Response to discussion paper on the role of environmental risk in the prudential framework
Go back
A “social taxonomy” would help with some of these definition and data questions, at least inside the EU. But a taxonomy classification alone would not be risk-based by nature and would face the same issue as the climate/environmental taxonomy within a prudential/risk-based framework.
Notwithstanding these caveats, we would like to briefly highlight one of our risk-related practices which considers social elements. Should the EBA decide to consult stakeholders on these social factors, we will elaborate in more detail. In our Environmental and Social Risk (“ESR”) framework, prospective clients of our wholesale banking business (notably excluding natural persons) are screened, including on social elements such as human rights. The principle is simple: if a prospective client does not “pass” the screening and is not willing to make required changes to their practices, financing is not granted. Whilst this system differs from the typical prudential treatment, particularly because the framework is ex ante, and is not based on ongoing exposures, it is geared towards mitigating social risk, including reputational risk for the bank. ESR screening also allows us to ex ante exclude the financing of certain activities or sectors. The details behind this policy can be found here.
Example: Human rights serve as one of the guiding principles of the ESR framework. In 2018, we identified the most severe human rights issues (called “saliencies” as per the United Nations Guiding Principles Reporting Framework) for both our lending operations and own operation. The methodology to define the human rights saliencies and how we monitoring these are included in our Human Rights Report. Following this initial report, we have been publishing Human rights updates on a yearly basis - most recently in 2022.
Whilst we believe social factors have their rightful place in the “ESR” framework and that a human rights policy is key to addressing a significant component of social factors, we currently do not have a firm view as to if or how this could be translated into either Pillar One prudential requirements. Social factors could, however, be acknowledged as part of the Business Model assessment within the SREP process (in terms of qualitative assessment, stress test results, etc.). If this was to happen, the EBA SREP GLs would need to be further assessed to ensure this is aligned with the ESG risk elements of the revised version published in March 2022.
By nature, social risks are different to environmental risks, and we believe a key purpose of assessing social risks is to actively attempt to mitigate, reduce or remove them, notably by proactively engaging with (prospective) clients. The incoming legislation related to supply chain due diligence may provide an additional tool for this.
An interesting distinction between environmental and social risk is that social risks are unlikely to be an inherent feature of a particular sector or activity. That said, we recognise some sectors are more prone to facing certain issues. Climate and environmental risk, on the other hand, will (at least for the time being) almost by definition be elevated in energy intensive industries where viable replacements are not yet in the market. These are the so called “hard-to-abate” (e.g., air travel, cement, steel) industries.
With regards to liquidity, and as duly acknowledged in the DP, liquidity ratios are not based on risk-sensitive measures, but rather measure resilience of the liquidity position through standardised approaches. Further, we agree that environmental risks are expected to materialise over longer time horizons than those of the liquidity ratios. At some point in the future, and to the extent environmental risks materialise and have a significant impact on the creditworthiness of our obligors, conceivably our liquidity ratios could be impacted by virtue of lower net inflows. This potential impact, however, does not warrant any changes to the prudential liquidity requirements.
With regards to the leverage ratio, this was introduced as a non-risk sensitive backstop measure, and we see no direct interaction with environmental risks. In the medium to long term, should our Tier 1 capital measure or our leverage ratio exposure measure increase or decrease by virtue of changes in either our own funds instruments or our asset values driven by environmental risks, this will be reflected in our leverage ratio. Like liquidity ratios, this potential impact does not warrant any changes to the prudential leverage ratio requirements.
In our view, environmental risks do not necessarily trigger the need for an increase in the optimal level of capital in the banking system, either through higher RWA or increased combined buffer requirements. This applies to both the micro- and macro-prudential approach. We believe the current high capital levels in the EU banking system sufficiently factor in the possibility of unexpected losses in the short- to medium term, provided banks actively manage their exposure to environmental risk through re-allocation where necessary.
Environmental risks will lead to reallocation of risk towards counterparties that poorly anticipate future risks, including within sectors. For example, counterparties in the same sectors will be distinguished based on their reported environmental metrics, which will in the future provide transparency about their anticipated risk mitigants against physical and transition risk. There may also be a reallocation of risk between sectors, especially if some sectors do not gradually adjust to consider the consequences for their business model of the coming transition; or if some sectors suffer from irreparable damage from physical risks (it is difficult to imagine an example of the latter though).
We appreciate the EU’s clear choice for double materiality in its reporting and disclosure rules – including for non-financial corporates, which differ significantly to other jurisdictions, where disclosures are typically based on single materiality (e.g., per the draft SEC rule in the US).
In the prudential framework, however, (excluding Pillar Three reporting), it seems counterproductive to factor in non-financial materiality of banks’ counterparties (and invested assets), given the prudential framework serves to deal with financial materiality on the bank itself. We do recognise that an entity’s non-financial materiality (“on the world”) has the potential to become financially material, which then brings it under the single materiality category.
Our reasoning:
• The prudential framework is built around financial materiality for the bank, with the goal to keep a bank solvent when unexpected losses occur, thus providing a level of protection against losses for bank depositors and creditors, and therefore indirectly for the wider economy. A focus on financial materiality is the best means of protecting the specific and crucial role the prudential framework has.
• That said, the line between single and double materiality can be blurry, which is why it is helpful that both are disclosed. We believe that banks’ impact on the world, the “double” materiality, is to an extent indirectly covered in the concept of transition risk, as the negative material impact of exposures to counterparties with a negative material impact on the world will likely translate into transition risks and therefore potentially into financial risk.
• Banks’ exposure to double materiality of its counterparties is also reflected in:
o operational risk management (both in Pillar One and Two), including through litigation risk and general considerations related to reputational impact; and
o market risk, as asset valuations, which impact equity, bond and derivatives linked to those valuations, consider double materiality where available (incl. through investor preferences).
• Outside the Pillar One and Pillar Two prudential framework, double materiality is addressed in reporting requirements, e.g., EUT art. 8disclosures, and (future) CSRD disclosures. Double materiality is also an important driver of business decisions as it can influence consumer and investor behaviour.
Secondly, where counterparty-driven data is not available, more prescriptive guidance around determining and using proxies and estimates would be useful.
'
Lastly, the emergence of a centrally maintained registry with structured data on counterparties, industries and economic activities would benefit all market participants. This centrally maintained registry could be publicly maintained; privately maintained by one or more consultancies, or a combination of both. We hope the ESAP has potential to become such hub.
In the case of a specific asset class like residential mortgages, which will clearly not fall under reporting requirements like CSRD, we have identified important data gaps as well. Mortgages are an interesting case study as EU legislation has developed an EU-wide label for energy efficiency (the EPC), which should make it easy to gather good data on energy efficiency of buildings (which is a good indicator for environmental sustainability, albeit not the same).
We however note a few issues with this labelling system:
• Uptake is limited – In many countries, only a small minority of buildings have an EPC attached to it. We would encourage policy makers at EU and national level to strongly incentive uptake (e.g., through subsidisation of cost and/or making it mandatory in case of same or renovation).
• EPCs are not comparable across borders – we hope the review of the European Directive on Energy Performance of Buildings will address this.
• In some countries, banks do not have access to a central database of EPCs.
The EBA rightly points out the dangers of non-risk-based approaches such as supporting/penalising factors (due to lack of transition considerations), and the propensity to lead to bad incentives. Most importantly, green supporting factors could introduce incentives for greenwashing as well as asset bubbles; whilst brown supporting factors would do nothing to facilitate the transition or to reward credible transition plans of companies currently categorised as brown.
Stimulating capital flows to sustainable, taxonomy-aligned activities can better be achieved through measures outside the prudential framework. This includes taxation, public financial support, and other public policy measures.
Should policy makers and/or regulators opt for approaches that for classification purposes require the identification of “high risk” counterparties (e.g., to set brown penalising factors, or concentration limits), it is crucial that from a risk perspective such identification is sufficiently granular and robust. Notably, in the identification of “high-risk” counterparties, the future transition planning of these counterparties should be considered, to avoid lock-in of certain counterparties in a punitive category from which they cannot escape. If international level (Basel) risk based environmental measures are introduced, these should be considered in such approaches.
As noted in the DP, under the IRB approach, the PD of an obligor is estimated in a one-year time horizon based on long-run average one-year default rates, whilst risk differentiating factors may be defined in a way that reflects longer-term characteristics of the obligor (e.g., in the LGD estimates).
