On 28th September 2021, GARP hosted Oliver Marchand, Global Head of ESG Research and Models at MSCI. In this webcast, Oliver delivered practical advice on understanding, creating, and deploying climate risk models with insights spanning core principles, data guidance, and best practices for financial firms. You can (re)watch the full webcast on-demand here.
Following the webcast, Oliver kindly took the time to respond to the most popular unanswered questions from the webinar’s live Q&A. In the article below, we have structured Oliver’s responses into the following groupings: data sources and models; the limitations of climate risk modeling; uses of climate risk modeling at financial firms; and technology-based solutions to climate change. (Note: GARP has added a few explanatory notes in italics).
Questions about available data sources and models
Q: We’d very much welcome your thoughts on data sources. In particular, where can risk professionals get data on ‘green premiums’ and hazards data? Note: Green premiums refer to the difference in cost between a product that involves emitting carbon and an alternative that doesn’t.
A: Green premium data is hard to get. I am not aware of a consistent dataset. Some numbers are available from Bill Gates’ book (‘How to Avoid a Climate Disaster’); Gartner also has some numbers. We have calculated a number of green premiums ourselves.
Hazard data is available from many international collaborations. I advise to use the Inter-Sectoral Impact Model Intercomparison Project (ISIMIP6) dataset. Check out http://isimip.org. Each local weather and climate service usually has some local hazard data available. Whether it is free to use or not is a matter of the local regulations.
Q: Are there any other useful data sets that are freely available for risk professionals that you would recommend?
A: That’s a very general question. It really depends on what you are trying to model:
- For countries, I would start with some of the freely available country-level data from the World Bank or some of the open-source projects on geospatial data:
- For companies, freely available data is harder to get – but I would start with some of the freely available exchange-traded fund (ETF) data:
- For sectors, there are usually trade organisations with data that is freely available.
Q: What is the name of the Potsdam climate model, which you recommended? Why is this one particularly good?
A: It’s called the Regional Model of Investment and Development (REMIND) model and I recommend it because it has the highest granularity. You can get their data from the Network for Greening the Financial System (NGFS) site.
Note: Developed by the Potsdam Institute for Climate Impact Research, the REMIND model is an “energy-economy general equilibrium model” used to inform investment strategy based on regional data.
Q: Can a non-financial company use the NGFS dataset? Will it impact the outcome of the analysis?
A: I can’t give legal advice on the usage of other parties’ data, but it’s my understanding that the usage rights are very broad.
Q: The graph you showed displayed estimates of the transient climate response to cumulative carbon emissions (TCRE). Are there any estimates of the "effective" TCRE that you are aware of?
A: To my best knowledge, the most recent studies do take into account both:
- A theoretical approach to the induced warming.
- Evidence from the currently observed warming (what I understand you mean by “effective”).
Note: TCRE is the ratio used in climate science to calculate the globally averaged surface temperature change per unit of CO2-equivalent greenhouse gas emissions.
Questions about the limitations of climate risk modelling and scenario analysis
Q. What are the limitations of integrated assessment models (IAMs) that risk professionals need to be aware of?
A: Let me try to name a few:
- IAMs are what-if scenario modelling, and we don’t really know the exact statistical probabilities of the modelled events occurring.
- It’s important to understand each IAM at a broad level to be able to understand the possibilities and the limitations.
- IAMs are usually not satisfactory in terms of the regional and sectoral breakdown of the outputs.
- Strong policy interventions can easily render the assumptions in IAMs useless.
- Different IAMs might “disagree” on the simulation of similar assumptions.
- Some IAMs are deliberately very, very simple to study one main aspect of climate modelling.
- In some IAMs, there is a heavy bet on negative carbon technologies – especially carbon capture utilization and storage (CCUS).
This is a great article.
Q: Given the non-linearities and irreversibility of climate change, as well as the catastrophic nature of impacts once certain climate thresholds are breached, shouldn't we be focusing more on tail risk? Perhaps we should be more focused on precautionary principles rather than settling on "average" scenarios, which might well be an underestimate of the "true" risk?
