Intesa Sanpaolo S.p.A.

According to our elaborations, the expected impact will be:
- Estimated implementation costs: 150 man days
- Recurring production costs: 10 man days

As far as implementation and estimations costs are concerned, the following remarks can increase our forecast:

1. Entry into force and frequency:

- As National Competent Authorities (NCAs) have already in place reporting standards different from those proposed by the EBA ITS, assurance is needed on the fact that no double implementation will be required;
- clarifications is sought about the frequency of reporting: due to burdensome computation, our understanding (e.g. the first flow with the new format is expected with reference date June 2018) is that the OpRisk loss sheets remain semi-annual.

2. Scope:

It is not clear:
- how the significance filter must be applied to small companies currently adopting Traditional Standardised Approach or Basic Indicator Approach belonging to an AMA or significant Group;
- if these subsidiaries contribute to Group OpRisk loss reports or not.

Therefore, further clarifications are needed:
- if these subsidiaries are individually not significant or they are not subject to Directive 2013/36/EU (CRD IV) and Regulation No 575/2013 (CRR) rules individually;
- on the exclusion of these non-significant and non-banking subsidiaries from the reporting scope: if the Supervisors expect them to be included, the possibility for them to collect loss data in a simplified way (compliant with Annex II instructions) should be made explicit (the local implementation of an OpRisk framework and loss data collection process could be very burdensome, costly and time expensive in comparison with the hypothetical losses these subsidiaries might face).
We find the threshold of 1% of the sum of total carrying amount for Debt securities and Loans and receivables quite low [Article 5 (b) 3 (a)] . We believe that applying a 10% or 20% threshold would be more appropriate.
First, the breakdown by accounting portfolio is not completely clear, especially in the light of the entry to force of IFRS 9. In particular, it is not clear how to fill both columns 030 “Financial assets held for trading” and 040 “Trading financial assets”.

Second, further clarification is sought on the treatment of direct short positions, especially about the netting. It is not clear whether netting has to be applied in the specific column or in each column.

Finally, it is not clear whether the accumulated changes in fair value should be limited to the reporting reference year or to all the historical changes related to the current exposures.
As a preliminary remark, a review of terms and definitions is deemed necessary to harmonize them throughout the documents: both in the Consultation paper and in the Annex II the terms “events”, “loss”, “loss event” and “event type” are often used indifferently, even if they refer to different entities: as the industry assigns specific meanings to each term (e.g. event is the occurrence, loss is one of the possible effects of the occurrence), this could cause misunderstandings in local implementation.
Our main concern related to the templates is driven by the huge computational effort required (especially to large Banks or Groups) by the proposed criteria to be implemented in downloading data from local databases, far more complicated than those currently in place (please refers also to the answer to question 1):

A. With reference to C 17.01:
• the insertion of rows 940-944 “Loss adjustments relating to previous reporting periods” and the deletions of events not in OpRisk scope cause high complexity and time consumption; the added requirements could transform the regulatory reporting flows into black boxes, hard to be investigated or checked manually. In particular, dealing with the original loss amount of the event (i.e. gross loss amount last time it was reported, in our interpretation) is further complicating the computation. Our suggestion is to remove every reference and burden related to the treatment of the “original loss amount”.
• assurance is also needed on what is meant for “last year” loss incidents to be reported: does this definition refer to “current period” events (included “fast closing” delayed ones as described in the following) or does it have a different meaning/scope?

B. With reference to C 17.02 - Information on large loss events, more details are required especially on the nature of largest loss incidents, to allow supervisors to capture their risk drivers: in our opinion, the request for a free text description is not the most effective solution to meet the goal.
As a minor remark, we suppose that a typo is in in paragraph 128, as the description and instructions referred to Maximum single loss are reported at the end of this paragraph and not in an independent one.

Please find below further detailed comments to the specific templates.

A. Template C 17.01: Operational risk losses and recoveries by business lines and event types in the last year

i. Column 080 Total event types
• the reported lists do not include the reference to the “Number of events related to loss adjustments”, even if the related rows are included in the format;
• whenever Banks own huge databases (per events, losses and recoveries figures) the points reported below could dramatically increase complexity and computational efforts:
- loss adjustments on one event can be more than once in a year and include adjustments referred to previous reporting years that could not be reported in advance due to “fast closing” delays. But data must be treated differently depending on the gross loss of the event when they were last time included in the EBA ITS templates. As Banks can hardly speed their data collection processes without weakening data quality and assurance, to comply with these requirements they will need to heavily intervene on and modify their database inherent structures, facing high costs and resources consumption to comply with this new regulatory reporting requirement;
- instruction is needed on the reporting of loss adjustments referred to events pertaining to companies or banks that are acquired by the Group and that were previously independent and/or part of a different banking Group: do they have to be treated as they were “new” events or as “already reported ones”? We would prefer the first option; in the second one, it could be very difficult to recover the gross loss amount of the event when it was last time reported by the incorporated company (and even whenever it was reported or not);
- cutting the historical length of time series (for example, signalling adjustments and recoveries only when referred to events with a first accounting date in last 10 years) could be a good solution to cope with computational problems due to huge numbers of events and losses, without seriously affecting the Regulators possibility to understand the risk profile of the Banks.

