Skip to main content

Integrating climate risk into risk management frameworks: Know your modelling blind spots

As the credit cycle turns and credit risk managers prepare for the impact of significant macro-economic challenges, there is another inescapable, rather ominous turn in the tide: The climate is changing and as a result severe weather events are becoming more frequent. These physical events are directly impacting communities, infrastructure, property, and economies. As physical risks rise, consumer behaviours and societal expectations change, urging the transition to a net zero economy.

Climate events are starting to have a real impact on bank’s and insurer’s financial losses and earnings volatility. From the devastating floods witnessed last year in Germany and Pakistan, to a lack of snow in the European Alps causing ski resorts to close during this winter. The great transition to a net zero economy is already under way, with nearly all governments taking action to encourage a green transition. However, the transition does not come without risk. The energy crisis (triggered by the Russian conflict which is not climate related) provides a concentrated example of the impact of a disorderly transition[1]. A disorderly transition assumes a sharp and sudden increase in carbon taxes. This could lead to an increase in energy costs, much like what we are experiencing in the current energy crisis.

Significantly more work needs to be done to integrate climate within credit risk management. Many firms have built first generation climate risk models to perform stress testing and scenario analysis. From our experience these models are based on overly simplified assumptions, with several limitations or “blind spots”. Model users and senior executives need to fully understand these blind spots to successfully integrate climate risk into risk management and across the credit life cycle, for example:

  • Enable product development and support the fair treatment of businesses through the transition phase: The results from climate scenario analysis can be used to identify customers who are vulnerable to climate change under different scenarios, and who will need to adapt their business models. The answer is not to immediately withdraw access to financial liquidity to businesses in high emitting sectors, as this will result in the transition effort halting. This presents an opportunity for banks to develop products that incentivise and support their customers to successfully transition to greener business models. Beyond risk management, and as discussed in our latest blog within this series, firms are increasingly expected to consider their social impact as well. Robust models can help firms identify vulnerable segments and support a transition that is fair to all sectors and customers.
  • Support risk identification and pricing within underwriting: The results from climate scenario analysis should be used to support risk identification within the underwriting process, and subsequently risk-based pricing. This will ensure that the true risk of holding long term credit risk assets in a portfolio are understood and priced appropriately to provide the return on capital required by a bank’s shareholders.
  • Enhance monitoring capabilities and early warning triggers: Robust climate models can support firms in the monitoring of their exposure to climate risk and ensure that they remain within their climate risk appetite limits. The results from climate models can also be used as early warning indicators to identify future climate risks that will impact the portfolio.

There are several different blind spots across different portfolios and examples based on our experience and are discussed below.

Lack of counterparty data

Deloitte and the Institute of International Finance (IIF) recently interviewed 135 financial industry executives around the world as part of a Net Zero survey; only 3% of firms are confident they can assess information about the climate risks posed by individual customers, spanning across all types of portfolios. Much of this inability to assess the true risk is driven by the lack of customer data.

Lack of customer data is a modelling blind spot across all portfolios as it may lead to simplified assumptions, as discussed in the next section. However, it is a major challenge for Small to Medium Enterprise (SME) portfolios due to the considerable lack of reliable data compared to wholesale portfolios. Firms are able to perform name-by-name assessments for wholesale customers given the availability disclosure reports (e.g. TCFD and CDP). Such assessments are not plausible for SME portfolios. Analysis performed by British Business Bank suggest that SMEs account for about half of the UK business emissions, exacerbating the potential impact of transition risk on these portfolios.

Across all portfolios firms need to improve the breadth and depth of counterparty data. SME borrowers are less likely to have the ability to produce accurate estimates on their emissions. Therefore, firms need to build robust outreach programmes which are innovative and should not be onerous. Technology can be used to reduce the burden on counterparties by limiting the number and frequency of touch points. Such programmes can be designed to help counterparties understand their own carbon footprint and exposure to climate risk, while at the same time collecting robust data for the firms’ modelling and risk management.

Simplifying assumptions

Climate risk modellers are required to make simplifying assumptions due to the complexity of climate change impacts, especially for corporate entities. Accurately quantifying business interruption due to climate change for each corporate borrower is challenging due to the complex nature of supply chains, limited availability of data relating to asset locations and business functions. This creates blind spots within the model. For example, supply chains often span over different geographies and can be disrupted due to both physical events and required transition changes. For simplicity, modellers may only consider the country of domicile (i.e., where the company is registered) and could therefore be underestimating the climate risk impacts.

Similarly, accurate industry classification is also an example of a potential modelling blind spot. Industry classification at company level (such as used by the NACE or ISIC system) may not capture all the nuances and could mask the true climate risk exposure. For example, a mining company may be classed within the Metals and Mining sector, with substantial operations in both coal and gold mining. The coal mining part of the business will be impacted differently compared to the gold mining business in a high transition risk scenario. Ideally, these two businesses should be modelled separately but in the absence of more information, such as financial information of the two separate businesses, simplifying assumptions need to be made. These may include modelling the company as a coal mining business only, given that this sector is expected to be impacted more materially in a Net Zero scenario.

Firms who are using the outputs from models based on simplifying assumptions in risk management need to consider mitigants to overcome these blind spots. These may include using additional metrics or model overlays while the firm improves its ability to understand customers better and in turn, its modelling capabilities.

Collecting data from counterparties (either through disclosure reports or through an outreach programme) is one way of improving modelling decisions or assumptions. As the impacts of climate change play out, banks should build the capability to capture the root cause of defaults beyond traditional default triggers. This will help them assess if climate change was a contributing factor in the default event and thereby improve their understanding of the risks and operations of the counterparty. For example, a manufacturing company may be unable to service its debt because of supply chain issues linked to a natural disaster in another geography. Identifying the real reason for the default will help the firm to better understand this company’s supply chain, for example, and therefore accurately model the exposure to climate risks.

Uncertainty of the accuracy of third-party models

Most firms are using third party vendors data to quantify the potential impact of physical risks. However, substantial discrepancies in physical risk scores between different vendors, assigned to the same asset locations have been noted. This could be driven by data vendors using different methodologies. But the fact remains: we are trying to predict events that have never been observed. To overcome this, we have seen more sophisticated data providers use artificial intelligence (AI) and machine learning, in combination with historical data. However, the absence of sufficient historical data continues to make the validation exercise of the physical risk model results challenging.

The inconsistency in the physical risk assessments creates uncertainty in the accuracy of model results and subsequently the risk management. Therefore, firms should carefully assess the underlying data and methodology of physical risk scores to determine its reliability. Some vendors, such as ClimateX, provide a reliability indicator that allows the user to understand the level of accuracy of the risk assessment. Such indicators can help risk managers apply appropriate mitigants to address the accuracy uncertainty.

Understanding and managing climate modelling blind spots

In summary, firms need to be clear on the limitations of their climate models. They should not be lulled into a false sense of security from tick box exercises or be blinded by data produced by third parties using black box models. There is significantly more work to be done before firms can understand, measure, and then build the controls required to manage the climate risk in their credit portfolios. While firms are building their climate risk modelling capabilities and improving their counterparty data, users of these models need to be fully aware of these blind spots.

Read the previous articles in this series here:

[1] A disorderly transition can be defined as a scenario where there are dramatic policy changes to limit global warming

As Peter Bernstein said: The essence of risk management lies in maximizing the areas where we have some control over the outcome while minimizing the areas where we have absolutely no control over the outcome and the linkage between effect and cause is hidden from us.