Scenario Testing for Climate Risk

Implementation approaches and challenges for emerging climate regulation Brian Kelly

Climate change continues to increase in priority for insurance companies. Almost every week there are news reports of another extreme weather event. At the same time, governments and regulators worldwide are introducing new requirements and extending existing guidelines—such as the ones from the Task Force on Climate-related Financial Disclosures (TCFD)—and making them mandatory. Although the details differ slightly by region, there is generally a requirement to quantify a company’s exposure to climate risk by modeling the impact of one or more climate scenarios on its profitability and balance sheet, referred to as scenario testing. Focusing on the life insurance industry, this article discusses key principles and challenges of scenario testing and implementation approaches.

Regulatory Environment

The basis for much of the existing climate-related reporting is the “Recommendations of the TCFD” report, issued in 2017. This report was one of the first proposals to focus on the link between climate outcomes and firms’ financial performance and decision-making.

The TCFD disclosures cover the following areas:

  • Governance—around climate-related risks and opportunities
  • Strategy—actual and potential impacts of climate-related risks and opportunities on the organization’s businesses, strategy and financial planning
  • Risk management—how the organization identifies, assesses and manages climate-related risks
  • Metrics and targets—used to assess and manage relevant climate-related risks and opportunities

The TCFD report recommends scenario analysis as part of the strategy disclosure to assess an organization’s climate resilience. Specifically, it mentions a scenario in which warming does not exceed 2°C above preindustrial levels.

Since its introduction, the TCFD recommendations have received support from about 4,000 companies and organizations. Companies supporting the TCFD span many sectors and geographies and have a market capitalization of $26 trillion (2022).1 However, the adoption of TCFD has yet to be universal. To my knowledge, some companies see it as irrelevant, while others, especially smaller ones, need more resources or expertise to implement the recommendations. Another challenge is partial implementation: Of the 11 disclosure requirements, the average disclosure rate across companies surveyed was 4.2 in 2021.2 Preparers reported many requirements as difficult to implement, most notably the climate resilience element of strategy (85% said it was somewhat difficult or very difficult).

These factors have led governments and regulators to push for a mandatory standard. The most wide-ranging of these is the International Sustainability Standards Board’s IFRS S2 Climate-related Disclosures. It was issued as a draft standard in March 2022, the final standard was delivered in June 2023, and it will go live in January 2024. Countries that have introduced regulations based on the TCFD principles include the United States, Singapore, Malaysia, Hong Kong and Japan.

The IFRS S2 standard has made scenario testing mandatory, with the definition of scenario testing being broadened to include qualitative discussion of particular scenarios where preparers cannot perform quantitative modeling of the scenarios.

Scenario Testing

Before looking in detail at the implementation of scenario testing, it is worth considering some of the benefits.

  • It deals well with issues that occur over the medium or long term where there is a high level of complexity and uncertainty.
  • It forces decision makers to consider, in a structured way, a broader range of plausible outcomes other than business as usual.
  • From the insurer’s point of view, the approach is likely to be in use for other purposes already.
  • It is possible to start simple, even from a qualitative consideration of scenarios, and layer in complexity and additional scenarios later.

The design and selection of scenarios are important parts of the process. The scenarios may be predefined for some applications, such as a field-testing exercise. Still, for others, such as an own risk and solvency assessment (ORSA), the organization will need to identify the scenarios and justify the decisions.

First, it is important to understand the nature of the climate risks to which the business is exposed, particularly whether physical or transition risk is more significant. Typically, life insurers are more exposed to transition risk while property and casualty (P&C) insurers are more exposed to physical risk. Although many exercises focus on one or the other, it is ideal to consider the interaction between the two.

Next, it is helpful to be aware of the existing scenarios that are available. From the practical implementation standpoint, the Network for Greening the Financial System (NGFS) scenarios are commonly used. The latest iteration provides six scenarios that cover different degrees of warming, the key dimensions being whether the transition is orderly or disorderly and whether or not global efforts to halt significant warming are sufficient.

The granularity of the scenario definition is another area that requires careful consideration. We can start from the scenario narrative, which only describes the key assumptions about the climate transition, the timing of shocks and the climate outcomes. This can be translated into specific climate outputs, such as temperature pathways, frequency and severity of perils, emissions, carbon price, energy prices and energy mix. We then move to broad economic outputs, such as gross domestic product (GDP), inflation and interest rate pathways. These impacts then can be disaggregated across sectors and regions based on sensitivity to climate risks.

Further disaggregation to individual firms or even economic activity levels is possible. Current guidance, such as that from the European Insurance and Occupational Pensions Authority (EIOPA), generally recommends looking at impacts at the economic sector level, with shocks calibrated at the country or regional level. It is particularly important to consider the economic sector for corporate bonds, equities and real estate, as transition risk will vary dramatically by sector. For sectors such as energy, more granularities may be considered—oil and gas will underperform in low-carbon scenarios while renewable energy will benefit.

Selecting an appropriate time horizon is another challenge. Stress testing for other purposes is often an instantaneous shock to the balance sheet in a one-to-three-year projection. Still, climate change risks materialize over a medium- to long-term time horizon, with the full impact of physical risks only becoming apparent beyond 30 years. This may be managed by different selections of the calculation time point (at the end of the projection horizon only or at intervals during the projection) and by choosing a static or dynamic balance sheet. A calculation at a single time point with a static balance sheet is relatively straightforward to implement. Still, it makes no allowance for management reactions to the scenario and thus may overstate the climate impact. On the other hand, multiple time point projections with dynamic balance sheets are more realistic but require a complex implementation of multiperiod stresses.

