by Mihir Tirodkar
Data plays an important part in investments. Through the analysis of data, investors can better understand the risks they are exposed to and are able to construct portfolios that are positioned relatively efficiently in their risk/return trade-off. Recent advances in data availability and analytics have allowed for improved analysis of risks and the setting of more accurate expectations. The advances of new datasets include ESG metrics and new analytical tools, such as machine learning.
Climate change is an example of a risk that is now being mapped with quantitative data. Climate change itself is a phenomenon comprised of many interlinked processes, each with their own risks and uncertainties. As a trend, investors are becoming more aware of the risks and opportunities that result from climate-related shocks and are starting to appreciate the importance of formal, quantitative risk metrics. To cater to this growing demand for quantitative climate change risk metrics, vendors have begun to explicitly model climate change risk, and some now include climate change risk data as a part of their overall offering.
The physical and transitional risks of climate change1 are also economic risks and are closely linked to long-term portfolio performance. Like the more traditional measures of portfolio risk, climate change risk is of relevance to risk-averse investors; however, climate data is different. As a relatively new area of risk analysis, climate data is still developing; consensus views in many aspects of risk modelling are yet to be established. Climate change also consists of many complicated sub-processes that interact with one another through flow-on effects, which increases modelling complexities. Many of the underlying processes that govern climatic variables are dynamically evolving over time,2 resulting in modelling difficulties when solely using historic climate data to forecast the future3. Despite these hurdles, there have been major increases in the sophistication of available climate change risk metrics, and we continue to see development in this area.
What are climate change risk metrics?
Initially, climate risks were proxied mainly using portfolio carbon exposure; this was typically limited to equity securities. Limitations of such an approach are that carbon emissions are predominantly backwards looking, while investment risk is concerned with the future. Furthermore, carbon emissions only reflect a portion of total transition risk, which is itself only a subset of total climate risk.
Developments in climate risk metrics now allow investors to consider multiple avenues of climate risk, in which multi-asset exposures to various climatic variables (i.e., both physical and transitional factors) can be used in risk analysis. As a result, investors are now able to use quantitative data that maps their portfolio sensitivity to climate change scenarios.
Climate change risk metrics are indicators of investment exposure to a variety of climate risks. These climate risks include various physical risks (broadly categorized by the acute and chronic physical processes that might play out in climate scenarios)4 and transitional risks (categorized by the Task Force on Climate-Related Financial Disclosures as policy and legal, technology, market and reputational risks).5 Both physical and transition risks are comprised of separate factors that investors should be concerned with, as each individually poses their own risk.
Providers of climate risk metrics vary in the individual drivers of climate risk that they model. There are differing levels of granularity in the sub-processes and climate risk factor exposures that are estimated by the vendors of climate metrics. For investors that care about specific sources of climate risk, there is value in spending some time searching for vendors that provide data that match the investor’s needs.
The granularity of vendor output varies significantly. Some data providers estimate risk metrics at the security level, while some estimate metrics at the asset class level. There is also some difference in coverage of securities and/or asset classes between providers.
Some providers of climate change risk metrics choose to generate separate climate risk metrics for several climate risk factors independently. An example of this would be a provider estimating sensitivity to temperature rise and carbon pricing shocks, separately to each other. This allows flexibility for data users, who can consider each source of risk and their respective investment exposures independently. Other data providers choose to amalgamate climate risk metrics into broader categories, such as total exposures to aggregate transition or physical shocks. The advantage of this approach is that it implicitly accounts for the interrelated processes of the sub-factors that make up the broader categories of climate risk.
Unlike conventional investment variables (such as factor betas or price multiples), climate risk metrics are typically not presented as continuous metrics. Instead, they are more often estimated as sensitivities to climate factor shocks through a scenario analysis framework. For example, a provider may estimate the expected losses of an equity portfolio in a hypothetical scenario of a 3°C temperature rise combined with a lack of regulatory response to carbon pricing. The advantage of such an approach is understandability— end users can easily form expectations of how their portfolio might behave in certain states of the world; the same users can then form beliefs on how likely these states are to happen, thereby creating expectations of future performance in light of climatic factors. Some climate data vendors allow users to create their own bespoke scenarios, where users can specify their beliefs on how individual processes will behave within given scenarios, therefore providing greater flexibility.
How are climate change risk metrics estimated?
