Acclimatizing to Climate Scenario Analysis

By Steven W. Easson, FCIA, FSA, CFA

Introduction

Climate change entails a complex suite of risks and opportunities. Various standards, such as the “S2 Standard” from the International Sustainability Standards Board (“ISSB”) and OSFI’s B-15 Guideline, categorize the issues into the pillars of: Governance; Strategy; Risk Management; and Metrics and Targets. In the absence of an even lengthier article, it is not possible to delve into all four more than superficially repeating concepts the reader is likely already familiar with. Accordingly, this article provides some deep dive perspectives for Canadian insurers on the vital component of scenario analysis, pertaining to potential transition risk impacts of climate change on financial variables[1].

To begin with though, a context to the importance of climate change. The following is a (very small) sample of eye-popping excerpts from Mark Carney’s 2021 book titled “Value(s)”

  • In relation to remaining carbon budgets, in 2018, the Intergovernmental Panel on Climate Change (“IPCC”) estimated a range of 420 Gigatons (to limit temperature increases to 1.5C above pre-industrial levels with 66% probability) to 1,500 Gigatons (to limit to 2OC with 50% probability). Based on current emissions rates these budgets would be exhausted respectively in less than a decade, and about 3.5 decades.
  • If the rise in GHG emissions does not level off until the latter half of this century, the IPCC estimates a temperature rise of 4OC (range 2.4C to 6.4C).
  • To meet the 1.5C target more than 80% of fossil fuel reserves would need to stay in the ground.
  • As early as 1990, the IPCC noted that the single biggest impact of climate change on our society and economy could be a mass of climate refugees.
  • It is self evident that the financial system cannot diversify itself out of this risk.

Scenario analysis

Overview of accounting and regulatory requirements of Canadian insurers:

One way to describe regulatory expectations is two broad buckets, described as “Pillar 1” (“quantitative” / ”regulatory requirements”) and “Pillar 2” (“supervisory” / ”own views”), or perhaps “entity has to do it this way” and “entity would like to do it that way”.

In the case of OSFI’s B-15 Guideline[2], Pillar 1 would be OSFI’s Standard Climate Scenario Exercise (“SCSE”) which, at the time of this writing, insurers will be expected to complete in 2024, and years beyond. “Pillar 2” includes the B-15 Guideline expectation of “Describe the resilience of the FRFI’s strategy, taking into consideration different climate-related scenarios, including a scenario which limits warming to the level aligned with the latest international agreement on climate change, or lower”[3], as well as consideration of climate impacts in Own Risk and Solvency Assessment (“ORSA”) analyses. Of note is that conceptually the B-15 Guideline expectation and the ORSA expectation are not one in the same. The former “resilience” expectation includes “short,” “medium”, and “long” term time horizons, typically, respectively, from the present to 2030, 2030-2050, and 2050-2100, whereas the ORSA time frame is short term. For 2024, at the time of this writing, insurers will not be required to complete the former, but OSFI does expect insurers to incorporate climate-related risks into its ORSA process, but under its Guideline E-19.

In relation to accounting requirements, the S2 Standard, similar to the B-15 Guideline, the explicit expectations include use of scenario analysis to assess “climate resilience”, with such assessment enabling stakeholders to understand how the entity would respond to the impacts identified in the scenario analysis, significant areas of uncertainty, and the entity’s ability to adjust / adapt.[4]

Other potential applications (“use cases”) of scenario analysis

Related to Principle 4 in the B-15 Guideline, namely, “use climate scenario analysis to assess the impact of climate-related risks on its risk profile, business strategy, and business model”, would be the use of scenario analysis to assist with assessing the effects on other concepts in the S2 Standard and the B-15 Guideline, such as the disclosure elements in the Strategy section in Annex 2-2 of the B-15 Guideline, “cash flows, access to finance or cost of capital”; “value chain”; “decision making”; and “financial position, financial performance, and cash flows”. One issue for insurers in this regard is to ensure they are clear on the scope for their organization of some of the terms, such as “value chain”. Outside of accounting and regulatory requirements, is the potential use of scenario analysis for functions such as pricing, product development and investment management.

Scenario generation

The remainder of this article delves into the nuances of scenario generation.

