Using data from the Current Population Survey from 1980 through 2010 we examine what drives variation and cyclicality in the growth rate of real wages over time. We employ a novel decomposition technique that allows us to divide the time series for median weekly earnings growth into the part associated with the wage growth of persons employed at the beginning and end of the period (the wage growth effect) and the part associated with changes in the composition of earners (the composition effect). The relative importance of these two effects varies widely over the business cycle. When the labor market is tight job switchers get high wage increases, making them account for half of the variation in median weekly earnings growth over our sample. Their wage growth, as well as that of job-stayers, is procyclical. During labor market downturns, this procyclicality is largely offset by the change in the composition of the workforce, leading aggregate real wages to be almost noncyclical.Most of this composition effect works through the part-time employment margin. Remarkably, the unemployment margin neither accounts for much of the variation nor for much of the cyclicality of median weekly earnings growth.