Climate Risk Stress Testing: Agenda 2024

Climate risk stress testing (CRST) for European banks’ credit portfolios in 2024 will have a widespread impact — across policy, risk modelling, and risk and regulatory related systems and business processes. However, current CRST scenario approaches, centered on the Network for Greening the Financial System (NGFS) scenarios applied in conjunction with the European Central Bank (ECB) climate approach for wholesale credit models, remain in their infancy. With the year just beginning, it is therefore helpful to outline some ideas for banks’ 2024 CRST agendas and beyond that relate to the rapidly evolving but still high-level CRST regulatory requirements.

Scott D. Aguais

There are distinct similarities between today and the beginning of 2003 when we launched the Barclays Capital Basel II credit modelling and implementation effort. To support Basel II, existing wholesale credit risk methodology and credit data in the industry were robust enough to develop satisfactory wholesale credit models. However, banks’ wholesale credit model implementations at that time weren’t robust, as most of their models were not validated, implemented in production systems or formally approved. Today, after 20 years of working on Basel regulations, credit models and infrastructure are substantially more robust.

Banks are expected to leverage existing Basel wholesale credit models in conjunction with bank’s internal capital adequacy assessment process (ICAAP) and traditional stress testing systems and processes to support current CRST efforts. While this may call to mind the beginning of Basel II, CRST is quite different than both Basel II and traditional stress testing. First, CRST requires a novel, scenario-based approach to assess potential stress climate-credit losses over much longer time horizons. The new CRST scenarios will also require banks to integrate detailed geolocation data for their physical assets together with detailed Scope 1-3 emissions data to also assess long-run transition risks that may affect firm-specific financial stability.

Secondly, climate change has only been observed in more recent years. Current industry research analyzing statistical links between rising temperatures and measures of financial risks suggest they are not yet observable. Objective credit measures reflecting climate sensitivity are therefore generally unavailable, making it mostly unfeasible to develop new climate-sensitive, wholesale credit models to support CRST efforts. Therefore, current CRST efforts, while seeking in principle to leverage long-term Basel credit model and capital stress testing investments, will face a new set of complexities.

Based on our extensive experience in both building credit models and designing and managing end-to-end (E2E) model implementation, we have organised a few ideas that may help inform banks' 2024 efforts in this space. Our short list of ideas below focuses on key model, process, and implementation design challenges that banks will face.

1. Climate-Sensitive Credit Models

Developing climate-sensitive models is a core 2024 model deliverable which will require combining NGFS climate scenarios with the climate data attributes of location and emissions.

Much of what banks will be required to implement is based on the methodology the ECB proposed that has been outlined in two, recent ECB methodology papers (one paper written in 2021, the other paper in 2023). In developing this approach, the ECB has made substantial progress, but it is fair to say this approach is still in flux, like much of CRST. The ECB has refined its approach to better support high-level regulatory requirements by assessing different time horizons, cost pass-through assumptions and different underlying credit model calibrations. Comparisons of the suggested climate probability of default (PD) impacts for wholesale credit models between the two main ECB papers shows they were quite different, indicating this approach is still quite novel.

Because it is not possible to develop revised, empirically based credit models that directly reflect past climate impacts, implementation of the ECB approach requires alternative implementation logistics. Banks will need to adapt existing credit models so that PDs can be assessed in a range of climate risk scenarios. These alternative scenarios will require several new financial assumptions on cost pass-through and carbon elasticity, carbon mitigation in the form of carbon taxes and complex links between physical assets and location. All these changes will require adaptation of existing credit models, not redevelopment.

Most banks typically utilize ‘model development sandboxes’ to flexibly assess various model sensitivities, evaluate alternative parameters and data and support model validation. Banks’ model development sandboxes have multiple versions of the models while the production systems implement the approved ‘golden source model.’ CRST models, developed as adaptations of existing Basel Internal Ratings Based (IRB) models and run with various NGFS scenarios and alternative financial assumptions, are well suited to a ‘sandbox’ style of model architecture. However, banks now need to move toward a more adaptable production-style of ‘model sandbox’ with substantial flexibility to apply alternative assumptions and to handle many scenarios. Planning for these types of hurdles will be helpful for 2024 and beyond.

2. Designing Climate Scenarios as ‘Objects’

After 20 years on the Basel regulatory journey, credit models which used to be implemented in spreadsheets have currently moved to the heart of banks’ regulatory and risk management production systems and their E2E risk and regulatory policies and business procedures. Credit models in this context can be thought of as ‘objects.’ In design terms, an object is a ‘data field or concept that has unique characteristics and behaviour’ and these objects’ usually have a ‘golden version’ at the core of production model servers.

The application of well-defined, single definition model ‘objects’ across complex bank data and systems can support more efficient implementation and more consistent decision-making across various bank stakeholders.