Insurer capabilities will evolve:

Extracting a small sample of concepts from various regulatory publications:

  • Scenario analysis is as much about building capabilities as it is about assessing the risks. Data gaps are a particular area of attention, as the quality of data has a direct bearing in the quality of results. Doing climate scenario analysis can help generate relevant data and address gaps, but this is an evolving process.
  • Ultimately, as climate scenario exercises develop, insights into the financial aspects from transition and physical risks will become increasingly comprehensive, based on a converging set of methodological practices, and will make use of more widely available data.
  • The climate transition scenarios are not meant to be forecasts or comprehensive. Rather, they explore different plausible transition pathways consistent with achieving climate targets

Models

The general process for deriving impacts on the financial variables that are used to assess the insurer’s “climate resiliency” is to start with an underlying model which essentially produces potential economic outcomes for climate change scenarios. There are a number of versions of such underlying models that are available for insurers to utilize, collectively referred to as Integrated Assessment Models (“IAM”) [5]. An IAM analyzes complex interactions between “human systems” drivers (e.g., societal and government policy actions, and technological developments), and “natural Earth systems” drivers (reflecting physics / chemistry / biology principles). These models are designed to integrate knowledge from various disciplines, including economics, environmental science, energy systems, and policy analysis, to provide insights into the impacts of different policy decisions and societal choices on the environment, economy, and society as a whole. They essentially combine models, including, but not always limited to, Climate[6], Energy Systems, and Macroeconomic models.

At the heart of the matter is GHG emissions levels / concentrations, under various climate scenarios, are translated into economic outputs that are subsequently translated into impacts (of climate) on financial variables, for use in scenario analysis. These IAMs can be categorized into more “simple” models (for example, the “Dynamic Integrated model of Climate and the Economy”, or “DICE”) and more sophisticated models (for example, the “Global Change Analysis Model”, or “GCAM”). Many are open source and written in languages such as R and Python. Regardless of whether a model is “simple”, obtaining a deep understanding of the IAM algorithms is truly a daunting task.

Deficiencies in marketplace models to date:

There have been numerous publications that have articulated deficiencies to date. One notable report is from the University of Exeter (“No Time To Lose – New Scenario Narratives for Action on Climate Change”, September 2023). Below is a small sample of excerpts:

  • Organizations need bespoke scenarios to embed the analysis of climate risk and opportunities into all their decision making.
  • The paradigm shift towards shorter horizons and business applications requires scenarios that focus less on the climate itself and more on the vicissitudes of politics, markets, and extreme weather events.
  • Global warming is not a major uncertainty over the next few years, but extreme weather events are rising rapidly, even if location and timing are uncertain. At the same time, the limitations of current official scenarios and methodologies, notably from the Network for Greening the Financial System (“NGFS”) are becoming increasingly apparent. They are failing to capture key aspects of the real world, including acute physical risk, politics and policy, unemployment, finance, asset prices, volatility, tipping points, path dependency and complex feedback loops.
  • The report presents a new set of four narrative global climate scenarios out to 2030 based on a framework which embraces the radical uncertainties surrounding the potential positive as well as negative tipping points. The four scenarios involve four different combinations of high or low policy activism and high or low market dynamism.

Other sources of deficiency relate to

  • Climate science models, for example, the lack of including non-linear changes in the climate, and insufficient damage functions.
  • Economic models, for example, the insufficiency of “general equilibrium models”, and
  • Insufficient consideration of many other factors, for example, but certainly not limited to, cumulative effects; cascade of events; ripple effects, tipping points, feedback loops; the potential for systemic risk, the impact of mitigation and adaptation; geopolitical shocks; mass migration; the timing and impact of climate policy decisions (e.g., carbon tax); policy levers beyond the reliance on carbon prices.

So how do Canadian insurers proceed with scenario generation?

In addition to determining the “use cases” for scenario analysis, as the first step in deriving scenarios, it would be constructive to utilize a visual aid to conceptualize the interconnectedness of drivers, feedback loops and causal relationships. Below is one such illustration[7].

diagram showing integrated framework of climate change for insurers.

That is the easy part. The really hard part is quantitatively forecasting the impact of the climate scenarios on financial variables that are directly used in scenario analysis, such as inflation, interest rates and equity returns.

This author suggests the following as potential ideas for guiding principles and techniques to assist with the task at hand. They can be collectively described as aligning the degree of precision to the degree of knowledge / degree of certainty (or, more saliently, the degree of lack of uncertainty), with such “degrees” expected to continually increase over time.

  • Alignment – It would be counterproductive to apply “precision on imprecision”. That is, avoid precision on quantifying assumptions for financial variables if there is imprecision arising from the plethora of sources of uncertainties of the underlying input variables related to data, assumptions, and / or methodologies. In the extreme, utilize qualitative assessments, not quantitative.

  • Sophistication – The greater the entity’s exposure to climate-related risks or opportunities, in relation to their other sources of risks and opportunities, the more likely it is that the entity would attempt to conduct a more technically sophisticated form of climate-related scenario analysis.

  • Uncertainty – The greater the uncertainty in the data, assumptions, and / or methodologies, the less likely it is that sophisticated stochastic analysis will provide more decision-useful disclosures over a suite of thoughtful plausible deterministic scenarios.

  • Conservatism – Given the literature on the current deficiencies in models, attempt to derive and incorporate, at least “back of the envelope”, adjustments for the deficiencies.

