Are climate change risks properly captured in the prudential framework?

After briefly presenting the DP, this article reviews the current Pillar 1 Capital (P1C) requirements. We limit ourselves to the P1C requirements for credit risk as this is by far the largest risk type for banks. Furthermore, we only discuss the interaction of the P1C with climate change risks (as opposed to broader environmental and/or social risk types). After establishing the extent to which the prudential framework takes climate change risks into account, possible amendments to the framework will be considered.

EBA’s Discussion Paper

In the introduction of the DP, the EBA mentions the increasing environmental risks – and their interaction with the traditional risk types – as the trigger for the review of the prudential framework. One of the main concerns is whether the current framework is sufficiently capturing the impact of transition risks and the more frequent and severe physical risks expected in the coming decades. In this context, they stress the special characteristics of environmental risks: compared to the traditional risk types, environmental risks tend to have a “multidimensional, non-linear, uncertain and forward-looking nature.”

The EBA also explains that the P1C requirements are not intended to cover all risks a financial institution is exposed to. The P1C represents a baseline capital requirement that is complemented by the Pillar 2 Capital requirement, which is more reflective of a financial institution’s specific business model and risks. Still, it is warranted to assess whether environmental risks are appropriately reflected in the P1C requirements, especially if these lead to systemic risks.

Even though the DP raises more questions than it provides answers, some starting points for the discussion are introduced. One is that the EBA takes a risk-based approach. Their standpoint is that changes to the prudential framework should reflect actual risk differentials compared to other risk types and that it should not be a tool to (unjustly) incentivize the transition to a sustainable economy. The latter lies “in the remit of political authorities.”

The DP also discusses some challenges related to environmental risks. One example is the lack of high-quality, granular historical data, which is needed to support the calibration of the prudential framework. The EBA also mentions the mismatch in the time horizon for the prudential framework (i.e., a business cycle) and the time horizon over which the environmental risks will unfold (i.e., several decades). They wonder whether “the business cycle concepts and assumptions that are used in estimating risk weights and capital requirements are sufficient to capture the emergence of these risks.”

Finally, the EBA does not favor supporting and/or penalizing factors, i.e., the introduction of adjustments to the existing risk weights based on a (green) taxonomy-based classification of the exposures¹. They are right to argue that there is no direct relationship between an exposure’s sustainability profile and its credit risk. In addition, there is a risk of double counting if environmental risk drivers have already been reflected in the current prudential framework. Consequently, the EBA concludes that targeted amendments to the framework may be more appropriate. An example would be to ensure that environmental risks are properly included in external credit ratings and the credit risk models of financial institutions. We explain this in more detail in the following paragraphs.

Pillar 1 Capital requirements

The assessment to what extent climate change risks are properly captured in the current prudential framework requires at least a high-level understanding of the framework. Figure 1 presents a schematic overview of the P1C requirements.

The P1C (at the top of Figure 1) depends on the total amount of Risk-Weighted Assets (RWAs; on the row below)². RWAs are determined separately for each (traditional) risk type. As mentioned, we only focus on credit risk in this article. The RWAs for credit risk are approximately 80% of the average bank’s total RWAs³. Financial institutions can choose between two methodologies for determining their credit risk RWAs: the Standardized Approach (SA)⁴ and the internal ratings-based (IRB) approach⁵ . In Europe, on average 40% of the total RWAs for credit risk are based on the SA, while the rest is based on the IRB approach⁶ :

 

Figure 1 – Schematic overview of the P1C requirements and the interaction with climate change risks

 

Interaction with climate change risks

The overview of the P1C requirements introduced in the previous section allows us to investigate the interaction between climate change risks and the P1C requirement. This is done separately for the SA and the IRB approach.