Exposures to each type of transition and physical risk are calculated differently and require different datasets. For example, exposures to some physical risks, such as sea level rise, are usually calculated using some form of geospatial analysis to determine potential losses for physical assets located near inundation/flood zones. In contrast, exposures to regulatory costs are a function of a firm’s emissions trajectory and ability to pass on regulatory costs. Depending on the scale and depth of the model, each individual climatic driver can be set as its own process that generates risk exposures for the asset being modelled.
As the field is still developing, there are multiple ways that individual climate processes are modeled. For example, some providers estimate risk exposures using statistical analysis on historic relationships, while others use theoretical, closed-end integrated assessment models6 as a basis for their predictions.
Underpinning any modelling is a set of assumptions. Because climate change is multifaceted, data providers often look to experts in each of these fields to develop their respective sections of the overall model. While some assumptions are made in-house by data providers, other assumptions are based on those in established climate change reports; for example, many assumptions on people’s behaviors are based on the representative concentration pathways built by the Intergovernmental Panel on Climate Change.
The need for making assumptions implicitly illustrates the difference between climate change risk and climate change uncertainty – while models can be custom made to show the distribution of climatic outcomes, there is a Knightian uncertainty7 in the setting of these assumptions themselves.
Ultimately, many models that estimate climate risk metrics rely on proprietary models that may seem like black boxes with all their interrelated, moving parts. However, this is often by construction. Due to the complicated interrelationships between economic and climate variables, there is a need for sophistication, which can result in modelling complexity. Translating economic and environmental shocks into portfolio outcomes (in return-space) adds an additional layer of complication, as assumptions must be made on how efficiently markets are pricing in these future shocks.8 If a baseline scenario is already priced efficiently, models would assume that climate-related costs and opportunities are already factored into prices, which would then be relatively insensitive once the expected climate events transpire. The impact under additional scenarios will be felt relative to the baseline scenario that the model assumes the market is expecting.
How are climate change risk metrics used?
Climate data is useful in various parts of the investment process. Currently, at Russell Investments, we are looking to incorporate these variables in a multifaceted approach. Firstly, projected climate change distributions are useful information when creating overall capital market forecasts. Using climate change risk metrics, Russell Investments is investigating how to better inform our expectations of risk and returns at the asset class level. Secondly, our in-house risk management practices are also exploring how to use these metrics in a more granular, portfolio-specific approach to estimate the climate risk. We also expect that these metrics will also play a big role within our managed funds, with managers being able to better identify risks and opportunities associated with the climate. We believe that asset managers will soon be able to use more sophisticated metrics within the reports generated by their management and consulting functions. Lastly, additional climate risk data will enable asset managers to engage and encourage companies to incorporate climate change measures into their management strategies.
Where is the industry headed?
As a nascent industry, climate change risk metrics are fast developing and will soon be more readily available to better inform the strategies and decisions of financial market participants. As awareness of climate risk increases among investors, we can expect to see greater sophistication in the modelling of these risks, more availability of metrics from a larger number of providers, a growing role of industry specialists in climate, better quality outputs and a greater number of use-cases of the final data.
Russell Investments, along with many other asset managers, is looking forward to, and actively taking part in, the evolution of climate change risk metrics. Hopefully, development in this field will lead to more considered strategies for investors concerned about climate risk.
ENDNOTES
1 See https://russellinvestments.com/us/blog/climate-change-impact-investments
2 In statistics, climate change and other processes which have moving distributions through time are referred to as ‘non-stationary’ processes.
3 For example, if one were to predict future long-term average temperatures using only historic temperatures, or the likelihood of future regulatory changes based only on historic climate change regulation, one might be creating biased forecasts due to structural regime shifts in processes.
4 Examples of acute physical risks include losses resulting from storm surges, heatwaves or hurricanes. Examples of chronic physical risks include sea level rise or higher average temperatures.
5 An example of policy and legal risks includes higher carbon prices, while an example of technology risk includes new developments in abatement technology. An example of market risk includes shocks resulting from changing demand and supply curves, and an example of reputational risk includes consumers’ preferences shifting towards ‘green’ substitutes.
6 Integrated Assessment Models, or IAMs, are complex models that integrate environmental, energy and economic variables to form climate pathway expectations for the future.
7 Knightian uncertainty refers to some degree of ignorance of the true probability parameters of a future event.
8 Anecdotally, we find that a number of climate risk data vendors assume that the market has under-priced the true costs of expected climate risks, and as a result, securities with exposure may be overpriced and exposed to a return shock once pricing becomes more ‘efficient’. Some investors also have adopted this viewpoint in their portfolio strategy.
