Large cross-publisher evidence base
The study analyzes 531,889 PLOS and BMC articles, extending earlier discipline-specific evidence across multiple journals and research areas.
↳ Abstract; Methods, Data; Introduction, final paragraphs
Abstract
Efforts to make research results open and reproducible are increasingly reflected by journal policies encouraging or mandating authors to provide data availability statements. As a consequence of this, there has been a strong uptake of data availability statements in recent literature. Nevertheless, it is still unclear what proportion of these statements actually contain well-formed links to data, for example via a URL or permanent identifier, and if there is an added value in providing such links. We consider 531, 889 journal articles published by PLOS and BMC, develop an automatic system for labelling their data availability statements according to four categories based on their content and the type of data availability they display, and finally analyze the citation advantage of different statement categories via regression. We find that, following mandated publisher policies, data availability statements become very common. In 2018 93.7% of 21,793 PLOS articles and 88.2% of 31,956 BMC articles had data availability statements. Data availability statements containing a link to data in a repository-rather than being available on request or included as supporting information files-are a fraction of the total. In 2017 and 2018, 20.8% of PLOS publications and 12.2% of BMC publications provided DAS containing a link to data in a repository. We also find an association between articles that include statements that link to data in a repository and up to 25.36% (± 1.07%) higher citation impact on average, using a citation prediction model. We discuss the potential implications of these results for authors (researchers) and journal publishers who make the effort of sharing their data in repositories. All our data and code are made available in order to reproduce and extend our results.
Key Claims
Primary findingSupported
We also find an association between articles that include statements that link to data in a repository and up to 25.36% (± 1.07%) higher citation impact on average, using a citation prediction model.
adjusted observational models consistently support the direction, but residual confounding prevents causal interpretation
Abstract; Results, Citation Prediction
SecondaryWell-supported
In 2018 93.7% of 21,793 PLOS articles and 88.2% of 31,956 BMC articles had data availability statements.
direct article counts support the reported publisher-specific uptake rates
Abstract; Results, The Presence of Data Availability Statements Over Time
SecondaryWell-supported
In 2017 and 2018, 20.8% of PLOS publications and 12.2% of BMC publications provided DAS containing a link to data in a repository.
validated DAS classification directly supports the reported repository-link proportions
Abstract; Results, The Presence of Data Availability Statements Over Time
Derived from the full evaluation — not a separate score.
Strengths · Limitations
Strengths
Validated automated DAS classification
The SVM classifier achieved 0.99 accuracy, precision, recall, and weighted F1 on the held-out set, with results reported by category in Table 2.
↳ Methods, Data Availability Statements: Classification; Table 2
Multiple robustness and validation checks
The citation finding was checked across several model families, two-, three-, and five-year windows, journal controls, and a Web of Science comparison reporting r=0.83.
↳ Results, Citation Prediction, Model and Dependent Variable sections; Table 6
Limitations
Residual confounding remains unresolved
Article quality, research resources, and group visibility may influence both repository sharing and citations. These alternatives are acknowledged in the Conclusion but are not addressed through formal sensitivity or causal-identification analysis.
↳ Conclusion, final paragraph
OA-only selection limits inference
The sample is limited to PLOS and BMC articles in PubMed OA. The Discussion acknowledges limited coverage but does not test whether this selection changes the estimated DAS–citation relationship.
↳ Methods, Data; Discussion, limitations paragraph
Policy language exceeds causal evidence
The title and policy discussion describe a citation “advantage,” “benefit,” and “incentive,” while the underlying analysis establishes correlation rather than causal gain.
↳ Title; Discussion, stakeholder implications; Conclusion
Nabu Commentary
The descriptive findings are directly supported by large-scale article counts and a DAS classifier reporting 0.99 held-out accuracy. The citation analysis adjusts for publication timing, article characteristics, author H-index, publisher, policy, field, and journal, and alternative model specifications corroborate the reported association. The conclusion also acknowledges that article quality and research-group resources may explain part of the relationship, so the evidence supports association rather than causal benefit. The strongest remaining methodological limitation is the restriction to PLOS and BMC articles in PubMed OA without testing how that selection affects the association itself.