  • Choice of scenarios – Choose a suitable number of the emissions / socioeconomic pathways that consider the different ways and speeds of reaching net zero, and accordingly the degrees of relative impacts of transition risks vs. physical risks over various time periods, for example the NGFS’s four scenarios of Orderly, Disorderly, Hot House World, and Too little too late. Consideration should be given to bespoke scenarios and emerging industry best practice scenario alternatives to those derived by regulatory bodies (e.g. OSFI / NGFS). Consider a range of short term, medium term, and long-term time horizons, for example as discussed above in the University of Exeter report.

  • Choice of models – Choose the model in the continuum of “simple” to “sophisticated” according to ability to understanding internally, including via the availability of documentation, the respective algorithm. Understand the key inputs and outputs and the “fan of possibilities” of outcomes of financial variables given the level of GHG emissions / concentration forecasts. At the extreme end of “simplicity” would be to use “bookend benchmarks”.

  • Bookend benchmarks – Given the underlying driver on financial variables is emissions / concentration levels (bottom left corner of the above chart), attempt to gauge, at least in ballpark terms, the extreme boundaries (“bookends”) of plausible impacts on financial variables. This involves ballparking sequential impacts, by starting with emissions / concentration levels and ending with ultimate impacts on GDP, and consequently the GDP impact on financial variables. Intermediate steps involve ballparking optimistic and pessimistic outcomes of the impact on “human system” components, a.k.a. mitigation and adaptation, of policy decisions and technological developments and incorporating judgments on adjustments to compensate for model deficiencies discussed above. The “bookend benchmarks” are the most “optimistic” and most “pessimistic” among the fan of possibilities generated from this brainstorming exercise. These bookend benchmarks can serve as a reality check of results from modeling software, or for direct use in assessing the continuum of potential climate impacts in the absence of using an IAM.

  • Utilize reverse stress testing – Determine the values of the key financial variables which cause various solvency levels. From there, work backwards from the process described above for bookend benchmarks to infer the underlying scenario of emissions / concentrations. Judgement on the likelihood of the resulting emissions / concentration levels informs the commensurate likelihood of insolvency risk.

  • Industry benchmarks – Although there is the premise of “no one size fits all”, insurers might consider a healthy balance of unique practices (that may seem like “outliers” to external stakeholders) with overall industry “best practices”.

Closing Remarks

Climate change is a vitally important and concerning evolving risk. The ten warmest years since 1850 have all occurred in the past decade and 2023 was the world’s warmest year on record, by far[8]. Climate change is overwhelming in complexity, but at the same time, a fascinating challenge. The insurance industry’s sophistication in dealing with the governance, strategy, risk management and metrics / targets pillars will evolve over time. Particularly for scenario analysis, given the uncertainty of many aspects of data, assumptions and modeling / methodologies, the principle of “proportionality” leads to implementing practical solutions, one in which the “benefits” (decision useful information) will exceed the costs. Solutions will increase in sophistication over time commensurate with the increasing reliability of data, assumptions, and modeling / methodologies. Continual industry benchmarking will greatly assist insurers in developing best practices related to all four pillars, notably with scenario analysis, and metrics / targets.

Consulting Actuary, Steven W. Easson, FCIA, FSA, CFA

Steven W. Easson, FCIA, FSA, CFA is Consulting Actuary with Valani Global. He is involved in the actuarial community in developing actuarial climate standards of practice, as a member of the (Canadian) Actuarial Standards Board (“ASB”), co-chair of the ASB’s Designated Group on Climate, and a member of the International Association of Actuaries Task Force to develop a Climate Standard of Practice.

[1] The perspectives provided in this article are only those of the author’s and not of any entity, notably actuarial associations globally.

[2]  As at the time of this writing, the latest version was released on March 20, 2024

[3]  As at the date of their publication, 1.5°C above pre-industrial levels, based on the 2015 Paris agreement.

[4] See the ISSB S2 Basis of Conclusions, paragraphs BC56-BC59, for clarification of expectations related to the concepts and interrelation of “scenario analysis”, “climate resilience / resilience assessment”, and “current and anticipated financial effects”.

[5] For example, see https://www.iamcdocumentation.eu/index.php/IAMC_wiki for descriptions and access to software.

[6] According to NASA, the global climate model from the National Center for Atmospheric Research (NCAR) can fill over 18,000 pages of printed text. Hundreds of scientists add hundreds of observations and millions of lines of code to build a model, requiring a supercomputer the size of a tennis court to process the information.

[7] Source: https://www.ipcc.ch/report/ar3/syr/summary-for-policy-makers/figspm-1/

[8] Source : https://www.noaa.gov/news/2023-was-worlds-warmest-year-on-record-by-far

Excerpted from MSA Research ‘s year-end 2023 Life/Health Quarterly Outlook Report. To obtain the full report please reach out to Nevina Kishun, it is free to Canadian Life (Re) Insurers for the balance of 2024.

Nevina Kishun

President – MSA Research
Phone: +1.416.953.3124
nevina.kishun@msaresearch.com