Handbook on supervisory benchmarking

Annex I is an excel worksheet which defines the benchmarking portfolios/counterparties to be used for submitting the data fields specified in the relevant templates in Annex III (for credit risk) and in the templates in Annex VIII (for IFRS9). The following table provides an overview on how to link the benchmarking portfolios defined in Annex I to the relevant templates:

Table 1: Link between Annex I and the Templates to be filled in

As an example, the IRB relevant benchmarking data that is to be submitted for the counterparties specified in sheet 101 of Annex I, is specified in template C 101.00 of Annex III (with the relevant instructions in Annex IV). The IFRS9 relevant benchmarking data that is to be submitted for the counterparties specified in sheet 101 of Annex I, is specified in templates C 111.00, C 112.00, C 113.00, C 114.00 of Annex VIII (with the relevant instructions in Annex IX).  

Next to the Supervisory Benchmarking exercise, credit institutions have several other reporting obligations, such as the common reporting framework (COREP) and the financial reporting framework (FINREP).

Within COREP, they report their solvency ratio to supervisory authorities under the Capital Requirements Directive (CRD) as well as other relevant supervisory metrics. Reporting guidelines for COREP can be found here.

Within FINREP, credit institutions report their Balance sheet, Income statement, Comprehensive Income and Equity, Disclosure of financial assets and liabilities, off balance sheet activities and non-financial instrument disclosure. Reporting guidelines for FINREP can be found here.

The data required in the benchmarking exercise frequently refers to data fields used in COREP and FINREP. The relevant references to COREP/FINREP for defining the benchmarking portfolios can be found in Annex II and the relevant references to COREP/FINREP for reporting the benchmarking data can be found in Annex IV.

Instructions on how to deal with fields that are not known or zero is provided in Part I of Annex IV.

The tables in Appendix A on this website provide an overview of the data fields and formats required for the templates of Annex III.

The deadline for data submission is specified in the ITS in Article 3. For the credit risk the deadline for data submission is usually the 11 April each year.

Competent authorities forward the relevant data to the EBA. Following the data submission, competent authorities check data quality and revert to institutions in case re-submissions become necessary. The final calculation of benchmarks is performed by the EBA and circulated to the competent authorities each year late in June or early in July.

The EBA does not collect any data directly from institutions, but receives them via the Competent Authorities. Therefore, participating institutions should contact their Competent authority to discuss any specific questions on the data submission.

All questions related to the ITS, the related rules and other guidance provided by the EBA should be addressed using the EBA Q&A Webform. Responses of general interest will be published in this Handbook.

The definition of the benchmarks and how they are computed is documented in the chart pack which is published on the EBA’s website after the annual exercise has been closed (i.e. usually in February or March of the following year). 

This section describes how the benchmarking portfolios are constructed. Figure 1 and 2 below illustrate the construction and hierarchy of the benchmark portfolios. There are 4 levels of granularity for which the credit risk benchmarking portfolios are specified, the first one is referred to as Level 1 split and represents the total exposure per exposure class which is divided according to the default status of the exposure (Default Split).  So each Level 1 portfolio is split into its defaulted and non-defaulted parts. There is no further break-down for defaulted exposure.

For the non-defaulted exposures, there are several characteristics defined in Annex I for which splits are introduced (level 2 split). Therefore, institutions should carefully read Annex I and check the defined characteristics for the different splits.

These characteristics partially differ between the Low Default Portfolios (LDP) and the High Default Portfolios (HDP) (see figure 1 and 2 below). Each combination of characteristics, that describes a benchmarking portfolio is connected to a unique ID, specified in Annex I. Consequently, the combination of exposure class, default status and characteristics from level 2 and level 3 split determines the portfolio ID.

For some characteristics there are further subsets marked as Level 3 split (i.e. for the secured/unsecured split, there is a subset with collateral types for the secured exposures).

There is one Q&A that deals with the assignment of retail exposures to benchmarking portfolios:

Q&A 2023_CR01
 

Figure 1: Portfolio Composition of low default portfolios

Figure 2: Portfolio composition of high default portfolios

In preparation of each exercise, the EBA publishes the list of validation rules used to ensure data quality in the institutions data submission for credit risk. The list is updated whenever a new release of the reporting framework is published and available for download on the EBA Website.

Competent authorities throughout the EU with the support of the EBA are applying the validation rules to the data submissions of institutions under their remit and approach institutions to resubmit whenever validation rules are failing.

