Increased focus on climate-related risks in new consulting paper on ECB’s revised “Guide to internal models”
On June 22, ECB published a consulting paper on their revised “Guide to internal models”. The public consultation period ends on September 15 and the updated Guide is expected to be finalized towards the end of the year.
“The striking difference in the revised version is that banks now need to take into consideration how climate-related risks could affect their models. Apart from this, there are, for example, changes around reversion to less sophisticated approaches, testing of rating philosophies and how to treat internal models in mergers”, says Patrik Scheele, Managing Director at FCG and expert within credit risk.
Can you share your thoughts on the changes concerning climate-related risk drivers?
“The main change in the revised version is concerning climate-related risks. It is obvious that climate-related risks are taken seriously now from a financial risk perspective. It is specified that climate-related risk drivers should be included in internal models (credit-and market risk) if they are found to be relevant. ECB writes that the materiality assessments should be performed in line with their ”Guide on management and reporting of climate-related and environmental risk”. In that guide, ECB mandates that both quantitative and qualitative information should be used for the materiality assessment, and institutions are “expected to document this judgement with the available qualitative and quantitative information underpinning its assessment”, says Patrik Scheele.
What would be the implications for a bank?
“It is specifically added in the Probability of Default (PD) and Loss Given Default (LGD) chapters in ECB’s revised version that climate-related risk drivers should be considered. We have already seen this requirement in EBA’s “Guidelines on loan origination and monitoring”. Excluding climate-related risk drivers from your models cannot only be done by testing possible risk drivers statistically in historic data, you also have to theoretically explain the possible lack of climate-related risk drivers as you should take into account risks that might be missing in the historical data but are relevant for the future. ECB also stresses in the Margin of Conservatism (MoC) chapter that “the MoC should consider any deficiencies stemming from missing or inaccurate information including, where relevant and material, any missing or inaccurate climate-related information”, says Patrik Scheele.
“This implies that if you think that there might be climate-related risks in your portfolio you need to include conservatism in your estimates if unable to prove it statistically based on your historical data. Climate-related risks have also been included in the revised guide in relation to overrides, stating the expectation to make conservative overrides if there is a suspicion of climate related risks that are not captured by the model.”
Reversion to less sophisticated approaches
In the revised guide there is a new chapter around what is needed from a bank that wants to go back to a less sophisticated approach, for example going from Advanced IRB to Foundation IRB or going from IRB to the Standardised approach. This is of course connected to the increased possibilities for banks to do so in the new Capital Requirement Regulation (CRR3) which also was formally decided by EU last week. Banks are required to explain why they want to revert with their reasons as this will be contrary to their original roll-out plan for IRB.
An analysis needs to be made including costs of retaining IRB, availability of data and capital requirement impact. According to Patrik Scheele, it is not probable that ECB will approve a reversion if the bank has enough data for internal models and if there is a major gain for the bank from the changed capital requirements.
Testing of rating philosophies
There are a number of new paragraphs around how to test rating philosophies in the PD dimension. As a background, banks’ PD rating systems can be;
Through-the-Cycle (TTC), meaning that we do not expect to see any migration between risk classes due to changes in the macroeconomic environment. In these systems migrations should only take place due to changes for a particular counterparty in relation to other counterparties, so called idiosyncratic risk changes. In these systems default frequencies are expected to vary within each risk grade due to macroeconomic environment, i e increased default frequencies in a downturn and vice versa.
Point-in-Time (PIT), meaning that we do expect to see a lot of migration between risk classes due to changes in the macroeconomic environment. In these systems default frequencies are expected to be stable within each risk grade in different macroeconomic environments, since these are reflected in migrations instead. However, at portfolio level it will vary due to the migrations between risk classes.
“Most PD rating systems are however a hybrid between TTC and PIT with the majority leaning towards the latter. Banks have a choice of calibrating their models on calibration segment level, i e portfolio level, or at risk grade level. The new part in the guide is stressing the importance of testing both calibration options and especially to review REA effects using both methods. These effects should also be calculated historically and in different macroeconomic environments. It seems obvious that ECB does not want banks to choose philosophy only to reduce capital requirements”, says Patrik Scheele.
