RISK MODELLING AND GOVERNANCE IN THE POST-COVID-19 WORLD

R&C: To what extent are existing risk models still relevant in the post-COVID-19 era?

Breed: When coronavirus (COVID-19) first became a reality, with lockdowns implemented across the world, some of the more extreme reactions in banks were that existing provisioning models should be abandoned and that they need to be replaced completely. This was especially true for models that use macroeconomic variables to incorporate forward-looking information. Although it is true that correlation assumptions were weakened and, in some cases, non-intuitive during the initial parts of the pandemic, the stimulus provided by governments did have a more pronounced effect on some consumers’ ability to meet payments than initially expected. We expect that original model assumptions will remain relevant once the economic impacts of COVID-19 subside. Some banks are therefore choosing not to disregard pre-pandemic models entirely. Instead, they are relying on more judgmental model overrides and adjustments during the COVID-19 period, with a gradual increase in reliance on models once things return to normal.

Zihan: Models basically are built on past data with the assumption that they are able to predict the future. This assumption can only hold true if the underlying circumstances remain more or less unchanged. We all can agree that the COVID-19 era over past year was an extremely unusual situation. To determine whether the model remains fit-for-use during and after the pandemic, each model needs to be diligently scrutinised. Even after the pandemic is over, the main challenge from a modelling perspective is a lack of data to rebuild the models. How do we treat the pandemic data? How long do we have to wait to see the true effect? As such, it may take some time to redevelop reliable models under the new operating environment post-COVID-19. In the meantime, most existing models will continue to be used with additional layers of judgmental overlay and subject to ongoing close monitoring.

Jul-Sep 2021 Issue

SAS Institute Inc.

Maybank

Absa Group