Standardized Approach

In the SA, there are two elements that allow for interaction between climate change risks and the resulting P1C. Climate change risks could be reflected in the P1C if the RW depends on an external credit rating, and this rating in turn properly accounts for climate change risks in the assessment of the counterparty’s creditworthiness (see 1 in Figure 1). The same holds if the RW depends on the LtV and in turn, the collateral valuation properly accounts for climate change risks (see 2 in Figure 1). This raises several concerns:

First, it can be questioned whether external credit ratings are properly capturing all climate change risks. In a report from the Network for Greening the Financial System (NGFS) [3], which was published at the same time as EBA’s DP, it is stated that credit rating agencies (CRAs) have so far not attempted to determine the credit impact of environmental risk factors (through back-testing for example). Also, the lack of high-quality historical data is mentioned as an explanation that statistical relationships between environmental risks and credit ratings have not been quantified. Further, a paper published by the ECB [4] concludes that, given the current level of disclosures, it is impossible for users of credit ratings to establish the magnitude of adjustments to the credit rating stemming from ESG-related risks. Nevertheless, they state that credit rating agencies “have made significant progress with their disclosures and methodologies around ESG in recent years.” The need for this is supported by academic research. An example is a study [5] from 2021 in which a correlation between credit default swap (CDS) spreads and ESG performance was demonstrated, and a study from 2020 [6] which demonstrated that high emitting companies have a shorter distance-to-default.

Secondly, the EBA has reported in the DP that less than 10% of the SA’s total RWAs is derived based on external credit ratings. This implies that a large share of the total RWAs is assigned a fixed RW. Obviously, in those cases there is no link between the P1C and the climate change risks involved in those exposures.

Finally, climate change risks only impact the P1C maintained for real estate exposures to the extent that these risks have been reflected in collateral valuations. Although climate change risks are priced in financial markets according to academic literature, many papers and institutions indicate that these risks are not (yet) fully reflected. In a survey held by Stroebel and Wurgler in 2021 [7], it is shown that a large majority of the respondents (consisting of finance academics, professionals and public sector regulators, among others) is of the opinion that climate change risks have insufficiently been priced in financial markets. A nice overview of this and related literature is presented in a publication from the Bank for International Settlements (BIS) [8]. The EBA DP itself lists some research papers in chapter 5.1 that indicate a relationship between a home’s sales price and its energy efficiency, or with the occurrence of physical risk events. It is unclear though if climate change risks are fully captured in the collateral valuations. For example, research is presented that information on flood risk is not priced into residential property prices. Recent research by ABN AMRO [9] also shows this.

Internal Ratings-Based Approach

In the IRB approach, financial institutions have more flexibility to include climate change risks in their internal models (see 3 in Figure 1). In the F-IRB approach this is limited to PD models, but in the A-IRB approach also LGD models can be adjusted.

A complicating factor is the forward-looking nature of climate change risks. In recent years, the competent authorities have pressured financial institutions to use historical data as much as possible in their model calibration and to back-test the performance of their models. As climate change risks will unfold over the next couple of decades, these are not (yet) reflected in historical data. To incorporate climate change risk, expert judgement would therefore be required. This has been discouraged over the past years (e.g., through the ECB’s Targeted Review of Internal Models (TRIM)) and it will probably trigger a discussion with the competent authorities. A possible deterioration of model performance (due to higher estimated risks compared to historically observations) is just one example that may attract attention.

Another complicating factor is that under the IRB approach, the PD of an obligor is estimated based on long-run average one-year default rates. While this may be an appropriate approach if there are no clear indications that the overall risk level will change, this does not hold if climate change risks increase in the future, and possibly increase systemic risks. By continuing to base a PD model on historical data only, especially for exposures with a time to maturity beyond a couple of years, the credit risk may be understated.

 

Are amendments to the prudential framework needed?

We have explained that there are several mechanisms in the prudential framework that allow environmental risks to be included in the P1C: the use of external credit ratings, the valuation of collateral, and the PD and LGD models used in the IRB approach. We have also seen, however, that it is questionable whether these mechanisms are fully effective. External credit ratings may not properly reflect all environmental risks and these risks may not be fully priced in on capital markets, leading to incorrect collateral values. Finally, a large share of the RWAs for credit risk depends on fixed RWs that are not (environmentally) risk-sensitive.