Internal models in the context of consolidations
There is a new chapter on how roll-out plans should be dealt with in the event of mergers or consolidations. Banks should submit a “return to compliance” plan for compliance issues related to the merger. This should include what the new model landscape will be directly after the merger and what the target model landscape looks like. The plan should include actions and timelines and how the bank plans to calculate risk-weighted assets (REA) until full compliance is achieved.
In the model use chapter, it is also mentioned that when two IRB banks merge it should have a consequent usage of the different IRB systems and not do cherry-picking to reduce capital requirements. When mergers are taking place combined default and loss history should be used as much as possible and appropriate adjustments (and MoCs) should be applied to cater for any differences in credit processes between the entities historically.
What other changes does the new Guide include?
“There is a whole new chapter on the Definition of Default (DoD), which is not part of the current guide. The chapter is following the EBA guideline on the definition of default but includes some specific detailed interpretations of the guideline.
There is a new section on treatment of massive disposals which explains in detail how banks may adjust its LGD estimates when they have had massive disposals of defaulted exposures in line with article 500 in CRR.
Implementation of models should be finalized within three months from the ECB approval. This will force most banks to start implementation in parallel with the ECB application process.
For all dimensions (PD, LGD and CCF), if the banks use statistical models for rank-ordering of exposures they should leave out observations for Out-of-Time and Out-of-Sample testing. It is only accepted to refrain from this if there is a scarcity of data available.
Banks should not use fixed time horizon for risk drivers in LGD estimation unless they can show proof of representativeness. This means that if a bank takes risk drivers’ values from, for example, exactly 6 months prior to default there can be representativity issues”, explains Patrik Scheele.
So, in summary, what are your most important observations so far?
According to Patrik Scheele, the inclusion of climate-related and environmental risks in the revised version verifies that these will be incorporated continuously into more financial regulation.
“With new requirements on climate-related risk drivers, institutions may need to reconsider their segmentation or how to structure the risk drivers as most climate-related and environmental risk drivers are only relevant for certain sectors or firms and of certain sizes. One thing to investigate could be combined risk drivers, combining climate-related or environmental risk drivers with e.g. corporate size or industry code. As an example of how risks vary, looking at commodities, (which are highly impacted by climate-related and environmental risks), price changes of wheat, energy or fuel have historical correlations with defaults, but only for certain industry codes and corporate sizes. On the other hand, many large firms may hedge their exposure towards price changes of wheat (drought risk), aluminum (environmental laws regulating mining), fuel (transition risk) or energy with complex derivate instruments. Some firms have vertically integrated supply chains while others have not set these firms up for completely different risk types which will show historically only with combined variables. As ECB now states that the Reference Data Set (RDS) should contain the data relevant for assessing climate-related and environmental risks, judgments should be made on what type of data to collect depending on the type of the corporation. Large amounts of data can be bought or collected but collecting the wrong data could become an expensive practice”, says Patrik Scheele.
“Making materiality assessments, identifying and structuring risk drivers in a correct manner, as well as documenting will require a tremendous amount of effort and knowledge development in many institutions, he continues. “As well as increased coordination between different parts of the bank with people who possess different types of knowledge. Model developers are used to certain type of risk drivers, but climate-related risk drivers are often different, although experience probably exists within other parts of the bank. Some can advise on how large animal-breeders and brewers buy wheat with futures contracts. Others can provide their insight on how upcoming EU laws and regulations will make certain business models unprofitable. Furthermore, model developers may need, as a last resort after everything else is tried and documented, to consider manual overlays, MoCs or overrides as solutions.”
“What the final version of the revised guide will look like remains to be seen after this consultation round. Climate-related risks will remain in the guide. There might be some small changes, but it will not likely disappear completely. What we have seen previously is also that local European FSAs follow in the footsteps of ECB in regard to their supervision practice. Hence, we can expect Nordic FSAs to require their banks to further include relevant climate-related risk drivers in their internal models for financial risks” Partik Scheele concludes.
