A general framework for distance-based consensus in ordinal ranking models

The problem of aggregating a set of ordinal rankings of n alternatives has given rise to a number of consensus models. Among the most common of these models are those due to Borda and Kendall, which amount to using average ranks, and the l(1) and l(2) distance models. A common criticism of these approaches is their use of ordinal rank position numbers directly as the values of being ranked at those levels. This paper presents a general framework for associating value or worth with ordinal ranks, and develops models for deriving a consensus based on this framework. It is shown that the l(p) distance models using this framework are equivalent to the conventional ordinal models for any p greater than or equal to 1. This observation can be seen as a form of validation of the practice of using ordinal data in a manner for which it was presumably not designed. In particular, it establishes the robustness of the simple Borda, Kendall and median ranking models.