With vaccinations the demand in certain services is likely to rise. This rise causes a form of non-stationarity in the demand data of these services.
My assumptions are that demand has a period on the COVID level, then a period of recovery, and then a stabilisation on a post COVID level.
The concept of level shift could help to model the downward trend at COVID onset, assuming a sharp decline. But would that also help during the recovery fase where the new level is not known, and the level is continuously increasing?
Would ARIMA be a way to model this data? Could the same ARIMA model work during all three periods? Are there alternative models to handle this kind of non-stationarity?