Revised ECB Guide to internal models – continuous alignment within a changing regulatory environment

On June 22, the ECB published its revised Guide to internal models. The revision considers recent regulatory changes as well as best practices from the ECB in supervising internal models. The revision is under public consultation until September 15, inviting comments from banks and stakeholders involved in internal models.

Internal models enable banks to determine their risk-weighted assets with the permission of the ECB. These assets reflect the risks on a bank's balance sheet and serve as the basis for calculating required capital. As of the end of 2022, approximately 60% of risk-weighted exposure amounts for credit risk among ECB-supervised banks were calculated using internal models, for which credit risk corresponded to 85% of total risk-weighted assets, amounting to €8.6 trillion in total.

Initially developed as an integral part of the targeted review of internal models (TRIM), the Guide aims to address inconsistencies and reduce variability resulting from the use of complex internal models and hence focuses on streamlining banks internal models to be in line with the requirements set forth in Regulation (EU) No 575/2013, better known as the Capital Requirements Regulation (CRR).

In comparison to the previous version of the ECB Guide released in November 2019, the revised version maintains the same structure. It starts with discussing overarching topics related to internal models before moving into three chapters dedicated to credit risk, market risk and counterparty credit risk respectively.

The main changes lie in the incorporation of material climate-related and environmental risks into internal models (1), providing additional assistance for institutions willing to convert their models to simplified approaches (SA / (F) IRB) (2), a common definition of default for credit risk (3), guidance for measuring default risk in the trading book (4) and an interpretation of the rules governing banks’ use of internal models for counterparty credit risk (5). These will be discussed in more detail in the following sections.

The general topics chapter newly includes the insights provided in the ECB Guide on climate-related and environmental risks, mandating financial institutions to assess materiality of such risk drivers and if deemed relevant, include them in their internal models for calculating own fund requirements for credit and market risk.

Section 1 also introduces a new subsection regarding the expectations for implementing changed or extended models, which should occur no later than 3 months after the formal notification of approval received by ECB.

Another section within the general topics chapter addresses the potential reversion to less sophisticated approaches, such as the standardised approach or the foundation internal ratings-based approach.

Institutions should document the rationale for reverting to a less sophisticated approach. Objective criteria should be defined for choosing the appropriate approach for calculating own funds requirements across the portfolio. The institution's internal model’s strategy should consider operational capability, cost, and the strategic nature of activities. Additionally, conditions for reverting include demonstrating no adverse impact on solvency or risk management and obtaining prior permission of the competent authority.

Consistent application of criteria is necessary across exposure classes and types. Requests to revert to a different approach should provide convincing evidence that it is not intended to reduce own funds requirements.

In the application process for initial model approval, significant model changes or the implementation of IRB approach, institutions are required to provide evidence of having integrated the proposed model into a live, or, if justifiable, a non-live production environment alongside the existing system.

This entails demonstrating various capabilities, such as generating risk parameter estimates for relevant exposures, passing IT user acceptance tests, calculating own funds requirements based on IRB risk parameters, submitting the required COREP reports, utilizing the model for internal risk management, and preparing a reporting system that incorporates the model's risk parameters. Additionally for significant model changes, institutions ensure a successful IT implementation by the designated implementation date of the model change.

A new section is added on the definition of default, providing the ECB’s interpretation of article 178 CRR and utilizing the EBA Guidelines on DoD.  It addresses the following topics:

• Consistency in application:When applying the default definition at the obligor level, institutions must assess both the days past due and unlikeness to pay criteria for all exposures across the institution, parent company, and subsidiaries.
   
• Days past due criterion:All exposures exceeding the materiality threshold and remaining past due for 91 consecutive days should be classified as defaulted. A credit obligation is considered material if it exceeds €100 (retail exposures) or €500 (non-retail exposures) and represents over 1% of all on-balance sheet exposures to that obligor. Unpaid principal, interest, or fees at the due date are considered past due obligations, except for write-offs (which are classified as UTP). Past due amounts should be calculated with a frequency that ensures timely default identification.
   
• Unlikeliness to pay: The unlikeness to pay criterion distinguishes between the sale of credit obligations and distressed restructuring. Institutions should analyse reasons and losses for credit obligation sales, while the guidelines provide a formula for calculating diminished financial obligations in distressed restructuring cases. Institutions should define additional indications of unlikeliness to pay, referring to the provided list in the EBA guidelines, while encouraged to incorporate external information into the default identification process.

• Return to non-defaulted status: For a return to non-defaulted status, a distinction is made between exposures subject to distressed restructuring and other exposures, both in terms of the probation period as well as the past due amounts. This depends on the definition of default set at level of the obligor or the facility. 
   
• Consistency of external data: Institutions using external data for risk quantification must ensure consistency with the definition of default used.
   
• Adjustments to risk estimates due to changes in DoD: Where changes have been made to the definition of default, institutions must ensure a proper risk differentiation based on realized default rates and compare the new definition of default with the observations used in the RDS used for risk quantification, ensuring representativeness. Furthermore, institutions should incorporate margin of conservatism to account for uncertainties resulting from deficiencies in the reference data set (RDS) used for risk quantification.

After the 2008 financial crisis, most European financial institutions saw an increase in their non-performing loans stock. Impeding access to capital markets, many banks disposed a large part of their portfolio of nonperforming loans. The revised guide provides the interpretation of the ECB on massive disposals following article 500 of the updated CRR ((EU) 2019/876) which provided an adjustment for LGD estimation in the case of massive disposals as the difference between the average estimated LGDs for comparable defaulted exposures that have not been fully liquidated and the average realized LGDs.

