The Effect of Forecasting Errors on Optimal Programming

Abstract

A programming method proposed by one of the authors is applied to a problem of antidepression policy. The method is applicable whenever the problem of a policy-maker can be formulated as that of maximizing a quadratic welfare or objective function subject to linear restrictions. A distinction is made between instrument variables which are controlled by the policy-maker and noncontrolled variables which form the subject of the policy-maker's preferences and can be influenced indirectly by means of the instruments.

Attention is paid to the problem of uncertainty arising from the fact that the constraints of the problem are only known imperfectly by the time the policy-maker has to take his decisions, and so have to be predicted. The “welfare” effect of errors in forecasting is studied in detail for the case of a government policy-maker whose restrictions are derived from an econometric model of his economy. The economy chosen for the application is that of the United States in the period 1933–1936. The object of the policy is supposed to be to end the depression. The uncertainty is thought of as restricted to the disturbances of the structural equations of the model.

This article first appeared in C.W. Churchman and M. Verhulst, eds., Management Sciences, Models and Techniques, 1 (1960), Pergamon Press PLC, Oxford, U.K., 295-322. Reprinted with the permission of Pergamon Press PLC.