ii. Rows 930-934 Number of loss events subject to loss adjustments
• Loss adjustments causing a change in the range of the related event are counted twice in the same way (+1), both in the “old” range and in the “new” one: even if this can enable the users to understand that some changes happened, it does not help to appreciate if there is an upward or a downward trend in the size of events of previous periods. In our opinion, the instruction should be removed by the proposal.

iii. Rows 940-944 Loss adjustments relating to previous reporting periods
• The proposed requirements (especially points c), d), e) of the Paper, page 6 of Annex II) are burdensome for Banks dealing with huge databases and with long time series; in addition, some banks erase from their databases events falling under the internal thresholds or that have been recognized as not belonging to the OpRisk scope.
The split between “previous” and “current” reporting periods will bring to a sensitive increase in the data points to be reported from (up to) 482 to (up to) 674 but it is expected to bring the following benefits:
- A clear allocation of impacts to root events could be established and the development of losses from previous years could be monitored;
- Reduce the risk to underestimate the current level of losses of an institution;
- Reveal potential under-provisioning for OpRisk events if there are large impacts in the current reporting period relating to events first accounted for in previous reporting periods;
- Allow to clearly distinguish between gross loss amounts, loss amounts net of direct recoveries and insurance recoveries without mixing direct and insurance recoveries which have to be treated differently.
The proposal is - as a whole - coherent with the goals set but it is not clear how banks should classify events related to previous periods whose losses have not been reported yet (not owing to a change in their inclusion in the OpRisk scope but due to shrinky “fast-closing” delivery date). In particular, shall these events be classified as “current” (for the sake of simplicity, this is the preferred option) or as “previous”?
Moreover, does the freezing date depend on the single event or on the last date when one of its losses (any recovery excluded) was reported to Supervisors?
Please note that, whatever the choice made, these events will affect the full understanding set in points a) and c) above. Moreover, it is not clear if they have to be included in C 17.02 sheet (e.g. last year greatest events list) or not, so affecting point d) above.

For further observations, see also Question 7.

B. Template C 17.02: Operational risk: Detailed information on the largest loss events in the last year

i. Confirmation is needed on the fact that loss adjustments on previous years’ events are not to be included. For example: if last time an event was reported based on a far lower amount and now, due to an adjustment, its amount has increased to a point such that it corresponds to one of the maximum losses reported in C 17.01 row 950, shall it be reported or not?

ii. Column 020 - Date of accounting
• Only events accounted in the last year are to be included: what about events never reported before (e.g. due to “fast-closing” delivery dates)?
Moreover, the use of a subsample of events reported in other parts of the report increase the burden in the flow constructions and control checks.

iii. Column 190 - Business Unit:
• The purpose of this column is not clear. As different Banks adopt different business unit definitions, discrepancies among institutions are to be expected, both in the names and in the level inside the organisation (Division, Department, sub-Department…).

iv. Columns 200 – Description
• Being the description a free text to be manually filled in by Banks (based on several different attributes of their own internal databases), without specific requirements, our expectation is to get weak information. Moreover, the language prescribed for filling in free text descriptions should be specified (local language or English); please note that, as local databases are usually in local language and most of the time automatically enriched through flows from further internal tools, this will cause “ad hoc” translations of these fields.
Clarification is needed for filling in rows 940-944. Our suggestion is to remove the instruction of dealing with the original loss amount of the event (e.g. gross loss amount last time it was reported, in our interpretation):
- the total loss amount to be reported in row 940 is defined as the simple aggregation of the loss adjustments related to previous reporting periods, but, in the following sentence, it is supposed to be amended by the original Gross loss amounts of the events that fell under the internal thresholds;
- due to the unclear use of terms “loss” (i.e. the consequence of an event) and “event” (i.e. the occurrence) in the Annex II, clarification is needed. As an example, we suggest to amend point c) as per below:
“If the loss adjustment entails a change in the range of the event, institutions shall reclassify the loss event by excluding the original loss amount of the event from the range applicable before the adjustment (report the amount with a negative sign in the respective row) and assigning the adjusted new gross loss amount of the event to the range applicable after the adjustment (report the amount with a positive sign)”.
Confirmation is required on the following points:
- figures reported in row 940 are expected not to be equal to the sum of rows x40 (e.g. they shall not be equal to the sum of Loss adjustments per business lines);
- “original loss amount” refers to the Gross Loss of the event, last time the event was reported at an year-end EBA ITS reports? Or check on interim figures too are to be applied? (Please consider that the second option would further enlarge complexity).
We believe that the proposed treatment on ranges is not appropriate in terms of cost/benefit and that the instruction should be removed by the proposal.

Major investment in developing internal databases will be needed and in a short time frame (one year), while also regular maintenance and data quality checks on flows to be sent to Regulators will be far more time and resources consuming.

Please refer to the answer to Question 6 for further details on the increased burden in loss data management and on the opportunity to cut historical series.
The less costly, clear and easy to implement option is “(b) 17 largest losses if above 100,000 € per loss event”. In fact, all institutions submitting template C 17.02 report the biggest loss event per event type plus the 10 largest loss events of the remaining events; in both cases subject to a minimum loss of 100,000 € per event.

We welcome the possibility of further increasing the 100.000 € threshold.
Francesca Passamonti