Source-linked AI summary
The decline in the concentration of citations, 1900-2007
Vincent Lariviere, Yves Gingras, Eric Archambault
TL;DR
The paper tests whether citation concentration has increased, as Evans (2008) reported, and addresses the limited long-term evidence on citedness and concentration. It analyzes Web of Science citations with three concentration measures across broad disciplines and finds that citations became increasingly dispersed rather than more concentrated. The authors conclude that Evans’s aggregate-level pattern does not hold in the biomedical, natural-science and engineering, or social-science fields examined.
Problem
Long-term evidence was lacking on changes in cited and uncited papers, amid conflicting claims about whether online availability concentrates citations on newer, fewer papers.
Method
The study analyzes Web of Science citations from 1900–2007 using cited-paper proportions, citation-share thresholds, and the Herfindahl-Hirschman index.
Results
Citations became increasingly dispersed across papers: cited-paper shares and the percentages needed to account for citation shares increased, while the HHI decreased.
Takeaways & Limitations
At the aggregate level, citations were not increasingly limited to fewer papers or journals in the biomedical, natural-science and engineering, or social-science fields.
Takeaways & Limitations
Because the data do not include online availability, the study cannot directly prove that online access is unrelated to increased citation concentration or drives the observed trends.
Abstract
from arXiv · showhide
This paper challenges recent research (Evans, 2008) reporting that the concentration of cited scientific literature increases with the online availability of articles and journals. Using Thomson Reuters' Web of Science, the present paper analyses changes in the concentration of citations received (two- and five-year citation windows) by papers published between 1900 and 2005. Three measures of concentration are used: the percentage of papers that received at least one citation (cited papers); the percentage of papers needed to account for 20, 50 and 80 percent of the citations; and, the Herfindahl-Hirschman index. These measures are used for four broad disciplines: natural sciences and engineering, medical fields, social sciences, and the humanities. All these measures converge and show that, contrary to what was reported by Evans, the dispersion of citations is actually increasing.
Introduction
The paper addresses conflicting claims about whether online access concentrates citations on newer, fewer papers. It responds with a long-term analysis designed to measure changes in citedness and citation concentration.
- Introduction: Evans (2008) reported that greater online availability made cited literature younger and concentrated citations on fewer papers.The claim challenged prevailing assumptions about how online access affects citation patterns.
- Introduction: Empirical studies instead found that researchers cite an increasingly older body of scientific literature.This observation was supported by both theoretical work and studies of researchers’ use patterns.
- Introduction: The belief that most scientific articles are never cited has been repeatedly challenged by empirical studies.The paper situates citation concentration within a longstanding debate about uncited scientific literature.
- Introduction: No previous study had measured long-term changes in the proportion of cited and uncited articles.Earlier evidence for the 1980s showed no clear trend in uncitedness.
- Introduction: Analyzing citations from 1900–2007, the paper finds that uncited papers and citation concentration decreased rather than increased.This directly contradicts the concentration trend reported by Evans.
Methods
The study measures citation concentration using three complementary indicators applied to Web of Science data across broad disciplines. Its dataset matches citations to eligible publications while excluding first-author self-citations.
- Methods: The study measures concentration using cited-paper percentages, citation-share thresholds, and the Herfindahl-Hirschman index.These indicators capture whether citations are distributed across more papers or concentrated among fewer papers.
- Methods: The cited-paper measure is the percentage of papers receiving at least one citation within two- and five-year windows.A higher cited-paper proportion indicates citations are dispersed across more published papers.
- Methods: The second indicator is the percentage of papers required to account for 20%, 50%, and 80% of citations received.A higher percentage needed for each citation share indicates decreasing concentration; uncited papers remain included.
- Methods: The HHI sums squared citation market shares, with higher values indicating greater concentration.For citations, each paper’s market share is its citations divided by the total citations received by papers published in the same year; uncited papers are excluded.
- Methods: Web of Science data cover 1900–2007 and classify journals into natural sciences and engineering, medical fields, social sciences, and humanities.The source combines SCIE, SSCI, and AHCI, with humanities classifications supplemented in-house.
- Methods: Citation matching used reference identifiers and bibliographic fields, included articles, notes, and reviews, and excluded first-author self-citations.The study retrieved citations to more than 27 million papers across the four broad categories.
Results
Across natural sciences and engineering, medical fields, and social sciences, citations became less concentrated over time, while humanities remained highly skewed and largely trendless. All three concentration measures support this pattern, contradicting Evans’s reported increase in citation concentration.
- By 2005, citations reached 80% of medical papers, 60% of natural-science and engineering papers, and 55% of social-science papers within two years.Using five-year windows, only 12% of medical, 27% of natural-science and engineering, and 32% of social-science papers remained uncited in 2002.
- 80% of citations required 33% of medical papers and 28% of natural-science and engineering papers in 2005, versus 24% and 23% in 1990.
- 28% of social-science papers accounted for 80% of citations in 2005, compared with 19% in 1990 and 14% in 1956.
- Humanities citations remained extremely concentrated: 0.5% of papers accounted for 20%, 2.6% for 50%, and 7.2% for 80% of citations in 2005.The humanities trend was essentially flat, aside from a small data-quality-related bump.
- Overall, citation dispersion increased in natural sciences and engineering, medical fields, and social sciences, contrary to Evans’s finding that citations became more concentrated.Medical and social-science papers published in 2005 had their lowest concentration in history.
- The Herfindahl-Hirschman index also decreased over the study period in all fields except the humanities, apart from wartime and brief late-1980s reversals.Medical and natural-science and engineering citations became more concentrated during the two World Wars, when fewer papers were published.
Discussion and Conclusion
At the macro level, multiple measures show that citation concentration has decreased in natural sciences and engineering, medical fields, and social sciences. The paper therefore rejects Evans’ aggregate-level picture of increasingly concentrated citation patterns, while acknowledging that its data do not directly test online availability as a causal factor.
- Citation concentration has decreased in natural sciences and engineering, medical fields, and social sciences.
- The percentage of papers receiving at least one citation has increased since the 1970s.
- The percentage of papers needed to account for 20%, 50% and 80% of citations has increased.
- The Herfindahl-Hirschman index has steadily decreased since the beginning of the twentieth century.
- These measures indicate increasingly dispersed citations rather than an increasingly exclusive and elitist scientific system.
- Because online availability was not measured, the study cannot directly prove that electronic publishing drives the observed trends.