To determine an optimum economic policy one needs: (a) a welfare function valuing "target" variables; (b) a model, describing the effect of policy "instruments" on the targets; and (c) limits within which the variables are allowed to vary. The welfare function is derived by "imaginary interviewing of the policy-makers." It is linearized in intervals or "facets." Linear programming indicates the optimum policy within each facet of the welfare function. A policy optimal with respect to all facets around it is the absolute optimum. By way of an appendix, a survey is given of the Multiplex Method of programming, developed by Professor R. Frisch, and particularly suited to this type of analysis.