The risks associated with transitioning to a low-carbon economy, or the climate change that comes with not transitioning, present a potentially significant threat to the global economy and financial system. The demand from regulators that asset managers, banks, pensions funds and insurance companies assess and monitor the potential impact of different climate risks on their organizations is gathering pace. With many central banks and regulators already launching consultations and stress tests; and many other countries committed to follow suit; it is vital that companies prepare now to meet the requirements that increased scrutiny of climate related risk will bring.
For insurance portfolio and risk managers, there are potentially multiple areas that need to be considered with respect to climate risks, and each poses certain questions that should be answered. These include:
- Market risk modeling – are the market risk models that we are using already fit for purpose based on the best information available today? How might climate factors affect credit worthiness?
- Stress testing and scenario analysis – do we need to add climate scenarios into our existing stress testing and scenario analysis frameworks? What should these scenarios look like?
- Internal and external risk reporting – What, how and where do we want to report climate risks?
- Analytics and data – What analytics should we look at or develop? What data do we need to create those analytics?
- Incorporation into Strategic Asset Allocation – How might our medium- and long-term strategy need to change?
- Benchmarking the risk – How do climate risks compare with other types of risk (e.g., operational, credit risk)? How does a portfolio compare to the market benchmark or peer group?
Assessing climate risk on a portfolio is however proving challenging to implement in practice. While the "what" is clear – what impact climate change and possible governmental policy action have on a firm's asset risk – the "how" has left many risk practitioners scrambling for an answer.
Stress testing using climate change scenarios is one part of the puzzle, and it is becoming the tool of choice for central banks and regulators. However, standard stress tests give few insights into climate impacts on an asset portfolio which can be used for strategic planning purposes. Currently gaining popularity is the technique of combining existing stochastic modeling techniques with climate scenarios; this is in part because it enables the development of a rich set of quantitative analyses which can then be used to develop a qualitative understanding of the risks posed in terms of already well-understood concepts such as value-at-risk. In this chapter we examine some of the approaches that are in use for measuring and monitoring climate risks on a portfolio, and summarize some of the most important considerations in the implementation of these methods.
Climate Risk and Scenario Analysis
Climate risk is somewhat different from the types of financial risk that we are normally concerned with as market risk practitioners. For most risks, we can define a model that broadly represents the dynamics of the market that we are considering and use historical data or analysis of other sources such as fundamentals to estimate the model parameters. The model can then be used to project a distribution of outcomes that might occur, effectively assigning probabilities to them.
With climate risk, however, we have little to no historical data that is of practical use, and many future developments in the world's response to climate change are both unknown and unknowable. For instance, we would need to know the names of the next five presidents of the United States of America, understand their position on climate change, and have certainty about their ability to pass laws which align with that position to have any hope of predicting the future or assigning a probability to a particular outcome.
In such cases where it is difficult to assign a probability to each outcome, scenario analysis is a useful tool and indeed the one that has been adopted by many organizations in analyzing climate risk. In the context of the risk on a portfolio of assets, scenario analysis is used to take what we believe the risk profile of our portfolio is based on current best estimates and then consider that risk conditioned on a particular scenario. Instead of assigning probabilities to each outcome, we ask what would happen if one particular set of actions were to occur in reality. The first question that needs to be considered when undertaking climate scenario analysis is, which scenarios might be significant to consider?
Fortunately, there is some convergence on which scenarios are most important, and much of the work in defining scenarios is already well developed and grounded in extensive research. Many central banks have or are in the process of developing concrete specifications around the regulatory requirements for climate scenario analysis. Much of this work is being harmonized globally, in part through their association with the Network for Greening the Financial System (NGFS), an umbrella organization for central banks and regulators concerned with environmental issues. The NGFS has defined a framework which is important within the context of portfolio risk management in terms of transition risk and physical risk, as well as whether a 2ºC temp erature rise target has been met or not. Example of the most commonly used scenarios are:
- Disorderly transition or late action – governments do little in the way of policy action in the next ten to fifteen years and then later are forced to make large adjustments to the cost of carbon with short-term but high- magnitude disruption to financial markets.
- Orderly transition (Early Action) – governments take early policy action, increasing the cost of carbon steadily over the long term.
