University leagues alongside rankings

Abstract

University rankings can both orient and disorient potential students. In rankings, universities with very different characteristics are compared, which makes interpretation difficult. We propose the application of a clustering method, which creates groups of universities that are close to each other with respect to a subset of indicators, but the indicators also show homogeneity with respect to the universities in that group. We call such groups leagues. These leagues are defined by the data themselves and are not based on subjective criteria. We demonstrate our proposition using one member of the family of the two-way clustering method, namely, biclustering. The case we present is based on the Round University Ranking (RUR) 2020 dataset. The use of leagues could provide better guidance not only for potential applicants but also for university funding organizations and policy-makers. Our case study led to a somewhat surprising observation. In the top league (based on the RUR data and indicators), the three most important indicators measure reputation, not scientific or educational performance.