This paper investigates the ability of several generalized Bayesian vector autoregressions to cope with the extreme COVID-19 observations and discusses their impact on prior calibration for inference and forecasting purposes. It shows that the preferred model interprets the pandemic episode as a...