- Hot house world – Sometime called a no- action scenario. Governments make little in the way of changes to climate policy, and global temperature rises in excess of 4o Celsius relative to pre-industrial levels are experienced over the next 50 to 80 years.
- Too little, too late – Governments make continual adjustments to the cost of carbon but do not act quickly enough to avert the worst effects of climate change.
- RCP based – Scenarios based directly on the representative concentration pathways (RCPs) defined by the Intergovernmental Panel on Climate Change.
An important distinction made in all of these scenarios is that between physical risk and transition risk, and analysts are usually asked to consider both under a given scenario. While no single definition of these risk sources exist, transition risks can be described as risks that arise from the transition to a low-carbon economy; these might include risks arising from policy changes, such as energy-efficiency regulations or carbon pricing mechanisms which increase the cost of fossil fuels. They may more broadly also include risks arising from disruptive technologies, which might rapidly replace a technology that is more damaging to the climate. Changes in consumer habits and reputational elements may also be considered to fall under the transition risk definition.
Physical risks arise from the physical effects of climate change and are perhaps more closely related with how we naturally perceive climate risk. These include acute physical risks, which might arise from weather-related events such as storms, floods, fires or heatwaves that damage physical property and disrupt the operations of companies that we invest in. More subtle are chronic physical risks, which manifest themselves over longer time horizons, such as rising sea levels, changes in temperature, and reduced water availability. In more detailed analyses, tertiary physical effects of climate change on biodiversity, soil quality, migration of people, higher incidence of conflict and other sources might also be considered.
One of the more fully developed examples of defining climate scenarios is the Climate Biennial Exploratory Scenarios (CBES) proposed by the UK Prudential Regulatory Authority (PRA) in June 2021. In this stress test, the PRA defined pathways for a broad range of financial, macroeconomic, physical and transition climate risk variables under three different scenarios. These scenarios were labeled Early Action, Late Action, and No Action, and defined in terms of the timing and magnitude of governmental policy actions that might reduce climate change (e.g., through increasing the cost of carbon).
Defined over a thirty-year time horizon, these CBES scenarios form a good basis for trying to understand the likely effects of each scenario on a portfolio of assets. While this stress test was defined by the UK regulator, it can be argued that it has global significance, because the scenarios are based heavily on the work of the NGFS, which is a global organization. The CBES also defines pathways for several different economies, including the United Kingdom, United States and Germany.
Given the many potential and complex sources of climate risk, a reduction of the problem to a small number of representative scenarios has significant advantages. By defining a standardized set of scenarios to consider, a good framework is in place to enable quantitative and qualitative comparisons of their impacts on an asset portfolio. In the next section we consider some of the main considerations, challenges and progress that has been made in more fully defining the impact of climate scenarios on asset valuations.
Figure 1- Climate scenarios for US A-rated credit spreads under an early action and late action scenario (left) and near surface air temperature scenarios under early action and no action scenarios (right)
Defining Climate Stresses for Financial Variables
Having a set of narratives for particular climate change scenarios is only the first step in analyzing and understanding the potential climate risks to which a portfolio is exposed. To make climate scenario analysis practicable, we must have a basis for defining what the risks of a particular set of investments are. The PRA, in its stress test of 2019 and subsequent CBES stress test (2021), provide a useful starting point for quantifying the impacts of different scenarios on financial variables and ultimately on asset returns.
The earlier stress test assigned deterministic shocks to be applied to the current market value of a portfolio based on asset-class-, sector-, and subsector-level information about the asset allocation. Further, a prescribed methodology was given for calculating the stresses on Sovereign bonds and United States Municipal Bonds. One of the big challenges in implementing this approach is mapping the securities held in the portfolio with the sectors and subsectors defined by the PRA. Another challenge is that for some firms the exposure spans multiple sector/sub sector classifications. For instance, a traditional oil and gas provider may have already partially diversified into renewables, and as such a stress should be calculated taking account of the proportion of revenue generated by each source (e.g., oil, gas, renewables).
