This article argues that to understand the behavior of productivity statistics, it is necessary to reexamine the basic assumptions underlying growth accounting. In particular, it offers theoretical and empirical support for the assertion that the elasticity of output with respect to an input like capital or labor might differ from the share of the input in total factor income. The theories offered in support of this possibility allow for spillovers of knowledge, specialization with monopolistic competition, and endogenous accumulation of labor-saving technological change. Evidence on the form of aggregate production is drawn from data for many countries and for long historical time periods. The specific interpretation of the productivity slowdown that is offered is that a low elasticity of output with respect to labor causes labor productivity growth rates to fall when labor growth speeds up.