A high-resolution three-dimensional chemistry transport model is used to evaluate the contributions of different mechanisms (in situ chemistry, import of chemically activated or ozone-depleted polar air) to the extravortex chemical ozone loss during the winter-spring period and to show how their efficiencies might have changed in the last 2 decades. Two extreme, but somewhat typical, present-day winters are considered: the 1999--2000 winter (cold with a stable and persistent vortex) and the 2001--2002 winter (warm with an weak and distorted vortex). An ozone budget analysis is performed for the partial ozone column between 350 and 600 K using a combination of geographical and chemical ozone tracers. The results suggest that the interannual variability in the seasonal ozone loss at midlatitudes is mainly driven by the import of ozone-depleted polar air. The contribution from the import of chemically activated polar air is not found to be significant. The magnitudes of the polar contributions (entirely driven by the halogen cycles) differ by about a factor 4 between the simulations. The polar contribution is found to be responsible for 60% of the total extravortex ozone loss for the exceptional 1999--2000 Arctic winter. In contrast, the magnitude of the in situ destruction (dominated by HOx and NOx cycles) varies little from one winter to the other one. The evolution of the atmospheric chemical composition since 1980 may account for an additional midlatitude ozone loss of 3 to 10 Dobson units depending on the winter considered (representing up to about 40% of the observed midlatitude ozone trends). The polar contribution is found to increase by a factor 2 to 3 in the last 2 decades, whereas the in situ destruction increases by about 20% only.