Although the time frame of June 28^th  2022 set forth in article 500 has passed, additional guidance is provided to institutions regarding what is meant by the disposal date (transfer of legal ownership of assets), how the threshold of 20% of cumulative defaulted exposures is set and the usage of the incomplete workout treatment as the date prior to disposal for calculating the average estimated LGD.

The credit risk chapter includes further clarifications on the management and quantification of climate-related and environmental risks within internal ratings-based models, building upon the generic expectations outlined in the general topics chapter. This includes:

Overrides: article 172 CRR

In cases where climate-related and environmental risk drivers are considered significant and not incorporated into the rating system, institutions should assess the appropriateness of applying a more cautious approach by overriding the final rating assignment output for the related facilities or obligors.

Structure of PD models: article 179(1)(a) CRR and paragraphs 57, 121-123 EBA Guidelines on PD and LGD Estimation

Estimates must be based on drivers of the risk parameters, including climate-related and environmental risk drivers affecting the PD (paragraph 57 EBA GL) and LGD (paragraph 121 EBA GL), where deemed relevant and material.

LGD reference dataset: paragraph 109 EBA Guidelines on PD and LGD Estimation

The RDS should contain all relevant information in relation to losses and recovery processes. This should also include climate-related and environmental information where considered relevant and material.

EL_BE and LGD-in-default: paragraph 177 EBA Guidelines on PD and LGD Estimation

For the purpose of EL_BE and LGD in-default estimation, institutions should analyse the potential risk drivers, not only until the moment of default but also after the date of default and until the date of termination of the recovery process. This should include climate-related and environmental risk drivers where those risk drivers are assessed as relevant and material.

Margin of Conservatism (MoC): paragraph 37 EBA Guidelines on PD and LGD Estimation

The MoC should consider any deficiencies stemming from missing or inaccurate information including, where relevant and material, any missing or inaccurate climate-related information considered in risk estimates.

The updated market risk chapter provides clarification on the following subset of topics.

Lent-out or repo ‘ed out instruments under the internal model approach should be included in the calculation of own funds requirements, opposed to instruments borrowed through securities lending as this transaction type does not transfer the market risk of the security.

Counterparty credit spread risk, the risk of a price change due to a change in the credit spread of the counterparty to a transaction, should not be included in the internal model approach as it does not fall under general or specific risk definitions and is not part of the hypothetical or actual profit and loss for backtesting.

Based on paragraphs 7.1 and 7.2 of the EBA Guidelines on Incremental Default and Migration Risk Charge (IRC) this chapter specifies that funding risk embedded in own liabilities held in the trading book should be modelled under the internal model approach, as this approach aims to capture all material price risks.

Institutions must perform sensitivity and scenario analyses to assess the reasonableness of their internal models, particularly regarding PDs and RRs, in the incremental default and migration risk charge model for trading book positions subject to specific interest rate risk requirements. Accurate capturing of material price risks is necessary for market risk calculations. The IRC model must be conceptually sound, providing meaningful risk differentiation and accurate estimates of default and migration risk. Institutions should ensure the statistical methodology for deriving PDs is robust and consistent across rating grades.

This chapter update includes clarifications on the ECB's understanding of various aspects, including the term "most recent exchange of collateral" in the margin period of risk definition, early termination clauses, and market value corrections in relation to back-testing. It also incorporates the ECB's perspective on non-easy replacements, concentration of trades or collateral in margined netting sets, and their impact on the margin period of risk.

In line with Article 289(2) of the CRR, the ECB recommends that institutions adopt a good practice of initially applying model changes or extensions for internal risk management purposes. This approach allows institutions to gain sufficient experience with the changes or extensions before full implementation. The upfront use should commence no later than the application date in the live production environment for exposure calculations, ensuring effective risk management. Institutions should determine the most suitable method of upfront use, considering their specific circumstances and the nature of the change or extension. The ECB suggests two options: implementing the changes in the live production environment used for daily limit utilization calculations or using a non-live production environment for testing purposes.

Another new subsection explains how institutions should address "risks not in effective expected positive exposure," covering immaterial risks outside the effective expected exposure framework. This aligns with the framework on "risks-not-in-the-model engines" outlined in the market risk chapter of the guide.

Based on recent regulatory changes and best practices in supervising internal models, the ECB has published a revised Guide on internal models. The public consultation period for the revised Guide to internal models is underway and will conclude on September 15, 2023, after which the ECB will publish received comments combined with a feedback statement and the updated Guide.

The most material change is stemming from the inclusion of climate-related and environmental risks into the internal model landscape, based on the 2020 publication of the C&E Guide of the ECB which was accompanied last year with the observations from the 2022 thematic review on C&E risks.

Furthermore, additional interpretations were provided on the usage of simplified approaches to internal models, a complete section describing the common definition of default as well as minor amendments in the chapters on market and counterparty credit risk. 

Future updates based on changes in supervisory criteria or regulatory obligations can occur, especially in light of the forthcoming amendment of the CRR which will emphasize credit risk, credit valuation adjustment risk, operational risk, market risk and the output floor.  The ECB will conduct a comprehensive assessment of the revised Guide on internal models once the regulation is finalized and implemented. These revisions might occur without public consultation to allow timely updates in line with evolving best practices and regulatory frameworks.