It is recommended that the absolute and relative thresholds are determined by considering the occurrence and materiality of technical defaults (and other similar conditions which may not be considered a real default), and the sizes of exposures within portfolios subject to this regulation. This will ensure that the materiality thresholds are fit for purpose, namely, effectively separating technical defaults and short-term cures from real defaults, and ensuring timely identification of real defaults.
Internal materiality thresholds within HSBC and other regulated institutions that drive the financial viability of collections and recovery activity, would affect the motivation for treating exposures in default, classification as real default, and thus the real default rates for different portfolios. Once sufficient insight has been obtained, the decision whether to choose from option 1 (either absolute or relative limit breached), or option 2 (absolute and relative limit) can be made effectively.
In the absence of such analysis, and due to variance across personal, wealth and business portfolios, it would be sensible for competent authorities to implement the thresholds as caps, allowing reasonable flexibility across different portfolios and regulated institutions. Furthermore, we support option 2 over option 1.
The percentage trigger for the relative threshold of 2% appears high for large exposures in the wholesale credit market. We would suggest reducing this threshold to a level that is meaningful for large exposures which will also maximise comparability of own funds requirements for credit risk; however, without proper analysis suggesting an alternative figure presents a challenge
For retail exposures, thresholds higher than 200 EUR and 2% would not be desirable. For smaller exposures with longer repayment terms, higher thresholds would delay the classification of delinquent accounts as 90 day default by a number of months; in some instances, accounts can have made zero payments for 180 days before the thresholds of 200 EUR and 2% are reached. This appears counter-intuitive given the CRDIV requirement to move from a 180 to 90 day default definition for the majority of retail exposures, and is also not consistent with internal triggers for debt recovery.
Any adjustment of the data history for the revised materiality thresholds would be a very time consuming manual task that would take several years.
Given that LGD models require at least seven years of data history a retrospective application of the new guidelines would require the manual review of seven years of data history for each customer. The task would be further complicated by the fact that some customers who were not identified as being in default might be classified as being in default once the revised thresholds are applied.
The change in the default definition will also potentially have impacts on future model developments, recalibrations, and reporting going forward not to mention subsequent supervisory approval of updated models going forward. The timings of this change in relation to these activities will need to be considered.
Considering all of these factors, as well as pre-existing and forthcoming changes to the A-IRB framework, we think it reasonable that these changes be given a sufficiently long phase-in where regulated institutions will have the flexibility to prioritise redevelopment efforts based on model materiality.
We agree that the benefit of establishing harmonised criteria for setting the materiality thresholds for past due exposure is greater comparability of own funds requirements for credit risk. It is unclear, however, how the proposal would reduce the administrative and operational burden to comply with different regulatory frameworks in different Member States, and indeed globally, given that each national regulator will have to set their own thresholds within the proposed EBA framework. Given that the proposal currently suggests that national regulators should determine fixed thresholds rather than caps it is unlikely that all regulators will define the same thresholds and the jurisdictional differences would therefore remain. Our response to Q02 therefore suggests that the thresholds should be used as caps and that the trigger for the relative threshold be set below 2% in order to minimise variability of own funds requirements for credit risk.
However it is recognised that harmonising criteria for setting materiality thresholds will increase comparability of own funds requirements for credit risk. As stated previously, it should be noted that some portfolios in HSBC are regulated by more than one competent authority, and in such cases, there is a need for alignment across a number of competent authorities for the same portfolio.
While the assessment states that the adjustment of data and recalibration of risk parameters may impose a significant operational burden on the banks that use the IRB approach we do not believe that this sufficiently expresses the excessive cost and operational burden in relation to the benefits achieved. Similarly, the re-adjustment of all historical data will be a lengthy and time-consuming task which adds to the operational burden intimated above.
The impact of this change cannot yet be fully determined. The size of the impact for retail portfolios would depend on how much any materiality thresholds used in current modelling, calibration and reporting processes, differ from those that will be enforced by competent authorities as a result of this consultation paper. Broadly speaking, the change in default definition due to this materiality threshold would be expected to have minimal impact on cost of own funds for credit risk and we do not generally expect that the proposal will have a significant impact on exposures classified as in default; however, this is very difficult to confirm in the absence of extensive manual checking.
It is estimated that the change to the default definition could have considerable operational impacts. Further clarification of the mandatory requirements and their ramifications for model development, model calibration and reporting is sought. A mandatory requirement for models to be developed and calibrated on this new definition and for reporting to be modified, could require the processing of huge amounts of data, significant re-modelling activity, a vast number of historic recalibrations, and changes to a number of other operational processes.