A: Good point. For that reason, in our physical risk model we model “average” risk and “aggressive” risk – which is exactly what you describe. But tail risks are hard to compute in general. Apart from the mathematical complexity, there is also a psychological factor. If the risk calculations yield very high values, practitioners will dismiss the dataset, or it will provide little insight or differentiation.
Questions about financial firms’ uses of climate risk modelling and scenario analysis
Q: What are the most important KPI's for climate risk modelling?
A: Check out the Task Force on Climate-Related Financial Disclosures’ (TCFD) chapter on metrics. You will find a lot of KPIs that you can use. Here’s an example.
Q: To facilitate institutional capital allocation, a strong understanding of transition and physical risks on credit impairment is needed. What are your thoughts on mapping scenarios and risks to credit recovery curves for risk opportunity identification?
A: We usually use a combination of existing credit risk research and simulations using the Merton model. The Merton model has default risk as an output. But the values that come out of the Merton model can’t be used on practice. These values need to be calibrated to sector and size of the company. This works in both ways: risk increase and risk reduction = opportunity modelling.
Note: Credit risk refers to potential losses arising from a borrower’s failure to repay outstanding financial obligations. The Merton model is a commonly used credit risk analysis model.
Q: When thinking about ‘alignment’ and carbon budgets, how do you allocate those budgets between sectors and then companies?
A: We do it using the Paris agreement. As part of the Paris agreement, the Nationally Determined Contributions (NDCs) of each country often detail the emission reduction plans on a sector level. To get to a certain temperature scenario, like 2°C or 1.5°C – we amplify the reduction requirement. There are obviously multiple ways of doing this. Think about your view on what the most likely decarbonization pathway is and model such a path. Or alternatively, model multiple pathways and compare them.
Q: Are there financial institutions that have worked the modelling end-to-end and figured out their Climate Value-at-Risk (CVaR)?
A: I know that some banks have created their own CVaR. I also know of some institutions mixing data from various providers with their own modelling.
Note: Climate Value-at-Risk is a “measure of the potential for asset-price corrections due to climate change”.
Q: How would you model transitional risks on corporate profits?
A: Please check the slides on transition risk modelling or consult this site. There’s also a link to a PDF that explains Climate Value-at-Risk.
Q: How do we go about incorporating climate risk into pricing? For example, how can we put a number on the price difference between an asset-backed security (ABS) of standard loans and an ABS of ‘green’ loans?
A: I am not an expert on ABS, but generally speaking I would think about how the pricing scheme differs with and without including climate change. With ABS, I would guess that the physical risk component is very dominant.
Q: Should decarbonization pathways use both short and long-term metrics for short-lived climate pollutants (SLCPs) like methane? For financial institutions with significant fossil fuel exposure, would you recommend a 20- or 100-year warming scenario timeline?
A: Theoretically, yes. In practice, I think there are two options:
- Using 2100 as the long-term time frame.
- Using 2030 or 2050 as intermediate timeframes with some level of certainty about the policy direction.
Questions about solutions to climate change
Q: In the list of the 80 existing solutions to climate change compiled by Project Drawdown, CCUS is not mentioned? Is there a specific reason for this? Note: Project Drawdown is a non-profit organization that assesses and promotes climate solutions to facilitate the transition towards global net-zero. They maintain a list of effective climate solutions.
A: You would have to ask the Project Drawdown. But I think it might either be because it is a relatively new concept, or alternatively because current applications of CCUS are not advanced enough to meet the standards of the project group.
Q: Are there metrics where companies are scored on how many of the 80 they have implemented? Or should we be thinking about this in a different way?
A: That’s a great idea. A few of these technologies you can derive from revenue data (e.g. wind power). For most, it will be hard to get good data on a meaningful universe of companies.
You can (re)watch Oliver’s webcast here at any time on-demand. You can also check out other climate risk webcasts, podcasts and articles here.
More
