[Paper Review] The Demise of Single-Authored Publications in Computer Science: A Citation Network Analysis

Brian K. Ryu|arXiv (Cornell University)|Jan 2, 2020

scientometrics and bibliometrics research28 references4 citations

TL;DR

This study analyzes the decline of single-authored publications in computer science using the DBLP citation network (1940–2019), finding that their share dropped from over 80% before 1950 to 7.4% by 2018, with citations and PageRank scores also declining significantly. Despite similar length and reference counts, single-author papers receive only 60% of the average citations of co-authored works, indicating systemic citation disadvantages beyond scope or volume.

ABSTRACT

In this study, I analyze the DBLP bibliographic database to study role of single author publications in the computer science literature between 1940 and 2019. I examine the demographics and reception by computing the population fraction, citation statistics, and PageRank scores of single author publications over the years. Both the population fraction and reception have been continuously declining since the 1940s. The overall decaying trend of single author publications is qualitatively consistent with those observed in other scientific disciplines, though the diminution is taking place several decades later than those in the natural sciences. Additionally, I analyze the scope and volume of single author publications, using page length and reference count as first-order approximations of the scope of publications. Although both metrics on average show positive correlations with citation count, single author papers show no significant difference in page or reference counts compared to the rest of the publications, suggesting that there exist other factors that impact the citations of single author publications.

Motivation & Objective

• To investigate the long-term trend in the proportion of single-authored publications in computer science from 1940 to 2019.
• To assess the citation impact and network significance of single-authored papers using citation counts and PageRank scores.
• To examine whether differences in paper scope—measured by page length and reference count—explain the declining citation performance of single-author works.
• To compare publication trends on arXiv versus peer-reviewed venues to understand the role of preprint sharing in single-author publication patterns.
• To determine whether the decline in single-author publications reflects broader shifts toward collaboration, interdisciplinary work, and changing researcher demographics.

Proposed method

• Utilized the full DBLP bibliographic database containing 4,107,340 publications and 36,624,464 citation edges, extracted on May 5, 2019.
• Calculated the population fraction of single-authored publications annually from 1940 to 2019 to track demographic trends.
• Computed mean citation counts and PageRank scores for single-authored versus co-authored publications to assess relative influence in the citation network.
• Used page length and reference count as proxies for paper scope and volume, comparing averages between single-author and multi-author works.
• Analyzed the fraction of single-authored papers uploaded to arXiv versus peer-reviewed journals and conferences to compare submission behaviors.
• Performed time-series analysis of arXiv and peer-reviewed venue data from 1992 to 2019 to assess trends in preprint vs. formal publication patterns.

Experimental results

Research questions

• RQ1 What has been the long-term trend in the proportion of single-authored publications in computer science from 1940 to 2019?
• RQ2 How has the citation impact of single-authored publications changed over time relative to co-authored works?
• RQ3 To what extent do differences in paper scope—measured by page length and reference count—explain the declining citation performance of single-author papers?
• RQ4 How do publication patterns on arXiv compare to peer-reviewed venues for single-authored works?
• RQ5 What factors explain the persistent decline in the influence and visibility of single-authored publications in computer science?

Key findings

• The fraction of single-authored publications in computer science declined from over 80% before 1950 to 7.4% in 2018, indicating a sustained and significant trend toward collaboration.
• Single-authored publications now receive only 60% of the average number of citations compared to all publications, showing a substantial decline in citation impact since the 1940s.
• PageRank scores for single-authored papers have declined over time, confirming their waning importance in the overall citation network structure.
• There is no significant difference in average page length or reference count between single-authored and co-authored papers, suggesting that scope or volume does not explain the citation gap.
• The fraction of single-authored papers on arXiv is significantly higher than in peer-reviewed venues, indicating a preference for preprint sharing among sole authors, possibly due to peer-review avoidance or lower quality standards.
• A slight increase in average references for single-authored papers published after 2010 was observed, but this did not correlate with improved citation outcomes, indicating other factors drive citation disparities.