Author Note / Field Feedback
Nabu’s assessment, alongside the field’s view.
Are you an author of this paper?
Dimension breakdown
Study Quality
Strong3.8
Contribution
3.8
The study meaningfully extends discipline-specific evidence through a 531,889-article, two-publisher analysis and an automated DAS classification system. Its main citation finding remains an observational extension of an existing association rather than a qualitatively new result.
Abstract; Introduction, final paragraphs; Methods, Data Availability Statements: Classification
“We also find an association between articles that include statements that link to data in a repository”
Methodological Rigour
3.7
The regression adjusts for publication timing, authorship, references, author H-index, publisher, policy, field, and journal, with multiple alternative models and citation-count validation. Restriction to PLOS and BMC articles in PubMed OA is disclosed but its possible influence on the DAS–citation association is not directly tested.
Results, Citation Prediction, Independent Variables and Model; Discussion, limitations paragraph
“While the PubMed OA collection is sizable, it includes only a fraction of all published literature.”
Reporting
4.0
The research questions, classification procedure, regression analysis, and findings form a readily traceable argument, and the central statistical result is usually described as an association or correlation. The title and policy discussion use “advantage” and “benefit,” which can suggest greater causal certainty than the design provides.
Introduction, research questions; Abstract; Discussion, paragraph beginning “There are several potential implications”; Conclusion
“Are different DAS categories correlated with an article’s citation impact?”
Positioning
3.8
The paper connects its analysis to prior discipline-specific citation studies and discusses OA coverage, citation measurement, DAS extraction, and the absence of repository-content verification. It acknowledges article quality and research-group resources as alternative explanations, although engagement with complicating or null findings is limited.
Introduction, paragraphs discussing prior studies; Discussion, limitations paragraph; Conclusion, final paragraph
“More efforts and resources are put into papers sharing data”
Reliability
Concerns4 of 4 checks
Internal CoherenceNoted
Reported headline numbers and methods–results alignment are internally consistent. Prominent benefit-oriented wording and minor production errors warrant a noted status but do not invalidate the descriptive or associational results.
Research ConductClean
Data and code are claimed available through a persistent Zenodo DOI, and the bibliometric design does not require human-subject ethics approval. No conduct concern is evident from the supplied text.
Flags: 2 declared / 5 total
Reference IntegrityConcern
58 of 66 checkable references verified
78 references in manuscript 12 are books, websites or datasets — counted, but not index-checkable
Post-Publication RecordClean
No retraction notice found in Retraction Watch.
Last checked: 14 Aug 2026
Sources: Retraction Watch ✓
Impact Potential
Medium3.5
Relevance
3.8
Publishers, funders, institutions, and researchers are explicitly identified as users of the findings for data-policy design, compliance monitoring, and incentives. The pathway is concrete, although the citation result alone cannot establish the effect of changing policy.
Discussion, paragraphs on stakeholder implications and DAS classification
“Our DAS classification approach, and release of the data and code, may be helpful for stakeholders interested in research data policy compliance”
Readiness
3.3
The classifier and descriptive findings were developed on operational publication records and the code and data are released for reuse. Readiness of the citation association into policy still requires causal triangulation and validation beyond the studied publishers.
Methods, Data Availability Statements: Classification; Discussion; Data and Code Availability
“Code and data can be found at: https://doi.org/10.5281/zenodo.3470062”
Generalizability
3.0
Coverage of two multidisciplinary publishers and journal-level variation supports transfer within open-access publishing contexts. The PubMed OA restriction and absence of subscription-publisher evidence limit broader generalisation.
Methods, Data; Results, Figure 2F–H; Discussion, limitations paragraph
“it includes only a fraction of all published literature”
Trajectory
3.8
The paper extends earlier discipline-specific findings to a larger cross-publisher corpus and releases resources that enable updates and external extensions. It therefore contributes clearly to cumulative research on data-sharing policy, though it does not resolve causal mechanisms.
Introduction, discussion of studies 24–29; Discussion; Data and Code Availability
“by sharing all our data and code, this study can be updated and built upon”
AI-generated, human-governed. Something look off? Contact us to request a review.