Estimates of the Nairu generally suer from a large uncertainty, which can be reduced by adopting a bivariate framework and assuming that shifts of the Phillips curve share a common trend with the unemployment rate. We consider in this paper if this common trend assumption is empirically relevant or not for seven economies over the sample 1973-2010. First, it appears that the Nairu can substantially dier from the unemployment trend. Second, relaxing the common trend assumption improves the t of the ination equation. Third, this assumption is necessary for