We examine monthly variation in weekly work hours using data for 2003-10 from the Current Population Survey (CPS) on hours/worker, from the Current Employment Survey (CES) on hours/job, and from the American Time Use Survey (ATUS) on both. The ATUS data minimize recall difficulties and constrain hours of work to accord with total available time. The ATUS hours/worker are less cyclical than the CPS series, but the hours/job are more cyclical than the CES series. We present alternative estimates of productivity based on ATUS data and find that it is more pro-cyclical than other productivity measures.