Correlating article citedness and journal impact: an empirical investigation by field on a large-scale dataset

In spite of previous research demonstrating the risks involved, and counsel against the practice as early as 1997, some research evaluations continue to use journal impact alone as a surrogate of the number of citations of hosted articles to assess the latter’s impact. Such usage is also taken up by research administrators and policy-makers, with very serious implications. The aim of this work is to investigate the correlation between the citedness of a publication and the impact of the host journal. We extend the analyses of previous literature to all STEM fields. Then we also aim to assess whether this correlation varies across fields and is stronger for highly cited authors than for lowly cited ones. Our dataset consists of a total of almost one million authorships of 2010–2019 publications authored by about 28,000 professors in 230 research fields. Results show a low correlation between the two indicators, more so for lowly cited authors as compared to highly cited ones, although differences occur across fields.