable boundaries is helpful for market and clients.”
However, Leadenhall’s Williams said that while each firm would have its own view on risk, “there needs to be some consensus on how to work together. I’m not saying we should have a strict policy. But there does need to be collaboration.”
For the most part, the scenarios diverge more strongly in the long-term – from 2040 to 2050 – than they do in the coming years.
Re-assessing current baselines
But it is this near term figure where there is also concern. ILS managers emphasized that the other, perhaps more pressing concern is how to apply climate science today, to help more accurately predict and price risk for the one to three year time-frames that the ILS market mostly deals in.
Since catastrophe models traditionally took historical data and extrapolated it into the future, some underwriters are concerned that their approach may not have baked in a reflection of the impact from climate change that is already happening.
At Integral ILS, the view is that the baseline risk has now changed relative to the observational record that is typically used to condition catastrophe models, and that climate change should be incorporated within the reference view rather than as a parallel forward-looking view.
“On balance we believe climate change already affects the extreme events we care about. So we can’t wait for the observational record to show this, we have to act now in our own view of risk”, said Integral ILS director Matthew Swann.
“It’s great that the model vendors are starting to respond, and we look forward to the continued development of these efforts. Ultimately we would like to see climate change within the vendors’ reference view of risk”, Swann said.
Ultimately we would like to see climate change within the vendors’ reference view of risk
Securis’s Wilson agrees that the latest models are less well refined for ILS purposes, than for fulfilling regulatory stress testing exercises. “We believe the tools vendors produce are very useful, primarily for the stress testing. They aren’t addressing very well the question, ‘how much has climate change already impacted on present day risks? How much is it already accounted for in the model? These are the questions that we’re striving to address in our use of models.”
RMS specialist Matthewman said that “where there is a strong consensus,” the firm has already built climate change into models, such as with wildfire and accounting for present-day sea levels.
“For perils where [the] impact is still unclear RMS climate change models have ‘what if’ views which can account for climate change to date.”
Karen Clark & Co has addressed this partly by evaluating different historical time periods—for example, if you look at the data going back to 1960 (near the start of the satellite era) the return period of a category five hurricane would appear to be 20 years versus 30 years when using the data back to 1900. Founder Karen Clark said: “Our scientists are continually looking at what’s currently happening.”
AIR's Sousounis said that its baseline models are designed to have a "useful life" of 10 years, and that it evaluates the historical record to establish changes that may be attributed to climate change from short-term climate variability in designing its models.
Ultimately, moves towards more transparency are underway, but investors and ILS managers are looking for quantitative demonstrations of how well and robustly climate change impacts are accounted for.