• Czech Banking Association strongly disagrees with the Option 1 proposed in the draft RTS, i.e. that default should be recognized as soon as one of the components of the threshold - absolute or relative limit - is breached.
• The application of the relative limit (2% of exposure) is problematic for small exposures as even a small technical overdraft will lead to a client’s default. In fact, for these cases, this rule does not set any practical materiality threshold.
• Czech Banking Association supports either the alternative option, i.e. recognition of default after both thresholds are breached (Option 2) or setting the absolute thresholds only (in the same values as proposed in the draft RTS).
• We suggest to increase the maximum level of the threshold for non-retail exposures to EUR 1000 (art. 2 par 3).
• In case the Option1 (from the Q1) is chosen, it will be necessary to recalibrate all PD, LGD models.
• This proposal effectively requires banks to re-simulate default events to the past as well as simulation of impact on risk parameters (PD,LGD). For default events recognized under current rules but no longer to be recognized under new rules the simulation is not much complicated. But much more complexity is to be used for default events not recognized under current rules, but to be recognized under new rules, especially when it comes to LGD / CCF estimation
o Most banks use workout approach to determine LGD – how could the LGD be estimated to accounts which were not subject to collection (as were not considered defaulted)?
o The issue is even more tricky if option 1 for default recognition (=either relative or absolute threshold fulfilled) is chosen and proposal to ban weighting approaches for LGD are simultaneously adopted. We are concerned that the technical defaults would be crucial driver for both PD and LGD estimation as
• For this reasons (technical impossibility to recalculate impacts to risk parameters precisely) and regulatory requested length of time series for risk parameters estimation (minimum of 5years) we propose postponement of effectiveness of this request by at least 3-5 years.
• In case the Option1 (from the Q1) is chosen, impacts will be quite important as relative threshold will change the existing dataset used for the development of PD and LGD models