For ICAAP and IFRS9 applications, most banks apply a small handful of regulatory or internal scenarios. Yet, CRST is evolving in a different direction with a higher number of climate scenarios expected to be required due to a growing scenario parameter dimensionality. Therefore, it may be helpful to think of CRST scenarios as ‘objects’ as well. This will facilitate developing the capability to efficiently run a large range of complex scenarios with different parameters and will be a key task to support various climate risk management objectives.

3. Alternative Industry Sector Climate (‘Brown/Green’) Segmentation

Currently, banks each develop their own internal, industry sector segments derived from standard NACE or SIC classifications for broad sectors. These segments usually number around 15-25 industry sectors for risk management and reporting purposes. Industry sectors are key to nearly all bank activity as front-office origination teams are usually organised around specialised industry expertise. Industry sectors are also a key part of the E2E risk management process, for managing credit exposures vs limits. Plus, they feed into managing loan origination, portfolio management and risk appetite setting.

In many cases, industry sector is also used to segment credit models and exposures, and industry sector is one of the main external financial reporting dimensions used by nearly all banks. The current industry segmentations used in banks are also embedded across nearly all front, middle and bank-office systems within a large bank.

Evolving CRST scenarios will require banks to differentiate each borrower’s sensitivity to carbon emissions (or carbon intensity) and future changes in carbon prices as well as other climate mitigation policies. The collection, processing and benchmarking of GHG emissions data by vendors and used by regulators in CRST scenario research is brand new in the last three-to-four years. CRST scenarios and models currently under development will most likely require the addition of a new risk segment dimension to the industry segmentation in current systems for managing individual borrowers, sectors and portfolios.

Given that current bank industry sectors have been well developed for a long time, adapting current industry sector segments to incorporate high (brown) and low-carbon (green) sectors will not be easy. Although it is not clear at this time how much banks will need to adapt their sector segmentation to better reflect ‘carbon differentiation,’ it is possible banks’ traditional industry sector segmentation will potentially need some degree of re-design. Therefore, beginning to think about how to redesign industry sector data, business processes and systems could also be on banks’ 2024 agenda.

4. Increased Use of Climate Scenarios (‘Narratives’) Instead of Models Will Not Be Easy

At most banks, the enhanced risk and capital management processes developed in recent years to support various ICAAP and IFRS9 business processes have moved into BAU. The general robustness of credit models after 20 years of Basel implementation also suggests that models are heavily relied upon and comprise the core of current bank E2E business processes for origination; risk and capital management; stress testing and IFRS9. As models become more robust and embedded, post-model adjustments tend to trend downward.

In contrast to current BAU risk and regulatory models and processes, newly developing CRST scenario approaches and processes are expected to be more qualitative and require a more iterative process because they have:

• Much less reliance on models as the data to link climate and credit directly does not really exist yet,
• A higher reliance on scenarios derived from detailed narratives, and,
• Climate stress testing and climate transition processes organised in conjunction with new ESG functions.

These key differences suggest that developing and applying climate scenarios will be a more complicated and qualitative ‘give and take’ process. If it is nearly impossible to build statistical climate/credit models for years, detailed narratives or structured, qualitative climate stories will remain at the core of CRST.

For example, in support of an increased role for narratives, the recent proposals from The Real World Climate Scenario (RWCS) effort in conjunction with the University of Exeter report “No Time to Lose” have strongly suggested that climate narratives will play a much greater role in managing climate change. And we agree with this.

In addition, banks will require multiple climate scenarios to assess complex uncertainty. For CRST, using multiple climate scenarios that are more narrative-based than model-based suggests credit risk decisioning in banks will become a more interactive, discussion-oriented process.

The implications are far reaching if a new cross-organisational function like ESG is needed to build and implement the multiple scenarios, and origination and risk decisions are filtered through the lens of Green Transition Financing.

It is probably a good idea for banks’ risk and finance change functions in 2024 to start thinking about climate business process complexity.

Parting Thoughts

Climate change has the potential to produce far reaching changes for financial institutions. including evolving CRST, which is still in it is early stages. In fact, we are already seeing major efforts in response to the rapidly moving regulatory requirements.

While the current focus is narrow — on scenarios and the adaptation of existing credit models to operationalise physical and transition risk credit adjustments — future implementation efforts are expected to be broader and create major data, systems and process changes.

For banks’ 2024 CRST agenda, responding to regulators urgent climate priorities will require modelling teams to get organised. Starting to plan for some of the more important strategic process and system implications to support future CRST is key to add to the 2024 CRST agenda.

Scott D. Aguais is managing director for the Z-Risk Engine solution and has led credit modelling teams for two global investment banks. He has been building and implementing wholesale credit models with his research partner Larry Forest for over 25 years and has published roughly 30 credit and climate risk modelling papers.