1.   The principles of the Internal MODEL Benchmarking exercise

1.1.   An EBA initiative in the context of the CRDIV

CRD IV requires competent authorities to conduct an annual assessment of the quality of internal models used by banks to calculate capital requirements.

EBA has set up the “EBA Internal Model benchmarking exercise” to assist the competent authorities in this assessment. EBA calculates and distributes benchmarks against which the risk parameters of each institution can be compared. The points of reference are based on data submitted by the institutions, which include portfolio and model information as well as standard definitions to be used in the annual exercises.

1.2.   EBA prudential benchmarking exercise with three maIN objectives

The first objective concerns the prudential assessment of the quality of internal models.

The second objective is to provide a powerful tool to explain and monitor the variability of RWAs over time and the resulting implications for prudential ratios. As such, the benchmarking exercise has led to the development of the EBA guidelines on the estimation of PD, LGD and the treatment of defaulted assets, published on 17 November 2017.

The third objective is the use of the results. Benchmarks provide banks with valuable information on their risk assessment in relation to other banks with comparable portfolios.

The achievement of these three objectives is dependent on the availability of information about the level of conservatism incorporated in the estimates of risk parameters or in the RWAs. In particular, it will be possible to analyse the amount of variability that arises from this.

2.   Planned developments in the market risk framework IN 2022

2.1.   Inclusion of sensitivities data required IN the FRTB alternative standard approach

2.1.1.    THE need to prepare the FRTB reporting framework in accordance with CRR2

EBA includes the collection of sensitivities data. This is seen as a natural step towards full implementation of the FRTB in the EU.

Thus, additional templates requesting information on the sensitivity-based method (SBM) of the FRTB alternative standard approach (ASA) are introduced. Although the currently applicable market risk framework and related existing reporting requirements will remain unchanged for the foreseeable future, CRR2 has already introduced a reporting obligation on FRTB. However, the FRTB is not yet mandatory for capital requirements. The European Commission has set 30 September 2021 as the first FRTB regulatory reporting reference date.

2.1.2.    Gradual changes to the market risk benchmarking framework in line with the phasing-in of FRTB requirements

The EBA ITS on market risk reporting requirements will apply from 1 September 2021. In this ITS, EBA has adopted a phased approach given the need for adjustment to FRTB reporting requirements. In fact, institutions will also continue to be subject to the current (non-FRTB) market risk management framework and associated reporting requirements. The regulatory requirements associated with market risk should not therefore be overly burdensome.

Following this gradual approach, the benchmarking exercise includes an extension of the market risk templates in line with the FRTB Alternative Standard Approach (ASA). This risk-sensitive approach is designed and calibrated to serve as a credible fall back to the internal model approach where the test results related to internal model use are unsatisfactory.

In line with a gradual approach, the other two elements of the ASA, the DRC and the RRAO, will be included in the subsequent benchmarking exercises.

2.1.3.    New benchmarking templates to challenge the capital levels derived from the ASA as a fall back solution in case of unsatisfactory internal models

The proposed templates for the collection of sensitivities data for the SBM approach are aligned with the regulatory definition of the relevant risk factors in the ASA and will only request the information required in the ASA calculation in a clear and concise manner. The information requested in these templates will assist the competent authorities in investigating sources of potential variability in capital requirements arising from the ASA. The ASA is the fall back to the internal model approach. This could be done for example by reconciling the reported risk factors with the calculated sensitivities.

2.1.4.    Targeted collection of sensitivity data at two specific points in time

It is proposed to collect SBM data at two points in time:

  • The first collection point is the IMV (Initial Market Value) reference date and includes the reporting of sensitivity information relating to the risk factors specified in the SBM at the instrument level. The collection with reference to the IMV reference date should allow competent authorities to perform an early analysis of the quality of sensitivities data and support the assessment of IMV submissions made by institutions.
  • The second collection point refers to the last day of the risk measurement period which lasts typically two weeks in the benchmarking exercise. It is proposed to collect both sensitivity data at the instrument/portfolio level and capital requirement data at the portfolio level.