The later CBES scenario definition was a lot less prescriptive and gave paths for broad financial, macroeconomic and climate-relevant variables. The insurer is then charged with taking that information and developing an approach to apply the scenarios to their business. This development was somewhat deliberately intended to ensure that users of the scenarios carefully considered their approach rather than just ticking a box. There was also no method prescribed for splitting the effects between transition and physical risk, and this again is left up to the user to decide and justify. Examples of CBES scenarios for some variables are shown in figure 1 for illustrative purposes.
Security-Level Risk Assessment
Regulator and externally defined scenarios only go so far in capturing the climate risks of a specific instance of a portfolio, however. This is because, as we have seen, they are typically defined at the higher strategic level rather than the individual ISIN or CUSIP level. For instance, in the credit spread scenarios shown in the last section, only the market average credit spread is defined. However, a corporate bond portfolio which is heavily invested in fossil fuel might be expected to have a markedly different risk profile under a given scenario than one heavily invested in green technology. The split between transition and physical risk for two such portfolios is also likely to look quite different. It is therefore necessary to consider the granular makeup of the portfolio in defining the magnitude of stresses to be incurred in scenario analysis or in assessing a portfolio's overall vulnerability level to climate risks.
This task usually requires some external data based on company-level analysis, and then an approach to process this data into usable metrics. This might, for example, involve each issuer being assigned a transition and physical risk score in the range of one to ten. This can then be used to score the aggregate portfolio and benchmark it relative to other portfolios or manage the exposure to be below some target threshold. Since data may not be available for all issuers, it is vital that particular focus is given to the most carbon-intensive subsectors of the portfolio. These might include manufacturing, mining, energy, utilities and automotive. Further, for each issuer it can be useful to consider climate risks as being composed of several subcategories of risk. For instance, we often split transition risk into two subcategories:
- Technology risk is the risks and opportunities associated with the technological shift in products and energy consumption patterns in moving toward a low-carbon economy.
- Policy risk encompasses the regulatory and legal aspects with which a firm will need to comply, as certain jurisdictions require behaviors consistent with climate-aware strategies.
Finally, it is important to consider at the individual issuer level the potential for "stranded assets," or assets that turn out to be worth less than expected as a result of changes associated with the energy transition. As an example, in Australia there are coal-fired power plants that are no longer economically viable.
Case Study: An Orderly Transition to a Low Carbon Economy
As more information emerges about what will be required with regards to climate risk reporting, we are seeing a convergence across sectors on many of the details. For instance, the UK pensions regulator, TPR, unveiled a new climate change strategy in 2021; this obliges trustees of big schemes to make climate change disclosures and commits the regulator to reviewing the resilience of schemes through a regular review of the pension funds' climate scenario analysis. For asset managers, elements of both TCFD and PRI disclosures encourage a consideration of different climate scenarios from a risk and return perspective.
We now show an example of how stochastic techniques combined with scenario analysis can be used to create a rich set of metrics to illustrate the effect that a transition to low- carbon economy might have on two portfolios. Portfolio 1 is made up of 70% fixed income investments and 30% risk assets, while Portfolio 2 is split 50%/50% between fixed income and risk assets. Fixed income includes developed world government and corporate bonds as well as United States municipal bond holdings. Risk assets are defined as global equity and real estate covering Europe, North America, and Asia Pacific. The models used include credit risk where appropriate.
We model each of these portfolios under an orderly transition to a low-carbon economy over a thirty-year future time horizon. The size of the shocks is based on a security-level analysis of the portfolio exposure to different carbon-intensive or at-risk subsectors taken from the 2019 PRA exercise. What is more, we shall consider two possible through-time paths that this scenario could take. In one we shall consider the transition risk impact on financial markets as increasing linearly with time, peaking in year 30 of the scenario. In the second we shall consider a more rapid decarbonization scenario in which the financial damage or impact to asset returns increases rapidly in the next 10 years and then falls as markets reorder to absorb the structural changes that have taken place over the subsequent 20 years of the scenario. For physical risk under this scenario, we assume impacts consistent with the shape of the Weitzman financial damage function, one of the commonly used paths from the financial literature.