Given the challenges in measuring environmental risk, and the longer time horizons usually associated with environmental risks, in our view the baseline appropriate time horizon in the Pillar One own funds requirements is one year. This should capture short term unexpected losses with a high level of confidence. Further analysis is required to determine if and to what extent medium term risks (e.g., in a time horizon of between one and five years) can be assessed with a high confidence level. For those that can, certain aspects of environmental risk (see response to question 17) could be incorporated into the assessment of risk drivers in the IRB calculations. Our PD estimates, for example, to some extent capture environmental risks via the credit rating / internal credit assessment process. Notably, these longer time horizons employed in Pillar One would signify a fundamental shift in methodology.
For medium term risks that cannot be determined with a high confidence level, and for all long-term risks (e.g., beyond 5 years), an idiosyncratic approach is warranted, via assessment in the SREP and stress testing processes, and Pillar Two add-ons applied if and to the extent warranted.
Lastly, the time horizon should give due consideration to short maturity exposures and to mitigants such as insurance, guarantees etc., so that risks are not assessed beyond the date of effective maturity.
As also suggested in the DP, we support the view that the Pillar One framework already includes mechanisms that allow the inclusion of new types of risk drivers, such as
environmental risks (see our response to question 17, which makes a distinction between different aspects of environmental risk that we determine can appropriately be captured in risk drivers). The use of additional risk drivers in existing risk categories is more appropriate than the introduction of new risk categories or adjustment factors.
And whilst we support these views, our methodologies to appropriately reflect the forward-looking nature of environmental risks into our existing risk categories is immature, albeit evolving. We are also redeveloping our Credit Risk economic capital models which reflect the economic (true) risks of an exposure, and as such our internal models are increasingly providing another source of insight.
On a macro-scale, one possibility that could be considered is the development of an industry-wide standardised approach to stress testing / scenario analysis, of which the outputs could be used to quantify environmental risk inputs as drivers in the existing risk categories.
With respect to the forward-looking nature, it is important to consider to what extent such environmental risks are “expected", versus those that are “unexpected". For the former, we anticipate that over the coming reporting periods, we will mature our capabilities in identifying and integrating environmental risks into our IFRS 9 expected credit loss (ECL) assessments. Any increases in ECL due to environmental risk drivers will result in lower profit for the year, and in turn lower contribution to the capital measure of the CET1 ratio.
We have included a climate risk appetite dimension within our credit risk appetite framework, applying the concept of sector heatmaps and associated haircuts on the sector credit risk appetites. A project to assess physical climate risks of the mortgage book based on third party data is in progress, and first results will be included in our 2022 climate risk report.
We would like to stress that identifying proper risk differentiation methods is challenging. It could be that certain sectors are more at risk; on the other hand, the companies could be well equipped to undergo a transition and may well easily survive. In our view, it is more an expert-based analysis on company/sector level that should be applied by financial institutions. This analysis could be based in line with several conditions to measure whether an exposure has higher risk criteria than others. If those risks are higher, then this transaction should be constrained by internal limits.
Another challenge is that ECAI scope in Europe is limited. Currently ECAI ratings are available predominantly for large, usually publicly listed corporates and institutions, applying for sizeable financing, while the regulatory intention stimulates environmentally favourable financing decisions in case of SMEs as well.
On top of our own internal analyses and methodologies, we make use of agencies specialised in issuing sustainability scores as a tool to assess customers in certain wholesale banking sectors and use in supporting transition through sustainable lending offering.
Lastly, in general, there are some severe data issues that relate to energy efficiency in buildings, including with the EU system of Energy Performance Certificates, the use of which we would like to further encourage (see answer to Q5 for detail). This type of data is important for the development of tailored products such as energy efficient, or green, mortgages.
Environmental criteria are already considered in the definition of the ISF in (per CRR2 art. 501(a)(o)). It is essential that a risk-based approach remains in place, and double counting must be avoided. As a due diligence criterion, there should not be an adjustment factor on top of other factors, as even high-quality project and object finance exposures are not completely risk-free. Tagging assets based on environmental factors has merit, but their prudential treatment should continue to be risk based.
For cases where ECAI ratings are available, the treatment of exposures is more like the general approach in the class of exposures to corporates and there is less risk of inconsistency. Additionally, ESMA approved rating agencies should have an environmental risk assessment embedded in their rating methodologies.
Finally, it is worth mentioning that for many infrastructure projects, even those with a positive environmental influence (such as investments in public transport), are not financing under a supporting factor regime. Criteria are too stringent for many important and relevant environmentally positive projects (like construction of a subway). Whilst we do not believe that financing “green” is a reason for a risk-weight decrease, it might be worth considering which criteria hold back the use of these tools for these types of projects. We believe that this is another indication that solutions in the prudential framework must be well balanced and prepared holistically.
If environmental risks are to be included in the prudential risk framework, it should be done in a way which incentivises proactive management of environmental risks and facilitates risk-based balance-sheet steering, rather than one which unnecessarily increases the cost of capital for legitimate client servicing activities.
Environmental risk is just entering to the scope of analysis of supervisory authorities and regulators. The CRD6 proposal points out that supervisors will be expected to analyse the risk stemming from environmental factors in Pillar Two and take appropriate steps to reduce the risk that is not covered appropriately by Pillar One regulatory capital (Pillar Two capital add-on for environmental risks or prohibition of certain activities). However, this differentiation is not always possible since financial institutions may either not have proper information on the current and expected environmental risk profile of the borrower (what environmental factors may have influence on their repayment capacity) or on the environmental footprint in assets or services being financed using credit granted.
On the distinction between retail and non-retail, while it may be easier to collect and process information related to borrowers’ environmental risk profile among large corporates, it may prove difficult and costly in case of small enterprises and private individuals (sometimes even impossible due to privacy concerns (see e.g., access to EPC labels for mortgages). A detailed investigation of environmental risk footprint in assets or services financed using credit may also be challenging in cases where loan covenants typically do not require an indication of the purpose of the loan or credit limits. Where loans are collateralised, most notably by real estate, it is crucial to have reliable access to trustworthy labels (see discussion on EPC labels in Q5).
Under SA-External Ratings Based approaches, improved risk weight differentiation should avoid double counting of environmental risks when these are incorporated in ECAI ratings methodologies.
• Risk of customers being affected by measures to reduce climate change (e.g., increase in CO2 pricing, ban on fossil fuel cars, etc.).
• Risk of customers being affected by extreme weather events such as floods, forest fires, storms, etc.
• Risk of customers being affected by climate change through land becoming inhabitable/barren.
The first aspect can be captured through the regular modelling process by including risk drivers that capture this effect (see also answer to question 18). The second aspect is more difficult to capture as the events are (to date) rare and not covered in the historical data used in the development of models. Such data should be either collected or obtained from an external source. The third aspect is also difficult to capture as the effect is slow and it would be difficult to observe the effect over a limited period.
With respect to the first aspect there is no need to change the IRB framework. With respect to the second and third aspects, the strong focus on historical data in the IRB framework limits the ability to include these risks in the IRB models. However, we do not see a need to change the Pillar One framework to accommodate this limitation, as those risk could also be assessed through stress testing and scenario analyses under Pillar Two. These are useful instruments for capturing these risks as they affect countries, regions, and sectors.
We see environmental risk assessment as a hybrid between operational and credit risk. The operational risk function or a dedicated function could deliver probability estimates of environmental events, which could feed into the ICAAP assessments.
The full review of Pillar One model risk drivers takes place periodically (according to internal monitoring and validation standards). Once suitable data is obtained, it is verified, and an assessment is made as to whether additional environmental factors would increase model predictiveness and discriminatory power. If this is the case, model refinement may be needed.
So far within our organisation we observe no proof of the historical correlation between negative environmental events and default for each the segments in our portfolio. Going forward, a more comprehensive dataset could e.g., identify whether after a flood the customer’s PD has increased and whether these two events are indeed related.
Under specific internally defined policy conditions, environmental risks can already be considered via overrides of parameters. These overrides are monitored and considered in the regular periodic calibration processes. Since banks are required by the EBA Guidelines on PD estimation, LGD estimation and the treatment of defaulted assets (p. 204) to monitor override reasons, the potential environmentally driven rating overrides are a valuable source of information.
Environmental factors are already reflected in the valuation of collateral for certain portfolios and therefore are included indirectly in the outcome of Pillar One capital requirements calculations. Elements like flood risk and energy efficiency are increasingly and implicitly considered through the valuation of physical collateral. Outside of the IRB modelling framework, corporate governance assessment partially addresses social risk, whilst customer due diligence verifies amongst others human rights inequalities or doing business with counterparties in countries of elevated reputational risk.
