Bibliometric performance and uncertainty in heterogeneous research groups: trends in authorship and fractional citation indicators

Abstract

Abstract Bibliometric indicators are widely used to evaluate scientific performance, despite well-known limitations related to disciplinary differences, collaboration practices, and strategic responses to evaluation systems. Traditional metrics such as the H-index are particularly sensitive to the number of co-authors and to the length of reference lists, both of which have exhibited strong historical growth. This study analyzes the temporal evolution of authorship, referencing, and productivity patterns in six research populations (four Italian Scientific Disciplinary Sectors in engineering and two international conference communities), covering 1,286 researchers and 86,351 publications. The results reveal systematic increases in authors per article, references per article, and documents per author, with structural changes that coincide with technological developments and the introduction of bibliometric criteria in national evaluation policies. Building on fractional counting, we propose a new citation-based indicator in which publication credit is shared among co-authors and each incoming citation is weighted by the inverse of the number of references in the citing article. Comparisons with the H-index show strong overall correlations, but indicate that the proposed indicator attenuates the advantage associated with hyper-authorship and inflated reference lists and reduces variability across heterogeneous research groups. Adopting a measurement-theoretic perspective, we further assess the uncertainty of bibliometric indicators by analyzing linear trends with researcher seniority and the dispersion of individual data around these trends. The proposed indicator exhibits a slightly more homogeneous trend across heterogeneous research groups and shows strong overall correlation with the H-index, suggesting that it may provide a more comparable representation of scientific performance across diverse research communities.

The authors' abstract, as published at the source. Scientometrics, 2026 · DOI ↗