Structural vector autoregressions give conflicting results on the effects of technology shocks on hours. The results depend crucially on the assumed data generating process for hours per capita. We show that the standard measure of hours per capita and productivity have significant low-frequency movements that are the source of the conflicting results. Hodrick–Prescott (HP)-filtered hours per capita produce results consistent with those obtained when hours are assumed to have a unit root. We show that important sources of the low-frequency movements in the standard measure are sectoral shifts in hours and the changing age composition of the working-age population. When we control for these low-frequency components to determine the effect of technology shocks on hours using long-run restrictions we get one consistent answer: hours decline in the short run in response to a positive technology shock. We further extend the analysis by examining the effects of demographic controls on the impulse responses to investment-specific technology shocks. Our results are less conclusive.