Is the best–worst method path dependent? Evidence from an empirical study

Abstract The Best–Worst method (BWM) is one of the latest contributions to pairwise comparisons methods. As its name suggests, it is based on pairwise comparisons of all criteria (or possibly other objects, such as alternatives, sub-criteria, etc.) with respect to the best (most important) and the worst (least important) criterion. The main aim of this study is to investigate the path and scale dependency of the BWM. Up to now, it is unknown whether the weights of compared objects obtained by the method differ when the objects are compared first with the best object, and then with the worst, or vice versa. It is also unknown if the outcomes of the method differ when compared objects are presented in a different order, or when different scales are applied. Therefore, an experiment in a laboratory setting is performed with more than 800 respondents university undergraduates from two countries in which the respondents compare areas of randomly generated figures and the relative size of objects is then estimated via the linearized version of the BWM. Last but not least, the accuracy of the BWM is examined with respect to different comparison scales, including Saaty’s nine-point linguistic scale, an integer scale from 1 to 9, and a continuous scale from 1 to infinity.