We propose a theory of low-frequency movements in unemployment based on asymmetric real wage rigidities. The theory generates two main predictions: long-run unemployment increases with (i) a fall in long-run productivity growth and (ii) a rise in the variance of productivity growth. Evidence based on U.S. time series and on an international panel strongly supports these predictions. The empirical specifications featuring the variance of productivity growth can account for two U.S. episodes which a linear model based only on long-run productivity growth cannot fully explain. These are the decline in long-run unemployment over the 1980s and its rise during the late 2000s.