EUROPEAN
JOURNAL
OF OPERATIONAL
RESEARCH
ELSEVIER European Journal of Operational Research 85 (1995) 625-635 I
Theory and Methodology
Average shadow price in a Mixed Integer Linear
programming problem
Alejandro Crema
Escuela de Computacidn, Facultad de Ciencias, Universidad Central de Venezuela, Apartado 47002, Caracas 1041-A, Venezuela
Received April 1993; revised July 1993
Abstract
Recently a new concept of shadow price, the average shadow price, based on the average and not on the marginal
contribution of a resource, has been developed for pure Integer Linear Programming problems. In this paper we
prove that average shadow prices can be used in a Mixed Integer Linear Programming problems and that some of its
properties are analogous with the properties of shadow prices in a Linear Programming problem. A natural link with
the right-hand-side parametric problem leads us to a proof of completeness for parametric analysis.
Keywords: Shadow prices; Langrangean relaxation; Integer Programming
1. Introduction
The concept of shadow price has been studied by many researches in Mathematical Programming and
Operations Research because of its economical interpretation and its relation with duality theory. In
Linear and Convex Programming the concept has been well studied (Akgull [1], Aucamp andi Steinberg
[3], Gal [4], Shapiro [10, pp. 36-38] and many others). All these studies are based on the Concept of
marginal contribution of a resource (or a 'package of resources') to the optimal objective value. The
marginal analysis may not be useful in (pure or mixed) Integer Linear Programming because the
objective function is neither concave nor convex when the availability of one or more resources changes.
In 1988, Kim and Cho [9] analyzed a new concept, which they called average shadow price (a.s.p), with
rich economical interpretation in pure Integer Linear Programming problems. The analysis of Kim and
Cho is based on the average and not on the marginal contribution of a resource. Previous work about
shadow prices in Integer Linear Programming have been published by Gomory and Baumol [6] and
Alcaly and Klevorick [2].
In this paper we prove that a.s.p can be used in Mixed Integer Linear Programming problems and that
some of its properties are analogous with the properties of the shadow prices in a Linear PrOgramming
problem. In Section 2 we review the concept and present a procedure that finds the exact value of the
a.s.p, by solving a finite sequence of Mixed Integer Linear Programming problems. Section 3 presents
some relations between a.s.p and optimal lagrangean multipliers (Geoffrion [5]). Section 4 presents a
procedure to obtain the net profit function. In Section 5 we establish a natural link with the
right-hand-side (r.h.s) parametric problem (see the paper of Jenkins [7] for a survey of parametric
problems in Integer Linear Programming). Concluding comments are given in Section 6.
0377-2217/95/$09.50 © 1995 Elsevier Science B.V. All rights reserved
SSD1 0377-2217(94)00003-U