Secondly, we participated in the first ECB climate risk stress test and noticed that environmental data collection for the use of incorporating environmental risk into Pillar One own funds calculation is still in its early stages. Financial institutions need time for developing the accurate methodology of capturing environmental factors into Pillar Two capital requirements.
From our perspective, one of the main reasons for the limited use of the ISF is the strict requirements on a corporate structure (SPV, Project Finance). These often exclude investments, that serve a public purpose and contribute to the achievement of environmental objectives from the ISF capital relief.
If an accurate estimate of the associated risks is the goal, then no overall adjustment factors should be used as the IRB models will already capture the risks to a varying degree. A ‘one size fits all’ adjustment will not lead to an accurate estimate and is therefore not appropriate. Existing IRB modelling (aspect 1 above in Q17) or stress testing (aspects 2 and 3 above in Q17) are better measures of achieving this purpose.
Environmental factors should only be integrated into Pillar One if and to the extent they lead to statistically relevant and feasible updates to parameter estimation for use in PD, EAD and/or LGD models.
Prudential risks stemming from environmental factors can be evaluated and steered across portfolios (geographies, client segments, sectors) based on existing credit risk metrics using our internal credit risk appetite framework and Pillar Two add-ons.
We believe that incorporating environmental risk into our Pillar One own funds calculation would be rather premature. At this moment, the focus should be on developing the correct methodologies of capturing environmental risk factors and including them in Pillar Two internal capital estimates. Such an approach should not disrupt existing credit processes and would still fulfil supervisory authorities’ expectations regarding environmental risk.
Double counting risk also exist between micro- (both Pillar 1 and Pillar 2) and macro-prudential requirements. This applies to both existing requirements and possible future requirements (e.g., if the sectoral Systemic Risk Buffer would ever be used to tackle climate-related risk. Close coordination between micro- and macro-prudential authorities is crucial.
a. Increase of the risk weight
The risk weights currently used in the FRTB sensitivity-based method (SbM) are defined in the regulation and are based on historic stress data. We remark that the calibration of these existing risk weights is already extremely conservative. The inclusion of forward-looking scenarios, on top of risks calibrated with historical data, would be a significant divergence from the existing approach. Furthermore, it would, given its macro design, insufficiently differentiate exposures which exhibit high levels of environmental risk from other market risk exposures. This would leave limited room for management steering of bank portfolios through limit setting.
If climate related risks are to be included in the prudential risk framework, it should be done in a way which incentivises proactive management of climate-related risks and so facilitates balance-sheet steering, rather than one which unnecessarily increases the cost of capital for legitimate client servicing activities.
b. Inclusion of an ESG component in the identification of the appropriate bucket
At present, equities are classified according to industry sector and economy (advanced or emerging). Similarly, bonds are categorised according to counterparty type and credit rating. Under this approach, an additional dimension would be included in the bucketing process to reflect environmental risks, which would lead to an increased risk weight for carbon-intense sectors. This would in effect introduce brown penalising and/or green supporting factors for securities in the trading book, an approach which EBA has rejected for credit risk.
This should not preclude the use of some form of sector-based assessment. Recent supervisory stress testing has focused on the sector specific risks associated with hypothetical shocks in carbon-intense sectors. A natural evolution of this initial materiality assessment would enable a better appreciation of the sensitivity of bank portfolios to climate-relevant factors such as increases in global carbon taxation. Pending finalisation of such analyses, we believe that Pillar Two stress testing and concentration risk management remains the natural home for this approach.
c. New risk factors
The establishment of new risk factors for products which explicitly reference climate and environmental risks, such as weather derivatives and Sustainability Improvement Derivatives, is in principle sensible and will enhance the associated risk management techniques. As these markets become more significant, market participants will benefit from the implied market-based pricing for these risks.
In more general terms, we consider that the development and universal application of environmental risk factors would be flawed, for the simple reason that environmental elements are not a risk factor but a risk driver: it is for this reason that we refer for example to climate-related market risks. We would not expect therefore to determine a specific ‘risk factor type’ in addition to delta, vega and curvature, but rather seek to understand the impact of physical and transition risk scenarios on the prospective evolution of these existing parameters.
d. Usage of the RRAO framework
EBA notes that the RRAO framework could in principle be used to determine capital requirements for environmental risks, so as not to tamper with the two main building blocks of the existing FRTB framework (SbM and JTD). We remark, consistent with EBA analysis, that the intended purpose of the RRAO is to address risks linked to complex products, as opposed to the climate related risks associated with conventional products.
The RRAO remains the correct place for the assessment of capital requirements for complex products whose payoff is highly sensitive to environmental risk considerations: for example, structured ESG products or complex weather derivatives. However, we believe that the inclusion of vanilla products within the scope of RRAO would be overly procrustean and would harm the integrity of the existing framework. Where vanilla products exhibit climate-related risks, these should be assessed in a manner more appropriate to the product.
The IMA expected shortfall measure relies on objective data in the form of financial market time series to include market stress events of the last 15 years, such as the financial crisis (2007-08), eurozone sovereign debt crisis (2010-13), the Covid-19 pandemic (2020-21), and the current Russia-Ukraine crisis. These crises collectively indicate the potential magnitude of a financial crisis, based on objective financial data. They do not however capture climate risks, which unlike traditional economic risks are solely forward-looking. This implies a high level of model risk in the computational process.
In the absence of historical data, Banks would need to consider multiple climate scenarios (e.g., IPCC RCPs) and translate these into sets of actionable stress parameters, making use of public sources such as the NGFS toolkit. The better place for such analyses is the Pillar Two framework. This allows for the development of good quality scenario models in conjunction with business model analysis, to determine future impact on risk appetite.
We believe that in general, the Pillar Two framework is best placed to assess environmental risks. This is for several reasons. First, the complexity associated with environmental risk quantification, associated with its relative immaturity, is not consistent with the high analytical standards demanded for Pillar One. Any proposal to incorporate this into Pillar One would increase model risk outside the accepted levels post-TRIM.
The same logic applies specifically to market risks. Whereas Pillar One is mostly focused on market risk in the trading book, which is mostly short term in nature, Pillar Two captures the longer-term climate related market risks, which are mostly in the Banking Book (for example, in the investment portfolio) and thus out of scope for Pillar One. Stress testing remains an acknowledged part of the Pillar Two of the Prudential framework and as such remains the most appropriate place for the assessment of forward-looking climate related risks, along with related Pillar Two measures such as the concentration risk management framework.
Finally, Pillar Two is a more flexible solution, which allows for a graduated enhancement of maturity over time without the need for repeated and extensive updates to legislative documentation.
Therefore, we consider the existing event types sufficient, as they cover types of events that climate and environmental risks may cause. To identify losses from environmental risks, an additional flag to the current event types could be introduced as proposed in the DP. For this, a standardized definition of environmental risks is required.
Therefore, at this stage, an environmental or broader ESG flag, as suggested in the DP, will be sufficient to enable more structured risk event reporting and analysis. We propose using existing industry bodies involved with Operational Risk, such as AFME and the ORX consortium, to develop an industry-standard/guidance for applying such a flag. This way, a widely accepted classification can be derived and used for internal risk management.
The current classification by event types focuses on the effect of operation risk (e.g., physical damage). One way to incorporate more transparency could be to add a cause dimension (e.g., weather effects). A challenge concerning climate risk will be the differentiation between regular weather events and climate risk, e.g., rainfall as a regular natural event and rainfall as a climate risk event.
Regarding physical risks, we see another challenge that needs to be carefully thought through. The DP points out that access to climate data and forecast possibilities is limited. We would encourage regulators to collaborate with other public authorities, such as environmental agencies, to ensure access to the required (standardised) information.
For these reasons, the Pillar One approach is accompanied by sound operational risk management in Pillar II. Any specific treatment of environmental risks should instead become part of the broader Pillar Two framework.
In this context, we note that it is challenging to predict the severity and frequency of physical environmental events in the medium-to long-term, resulting in predictions with a high level of uncertainty. An annual evaluation of scenarios is more efficient and reliable because structural changes in severity and physical environmental events do not happen overnight.
There may be merit in standardised transparency towards competent authorities regarding exposures to individual and/or connected counterparties with high environmental risk. This could help foster meaningful supervisory dialogue regarding concentration risk issues.
A clear challenge when developing such an approach would be to identify precisely which exposures are considered high-risk, for which the current EU Taxonomy Framework does not provide a definitive answer. In the short term, focus should in any case be on climate change related metrics only, not wider environmental metrics more broadly.
This classification issue should be considered carefully and would likely need to be developed separately from the taxonomy. If this approach were to be pursued, we believe it is important to distinguish entities that are:
• High-risk counterparties (e.g., polluting industries), but that have credible transition plans in place through which they demonstrate how they will transform their activities by 2050. This should include measurable intermediary milestones, possibly with 3rd party control mechanisms.
