This paper investigates the source of historical fluctuations in annual US data extending back to the late 19th century. Long-run identifying restrictions are used to decompose productivity, hours, and output into technology shocks and non-technology shocks. A variety of models with differing...