The sensitivity of the analytic hierarchy process to alternative scale and cue presentations

The results offer some evidence that the scale used or the format order of paired comparisons yields significant differences in the AHP models, but the results for the scale and format effects are not evident across all experiments.

Abstract

The Analytic Hierarchy Process (AHP) has gained prominence in the accounting literature as a method to model the decisions of experts. Also, the AHP is used to help individuals structure decisions. AHP researchers have several choices when constructing AHP instruments that elicit judgments from participants; however, little guidance is available regarding the ‘best’ choice. In particular, the AHP response scale can be numerical, verbal, or graphical. Paired comparisons can be presented in a random or nonrandom format, or in a top-down or bottom-up order. This paper reports the results of three related experiments investigating whether differences in the scale used or the format order of paired comparisons yields significant differences in the AHP models. The results offer some evidence that the scale used is associated with different AHP models. Also, some evidence is provided that the random versus nonrandom format of the paired-comparison presentation is associated with different AHP models. However, the results for the scale and format effects are not evident across all experiments.