NBB Circular NBB_2021_31: EBA Guidelines on Input Data Criteria for Risk Measurement Models under Article 325a of Regulation (EU) No 575/2013

NBB_2021_31 – 15 December 2021 Circular – Page 1/3 boulevard de Berlaimont 14 – BE-1000 Brussels tel. +32 2 221 30 17 company number: 0203.201.340 RPM Brussels www.bnb.be Circular Brussels, 15 December 2021 Reference: NBB_2021_31 your contact:
Gaëtan Doucet tel. +32 2 221 38 10 gaetan.doucet@nbb.be EBA Guidelines of 13 July 2021 specifying the criteria relating to the use of input data in the risk measurement model referred to in Article 325a of Regulation (EU) No 575/2013 Scope This circular applies to Belgian credit institutions and approved or designated Belgian financial companies (hereinafter 'institutions'), and to Belgian stockbroking companies referred to in Article 1, paragraphs 2 and 5 of Regulation 2019/2033 (hereinafter, 'significant Belgian stockbroking companies'). Summary/Objectives The EBA Guidelines EBA/GL/2021/07 (hereinafter 'the EBA guidelines specifying the criteria relating to the use of input data in the risk measurement model referred to in Article 325a of Regulation (EU) No 575/2013') define the characteristics that the input data used in the risk model developed by a Belgian institution or significant Belgian stockbroking company should possess. These guidelines apply from 1 January 2022. This circular clarifies the NBB's approach regarding less significant institutions (LSI) and significant Belgian stockbroking companies. It is recommended that significant institutions subject to direct ECB supervision follow this circular to the extent that the ECB does not publish guidelines on this matter.

Circular – Page 2/3 NBB_2021_31 – 15 December 2021 Madam, Sir, The National Bank of Belgium wishes to indicate by this circular that the EBA guidelines specifying the criteria relating to the use of input data in the risk measurement model referred to in Article 325a of Regulation (EU) No 575/2013 are integrated into its supervisory practice. These guidelines apply from 1 January 2022. In order to be used in the risk measurement model referred to in Article 325a, input data should possess the following characteristics, some important elements of which are recalled below:

• they should be accurate,
• they should be appropriate,
• they should be updated with sufficient frequency,
• they should be complete.
  The accuracy of input data covers various dimensions detailed in the guidelines.
  In particular, institutions should reconcile the input data of the current period either with the corresponding data derived from verifiable prices, or with the corresponding data used by trading room or post-trade systems. The reconciliation can also be performed on the prices produced by their internal risk measurement model and based on the input data of the current period with, either verifiable prices, or prices produced by trading room or post-trade systems. When these techniques are not possible, it is authorized to use historical data used to calibrate the input data or prices resulting from these historical data. Furthermore, input data should reflect the properties of the distributions of the corresponding risk factors as well as the dependency structure thereof. The guidelines also specify the conditions under which substitute data from the period of financial stress can be used to take into account the effect of fundamental changes. Finally, accuracy is also emphasized in the case of the use of beta approximations or other random data generation methods for the determination of future shock scenarios, as well as in the case of the use of combinations of risk factors. The appropriateness of input data refers to the coverage, where applicable, of general market risks and specific market risks, for which institutions are expected to produce detailed analyses. Institutions should update the historical data used to calibrate the input data of the current period more frequently than once a month, when a monthly update could lead to an inaccurate estimate. When institutions use beta approximations or other random data generation approaches for the determination of future shock scenarios, they should recalibrate at least once a month. Finally, the guidelines specify the conditions for the completeness of input data, particularly in the case of missing or inconsistent values, as well as in terms of alternative data sources (for example, to deal with the unavailability of usual sources).

NBB_2021_31 – 15 December 2021 Circular – Page 3/3 The EBA guidelines specifying the criteria relating to the use of input data in the risk measurement model referred to in Article 325a of Regulation (EU) No 575/2013 can be consulted on the EBA website:
Guidelines on use of data inputs in the IMA_FR.pdf (europa.eu) or on the NBB website.
A copy of this circular is sent to the commissioner(s), approved auditor(s) of your institution.
Please accept, Madam, Sir, the assurance of my distinguished consideration.
Pierre Wunsch
Governor