• High-risk counterparties that do not have such plans in place. Such exposures clearly pose a more severe concentration risk in the medium and longer terms.
From a Pillar Three reporting perspective, some elements of the current framework may already cover some aspects needed for this regime. See also AFME’s response to Q30 for more detail.
In the short term, we believe it is better for such a limit to not automatically trigger RWA or buffer consequences. Instead, limits could be used to trigger additional disclosure requirements, and/or to inform supervisory dialogue, which would allow supervisors to take concentration risks - if left unexplained and unaddressed - and incorporate them into the Pillar Two requirements. One of the reasons for this more discretionary approach is that there may be temporary challenges in hard-to-abate sectors that warrant a recalibration of sector decarbonisation pathways. Examples of this include temporary supply chain issues, geopolitical events, or other idiosyncratic events (the pandemic being a recent example).
In terms of scope, we currently do not see a credible means of measuring an exact level of climate and/or environmental risk across all exposures, as we lack historical data on both physical and transition risk. A more realistic approach might be to focus on high-risk sectors, which would require – as stressed in Q30 – a clear distinction between counterparties with or without credible transition plans. As the EBA rightly highlights in the DP, a concentration limit should not disincentivise counterparties currently in polluting sectors from transitioning to low-carbon activities.
The DP raises several other important considerations on the imposition of a concentration limit, the most important of which is that of a possible indiscriminate shift of environmental risk to non-bank financial intermediaries. This must be avoided, as it would increase concentration risk to the overall financial system. This last point illustrates that large exposure or concentration limits may be better addressed in a more general way, for example, the context of macro-prudential considerations affecting both bank and non-bank financial intermediaries.
Q1: In your view, how could exposures associated with social objectives and/or subject to social impacts, which are outside the scope of this DP, be considered in the prudential framework? Please provide available evidence and methodologies which could inform further assessment in that regard.
As highlighted in the Discussion paper (DP), less progress has been made on determining a common European definition of social factors in finance. This is partly due to the understandable prioritisation of climate and environmental factors but is also due to the inherent difficulty of defining and measuring social factors across jurisdictions. A notable exemption is human rights, for which there is strong international guidance available.A “social taxonomy” would help with some of these definition and data questions, at least inside the EU. But a taxonomy classification alone would not be risk-based by nature and would face the same issue as the climate/environmental taxonomy within a prudential/risk-based framework.
Notwithstanding these caveats, we would like to briefly highlight one of our risk-related practices which considers social elements. Should the EBA decide to consult stakeholders on these social factors, we will elaborate in more detail. In our Environmental and Social Risk (“ESR”) framework, prospective clients of our wholesale banking business (notably excluding natural persons) are screened, including on social elements such as human rights. The principle is simple: if a prospective client does not “pass” the screening and is not willing to make required changes to their practices, financing is not granted. Whilst this system differs from the typical prudential treatment, particularly because the framework is ex ante, and is not based on ongoing exposures, it is geared towards mitigating social risk, including reputational risk for the bank. ESR screening also allows us to ex ante exclude the financing of certain activities or sectors. The details behind this policy can be found here.
Example: Human rights serve as one of the guiding principles of the ESR framework. In 2018, we identified the most severe human rights issues (called “saliencies” as per the United Nations Guiding Principles Reporting Framework) for both our lending operations and own operation. The methodology to define the human rights saliencies and how we monitoring these are included in our Human Rights Report. Following this initial report, we have been publishing Human rights updates on a yearly basis - most recently in 2022.
Whilst we believe social factors have their rightful place in the “ESR” framework and that a human rights policy is key to addressing a significant component of social factors, we currently do not have a firm view as to if or how this could be translated into either Pillar One prudential requirements. Social factors could, however, be acknowledged as part of the Business Model assessment within the SREP process (in terms of qualitative assessment, stress test results, etc.). If this was to happen, the EBA SREP GLs would need to be further assessed to ensure this is aligned with the ESG risk elements of the revised version published in March 2022.
By nature, social risks are different to environmental risks, and we believe a key purpose of assessing social risks is to actively attempt to mitigate, reduce or remove them, notably by proactively engaging with (prospective) clients. The incoming legislation related to supply chain due diligence may provide an additional tool for this.
An interesting distinction between environmental and social risk is that social risks are unlikely to be an inherent feature of a particular sector or activity. That said, we recognise some sectors are more prone to facing certain issues. Climate and environmental risk, on the other hand, will (at least for the time being) almost by definition be elevated in energy intensive industries where viable replacements are not yet in the market. These are the so called “hard-to-abate” (e.g., air travel, cement, steel) industries.
Q2: Do you agree with the EBA’s assessment that liquidity and leverage ratios will not be significantly affected by environmental risks? If not, how should these parts of the framework be included in the analysis?
Yes, we agree.With regards to liquidity, and as duly acknowledged in the DP, liquidity ratios are not based on risk-sensitive measures, but rather measure resilience of the liquidity position through standardised approaches. Further, we agree that environmental risks are expected to materialise over longer time horizons than those of the liquidity ratios. At some point in the future, and to the extent environmental risks materialise and have a significant impact on the creditworthiness of our obligors, conceivably our liquidity ratios could be impacted by virtue of lower net inflows. This potential impact, however, does not warrant any changes to the prudential liquidity requirements.
With regards to the leverage ratio, this was introduced as a non-risk sensitive backstop measure, and we see no direct interaction with environmental risks. In the medium to long term, should our Tier 1 capital measure or our leverage ratio exposure measure increase or decrease by virtue of changes in either our own funds instruments or our asset values driven by environmental risks, this will be reflected in our leverage ratio. Like liquidity ratios, this potential impact does not warrant any changes to the prudential leverage ratio requirements.
Q3: In your view, are environmental risks likely to be predominantly about reallocation of risk between sectors, or does it imply an increase in overall risk to the system as a whole? What are the implications for optimum levels of bank capital?
For the purpose of answering this question, we assume that environmental risks are material for at least part of banks’ exposures.In our view, environmental risks do not necessarily trigger the need for an increase in the optimal level of capital in the banking system, either through higher RWA or increased combined buffer requirements. This applies to both the micro- and macro-prudential approach. We believe the current high capital levels in the EU banking system sufficiently factor in the possibility of unexpected losses in the short- to medium term, provided banks actively manage their exposure to environmental risk through re-allocation where necessary.
Environmental risks will lead to reallocation of risk towards counterparties that poorly anticipate future risks, including within sectors. For example, counterparties in the same sectors will be distinguished based on their reported environmental metrics, which will in the future provide transparency about their anticipated risk mitigants against physical and transition risk. There may also be a reallocation of risk between sectors, especially if some sectors do not gradually adjust to consider the consequences for their business model of the coming transition; or if some sectors suffer from irreparable damage from physical risks (it is difficult to imagine an example of the latter though).
Q4: Should the ‘double materiality’ concept be incorporated within the prudential framework? If so, how could it be addressed?
While we follow the EBA's reasoning in para 23 of the DP, we generally disagree that double materiality should be incorporated into the prudential framework. Our different opinion may in part be due to semantics.We appreciate the EU’s clear choice for double materiality in its reporting and disclosure rules – including for non-financial corporates, which differ significantly to other jurisdictions, where disclosures are typically based on single materiality (e.g., per the draft SEC rule in the US).
In the prudential framework, however, (excluding Pillar Three reporting), it seems counterproductive to factor in non-financial materiality of banks’ counterparties (and invested assets), given the prudential framework serves to deal with financial materiality on the bank itself. We do recognise that an entity’s non-financial materiality (“on the world”) has the potential to become financially material, which then brings it under the single materiality category.
Our reasoning:
• The prudential framework is built around financial materiality for the bank, with the goal to keep a bank solvent when unexpected losses occur, thus providing a level of protection against losses for bank depositors and creditors, and therefore indirectly for the wider economy. A focus on financial materiality is the best means of protecting the specific and crucial role the prudential framework has.
• That said, the line between single and double materiality can be blurry, which is why it is helpful that both are disclosed. We believe that banks’ impact on the world, the “double” materiality, is to an extent indirectly covered in the concept of transition risk, as the negative material impact of exposures to counterparties with a negative material impact on the world will likely translate into transition risks and therefore potentially into financial risk.
• Banks’ exposure to double materiality of its counterparties is also reflected in:
o operational risk management (both in Pillar One and Two), including through litigation risk and general considerations related to reputational impact; and
o market risk, as asset valuations, which impact equity, bond and derivatives linked to those valuations, consider double materiality where available (incl. through investor preferences).
