News

EBA principles on the use of COVID-19 impacted data for internal credit risk models

On 21 July, 2022 the EBA published four draft principles on how to handle historical data affected by COVID-19. Impact can be direct, through moratoria or guarantees, or indirect when the input parameters of a rating model are affected by COVID-19 support measures such as employment support. As described in our white paper published in 2021, the COVID-19 pandemic affects trends in the historical data used in modelsraising questions about representativeness and requiring action by credit risk modelers. The EBA principles hint that the set of guidelines and principles will be part of the supervisory handbook that will be published by EBA later in 2022 and that these guidelines are of general interest, applicable for all credit risk models, including IFRS9. However, there is an increased attention to how financial institutions account for the potential impact of the COVID-19 crisis and the countering measures on the IRB risk parameters and related measures (PD, LGD, ELBE and CCF).

During recalibration of credit risk models, the following principles should be kept in mind ensuring representativeness of the IRB-relevant data affected by the crisis:

1. Based on EBA-GL-2017-16, data underlying risk quantification should not be excluded based on non-comparability, but should lead to Appropriate Adjustment (AA) and increased margin of conservatism (MoC) for IRB risk parameter estimates.
2. Risk drivers should be analyzed more thoroughly ifa significant decrease is observed in the average value of a relevant IRB risk parameter estimate compared to the pre-COVID-19 period (i.e. end 2019). However, if it can be shown that the observed decrease of the average IRB risk parameter estimate is due to improvements in the idiosyncratic risk of the obligors/exposures, no further action is required.
3. Recalibration of default and loss rates should be postponed in case the observed risk parameter levels during the COVID-19 crisis lack representativeness. Any recalibration should be postponed up until the data demonstrate with sufficient certainty that trends observed during the COVID-19 crisis are sustainable.
4. Like principle 3, a downward recalibration of the downturn LGD should be postponed until the impact of COVID-19 has fully materialized in the observed loss rates.

Read the full publication of the EBA here.