Response to consultation on ITS amending Commission Implementing Regulation (EU) 2016/2070 with regard to benchmarking of internal models
Go back
However, these requirements are not yet required to be applied by all institutions at the reference date for the benchmarking exercise for 2022 (31 December 2021). We therefore reject the required classification of MoCs into categories A to C.
In order to estimate conservatism and the effect of the margin of conservatism, we believe that it is sufficient to only add “PD without MoC”. A further breakdown into the MoC components in each case without A, B and C as well as “PD without supervisory measures” would represent a disproportionately high effort, since these components are currently not part of the reportable figures.
An alternative would be to provide the parameters without any MoC and with all MoCs. Intermediate parameters with part of the MoCs are not possible. In addition, the required detailed disclosure of MoCs of each category unnecessarily multiplies the number of parameters.
The individual components do not provide any information on the overall margin of conservatism, but only the aggregated values, which can already be calculated accordingly from a comparison of the reported PD and the “PD without MoC” this is now additionally required. However, quantifying the detailed MoCs in isolation and disclosing them in aggregated form in the reports does not generally appear to offer any added value due to the specifics of the individual institutions, their data and the aggregated reporting.
Last but not least, we would like to point out that, in its Final Report on the Draft ITS on Supervisory Reporting of June 2020, the EBA did not expect the institutions to provide information on MoCs as at the 31 December 2021 reference date. It gave the late publication of the Guidelines (p. 75) as the reason for this. We believe that the same policy should apply to benchmarking reports.
Furthermore, the difference between capital add-ons (Other relevant risk exposure) and RWA add-ons is not clear from the explanations in the Consultation Paper (Annex 4). Please provide clarification here.
Please clarify whether we can assume that the approved scope of notification continues to apply. The reason is that extending it to all types of risk would entail considerable additional effort.
The market risk instruments specified in the EBA benchmark study are subclassified into the equity, IR, FX, commodities and credit spread risk types. The participating institutions need only take into account the instruments of the risk types that form part of the scope of the notified internal model. The entire process sequence for performing the benchmark study generally refers to the approved scope of notification. The addition of templates 106.01 and 120.1-3 regarding information relating to the sensitivities-based method of the alternative standardised approach under FRTB now raises the question of whether the associated metrics have to be determined for all risk types, or only for the risk types that are relevant for the notified internal model.
Q1.1: Do you have any concerns on the proposed collection of data on conservatism in the PD and LGD estimates? In particular as regards the breakdown into Moc A, B and C?
The EBA Guidelines on PD and LGD estimation, which are required to be applied from 1 January 2022, stipulate that any deficiency related to the estimation of risk parameters must be classified into one of the two categories A or B (paragraph 36 of the Guidelines). The general estimation error is classified into category C (paragraph 42). Each of the categories has its own MoC to be quantified individually at the calibration segment level.However, these requirements are not yet required to be applied by all institutions at the reference date for the benchmarking exercise for 2022 (31 December 2021). We therefore reject the required classification of MoCs into categories A to C.
In order to estimate conservatism and the effect of the margin of conservatism, we believe that it is sufficient to only add “PD without MoC”. A further breakdown into the MoC components in each case without A, B and C as well as “PD without supervisory measures” would represent a disproportionately high effort, since these components are currently not part of the reportable figures.
An alternative would be to provide the parameters without any MoC and with all MoCs. Intermediate parameters with part of the MoCs are not possible. In addition, the required detailed disclosure of MoCs of each category unnecessarily multiplies the number of parameters.
The individual components do not provide any information on the overall margin of conservatism, but only the aggregated values, which can already be calculated accordingly from a comparison of the reported PD and the “PD without MoC” this is now additionally required. However, quantifying the detailed MoCs in isolation and disclosing them in aggregated form in the reports does not generally appear to offer any added value due to the specifics of the individual institutions, their data and the aggregated reporting.
Last but not least, we would like to point out that, in its Final Report on the Draft ITS on Supervisory Reporting of June 2020, the EBA did not expect the institutions to provide information on MoCs as at the 31 December 2021 reference date. It gave the late publication of the Guidelines (p. 75) as the reason for this. We believe that the same policy should apply to benchmarking reports.
Q1.3: Do you agree to the voluntary collection of the information for LDP portfolios?
We also believe that voluntary reporting for LDP portfolios should not begin until the 31 December 2022 deadline at the earliest, as there will not yet be any supervisory approval for 2022. The use of models based on the Guidelines on PD/LGD estimation (EBA/GL/2017/16) will start at the beginning of 2022 at the earliest. The values may also not yet be used in the bank for this reason. An earlier deadline would effectively result in the reporting of figures that are not yet approved by the banking supervisory authorities. This in turn results in the development of parallel structures, which would involve an extremely high effort.Q1.4: What are the main challenges for institutions in this regard?
From our perspective, global add-ons to portfolios should be avoided, since the result will depend heavily on the calculation methodology used. However, the methodology is not specified, which will therefore lead to results being distorted if each institution chooses different approaches.Furthermore, the difference between capital add-ons (Other relevant risk exposure) and RWA add-ons is not clear from the explanations in the Consultation Paper (Annex 4). Please provide clarification here.
Q2.1: For which kind of portfolios would you expect that outdated ratings (or other missing information hindering the annual re-rating) are a material driver of variability when comparing institutions RWA on homogeneous benchmarking portfolios?
As outdated/delayed ratings are extremely rare due to the institution’s internal processes, we do not consider them to be a material driver.Q7: Do you see the need to collect weights of economic scenario per time horizon?
Since the scenarios weights are generally constant, we do not see any need here to collect weights of the economic scenario for the relevant time horizon.Q12: Do you see any issues or lack of clarity in the definition in the changes and updates introduced in the list of instruments and portfolio of Annex 5?
The EBA contemplates a portfolio overhaul for future BM update, which would aim to better align the range of instruments and risk considered in the benchmarking portfolio with banks’ actual trading book portfolios and aim to better capture specific features of ASA.Please clarify whether we can assume that the approved scope of notification continues to apply. The reason is that extending it to all types of risk would entail considerable additional effort.
The market risk instruments specified in the EBA benchmark study are subclassified into the equity, IR, FX, commodities and credit spread risk types. The participating institutions need only take into account the instruments of the risk types that form part of the scope of the notified internal model. The entire process sequence for performing the benchmark study generally refers to the approved scope of notification. The addition of templates 106.01 and 120.1-3 regarding information relating to the sensitivities-based method of the alternative standardised approach under FRTB now raises the question of whether the associated metrics have to be determined for all risk types, or only for the risk types that are relevant for the notified internal model.