How co-authorship affects the H-index?

Abstract

H-Index is a widely used metric for measuring scientific output. In this paper we showcase the weakness of this index as regards co-authorship. By ignoring the number of co-authors, each author gets the full credit of a joint work, something that is not fair for evaluation purposes. For this purpose we report the results of simulation scenarios that demonstrate the impact that co-authorship can have. To tackle this weakness, and achieve a more fair evaluation, we propose a few simple variations of H-index that consider the number of co-authors, as well as the active time period of a researcher. In particular we propose using HI/co and HI/(coy), two metrics that are simple to understand and compute, and thus they are convenient for decision making. The simulation shows that they can tackle well co-authorship. Subsequently we report measurements over real data of researchers coming from five universities (Cambridge, Crete, Harvard, Oxford and Ziauddin), as well as other datasets, that reveal big variations in the average number of co-authors. In total, we analyzed 526 authors, having in total more than 127 thousands publications, and 16.7 million citations. These measurements revealed big variations of the number of co-authors. Consequently, by including the number of co-authors in the measures for scientific output (e.g. through the proposed HI/co) we get rankings that differ significantly from the rankings obtained by citations, or by the plain H-Index. The normalized Kendall’s tau distance of these rankings ranged from 0.28 to 0.46, which is quite high.