Figure 2- Climate Risk Analysis using the Conning Climate Risk Analyzer software. The effect of scenario B on the mean market value (left) and 1% annual VaR(right) of the portfolio are shown
Figure 2 shows some analysis of the two portfolios conditional on this orderly transition scenario according to the two paths defined. Shown is the effect through time on the mean market value of each portfolio (top left) and the 1% annual value-at-risk (VaR) (top right) of each portfolio. The effect of the scenario is expressed as a change in the statistic relative to the current base or best estimate of risk and reward. A value of -1% in the mean, for instance, can be interpreted as saying that conditional on this scenario playing out, we would expect that the portfolio market value is 1% lower than our current best estimate. We refer to this additional risk as Excess-Climate-Risk (ECR). Also shown is the attribution of the ECR to the different sources of risk, transition (middle) and physical (bottom), which is another useful feature of the algorithm used.
The analysis reveals some interesting results and conclusions. For both portfolios we observe that, as expected, a more rapid decarbonization scenario would lead to larger impacts on risk and return at the shorter and medium time horizons used for strategic planning. Over the longer term and full 30-year horizon of the scenario, we can see that the portfolios are relatively insensitive to the path that the climate scenario takes through time. Because the impact of the rapid decarbonization scenario is "front loaded" in time, markets later reorder and reprice the risk, allowing the portfolios to recover some of the drawdown relative to the linear transition scenario.
We also see that the scenario has a larger impact on risk, defined here as the 1% annual VaR, than on return. For Portfolio 1 the mean return can be expected to be 5% lower under the rapid decarbonization scenario than would otherwise have occurred, while the 1%-VaR has increased by nearly 6%. For Portfolio 2 the differences are even larger, with the mean return decreasing by approximately 7.5% while the 1%-VaR increases by 9%. If we consider the 5-year horizon, which might be used for ORSA reporting or analysis to form part of the TCFD disclosures, we observe an increase in risk of only 1.3% for Portfolio 1 and 1.9% for Portfolio 2 under the rapid decarbonization scenario.
We might conclude as part of these analyses that the absolute value of the increased risk and shocks to return are not particularly large at this horizon relative to other types of risk that we might consider (e.g. operational risks, sovereign default scenario etc.). However, at the longer horizons these risks clearly have the potential to become significant, and we may wish to act to mitigate or plan to mitigate them on a shorter time scale.
This serves to demonstrate the practical way in which we might go from a defined scenario to an understanding of the effect of that scenario on our asset allocation and portfolio risk.
Requirements for climate risk disclosure and reporting are gathering pace on all fronts. Many financial services firms, however, are quickly finding difficulties in the implementation stage of scenario analysis that is crucial to quantifying and understanding these risks. This is especially true of smaller and mid-sized organizations where the cost of data, resource requirements and expertise all act to impede meaningful analysis of climate risks. At the time of writing (2021) the European regulatory body EIOPA reported that only one in eight ORSAs included climate scenarios. This highlights a significant gap between the requirements as laid out and the ability to practically implement them within a risk management framework. In this chapter, we showed an example of how existing stochastic risk management platforms can be used to bridge the gap between the well-defined concept of a climate scenario and the ill-defined methodological problem of implementing that scenario for a real portfolio.
We have also introduced some of the most important concepts that an analyst should be familiar with on this topic. We have described some of the considerations that are needed in defining an instantaneous stress on a portfolio of securities based on the portfolio exposure to carbon-intensive or particularly at-risk sectors as well as security-level information. Finally, we have shown how stochastic projections conditioned on climate scenarios can be used to create a rich analysis of the potential impacts on risk and return. This type of analysis is an important part of the puzzle in understanding climate risks, and it can act as a driver for creating the types of informed narratives that are key in responding to the ORSA, PRI, TCFD and other requirements.
This piece contains forward-looking statements. Readers should not place undue reliance on forward-looking statements. Actual results could differ materially from those referenced in forward-looking statements for many reasons. Forward-looking statements are necessarily speculative in nature, and it can be expected that some or all of the assumptions underlying any forward-looking statements will not materialize or will vary significantly from actual results. Variations of assumptions and results may be material. Without limiting the generality of the foregoing, the inclusion of forward-looking statements herein should not be regarded as a representation by the Investment Manager or any of their respective affiliates or any other person of the results that will actually be achieved as presented. None of the foregoing persons has any obligation to update or otherwise revise any forward-looking statements, including any revision to reflect changes in any circumstances arising after the date hereof relating to any assumptions or otherwise.