Reviewing model reaction to COVID-19 and assessing the impact of volatility in 2021
The views and opinions expressed in this article are those of the thought leader and not those of CeFPro.
By Mohit Dhillion, Managing Director, Quantitative Analytics, Barclays
Can you provide an overview of your experience in how models reacted to Covid-19 and the severity of volatility?
Many credit risk models used for impairment, planning and stress testing produced unintuitive estimates during COVID-19. This is primarily due to the fact that historical economic data used to calibrate models do not include rates and levels of volatility as observed with recent forecasts.
The credit loss rates produced monthly or quarterly over the last year have been volatile given the sharp acceleration in economic variables as well as the sharp forecasted recovery. Many models use transformations including ‘percentage change over the last year’ causing the values fed into the model to be extremely high.
Models assume linear relationships between losses and economic variables however very few are calibrated on multiple credit cycles and none are calibrated over periods where the impact of economic stress is offset by government support.