There are 2 different severity categories of validation rules: “Error” and “Warning”. Not meeting "Error" validation rules when submitting data will cause the file to be rejected immediately, therefore the submission is not accomplished. Where reporting entities are of the opinion that they can only comply with the reporting requirements set out in the applicable reporting standard if they breach a validation rule of the type “Error”, they should contact their competent authority.

Validation rules with the severity status “Warning” are not causing an immediate rejection of the submission but should still be investigated and closely monitored by the reporters. Reporters are expected to analyze the results of “Warning” validation rules and verify and revise the data, if necessary. Some warnings may not be applicable to all reporters, in all circumstances, or to all reference dates. As such, not meeting them will not necessarily prevent the successful submission of data. Where a reporter believes that it is not possible to meet the validation rule, the reasons for the inability to comply should be explained to the competent authority.

Warnings may become “Error” validation rules in the future. Except for validation rules reflecting technical requirements, new validation rules are introduced as warnings initially.

Please see Q&A 2018_3934 for further elaboration on a specific validation rule.

In addition, Competent Authorities together with the EBA perform benchmarking specific Data Quality checks. A failing of these checks may trigger questions addressed to the submitting institution and may eventually lead to resubmissions.

2.1.3.1    Allocation of exposures to the secured/unsecured portfolio, collateral split portfolios

Annex I and II of the ITS provide clear instructions on the assignments of collateralization status and collateral types. For a single exposure for one type of collateral, the secured and unsecured share of the exposure should add up to the original exposure value.

There are three Q&As that cover the collateralization split and how exposures should be assigned.

Q&A 2018_4093

Q&A 2018_4091

Q&A 2018_2999

2.1.3.2    Reporting of risk parameter and exposure values on secured/unsecured portfolio and collateral split portfolios

Instructions on the metrics for the secured/unsecured portfolios and the collateral split are included in Annex IV of the ITS.  

The reporting of CRM techniques with substitution effect is covered in Q&A 2018_4093 and Q&A 2018_4091.  

There are two different ways how the substitution approach can be applied (RWA substitution and risk parameter substitution). With both options, several data points in the reporting are affected.

In template C102 and C103, CCF values should only be reported for off balance sheet exposures. The data point ‘CCF’ should therefore be left blank for on balance sheet exposures denoted by _ONX in the portfolio ID.

For template C101, the scope of CCF data to be collected is specified in Annex IV, Part II of the ITS.

Q&A 2020_5368 covers questions with the weighted average of the CCF.

The data points ‘DR1YR’ and ‘DR5YR’ (1 year and 5 year default rates) refer to observed parameters which can be used in the exercise to explain outlier observations e.g. where the empirically observed default risk in a portfolio is very low this may explain a lower PD estimate compared to peers for a considered portfolio. Instructions for DR1YR and DR5YR can be found in Annex IV of the ITS. This information shall only be reported for portfolio IDs relating to non-defaulted exposures.

Instructions on the metrics for the data points ‘LR1YR’ and ‘LR5YR’ can be found in Annex IV of the ITS.

The loss rate of the latest year shall be reported for portfolio IDs relating to ‘non-defaulted’ and ‘defaulted’ exposures only.

Where the institution is not able to calculate a loss rate for the past five years it shall develop a proxy using its longest history up to 5 years and provide documentation detailing the calculation to its competent authority.

Annex IV of the ITS specifies how the data points ‘RWA--/-/+/++’ need to be reported. Furthermore, Q&A 2016_2782 and Q&A 2020_5339 take up information on these fields.

While for templates C101-C103 portfolios are defined particularly for the Benchmarking exercise to make bank data more comparable, the aim of template C105 is to collect more granular internal model data for which portfolios are submitted. This implies that template C105 conveys the internal model perspective, while template C101-C103 relate to the portfolio perspective.

Generally, it is possible that institutions have several PD models but only 1 LGD model because the models have different segmentations in the calibration.

To facilitate the mapping of benchmarking portfolios to models, institutio shall only indicate ‘PD’, ‘LGD’ or ‘CCF’ for the data point ‘IRBA Risk parameter’ rather than ‘PD estimate’ in template C105.01.

Institution should make sure to report an Internal model ID (c0010) for each risk parameter that is used in the own funds calculation. In other words, for benchmarking portfolios covering FIRB exposures, only Internal model IDs related to PD models shall be reported and for benchmarking portfolios covering AIRB exposures at least 2 Internal Model IDs should be reported for the respective PD and LGD models and where relevant CCF models.