• Outside the Pillar One and Pillar Two prudential framework, double materiality is addressed in reporting requirements, e.g., EUT art. 8disclosures, and (future) CSRD disclosures. Double materiality is also an important driver of business decisions as it can influence consumer and investor behaviour.
Q5: How can availability of meaningful and comparable data be improved? What specific actions are you planning or would you suggest to achieve this improvement?
Firstly, availability of meaningful and comparable data can be improved by European-level coordination of application dates, so that counterparties (e.g., Corporates subject to CSRD) make relevant, high-quality, and standardised disclosures ahead of when they are required for disclosure or other purposes by financial market participants. Furthermore, inconsistencies in data definitions need to be identified and addressed, such as those that arise between the non-risk-based (e.g., EUT Art. 8) disclosures and risk-based (e.g., Pillar 3 ESG risks) disclosure requirements. Additionally, any data-related differences between the in-progress EFRAG standards (on a European level), and the in-progress ISSB Exposure Drafts (on a global level) need to be acknowledged and factored into future requirements.Secondly, where counterparty-driven data is not available, more prescriptive guidance around determining and using proxies and estimates would be useful.
'
Lastly, the emergence of a centrally maintained registry with structured data on counterparties, industries and economic activities would benefit all market participants. This centrally maintained registry could be publicly maintained; privately maintained by one or more consultancies, or a combination of both. We hope the ESAP has potential to become such hub.
In the case of a specific asset class like residential mortgages, which will clearly not fall under reporting requirements like CSRD, we have identified important data gaps as well. Mortgages are an interesting case study as EU legislation has developed an EU-wide label for energy efficiency (the EPC), which should make it easy to gather good data on energy efficiency of buildings (which is a good indicator for environmental sustainability, albeit not the same).
We however note a few issues with this labelling system:
• Uptake is limited – In many countries, only a small minority of buildings have an EPC attached to it. We would encourage policy makers at EU and national level to strongly incentive uptake (e.g., through subsidisation of cost and/or making it mandatory in case of same or renovation).
• EPCs are not comparable across borders – we hope the review of the European Directive on Energy Performance of Buildings will address this.
• In some countries, banks do not have access to a central database of EPCs.
Q6: Do you agree with the risk-based approach adopted by the EBA for assessing the prudential treatment of exposures associated with environmental objectives / subject to environmental impacts? Please provide a rationale for your view.
We agree with the EBA’s general direction to opt for a risk-based approach, in contrast with approaches that consider other policy objectives. At the core, the prudential framework serves to ensure banking resilience through risk management and appropriate capital requirements based on risk. This is the primary, and only, objective of prudential regulation.The EBA rightly points out the dangers of non-risk-based approaches such as supporting/penalising factors (due to lack of transition considerations), and the propensity to lead to bad incentives. Most importantly, green supporting factors could introduce incentives for greenwashing as well as asset bubbles; whilst brown supporting factors would do nothing to facilitate the transition or to reward credible transition plans of companies currently categorised as brown.
Stimulating capital flows to sustainable, taxonomy-aligned activities can better be achieved through measures outside the prudential framework. This includes taxation, public financial support, and other public policy measures.
Should policy makers and/or regulators opt for approaches that for classification purposes require the identification of “high risk” counterparties (e.g., to set brown penalising factors, or concentration limits), it is crucial that from a risk perspective such identification is sufficiently granular and robust. Notably, in the identification of “high-risk” counterparties, the future transition planning of these counterparties should be considered, to avoid lock-in of certain counterparties in a punitive category from which they cannot escape. If international level (Basel) risk based environmental measures are introduced, these should be considered in such approaches.
Q7: What is your view on the appropriate time horizon (s) to be reflected in the Pillar 1 own funds requirements?
Noting that the capital requirements are designed to protect institutions (and indirectly the financial system as a whole) against unexpected losses with a high confidence level (>99%), the appropriate time horizon is that which enables this confidence level to be achieved.As noted in the DP, under the IRB approach, the PD of an obligor is estimated in a one-year time horizon based on long-run average one-year default rates, whilst risk differentiating factors may be defined in a way that reflects longer-term characteristics of the obligor (e.g., in the LGD estimates).
Given the challenges in measuring environmental risk, and the longer time horizons usually associated with environmental risks, in our view the baseline appropriate time horizon in the Pillar One own funds requirements is one year. This should capture short term unexpected losses with a high level of confidence. Further analysis is required to determine if and to what extent medium term risks (e.g., in a time horizon of between one and five years) can be assessed with a high confidence level. For those that can, certain aspects of environmental risk (see response to question 17) could be incorporated into the assessment of risk drivers in the IRB calculations. Our PD estimates, for example, to some extent capture environmental risks via the credit rating / internal credit assessment process. Notably, these longer time horizons employed in Pillar One would signify a fundamental shift in methodology.
For medium term risks that cannot be determined with a high confidence level, and for all long-term risks (e.g., beyond 5 years), an idiosyncratic approach is warranted, via assessment in the SREP and stress testing processes, and Pillar Two add-ons applied if and to the extent warranted.
Lastly, the time horizon should give due consideration to short maturity exposures and to mitigants such as insurance, guarantees etc., so that risks are not assessed beyond the date of effective maturity.
Q8: Do you have concrete suggestions on how the forwardlooking nature of environmental risks could be reflected across the risk categories in the Pillar 1 framework?
We appreciate both the importance and the challenge of this question. In the DP's concluding remarks, the point made that “the prudential framework has historically been built on backward-looking risk assessments, which does not align well with the forward-looking nature of environmental risks” is significant.As also suggested in the DP, we support the view that the Pillar One framework already includes mechanisms that allow the inclusion of new types of risk drivers, such as
environmental risks (see our response to question 17, which makes a distinction between different aspects of environmental risk that we determine can appropriately be captured in risk drivers). The use of additional risk drivers in existing risk categories is more appropriate than the introduction of new risk categories or adjustment factors.
And whilst we support these views, our methodologies to appropriately reflect the forward-looking nature of environmental risks into our existing risk categories is immature, albeit evolving. We are also redeveloping our Credit Risk economic capital models which reflect the economic (true) risks of an exposure, and as such our internal models are increasingly providing another source of insight.
On a macro-scale, one possibility that could be considered is the development of an industry-wide standardised approach to stress testing / scenario analysis, of which the outputs could be used to quantify environmental risk inputs as drivers in the existing risk categories.
With respect to the forward-looking nature, it is important to consider to what extent such environmental risks are “expected", versus those that are “unexpected". For the former, we anticipate that over the coming reporting periods, we will mature our capabilities in identifying and integrating environmental risks into our IFRS 9 expected credit loss (ECL) assessments. Any increases in ECL due to environmental risk drivers will result in lower profit for the year, and in turn lower contribution to the capital measure of the CET1 ratio.
Q9: Have you performed any further studies or are you already using any specific ESG dimensions to differentiate within credit risk? If so, would you be willing to share your results?
We have performed studies and implemented elements in our sustainable lending offering covering certain business areas and sectors. Furthermore, we are conducting studies in other areas like retail banking, but these are not yet conclusive.We have included a climate risk appetite dimension within our credit risk appetite framework, applying the concept of sector heatmaps and associated haircuts on the sector credit risk appetites. A project to assess physical climate risks of the mortgage book based on third party data is in progress, and first results will be included in our 2022 climate risk report.
We would like to stress that identifying proper risk differentiation methods is challenging. It could be that certain sectors are more at risk; on the other hand, the companies could be well equipped to undergo a transition and may well easily survive. In our view, it is more an expert-based analysis on company/sector level that should be applied by financial institutions. This analysis could be based in line with several conditions to measure whether an exposure has higher risk criteria than others. If those risks are higher, then this transaction should be constrained by internal limits.
Q10: What are the main challenges that credit rating agencies face in incorporating environmental considerations into credit risk assessments? Do you make use of external ratings when performing an assessment of environmental risks?
From the perspective of a ECAI ratings user, credit rating agencies must ensure that banks are able to understand the structure and nature of their scores or ratings, and their key drivers (as per ECB TRIM Guide). Banks should also regularly verify that the results of their credit bureau score continues to be an appropriate input variable in their credit rating process, for example by reviewing any changes in the credit bureau score methodology. To a certain degree there is standardisation as per ESMA regulations. However, the use of drivers and their weights to arrive at sub-scores, including those considering ESG risks, is lacking sufficient transparency.Another challenge is that ECAI scope in Europe is limited. Currently ECAI ratings are available predominantly for large, usually publicly listed corporates and institutions, applying for sizeable financing, while the regulatory intention stimulates environmentally favourable financing decisions in case of SMEs as well.
