Climate risk models and metrics: what works and what doesn’t?

Panellists at Risk Live Europe 2023 discussed what has been achieved so far in climate risk stress-testing and modelling, alongside what needs tackling next

The panel

• Prerna Divecha, Global lead, market strategy and development, climate-linked credit and risk solutions, S&P Global Market Intelligence
• David Carlin, Climate risk lead, UN Environment Programme Finance Initiative
• Navin Rauniar, Co-chair, ESG working group, and UK SteerCo member, Professional Risk Managers’ International Association

The importance of climate risk stress-testing has grown significantly in recent years, driven by growing regulatory requirements. Approximately 30 central banks and supervisors worldwide now require banks to carry out some form of climate risk stress test or scenario analysis. 

However, financial institutions conducting this nascent analysis now face an array of challenges: from the lack of available data to the difficulties of integrating the more time-consuming horizon of climate risk into traditional credit risk modelling. Additionally, while organisations such as the Network for Greening the Financial System (NGFS) and the Task Force on Climate-Related Financial Disclosures (TCFD) are helping to create some standardisation, much work remains to be done. Regulatory requirements differ across jurisdictions, while the methodology and underlying economic and climate models used by different financial institutions can vary significantly. 

Talking at Risk Live Europe on June 7, David Carlin, climate risk lead at the UN Environment Programme Finance Initiative, said the output of these early stress tests is still fairly limited in its ability to estimate climate risk in a portfolio. 

“In the summary findings from most of these stress tests, the biggest value has not been the uncovering of new information or even the sizing of the risk … but rather it has been a valuable learning exercise on the topic of the readiness of institutions, what resources and parts of the team they need to marshal to do this, and where the data and data gaps are,” he said.

When it comes to individual institutions understanding their idiosyncratic risks, and supervisors understanding systemic risk, there is a long way to go, said Carlin. “In terms of the rank ordering by scenario and exposure by sector, most of the assessments [so far] have returned fairly intuitive and expected results,” he said. “The next step is to really begin refining the sizing, refining what this means when it comes to decision-making in the near term and where the potential risks lie that may be missed or underestimated by these initial runs.”

Prerna Divecha, S&P Global Market Intelligence

The panellists agreed that reaching this stage will be challenging. Prerna Divecha of S&P Global Market Intelligence has worked closely in recent years with banks on their climate stress tests, acknowledging that resources and data “are very real challenges faced by our clients and financial institutions when it comes to climate stress-testing”. 

Not only is there a dearth of data on non‑listed firms but, even where there are disclosures, a lack of harmonisation can make comparability difficult, although initiatives such as the TCFD are improving this, she noted.

Inevitably, much of her work has involved reconciling the difference in time horizons between climate scenario analysis and traditional credit stress tests or models. 

In traditional analysis, when events such as stock price moves or an oil price shock are being stressed, it’s much more measurable, she said. “[With] climate-related stress-testing, there are a whole number of dynamics coming into play, and fitting them into traditional stress tests has been quite a challenge.” According to Divecha, working out how to quantify these risks in a shorter time period and bringing them into current portfolio analysis has been one of the most important lessons of her work.

Additionally, another area of work has centred around how to adequately capture sectoral differentiation, in a way that will be meaningful for investment and lending portfolios. This applies not only to risks, but also to opportunities in transition financing, she stressed. “[It is important] to be able to distinguish – even within a particular sector – corporate counterparties, which are, for example, first-movers in [transition] technology. Even within sectors, there are these pockets of opportunities that need to be identified. Capturing this during stress‑testing has been another challenge.”
 

Disparity concerns

Navin Rauniar of the Professional Risk Managers’ International Association described the lack of standardisation of climate stress-testing as “very, very worrying”. 

“Central banks are asking for different things in their stress tests,” he noted. “Some are looking at static balance sheets, others at dynamic balance sheets. Time horizons vary. Even between the European Central Bank and the Bank of England, the scope is completely different in terms of sizing, the number of counterparties, and also the data and the models,” he said.

According to Rauniar, this risks inaccurate output and bad decisions, evoking memories of the causes of the financial crisis that begain in 2007–08.

“If we’re going to use the output of climate risk stress-testing to make decisions on which sectors to invest in for the real economy, especially in these very challenging economic times, [it isn’t a good idea to do it] with poor data and models,” he said.

Rauniar noted that, although the NGFS is “doing a good job” refining the models, more needs to be done. For example, the NGFS’s National Institute Global Econometric Model is one of the most widely used econometric models, but it has some major shortcomings, he said. For example, the model doesn’t account for the possibility of “selloff” behaviour or volatility between five‑year time steps and, therefore, cannot model the impact of the war in Ukraine, or a scenario around China and Taiwan that would impact commodity prices. As a result, many banks are turning to their own economic models.  

This difference in scopes across supervisors and banks needs to be addressed, said Rauniar. “We should not let this again lead to banks, financial services, insurance companies, and so on, withdrawing from the real economy and potentially causing systemic risk in the system.”
 

Causes of climate risk

A lot more work still needs to be done to gain insight into the drivers of climate risk, said Carlin. “Stress-testing may not always be the most useful thing for business decision-making. What we’re struggling with right now is how we are using these results. Is it to help us set decarbonisation targets? Is it to help us to find opportunities? Is it to manage short-term risks? Is it to manage long-term risks?”

If you want answers to all of these questions, then “individual models and individual types of assessments are going to prove themselves inadequate”, he said. “A greater focus and a greater notion of transparency in the approaches are going to be needed.”

Greater transparency is also needed around methodologies, commented Divecha. Modellers need to be very clear on what it is they are trying to measure before, for example, using proxy data.  

If the aim is to quantify the risk of climate change, the methodology needs to consider things such as market dynamics, supply/demand elasticity and expected changes to technology, she said. “[This will reveal] some drivers of financial impact, like the expected changes in commodity prices, volumes, what kind of assets could get stranded and the following direct implication on corporate balance sheets.”
 

Pricing climate risk

As the industry progresses slowly towards quantifying and pricing climate risk within a portfolio, or in individual asset prices, discussions are taking place around the use of a climate value‑at‑risk (VAR) metric. While Carlin warned that distilling anything into a single number has the potential for misinterpretation, he acknowledged that the desire to estimate the asset implications of climate and transition risk was an important aim of the TCFD.

“That said, [although] the TCFD has been a massive success in making climate a mainstream topic for the financial sector, it has not necessarily been as definitive in terms of the impacts on pricing,” he noted.

While trying to quantify the climate risk impact on specific asset valuations is “absolutely valuable”, Carlin said we should also be considering the impact of climate itself on being a multiplier and extender of the volatility of asset values “How does it bring about unusual or unexpected asset values, either more frequently or at more extreme levels, and how are we changing that distribution and the variance of it?” he asked.

“Additionally, we’re not just trying to deal with the simplicity of a single asset, but how the overall movements are going to change and behave differently when climate is in the mix, not with climate as a separate climate VAR factor,” Carlin said.

While there are still many challenges and areas to think through regarding climate risk, the fact that work has started already is very valuable, even though there is a mismatch between the time horizons of long-term climate scenarios and shorter‑term risk models, Divecha noted. 

“The analysis we conduct today and the outcome will be useful in informing long-term underwriting and lending policies, and the shape those policies need to start taking today,” she said. 

Some firms are dealing with the mismatch in scenario timeframes by using shorter-term climate scenarios to analyse regulatory stresses and longer‑duration ones for long-term portfolio risk analysis. “Both of these can go hand-in-hand in helping inform credit strategies, in policies, compliance and aspects of governance as well,” Divecha said.

Sponsored content