2.2.   minor updates to the list of financial instruments

The overall structure of the portfolios, as required in prior years, has not changed. However, there are some minor changes:

  • A number of instruments, had reached maturity after three financial years (2019-2021). Therefore, the ITS 2022 replaces a number of instruments with comparable instruments of longer maturity.
  • The range of instruments and portfolios has been slightly expanded to include dated sovereign bonds and CDS instruments.
  • The portfolio structure has been simplified by setting the number of instruments in a given portfolio to one and by modifying the instrument definitions to include quantity information.

2.3.   A need to harmonise the accounting framework for financial instruments in order to improve the justification of the variability of RWA

In order to simplify the use of the different terminologies used in the processing of instruments, the use of a standard language regardless of chosen format, can be applied to the instruments of the benchmarking exercise. This will allow:

  • A better definition of the instruments in a granular way.
  • A reduction of the ambiguity around the specification of instruments.

The reduction of ambiguity around the recognition of instruments will lead mechanically to a reduction in the variability of model outputs. In particular, this variability will henceforth depend mainly on the real differences in model rather than interpretation of the accounting methods for the various instruments.

3.   Planned developments in the credit risk framework for 2022 exercice

3.1.   Proposed changes for transparency on the level of conservatism included in the risk parameter estimates

3.1.1.    the objectives of the irb benchmarking exercise

IRB benchmarking aims to reveal unwarranted variability in the estimates of risk parameters (PD, LGD, CCF) and related IRB RWAs.

It appears that a significant part of the observed variability is due to the different levels of conservatism that institutions incorporate in the estimates of risk parameters or the calculation of RWAs. The inclusion of a margin of conservatism in the models is consistent with the requirements of the CRR.

More specifically, according to the CRR, it is necessary to add a margin of conservatism linked to the expected range of estimation errors to the banks’ estimates. When the methods and data are deemed less satisfactory, the expected margin of error is wider. This inevitably leads to a larger margin of conservatism.

This ITS draft therefore proposes to collect information in this regard.

3.1.2.    Clarification of reporting requirements on Margin of conservatism

The EBA guideline on PD and LGD introduced a framework for estimating the MoC. The latter applies to HDP IRB portfolios from 1 January 2022 and to LDP IRB portfolios from 1 January 2024. Thus, the information on MoC included in the estimates of the risk parameters of the HDP portfolios must be collected based on data as at 31 December 2021 from the 2022 benchmarking exercise onwards,.

The collection of MoC information will be voluntary for LDP portfolios until 31 December 2023.

3.1.3.    Data collection to improve the justification of the variability of RWAs between banks

The data collection will help justify the variability observed in the risk estimates used to calculate RWAs and the variability observed in the risk parameters without any margin of conservatism as well as without any conservative floors resulting from supervisory or regulatory measures.

Furthermore, there are divergent supervisory practices regarding conservatism. In particular, some supervisors prefer to prescribe supervisory add –on directly on capital levels rather than risk parameter estimates or RWAs. In order to capture all sources of conservatism, information on potentially existing RWAs or capital add-ons is also collected.  However, it may be this information collection is only for specific benchmark portfolios for reasons of simplicity of implementation.

3.1.4.    New information requested in the Benchmark templates

New columns have been added to the benchmark templates to collect information on the risk parameters taken into account in the calculation of RWAs. These depend on one of the following elements:

  • Conservative floors or multipliers imposed by prudential supervision teams.
  • The MoC component incorporated in the banks internal estimates.

As the CCF was not the main focus of the benchmarking exercise, information on CCF MoCs is not collected.

Additional columns have also been added to the templates in order to collect information on capital add-ons prescribed by supervisory authorities.  This is as a result of shortcomings in the IRB approach and which are not reflected in the risk parameter estimates analysed in the benchmarking exercise. This information on capital add-ons will be requested at a global level due to the operational difficulties of allocating these add-ons by sub-portfolio.