On top of our own internal analyses and methodologies, we make use of agencies specialised in issuing sustainability scores as a tool to assess customers in certain wholesale banking sectors and use in supporting transition through sustainable lending offering.
Q11: Do you see any challenge in broadening due diligence requirements to explicitly integrate environmental risks?
The key challenge in onboarding environmental risk assessments at a customer level is the limited availability of consistently high-quality data. We observe that customer awareness and preparedness to provide or publish such data differs per geography and sector. In areas where the need to address environmental risks and making a sustainable transition is recognised, clients are adapting faster than in other areas.Q12: Do you see any specific aspects of the CRM framework that may warrant a revision to further account for environmental risks?
The current CRM framework principles do not need major modifications to further account for environmental factors. We deem the current framework sufficient to also capture environmental risks. Collateral re-valuation requirements and data collection will implicitly enhance the quality of the valuation process going forward, also taking account of environmental risks.Q13: Does the CRR3 proposal’s clarification on energy efficiency improvements bring enough risk sensitiveness to the framework for exposures secured by immovable properties? Should further granularity of risk weights be introduced, considering energy-efficient mortgages? Please substantiate your view.
Based on the current valuation standards considering average levels of transaction prices for comparable properties in a region, we have doubts on whether it will be possible to distinguish in a transparent manner the increase of the value of the property resulting from energy efficiency improvements and from many possible other factors. We are concerned that market practice may arbitrage these different factors.Lastly, in general, there are some severe data issues that relate to energy efficiency in buildings, including with the EU system of Energy Performance Certificates, the use of which we would like to further encourage (see answer to Q5 for detail). This type of data is important for the development of tailored products such as energy efficient, or green, mortgages.
Q14: Do you consider that high-quality project finance and high-quality object finance exposures introduced in the CRR3 proposal should potentially consider environmental criteria? If so, please provide the rationale for this and potential implementation issues.
Currently, CRR2 does not (nor does the CRR3 proposal) include a precise definition of high-quality project and object finance. EBA is to prepare a draft of regulatory technical standards specifying further details of the definition of high-quality project and object finance. Environmental considerations can be considered here, however, a sound calibration of expectations is necessary to avoid customers resigning from this kind of lending and moving towards general purpose lending. This would hamper the general idea behind the finalisation of Basel3 / CRR3 to increase the risk sensitivity of capital requirements calculated using the standardised approach. In case of projects without ECAI ratings indicating credit quality, moving towards general purpose lending may be more beneficial due to the possibility to use the infrastructure supporting factor (ISF) (art. 501a CRR).Environmental criteria are already considered in the definition of the ISF in (per CRR2 art. 501(a)(o)). It is essential that a risk-based approach remains in place, and double counting must be avoided. As a due diligence criterion, there should not be an adjustment factor on top of other factors, as even high-quality project and object finance exposures are not completely risk-free. Tagging assets based on environmental factors has merit, but their prudential treatment should continue to be risk based.
For cases where ECAI ratings are available, the treatment of exposures is more like the general approach in the class of exposures to corporates and there is less risk of inconsistency. Additionally, ESMA approved rating agencies should have an environmental risk assessment embedded in their rating methodologies.
Finally, it is worth mentioning that for many infrastructure projects, even those with a positive environmental influence (such as investments in public transport), are not financing under a supporting factor regime. Criteria are too stringent for many important and relevant environmentally positive projects (like construction of a subway). Whilst we do not believe that financing “green” is a reason for a risk-weight decrease, it might be worth considering which criteria hold back the use of these tools for these types of projects. We believe that this is another indication that solutions in the prudential framework must be well balanced and prepared holistically.
Q15: Do you consider that further risk differentiation in the corporate, retail and/or other exposure classes would be justified? Which criteria could be used for that purpose? In particular, would you support risk differentiation based on forward-looking analytical tools?
In general, we support differentiation based on forward-looking analytical tools, but to a limited extent in Pillar One. Methodologies for forward-looking analysis towards environmental risks are not mature enough, and data availability to support risk differentiation for prudential treatment is limited.If environmental risks are to be included in the prudential risk framework, it should be done in a way which incentivises proactive management of environmental risks and facilitates risk-based balance-sheet steering, rather than one which unnecessarily increases the cost of capital for legitimate client servicing activities.
Environmental risk is just entering to the scope of analysis of supervisory authorities and regulators. The CRD6 proposal points out that supervisors will be expected to analyse the risk stemming from environmental factors in Pillar Two and take appropriate steps to reduce the risk that is not covered appropriately by Pillar One regulatory capital (Pillar Two capital add-on for environmental risks or prohibition of certain activities). However, this differentiation is not always possible since financial institutions may either not have proper information on the current and expected environmental risk profile of the borrower (what environmental factors may have influence on their repayment capacity) or on the environmental footprint in assets or services being financed using credit granted.
On the distinction between retail and non-retail, while it may be easier to collect and process information related to borrowers’ environmental risk profile among large corporates, it may prove difficult and costly in case of small enterprises and private individuals (sometimes even impossible due to privacy concerns (see e.g., access to EPC labels for mortgages). A detailed investigation of environmental risk footprint in assets or services financed using credit may also be challenging in cases where loan covenants typically do not require an indication of the purpose of the loan or credit limits. Where loans are collateralised, most notably by real estate, it is crucial to have reliable access to trustworthy labels (see discussion on EPC labels in Q5).
Q16: Do you have any other proposals on integrating environmental risks within the SA framework?
Risk differentiation based on forward-looking analytical tools in the SA does not seem sufficiently prudent since this approach is already conservative and is based on risk weighting schemes provided by supervisors using historical results of quantitative impact studies. We would support this integration to the extent quantitative studies demonstrate links between environmental and credit (or market / operational) risks.Under SA-External Ratings Based approaches, improved risk weight differentiation should avoid double counting of environmental risks when these are incorporated in ECAI ratings methodologies.
Q17: What are your views on the need for revisions to the IRB framework or additional guidance to better capture environmental risks? Which part of the IRB framework is, in your view, the most appropriate to reflect environmental risk drivers?
In general, environmental risk drivers should only be incorporated into the IRB models to the extent they statistically improve model performance and predictive power. This also depends on the aspects of environmental risk that should be captured, such as:• Risk of customers being affected by measures to reduce climate change (e.g., increase in CO2 pricing, ban on fossil fuel cars, etc.).
• Risk of customers being affected by extreme weather events such as floods, forest fires, storms, etc.
• Risk of customers being affected by climate change through land becoming inhabitable/barren.
The first aspect can be captured through the regular modelling process by including risk drivers that capture this effect (see also answer to question 18). The second aspect is more difficult to capture as the events are (to date) rare and not covered in the historical data used in the development of models. Such data should be either collected or obtained from an external source. The third aspect is also difficult to capture as the effect is slow and it would be difficult to observe the effect over a limited period.
With respect to the first aspect there is no need to change the IRB framework. With respect to the second and third aspects, the strong focus on historical data in the IRB framework limits the ability to include these risks in the IRB models. However, we do not see a need to change the Pillar One framework to accommodate this limitation, as those risk could also be assessed through stress testing and scenario analyses under Pillar Two. These are useful instruments for capturing these risks as they affect countries, regions, and sectors.
We see environmental risk assessment as a hybrid between operational and credit risk. The operational risk function or a dedicated function could deliver probability estimates of environmental events, which could feed into the ICAAP assessments.
The full review of Pillar One model risk drivers takes place periodically (according to internal monitoring and validation standards). Once suitable data is obtained, it is verified, and an assessment is made as to whether additional environmental factors would increase model predictiveness and discriminatory power. If this is the case, model refinement may be needed.
So far within our organisation we observe no proof of the historical correlation between negative environmental events and default for each the segments in our portfolio. Going forward, a more comprehensive dataset could e.g., identify whether after a flood the customer’s PD has increased and whether these two events are indeed related.
Under specific internally defined policy conditions, environmental risks can already be considered via overrides of parameters. These overrides are monitored and considered in the regular periodic calibration processes. Since banks are required by the EBA Guidelines on PD estimation, LGD estimation and the treatment of defaulted assets (p. 204) to monitor override reasons, the potential environmentally driven rating overrides are a valuable source of information.
Q18: Have you incorporated the environmental risks or broader ESG risk factors in your IRB models? If so, can you share your insight on the risk drivers and modelling techniques that you are using?
