PD Model Calibration Post COVID Pandemic: Balancing Representativeness of Current Portfolio and Likely Range of DR Variability

Abstract

The COVID-19 pandemic placed many challenges to everyone in terms of wellbeing and economic activities. As for banking and finance, while many rely on re-calibration of probability of default models to adapt own portfolio to the latest reality, it is worthwhile to bring to reader's attention that common practice of model calibration in the industry struggles to meeting regulatory requirements, particularly at the time of this paper when players in the filed is about ready to conclude the 2020 annual observation of portfolio default rate and potentially facing an even tougher forthcoming market condition.

In this paper, the observed gap is first illustrated and discussed in detail with a numerical example. Next, we propose a novel methodology for model calibration where the specified gap is addressed. Lastly, methodological properties shown with numerical results encourage the adoption of the proposed approach where pandemic impact is sought in consideration of regulatory compliance.