Consequently, it can be argued that amendments or enhancements to the prudential framework are needed. One must be careful, however, as the risk of double counting is just around the corner. Therefore, the following amendments or actions should be considered:

• Further research should be undertaken to investigate the relationship between climate change risk and the creditworthiness of counterparties. If there is more clarity on this relationship, it should also be assessed to what extent this relationship is sufficiently reflected in external ratings. Requiring more advanced disclosures from credit rating agencies could help to understand whether these risks are sufficiently captured in the prudential framework. One should be cautious to amend the ratings-based RWs in the SA, since credit rating agencies are continuously working on the inclusion of environmental risks into their credit assessments; there would be a real risk of double counting.
• The potential negative impact of climate change risks on collateral value should be further investigated. Financial institutions are already required by the ECB⁹ to consider environmental risks in their collateral valuations but this is not at a sufficient level yet. It will be important to consider the possibility of sudden value changes due to transition risks like shifting consumer sentiment or awareness.
• To improve the risk-sensitivity of the framework, a dependency on the carbon emissions of the counterparty could be introduced in the fixed RWs, possibly only for the most carbon-intensive sectors. It could be argued that there are other factors that have a more significant relationship with the default risk of a certain counterparty that could be included in the SA. Climate change risks, however, differ in the sense that they can lead to a systemic risk (as opposed to an idiosyncratic risk) that is currently not captured in the overall level of the RWs.
• In the SA, a distinction could be introduced based on the exposure’s time to maturity. For relatively short-term exposures, the current calibrations are probably fine. For longer-term exposures, however, the risks stemming from climate change may be underestimated as these are expected to increase over time.
• In the IRB approach, a reflection of climate change risk would require the regulator to allow for forward-looking expert judgment in the (re)calibration of PD and LGD models. Further guidance from the competent authorities on the potentially negative impact on model performance based on historical data would also be useful.

Conclusion

Based on the schematic overview of the P1C requirements and the (potential) interaction with climate change risks, we conclude that several mechanisms in the prudential framework allow for climate change risks to be incorporated into the P1C. At the same time, we conclude that this interaction is limited to specific parts of the portfolio, and that in those cases it remains to be seen to what extent this is properly accounted for. To remedy this, amendments to the prudential framework could be considered. It is important, however, to avoid double counting issues and to be mindful of time horizon differences.

It is expected that the EBA will publish a final report on the prudential treatment of environmental risks in the first half of this year. However, especially financial institutions that are using the IRB approach should not take a wait-and-see approach. Given the complexity of modeling climate change risks, it is prudent to start incorporating climate change risks into PD and LGD models sooner rather than later.

With Zanders’ extensive experience covering both credit risk modeling and climate change risk, we are well suited to support with this process. If you are looking for support, please reach out to us.

 

¹ Supporting factors are currently in place for SMEs and infrastructure projects, but the EBA advocated their removal.

² See RBC20.1 in the Basel Framework.

³ See for example the results from the EBA’s EU-wide transparency exercise. This is reflected in Figure 1 by the percentage in the grey link between P1C and RWAs for credit risk.

⁴ See CRE20 to CRE22 in the Basel Framework.

⁵ See CRE30 to CRE36 in the Basel Framework.

⁶ In the Netherlands, less than 20% of the total RWAs is based on the SA. See the EBA’s EU-wide transparency exercise for more information. The percentages in the grey link between ‘Risk-weighted assets’ and ‘Methodology’ in Figure 1 are based on the European average.

⁷ See the EBA’s Risk assessment of the European banking system [2]. The percentages in the grey link between ‘Standardized Approach’ and the ‘Exposure class’ in Figure 1 reflect the share of RWAs in the SA for each of the different exposure classes.

⁸ The LtV is defined as the ratio between the loan amount and the value of the property that serves as collateral.

⁹ See expectation 8.3 in the ECB’s Guide on climate-related and environmental risks.