African Journal of Business Management Vol. 6(1), pp. 459-473,11 January, 2012
Available online at http://www.academicjournals.org/AJBM
DOI: 10.5897/AJBM11.060
ISSN 1993-8233 ©2012 Academic Journals
Full Length Research Paper
Comparison of the REMBRANDT system with the Wang
and Elhag approach: A practical example of the rank
reversal problem
Hamed Maleki
1
* and Amir Hassan Zadeh
2
1
Department of Industrial Engineering, Islamic Azad University, South Tehran Branch, Tehran, Iran.
2
Department of Industrial Engineering, Amir Kabir University of Technology, Tehran, Iran.
Accepted 24 October, 2011
Analytic hierarchy process (AHP) has been considerably criticized for possible rank reversal
phenomenon caused by: i, when new alternatives are added or old ones deleted; and ii, when new
criteria are added or old ones deleted with the caveat that the priorities of alternatives would be tied
under these criteria and hence argued that the criteria should be irrelevant when ranking the
alternatives. While in many cases this is a perfectly valid phenomenon, there are also many cases
where rank should be preserved. This paper deals with rank reversal due to the inconsistency of the
inputs. The preference intensities on REMBRANDT scale are more compatible than AHP scale. The
REMBRANDT system is therefore proposed to avoid rank reversal phenomenon. We have provided a
practical example to show that the rank reversal phenomenon did not occur with the REMBRANDT
system, but did occur with the Wang and Elhag approach.
Key words: Analytic hierarchy process (AHP), REMBRANDT system, rank reversal, multiple-attribute decision
making (MADM), uncertainty.
INTRODUCTION
Analytic hierarchy process (AHP), as a very popular
multiple criteria decision making (MCDM) tool, has been
considerably criticized for its possible rank reversal
phenomenon, which means changes of the relative
rankings of the other alternatives after an alternative is
added or deleted. Such a phenomenon was first noticed
and pointed out by Belton and Gear (1983), which leads
to a long-lasting debate about the validity of AHP (Dyer,
1990a, b; Harker and Vargas, 1987; Leung and Cao,
2001; Perez, 1995; Saaty, 1990a; Saaty, 1986,
1990,1994; Saaty et al., 1983; Schoner et al., 1989;
Stewart, 1992; Troutt, 1988; Vargas, 1994; Wang and
Elhag, 2006; Watson and Freeling, 1982, 1983),
especially about the legitimacy of rank reversal (Forman,
1990; Millet and Saaty, 2000; Saaty, 1987a,b; Saaty and
Vargas, 1984; Schoner et al.,1992).
*Corresponding author. E-mail: st_h_maleki@azad.ac.ir.
In order to avoid the rank reversal, Belton and Gear
(1983) suggested normalizing the eigenvector weights of
alternatives using their maximum rather than their sum,
which was usually called B–G modified AHP. Saaty and
Vargas (1984) provided a counterexample to show that
B– G modified AHP was also subject to rank reversal.
Belton and Gear (1985) argued that their procedure was
misunderstood and insisted that their approach would not
result in any rank reversal if criteria weights were
changed accordingly.
Schoner and Wedley (1989) presented a referenced
AHP to avoid rank reversal phenomenon, which requires
the modification of criteria weights when an alternative is
added or deleted.
Schoner et al. (1993) also suggested a method of
normalization to the minimum and a linking pin AHP [see
also (Schoner et al., 1997)], in which one of the
alternatives under each criterion is chosen as the link for
criteria comparisons and the values in the linking cells
are assigned a value of one, with proportional values in