There is broad consensus among EBF members that Option 2 is preferable for the reasons expressed below. The best solution would be a relative threshold in combination with a minimum absolute threshold.
This solution would minimise unintended cases of technical default computed as actual default, i.e. false positives. Option 2 is the preferred one because it really leaves out virtually all cases of technical default, as intended, without representing any material departure from the general principle of computing all true defaults for regulatory modelling purposes. Expert judgment could complement the quantitative thresholds to identify remaining cases of technical default.
Option 1 does not seem to be a plausible solution as it would flag defaults for residual negligible due amounts of smaller clients (e.g. €25 past due for technical reasons on a €600 outstanding debt) and it would also mark as default relatively small hitches like a technical overdraft of €600 in a large corporate with €10 million exposure.
Option 1 would lead to a vast number of technical defaults being computed as true defaults thus distorting the meaning of the internal risk model results mainly because:
Technical defaults add to the number of defaults used for Probability of Default (PD) purposes in internal models as if they were true and full defaults;
Technical defaults are usually resolved in a short time and give rise to recoveries thus distorting as well the meaning of the Loss Given Default (LGD) parameter.
In addition, the use of option 1 implies that the mere breach of an absolute threshold would trigger the default mark with no regard to the relative materiality. In consequence, it would lead to the following problems:
Differences across countries and markets (e.g. in relative prices) would be disregarded;
FX volatility would compound the cost of computation for debts denominated in currencies other than Euro notably in EU banking group branches and subsidiaries in third countries. It seems to make little economic sense that a bank has to readjust models and parameters only due to a breach of an absolute threshold after FX fluctuations if the defaulted amount in local currency remains unchanged.
The relative threshold should be placed, in the EBF opinion, at a level of 4% instead of the 2% proposed in the draft EBA paper. We note that there are currently varying levels of relative threshold in different countries up to 5%; therefore a 4% threshold could be an intermediate level.
Also importantly, the following aspects of the calculation should be considered:
The compensation of overdue amounts with unused general credit lines for the same debtor;
The count of past due days could start only after the materiality threshold is exceeded;
The amount of reference should be the actual past due balance.
We note that a 2-year period is proposed in the RTS but institutions would have to collect 5 years of data for the calculations of risk parameters with the new thresholds. Therefore, the period proposed is not considered a feasible scenario in practice.
We also note that changes in the definition of default is one of the measures that requires competent authorities' approval according to the Commission’s Delegated Regulation . For the sake of clarity, the EBA Regulatory Technical Standards (RTS) should specify that changes due to regulatory amendments do not lead to the abovementioned approval process. But if the final RTS requires changes to be signed-off by competent authorities then the time for supervisory review and approval should be factored into the implementation period, taking into account that the supervisory review could take almost one year in some jurisdictions.
It is also important to consider that the Basel Committee is known to be preparing a review of the IRB approach for credit risk. Should the EBA’s RTS be adopted prior to the Basel Committee proposals, EBF members believe it is crucial that the European approach is recognised as being compatible within the BCBS framework given its potentially far-reaching implications for banks' internal models and data series.
In general, the adjustments to datasets, the re-calibration of risk parameters and the implementation of model changes will mean a significant burden for banks. These costs would be exacerbated if the change is intended to be applied retrospectively and, in many cases, it could be unfeasible due to unavailability of data.
Against this background, the most cost-efficient solution is to grant a sufficiently long transition period during which legacy data bases could still be used.
Option 2 would be more affordable as it is supposed to produce less changes to existing data.
The overall impact of these changes will be considerably high. However it can be reduced as long as the following points are observed. There are two main sources of cost:
(i) Data systems and IT adaptations as well as human resources devoted to the change:
A streamlined process for supervisory model approval for changes due to the adaptation of the materiality threshold, or the absence thereof, would partially alleviate the burden.
(ii) Variations in Risk Weighted Assets (RWA):
Option 1 would leave more cases of true technical default out of scope. This could potentially lead to variations in risk weighted assets which impact is not easy to estimate with precision. Whereas option 2 captures a wider range of true technical defaults therefore it is more accurate for modelling purposes.