The problem of scheduling production and distribution activities in a multi-product, multi-period, two-stage distribution network is formulated as a concave programming problem. Incorporated are several characteristics of real distribution systems such as economies of scale in production, transportation and handling, limitations in production and inventory capacities, direct shipments from plants to customers, and maintenance of stable production and handling work forces. The iterative algorithm developed for this problem sequentially represents the concave objective function with a linear approximation (extrapolation), continuously improving the objective function value by solving a number of linear programming subproblems. The quality of the obtained solutions is assessed using a statistical procedure, a lower bound, and a comparison with solutions obtained empirically. Computational experience on realistically sized problems is reported.