Modeling Framework

Let’s turn to the modeling approach for scenario testing for a life insurer. The TCFD highlighted this as an area where insurers need more technical capability. Given a choice between introducing specialist tools for scenario analysis of physical or transition risk or integrating climate factors into existing risk models, the TCFD indicated the latter was often more effective. In many regions, regulatory requirements have already driven insurers to develop sophisticated company-level models for economic capital, regulatory capital or other asset-liability management (ALM) purposes. These models generally support scenario or stochastic modeling, management actions and dynamic policyholder behavior (see Figure 1 for an example).

Figure 1: Standard ALM ModelFigure 1

Figure 1 shows a typical model layout and the interdependencies among the different elements. We may consider the following adjustments:

  • Asset transition risk—assets can be revalued under different scenarios by applying economic data consistent with climate scenarios.
  • Liability transition risk—the best-estimate liabilities can be recalculated by discounting yield curves consistent with the different climate scenarios, considering any illiquidity premium adjustments. Also, where account balances depend on the value of assets, the balances should be adjusted to match the asset side. Differences in inflation rates may affect the level of expenses modeled.
  • Liability physical risk—changes to mortality, morbidity or other assumptions directly affected by climate will be applied as shocks, which may be product and time dependent.
  • Asset physical risk—for assets held, the physical risk may affect the issuer or counterparty’s business directly, potentially reducing the value of their equity or creditworthiness. This risk would require an in-depth investigation of the business of all counterparties and issuers, so it is likely to be impractical.
  • Management actions—any additional management actions that might be taken in response to unanticipated climate events (e.g., a sudden transition from business as usual to low-carbon transition, as in the NGFS’s Delayed Transition scenario) may be built in. General mitigations reflecting planned changes in strategy, such as changing strategic asset allocations to reduce exposure to high-carbon industries or assets with poor environmental, social and governance (ESG) scores, also may be incorporated.

Data

The most significant challenge facing most companies in the implementation phase is data. Although existing calculation capabilities may be leveraged, the data requirements will change considerably, with data availability, reliability and level of granularity all being potential issues.

Data elements that need to be adapted compared to existing processes include economic scenario data projected to the scenario horizon and future mortality and morbidity rates.

Economic Scenario Data Projected to the Scenario Horizon

Economic scenario data that need to be adapted include:

  • Yield curve
  • Equity returns by economic sector
  • Credit spreads by economic sector
  • Inflation

The relative performance and economy share of different sectors will depend on the details of the scenario, not just the degree of warming. For instance, a scenario in which emissions targets are met using a high level of carbon capture and storage technology would be better for oil and gas companies than one in which regulation forced a rapid shift away from fossil fuels in the energy mix. There is also no established methodology for converting sector-level information about the growth or contraction of an economic sector into an equity return projection, which would depend heavily on the initial assumptions about which climate outcomes are priced into the initial equity costs. Also, the emergence of stranded assets in certain industries will be highly sensitive to the socioeconomic pathways selected.

Preparation of the economic data becomes more complex if the model involves stochastic valuation. In this case, we may require a set of simulations at one or more future time points under the different climate scenarios.

Future Mortality and Morbidity Rates

There is no clear scientific consensus on climate change’s effect on future mortality rates. Reduced deaths from cold weather likely will be offset by an increase in heat-related deaths, subject to a degree of adaptation that will be influenced by increasing wealth over the course of the next century.

A modeling exercise in 2021 suggested that for developed countries with cooler climates, the overall impact after adaptation will be a slight reduction in mortality, while in less developed countries with warmer climates, mortality will increase. The picture seems even less clear for morbidity, where in addition to the effect of temperature, the socioeconomic pathway will affect air pollution (e.g., in a rapid decarbonization scenario). Some exercises, such as the 2021 Bank of England stress test, have assumed no impact on mortality or morbidity. This assumption seems reasonable for an initial model iteration and may be refined later.

Availability of Expertise

Although awareness and understanding of climate risk have grown dramatically in the last few years, I believe there still needs to be more resources and actuaries who have developed the necessary scenario testing skills. In establishing an approach to scenario analysis in this article, we discussed climate science, social and technological policy, and macroeconomics in addition to the more traditional actuarial toolkit. Demand for consulting resources and expertise will be high as companies look to meet their reporting obligations and build their internal capabilities.

In addition, policymakers and regulators are imposing short implementation timelines on the industry. For example, the IFRS S2 standard is planned to go live only six months after it was issued. This will place a further strain on the available resources. A consequence for actuaries is that climate change provides opportunities to move into a nontraditional field. Many actuarial associations, including the Society of Actuaries (SOA), have plans to introduce training programs that cover the fundamentals of climate risk. From my experience, this is an excellent starting point for anyone considering moving into climate risk work.

Conclusion

Scenario testing is a powerful tool for understanding climate risk and is expected to be mandatory when the IFRS S2 standard is released. It is flexible in that it can support various applications, from investigating climate resilience to ORSA. However, companies consider it to be one of the most challenging parts of TCFD compliance. Companies that already have developed a company-level model will benefit, as it is natural to make use of this as a starting point. However, some key challenges remain, such as the following:

  • Selecting and defining scenarios appropriate for the exercise
  • Data availability, reliability and granularity, especially regarding economic inputs to the model
  • Lack of expertise available in the market

The rapidly increasing demand creates new opportunities for actuaries, so moving into climate risk could be an excellent career change.

Brian Kelly, FIA, FASHK, is a senior actuarial solutions manager at FIS in Hong Kong.
Masu Ma, chair of SOA Asia Editorial Sub-Committee, was a contributing editor for this article.

Statements of fact and opinions expressed herein are those of the individual authors and are not necessarily those of the Society of Actuaries or the respective authors’ employers.

Copyright © 2023 by the Society of Actuaries, Chicago, Illinois.