The internal Environmental and Social Risk (ESR) score – an indicator of broader ESG risk of a counterparty - is used as a candidate risk driver in our IRB model development process. The regular risk driver selection process is followed, and risk drivers may be included in the final model if it provides sufficient risk discriminatory power.Environmental factors are already reflected in the valuation of collateral for certain portfolios and therefore are included indirectly in the outcome of Pillar One capital requirements calculations. Elements like flood risk and energy efficiency are increasingly and implicitly considered through the valuation of physical collateral. Outside of the IRB modelling framework, corporate governance assessment partially addresses social risk, whilst customer due diligence verifies amongst others human rights inequalities or doing business with counterparties in countries of elevated reputational risk.
Q19: Do you have any other proposals on integrating environmental risks within the IRB framework?
Firstly, downturn (DT) LGD quantification in line with the EBA GLs on DT estimation currently focuses solely on historically observed DT periods. Increased LGDs due to e.g., a severe weather event (flood, fire, storm, etc.) could be biased consequently. This should, however, not necessarily be seen as a shortcoming as the risk could be assessed through the stress testing framework (using for example a pre-defined scenario).Secondly, we participated in the first ECB climate risk stress test and noticed that environmental data collection for the use of incorporating environmental risk into Pillar One own funds calculation is still in its early stages. Financial institutions need time for developing the accurate methodology of capturing environmental factors into Pillar Two capital requirements.
Q20: What are your views on potential strengthening of the environmental criterion for the infrastructure supporting factor? How could this criterion be strengthened?
Infrastructure financing that stimulates the transition, amongst others in renewable energy, should receive strong support of the financial sector and public-private partnerships. It should be noted, however, that the ISF is not per se a discriminating indicator for risk. If the objective is to have accurate Pillar One estimates that adequately discriminate risk (option 1 in Q21), an overall adjustment factor is less desirable.From our perspective, one of the main reasons for the limited use of the ISF is the strict requirements on a corporate structure (SPV, Project Finance). These often exclude investments, that serve a public purpose and contribute to the achievement of environmental objectives from the ISF capital relief.
Q21: What would in your view be the most appropriate from a prudential perspective: aiming at integrating environmental risks into existing Pillar 1 instruments, or a dedicated adjustment factor for one, several or across exposure classes? Please elaborate.
We are generally not in favour of the introduction of green supporting and/or brown penalising factors, as they are not risk based. The prudential framework should not serve as a tool to replace other, more appropriate policy tools.If an accurate estimate of the associated risks is the goal, then no overall adjustment factors should be used as the IRB models will already capture the risks to a varying degree. A ‘one size fits all’ adjustment will not lead to an accurate estimate and is therefore not appropriate. Existing IRB modelling (aspect 1 above in Q17) or stress testing (aspects 2 and 3 above in Q17) are better measures of achieving this purpose.
Environmental factors should only be integrated into Pillar One if and to the extent they lead to statistically relevant and feasible updates to parameter estimation for use in PD, EAD and/or LGD models.
Prudential risks stemming from environmental factors can be evaluated and steered across portfolios (geographies, client segments, sectors) based on existing credit risk metrics using our internal credit risk appetite framework and Pillar Two add-ons.
We believe that incorporating environmental risk into our Pillar One own funds calculation would be rather premature. At this moment, the focus should be on developing the correct methodologies of capturing environmental risk factors and including them in Pillar Two internal capital estimates. Such an approach should not disrupt existing credit processes and would still fulfil supervisory authorities’ expectations regarding environmental risk.
Q22: If you support the introduction of adjustment factors to tackle environmental risks, in your view how can double counting be avoided and how can it be ensured that those adjustment factors remain risk-based over time?
For the purpose of having accurate Pillar One estimates that adequately discriminate risk (option 2 in Q21), an overall adjustment factor is less desirable as effects from double counting will be inevitable.Double counting risk also exist between micro- (both Pillar 1 and Pillar 2) and macro-prudential requirements. This applies to both existing requirements and possible future requirements (e.g., if the sectoral Systemic Risk Buffer would ever be used to tackle climate-related risk. Close coordination between micro- and macro-prudential authorities is crucial.
Q23: What are your views on possible approaches to incorporating environmental risks into the FRTB Standardised Approach? In particular, what are your views with respect to the various options presented: increase of the risk-weight, inclusion of an ESG component in the identification of the appropriate bucket, a new risk factor, and usage of the RRAO framework?
In summary, we welcome the acknowledgement that there is a growing market in products which explicitly reference climate and environmental risks and welcome EBA initiatives to integrate these into the FRTB framework. Where an environmental factor acts as a risk driver rather than a risk factor, the required analysis is best accommodated through the Pillar Two framework. The different presented options are discussed below.a. Increase of the risk weight
The risk weights currently used in the FRTB sensitivity-based method (SbM) are defined in the regulation and are based on historic stress data. We remark that the calibration of these existing risk weights is already extremely conservative. The inclusion of forward-looking scenarios, on top of risks calibrated with historical data, would be a significant divergence from the existing approach. Furthermore, it would, given its macro design, insufficiently differentiate exposures which exhibit high levels of environmental risk from other market risk exposures. This would leave limited room for management steering of bank portfolios through limit setting.
If climate related risks are to be included in the prudential risk framework, it should be done in a way which incentivises proactive management of climate-related risks and so facilitates balance-sheet steering, rather than one which unnecessarily increases the cost of capital for legitimate client servicing activities.
b. Inclusion of an ESG component in the identification of the appropriate bucket
At present, equities are classified according to industry sector and economy (advanced or emerging). Similarly, bonds are categorised according to counterparty type and credit rating. Under this approach, an additional dimension would be included in the bucketing process to reflect environmental risks, which would lead to an increased risk weight for carbon-intense sectors. This would in effect introduce brown penalising and/or green supporting factors for securities in the trading book, an approach which EBA has rejected for credit risk.
This should not preclude the use of some form of sector-based assessment. Recent supervisory stress testing has focused on the sector specific risks associated with hypothetical shocks in carbon-intense sectors. A natural evolution of this initial materiality assessment would enable a better appreciation of the sensitivity of bank portfolios to climate-relevant factors such as increases in global carbon taxation. Pending finalisation of such analyses, we believe that Pillar Two stress testing and concentration risk management remains the natural home for this approach.
c. New risk factors
The establishment of new risk factors for products which explicitly reference climate and environmental risks, such as weather derivatives and Sustainability Improvement Derivatives, is in principle sensible and will enhance the associated risk management techniques. As these markets become more significant, market participants will benefit from the implied market-based pricing for these risks.
In more general terms, we consider that the development and universal application of environmental risk factors would be flawed, for the simple reason that environmental elements are not a risk factor but a risk driver: it is for this reason that we refer for example to climate-related market risks. We would not expect therefore to determine a specific ‘risk factor type’ in addition to delta, vega and curvature, but rather seek to understand the impact of physical and transition risk scenarios on the prospective evolution of these existing parameters.
d. Usage of the RRAO framework
EBA notes that the RRAO framework could in principle be used to determine capital requirements for environmental risks, so as not to tamper with the two main building blocks of the existing FRTB framework (SbM and JTD). We remark, consistent with EBA analysis, that the intended purpose of the RRAO is to address risks linked to complex products, as opposed to the climate related risks associated with conventional products.
The RRAO remains the correct place for the assessment of capital requirements for complex products whose payoff is highly sensitive to environmental risk considerations: for example, structured ESG products or complex weather derivatives. However, we believe that the inclusion of vanilla products within the scope of RRAO would be overly procrustean and would harm the integrity of the existing framework. Where vanilla products exhibit climate-related risks, these should be assessed in a manner more appropriate to the product.
Q24: For the Internal Model Approach, do you think that environmental risks could be better captured outside of the model or within it? What would be the challenges of modelling environmental risks directly in the model as compared to modelling it outside of the internal model? Please describe modelling techniques that you think could be used to model ESG risk either within or outside of the model.
We believe that environmental risks would be better captured outside the market risk IMA models, which are based on historical data. For this reason, Pillar Two is better suited.The IMA expected shortfall measure relies on objective data in the form of financial market time series to include market stress events of the last 15 years, such as the financial crisis (2007-08), eurozone sovereign debt crisis (2010-13), the Covid-19 pandemic (2020-21), and the current Russia-Ukraine crisis. These crises collectively indicate the potential magnitude of a financial crisis, based on objective financial data. They do not however capture climate risks, which unlike traditional economic risks are solely forward-looking. This implies a high level of model risk in the computational process.
