International university rankings: composite indicators and methodological approaches

This article explores the methodological landscape of international university rankings by reconstructing and formalizing the computational pipelines of five widely used global systems (ARWU-ShanghaiRanking, Times Higher Education (THE), U.S. News Best Global Universities, QS World University Rankings, CWUR) and reviews mathematically principled alternatives to fixed-weight composite rank tables. For the major ranking systems, the study details indicator design, normalization and transformation mechanisms and weighted aggregation into final composite scores, highlighting how these design choices encode normative assumptions and drive cross-system discrepancies. Responding to critiques of single-score rank tables, the paper reviews non-aggregative and minimally aggregative paradigms that preserve multidimensionality and reduce misleading ordinal precision, including dominance-based methods (Pareto front, skyline and discriminative skyline ranking), categorical layering (Rainbow Ranking and RR-index), dashboard and group-based multidimensional systems (U-Multirank, CHE-style reporting, and CWTS Leiden Ranking), frontier efficiency models (DEA/FDH), and data-driven latent-dimension approaches (PCA, factor and Bayesian latent-trait models). The article also situates aggregated-ranking voting rules and weighted voting power indices as complementary formal tools for preference aggregation and influence analysis. Overall, it argues that global rankings should be interpreted as algorithmic evaluative frameworks rather than neutral measurements, and that dominance-, profile-, frontier-, and latent-variable approaches offer transparent and conceptually robust complements for multidimensional university evaluation. We aim to identify ranking methods that operate without human-chosen weights or subjective parameter choices, because such choices influence and often distort the resulting rankings.