AVERAGE SHADOW PRICES IN MATHEMATICAL-PROGRAMMING

Abstract

We develop the concept of average shadow price in mathematical programming. This concept measures the value of resources along a direction in an average sense, in contrast to traditional marginal analysis; it serves as a useful standard price for management decisions about resources, particularly when there are nonconvexities. We give it an economic interpretation. We also develop simple computational schemes for obtaining and improving the bounds of the average shadow prices and illustrate them in two important classes of nonconvex programs: convex maximization problems and mixed integer programs.