Climate models that do not simulate changes in stratospheric ozone concentrations require ozone input fields to accurately calculate UV fluxes and stratospheric heating rates. In this study, three different global ozone time series that are available for this purpose are compared: the data set of Randel and Wu (2007) (RW07), Cionni et al. (2011) (SPARC), and Bodeker et al. (2012) (BDBP). The latter is a very recent data set, based on the comprehensive ozone measurement database described by Hassler et al. (2008). All three data sets represent multiple-linear regression fits to vertically resolved ozone observations, resulting in a patially and temporally continuous stratospheric ozone field covering at least the period from 1979 to 2005. The main difference between the data sets result from using different observations and including different basis functions for the regression model fits. These three regression-based data sets are compared against observations from ozonesondes and satellites to compare how the data sets represent concentrations, trends, and interannual variability. In the Southern Hemisphere polar region, RW07 and SPARC underestimate the ozone depletion in spring as seen in ozonesonde measurements. A piecewise linear trend regression is performed to estimate the 1979–1996 ozone decrease globally, covering a period of extreme depletion in most regions. BDBP seems to overestimate Arctic and tropical ozone loss over this period somewhat relative to the available measurements, whereas these appear to be underestimated in RW07 and SPARC. In most regions, the three data sets yield ozone values that are within the range of the different observations that serve as input to the regressions. However, the differences among the three suggest that there are large uncertainties in ozone trends. These result in differences of almost a factor of four in radiative forcing, which is important for the resulting climate changes.