Single equation estimates of production models usually are justified by the assumption that production inputs are chosen as part of a one-period decision problem. Yet, most production decisions in agriculture are made sequentially. In this paper the farmer's optimal input choices are modeled as optimal controls in a stochastic control problem. A two-period Cobb-Douglas example is used to show that sequential solutions to production problems may yield models which require either single equation or simultaneous equation estimators. Functional separability, stochastic specification, and behavior under uncertainty are discussed in the context of dynamic production models.