We estimate the effect of sleep on labor productivity addressing the two main challenges in time use research: the unavoidable substitutions among activities implied by the time budget constraint and the endogeneity of the allocation of time. We use complete time diary data to identify the relative effect of sleep vs. non-work activities among employees working the same number of hours, and account for the endogeneity of time use choices by leveraging longitudinal information on productivity in a value-added specification. We show that, when work hours are held constant, substituting sleep with other non-work activities does not affect labor productivity.