In the absence of historical data, Banks would need to consider multiple climate scenarios (e.g., IPCC RCPs) and translate these into sets of actionable stress parameters, making use of public sources such as the NGFS toolkit. The better place for such analyses is the Pillar Two framework. This allows for the development of good quality scenario models in conjunction with business model analysis, to determine future impact on risk appetite.
Q25: Do you have any other proposals on integrating environmental risks within the market risk framework?
Significant changes to Pillar One should be globally co-ordinated through the Basel Committee. Uncoordinated policy initiatives leading to divergent Pillar One regimes will shift environmentally damaging activities to non-EU jurisdictions and would not address underlying concerns. This will not mitigate the systemic risks associated with potentially disorderly future market transitions.We believe that in general, the Pillar Two framework is best placed to assess environmental risks. This is for several reasons. First, the complexity associated with environmental risk quantification, associated with its relative immaturity, is not consistent with the high analytical standards demanded for Pillar One. Any proposal to incorporate this into Pillar One would increase model risk outside the accepted levels post-TRIM.
The same logic applies specifically to market risks. Whereas Pillar One is mostly focused on market risk in the trading book, which is mostly short term in nature, Pillar Two captures the longer-term climate related market risks, which are mostly in the Banking Book (for example, in the investment portfolio) and thus out of scope for Pillar One. Stress testing remains an acknowledged part of the Pillar Two of the Prudential framework and as such remains the most appropriate place for the assessment of forward-looking climate related risks, along with related Pillar Two measures such as the concentration risk management framework.
Finally, Pillar Two is a more flexible solution, which allows for a graduated enhancement of maturity over time without the need for repeated and extensive updates to legislative documentation.
Q26: What additional information would need to be collected in order to understand how environmental risks impact banks’ operational risk? What are the practical challenges to identifying environmental risk losses on top of the existing loss event type classification?
First, it would be appropriate to clarify that environmental, but also social and governance phenomena, do not represent new forms of risk but risk drivers that could increase the probability of occurrence and/or the severity of existing risk categories (e.g., Credit, Market, Operational and Reputational).Therefore, we consider the existing event types sufficient, as they cover types of events that climate and environmental risks may cause. To identify losses from environmental risks, an additional flag to the current event types could be introduced as proposed in the DP. For this, a standardized definition of environmental risks is required.
Therefore, at this stage, an environmental or broader ESG flag, as suggested in the DP, will be sufficient to enable more structured risk event reporting and analysis. We propose using existing industry bodies involved with Operational Risk, such as AFME and the ORX consortium, to develop an industry-standard/guidance for applying such a flag. This way, a widely accepted classification can be derived and used for internal risk management.
The current classification by event types focuses on the effect of operation risk (e.g., physical damage). One way to incorporate more transparency could be to add a cause dimension (e.g., weather effects). A challenge concerning climate risk will be the differentiation between regular weather events and climate risk, e.g., rainfall as a regular natural event and rainfall as a climate risk event.
Regarding physical risks, we see another challenge that needs to be carefully thought through. The DP points out that access to climate data and forecast possibilities is limited. We would encourage regulators to collaborate with other public authorities, such as environmental agencies, to ensure access to the required (standardised) information.
Q27: What is your view on potential integration of a forward-looking perspective into the operational risk framework to account for the increasing severity and frequency of physical environmental events? What are the theoretical and practical challenges of introducing such a perspective in the Standardised Approach?
As correctly summarised in the DP, SA-OR is not designed to give this forward-looking perspective. This applies to all operational risks, not just the environmentally driven ones. The BCBS decided on a simple approach after the sub-optimal experience with more complex approaches under the previous framework. The regulatory framework in the Basel III accord relies solely on a size-based metric (Business Indicator), especially if the Internal Loss Multiplier (ILM) will be set equal to 1 in the final CRR3 text. Such an approach does not seem capable of capturing the increasing severity and frequency of physical environmental events. Furthermore, should the ILM not be neutralised at 1, it will not be possible to extrapolate the portion of the ILM connected to environmental issues as the ILM is based on aggregate historical losses. Therefore, it implies that such an approach is not forward-looking and is not very risk-sensitive (estimates the historic average rather than the proper risk profile).For these reasons, the Pillar One approach is accompanied by sound operational risk management in Pillar II. Any specific treatment of environmental risks should instead become part of the broader Pillar Two framework.
In this context, we note that it is challenging to predict the severity and frequency of physical environmental events in the medium-to long-term, resulting in predictions with a high level of uncertainty. An annual evaluation of scenarios is more efficient and reliable because structural changes in severity and physical environmental events do not happen overnight.
Q28: Do you agree that the impact of environmental risk factors on strategic and reputational risk should remain under the scope of the Pillar 2 framework?
We agree the environmental risk factor on strategic and reputational risks should remain under Pillar Two. This allows applying banks’ specific situations to the assessment rather than relying on the more standardized approach under Pillar One to provide appropriate outcomes. At the same time, supervisory/regulatory guidance is needed to prevent the development of an unlevel playing field.Q29: Do you have any other proposals on integrating environmental risks within the operational risk framework?
Any regulatory framework must treat environmental risk as a risk driver rather than as an isolated new risk type. It is difficult to recommend an industry-wide approach to integrate climate and environmental risks into the operational risk framework and, more specifically, the risk taxonomy, as the choices made would invariably depend on individual banks’ internal risk organisation and enterprise risk-wide structures. Nevertheless, we agree with adding environmental risks to the Pillar Two toolbox.Q30: What, in your view, are the best ways to address concentration risks stemming from environmental risk drivers?
We generally agree with the EBA’s assessment that the CRR large exposure limits (LEX) regime is not designed to limit exposures to certain sectors or specific risk drivers. The current backstop nature of the LEX regime should be preserved, and its use should remain limited. This is especially so as sectoral or geographical concentration risks can already be captured in Pillar Two.There may be merit in standardised transparency towards competent authorities regarding exposures to individual and/or connected counterparties with high environmental risk. This could help foster meaningful supervisory dialogue regarding concentration risk issues.
A clear challenge when developing such an approach would be to identify precisely which exposures are considered high-risk, for which the current EU Taxonomy Framework does not provide a definitive answer. In the short term, focus should in any case be on climate change related metrics only, not wider environmental metrics more broadly.
This classification issue should be considered carefully and would likely need to be developed separately from the taxonomy. If this approach were to be pursued, we believe it is important to distinguish entities that are:
• High-risk counterparties (e.g., polluting industries), but that have credible transition plans in place through which they demonstrate how they will transform their activities by 2050. This should include measurable intermediary milestones, possibly with 3rd party control mechanisms.
• High-risk counterparties that do not have such plans in place. Such exposures clearly pose a more severe concentration risk in the medium and longer terms.
From a Pillar Three reporting perspective, some elements of the current framework may already cover some aspects needed for this regime. See also AFME’s response to Q30 for more detail.
Q31: What is your view on the potential new concentration limit? Do you identify other considerations related to such a limit? How should such a limit be designed to avoid the risk of disincentivising the transition?
We believe there is merit in exploring the idea of a dynamic concentration limit linked to the gradual decarbonisation of the economy. This is based on the idea that for banks to align their balance sheet with the Net Zero target by 2050, it is paramount that their “carbon exposures” gradually decline over time. The calibration of such a limit would require thorough investigation about appropriate levels, which should be informed by risk-based data.In the short term, we believe it is better for such a limit to not automatically trigger RWA or buffer consequences. Instead, limits could be used to trigger additional disclosure requirements, and/or to inform supervisory dialogue, which would allow supervisors to take concentration risks - if left unexplained and unaddressed - and incorporate them into the Pillar Two requirements. One of the reasons for this more discretionary approach is that there may be temporary challenges in hard-to-abate sectors that warrant a recalibration of sector decarbonisation pathways. Examples of this include temporary supply chain issues, geopolitical events, or other idiosyncratic events (the pandemic being a recent example).
In terms of scope, we currently do not see a credible means of measuring an exact level of climate and/or environmental risk across all exposures, as we lack historical data on both physical and transition risk. A more realistic approach might be to focus on high-risk sectors, which would require – as stressed in Q30 – a clear distinction between counterparties with or without credible transition plans. As the EBA rightly highlights in the DP, a concentration limit should not disincentivise counterparties currently in polluting sectors from transitioning to low-carbon activities.
The DP raises several other important considerations on the imposition of a concentration limit, the most important of which is that of a possible indiscriminate shift of environmental risk to non-bank financial intermediaries. This must be avoided, as it would increase concentration risk to the overall financial system. This last point illustrates that large exposure or concentration limits may be better addressed in a more general way, for example, the context of macro-prudential considerations affecting both bank and non-bank financial intermediaries.