DPM 4.0 introduced a new portfolio where all the exposures towards CIU treated with internal models should be reported. In Template C_02 row 425 (Collective investments undertakings (CIU)) was introduced and, on the basis of the item number 1.1.2.4, this row should not be included neither in row  250 (item 1.1.2.1) Total FIRB nor in row 310 (item 1.1.2.2) Total AIRB, but directly in row 240 (item 1.1.2) Internal ratings based Approach (IRB). This representation seems to be in contrast with the instructions available in paragraph 3.3 of Annex 2 of Template C_08.01 / C_08.02 / C_08.03 / C_08.05 (applicable by extension to Template C_34.07 paragraph 3.9 even if with reference to C_34.07 it is not clear why in the annotated table the domain “qAE:qAE1” is used and it differs from the one used for C08.xx templates which is “qAE:qAE0”). Specifically, i) the new CIU portfolio (qx2082) is mentioned under point B.7) which is included under paragraph B.1) Total IRB approaches when own estimates of LGD and/or Conversion Factors are used (I.e. qx2023) and ii) instructions for rows 190/200/210 specifically state that “This row shall be reported for the exposure class 'Collective investments undertakings (CIU)'. and in the ‘Total’ sheet.”.

In addition, on April 2025 ECB published a series of EGDQs (e.g. EGDQ_0804 and EGDQ_0806) which initially corroborated this representation (The total assigned to obligor grades or pools with own estimates of LGD and/or conversion factors ({qx2023}) should be equal to the sum of all portfolios with own estimates of LGD and/or conversion factors ({qx2009+qx2013+qx2015+qx2021+qx2055+qx2068+qx2071+qx2073+qx2075+qx2081+qx2082}).

This check was later amended with the exclusion of the qx2082 from the calculation of qx2023, amendment accompanied by the comment that the instructions of Annex 2 need to be corrected to remove CIUs from the section B.1) ‘IRB approaches when own estimates of LGD and/or Conversion Factors are used’ and should be treated as a separate portfolio in the numbering of sections in the step. For CIUs it should not be B.7) but C) and thus treated as a separate Total portfolio in addition to Total AIRB and FIRB portfolios.

On the Pillar3 side, we would like to point out: 

• the inconsistency between the ITS instructions “22 Annex XXII - Disclosure of credit risk IRB” which specify the exclusion of CIUs from templates CR6, CR7, CR7-A (“This template excludes counterparty credit risk (CCR) exposures, securitization exposures, other non-credit obligation assets, collective investment undertakings and equity exposures”) and the annotated tables DPM4.1/DPM4.2. In the latter, in fact, for CR6 the domain is “qAE:qAE2” (which includes qx2082) and in the totals the reference is to qx2023/qx2022; for CR7 and CR7-A the explicit CIU row is not provided while the total rows refer to qx2022 and qx2023: it is therefore not clear whether any CIUs should be included in the totals or not (shown in the table in the case of CR6, not shown in the table in the case of CR7 and CR7-A) and whether priority should be given to the ITS instructions or the annotated tables;

• For CR9 table, unlike CR6, the exclusion of CIUs in CR9 is not mentioned in the ITS instructions cited above and this is confirmed by the fact that the domain used is “qAE:qAE1”. In the Totals table, the annotated tables refer to qx2023 and qx2022, therefore it is not clear whether these total classes include CIUs or not (which are instead required in the table breakdown).

• CCR4 (K_04.00.a) DPM4.1 refers to the domain “qAE:qAE2”, DPM4.2 instead refers to the domain “qAE:qAE5” which aggregates retail and corporates exposure classes and does not refer to CIUs qx2082, Regional qx2071/qx2072 and PSE qx2073/qx2074. The reason for the exclusion of these exposure classes, reported in the corresponding table of P1 C34.07, is unclear. Furthermore, in the total CCR4 template (K_04.00.b) reference is made to the domain “qEC:qx4 Exposure classes excluding equities, securitisations and other non-credit-obligation assets”, therefore CIUs appear to be included.