The computation of large-scale nonlinear intertemporal optimization problems requires time aggregation. A procedure generally adopted is shown to introduce a dependency of the solution steady state to a specific choice of sequence of time intervals. The authors establish necessary and sufficient conditions to avoid this dependency. The result is a considerable improvement in the numerical accuracy of the time-aggregated approximation. The conditions apply to a broad class of models and prove useful in large-scale applied general-equilibrium modeling. This conclusion is highlighted by a comparison with a spectral-projection method using optimal orthogonal collocation as in K. L. Judd (1992). Copyright 1994 by The Econometric Society.