This paper compares models used to explain OECD unemployment. The models suggest that the "natural rate of unemployment" has been driven up mainly by wage push factors. Panel data on twenty-two OECD countries are used to investigate the explanatory power of these models over the past two decades. Our estimates reveal that coefficients on key variables often turn out with signs which are at odds with the theories or are insignificant and that a second order autoregressive model performs nearly as well as all the other models. The conclusion offers some directions for future research.