3.1.5.    Collection of information on conservatism targeted at the portfolio level for pragmatic reasons

The advantage of collecting prudential information at a granular level of benchmark portfolios lies in the simplicity of the associated analysis. For example, the analysis of variance with and without conservatism will reveal outliers more easily due to variability at this level of granularity.

The main complexity in dealing with the newly added information lies in its aggregated nature, as institutions will have to aggregate risk parameters with and without conservatism and thus calculate regulatory capital several times for the relevant benchmark portfolios.

3.2.   Minor changes to IRB templates

The creation of the following additional portfolios is proposed:

  • Four portfolios for the non-financial companies and other financial companies exposure class.
  • Two portfolios for the non-financial corporate exposure class.
  • Four portfolios for the credit institution and sovereign exposure class.
  • Four portfolios for the credit institution and sovereign exposure class.

There has been a significant increase in the use of the risk weight substitution principle due to government guarantees during the COVID crisis. Portfolios for which risk weight substitution is applied so that the exposure at risk after taking into account guarantees and RWAs are zero or blank must be reported in a dedicated template.

This will facilitate better transparency on default rates which may be biased in the coming years due to governmental guarantees.

4.   Developments under the IFRS9 templates for 2022 exercice

4.1.   IFRS9 PD templates included in the Benchmarking exercise from 2021

In line with the EBA communications on monitoring the effective implementation of IFRS 9 in the EU, a new set of templates related to the risk parameters of IFRS 9 have been included in the 2021 benchmarking exercise.

Based on the phased approach developed in the IFRS 9 roadmap, these templates for the 2021 financial year focused on the collection of PD metrics (including metrics related to the significant increase in credit risk).

4.2.   An extension of the templates for the collection of IFRS9 ECLs for the 2022 benchmarking exercise

4.2.1.    Purpose of the new templates

The main objective of the current set of templates is to collect quantitative data on IFRS 9 ECL parameters and other relevant information that may lead to significant inconsistencies in ECL results, thereby affecting regulatory capital and ratios.

Banks will also be asked to complete a complementary qualitative questionnaire.

All the information collected will provide a good understanding of the different methodologies, models, inputs and scenarios used under the IFRS9 framework.

4.2.2.    New templates to improve the challenge of the IFRS9 LGD and improve comparability between banks

The new templates are mainly related to the LGD risk parameter.

The collection of LGD data is extended to ensure comparability across institutions. In particular, while PDs for the same debtors are directly comparable with respect to the same default risk, the same is not true for LGD, as the characteristics of the exposure (e.g. the existence of collateral or a guarantee) could lead to different values for the same debtor.

Accordingly, the data collection is based on the logic of “hypothetical LGD”. Unlike PD values, which are based only on parameters actually assigned by institutions (and thus only for debtors for whom the institution incurs an exposure), the hypothetical LGD may differ from the parameters actually used in the ECL calculation.

In practice, the hypothetical LGD values are those that would be applied assuming that the exposure to the counterparty was senior exposure and unsecured and without a negative collateral clause in place. This concept is used for :

  • LGD IFRS 9 not guaranteed at 12 months (hypothetical).
  • LGD IRB Unsecured and without MoC (hypothetical). The latter is defined on the basis of hypothetical estimates of LGD that would be applied by the institution to the counterparty, without any:
    • margin of conservatism (MoC).
    • regulatory floor.
    • downward adjustment.
    • add-ons from supervisory authorities.
    • other conservative measures or adjustments, except those used for the IFRS 9 12-month LGD estimate.

5.   References

Abbreviations and glossary

EBA: European Banking Authority

MoC: Margin of Conservatism

LGD: Loss Given Default

PD: Probability of Default

CCF: credit conversion factor

HDP: High Default Portfolio

LDP: Low Default Portfolio

IRB: Internal Rating Based

IMV: Initial Market Valuation

ASA: Alternative Standard Approach

SBM: Sensitivity Based Method

DRC: default risk charge

RRAO: the residual risk add-on

RM: Risk Measure

FRTB: Fundamental Review of Trading Book

ECL: Expected Credit Loss