Co-Authorship Network Analysis
Also known as: collaboration network, authorship network, research collaboration mapping
Co-authorship network analysis is a method that maps research collaboration patterns by treating authors as nodes and co-authored papers as edges in a network graph. The structure, density, and centrality patterns of this network reveal how researchers connect, collaborate across institutions and disciplines, and form research communities. Pioneered formally by Newman (2001), co-authorship analysis provides quantitative insights into the social fabric of science, revealing collaboration patterns, identifying scientific leaders, and detecting institutional or disciplinary boundaries.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use co-authorship analysis to assess the state of team science in a discipline, identify key researchers and research teams, understand institutional collaboration patterns, detect interdisciplinary research activity, map the research landscape of a funding program or journal, or measure the effects of policy changes (e.g., funding incentives for collaboration) on research organization. It is particularly valuable for institutional benchmarking (How collaborative is our research compared to peer institutions?) and for studying the social determinants of research productivity (Does collaboration increase output quality/quantity?). Co-authorship analysis is also useful for identifying potential collaborators based on complementary expertise clusters.
Strengths & limitations
- Objective and unambiguous: co-authorship is explicitly recorded in publications; no interpretation required.
- Large-scale and scalable: modern networks contain tens of thousands of researchers; network algorithms process these efficiently.
- Temporal dynamics: comparing networks across years reveals collaboration trends and generational shifts.
- Rich metadata: author names can be linked to institution, department, country, and funding data, enabling multidimensional analysis.
- Practical utility: identifies potential research partners, assesses lab productivity, and reveals structural collaboration barriers.
- Author name disambiguation: variations in names (initials, name changes, transliteration, nicknames) create false nodes or merge distinct authors; this is labor-intensive to correct.
- Missing data: not all research produces publications (internal reports, software); authors may have gaps in publication records due to career breaks.
- Discipline-specific norms: co-authorship prevalence varies dramatically by field (rare in mathematics/philosophy, universal in biophysics); networks are not comparable across disciplines.
- Changing team size over time: comparing author networks across decades may reflect publication growth rather than collaboration growth.
Frequently asked
How do I handle author name disambiguation?
This is a major challenge. Best practices: (1) Use author identifiers where available (ORCID, Scopus Author ID, Web of Science ResearcherID). (2) For datasets without identifiers, use automated disambiguation tools (name parsing, affiliation matching, publication history). (3) For small networks (< 500 authors), manual review of common/ambiguous names is feasible. (4) Always report your disambiguation method and estimate error rates. Some researchers accept 10–15% disambiguation error as acceptable for large-scale analysis; for precision-critical work, manual curation is necessary.
Should I weight edges by the number of co-authored papers?
Yes. An unweighted network treats one co-authored paper the same as ten; weighting by paper count reveals collaboration depth. Use weighted edges when analyzing collaboration intensity or identifying strong research partnerships. Use unweighted edges when analyzing network connectivity or identifying structural bridges. Report both: 'degree centrality' (unweighted: number of collaborators) and 'weighted degree' (weighted: total collaboration intensity).
How do I account for multi-authored papers where all authors are equally credited?
Standard co-authorship networks connect all author pairs, treating a 5-author paper as a complete graph with 10 edges (5 choose 2). This is the norm. Alternatively, some researchers weight edges inversely by the number of authors on a paper (fractional counting), but this is less common and not necessary unless you have specific theoretical reasons. Report which method you use.
Can I use co-authorship networks to predict future collaboration or research breakthroughs?
Not directly. Co-authorship networks are descriptive; they show who has collaborated but not why or what impact collaboration had. However, combined with citation analysis and metadata, networks can suggest where new collaborations might be fruitful (authors in separate clusters with similar research interests). Predictive claims require causal validation—does collaboration improve outcomes?—which is beyond the scope of network analysis alone.
Sources
- Newman, M. E. J. (2001). The structure of scientific collaboration networks. Proceedings of the National Academy of Sciences, 98(2), 404–409. DOI: 10.1073/pnas.021544898 ↗
- Braun, T., Glänzel, W., & Schubert, A. (2001). Dynamic scientometric relations: Citation and collaboration patterns in selected research areas. Scientometrics, 51(3), 487–502. link ↗
How to cite this page
ScholarGate. (2026, June 4). Co-Authorship Network Analysis. ScholarGate. https://scholargate.app/en/bibliometrics/co-authorship-analysis
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Bibliographic CouplingBibliometrics↔ compare
- Co-Citation AnalysisBibliometrics↔ compare
- Keyword Co-Occurrence AnalysisBibliometrics↔ compare
- Science MappingBibliometrics↔ compare