Loading evaluation data…
Abstract In this article, we analyze the citations to articles published in 11 biological and medical journals from 2003 to 2007 that employ author‐choice open‐access models. Controlling for known explanatory predictors of citations, only 2 of the 11 journals show positive and significant open‐access effects. Analyzing all journals together, we report a small but significant increase in article citations of 17%. In addition, there is strong evidence to suggest that the open‐access advantage is declining by about 7% per year, from 32% in 2004 to 11% in 2007.
Analyzing all journals together, we report a small but significant increase in article citations of 17%.
the pooled adjusted association is precise, but author-choice confounding prevents causal identification
Controlling for known explanatory predictors of citations, only 2 of the 11 journals show positive and significant open access effects.
Table 2 reports three significant positive full-model journal estimates rather than two
There is strong evidence to suggest that the open access advantage is declining by about 7% per year, from 32% in 2004 to 11% in 2007.
single-date cumulative citations cannot directly identify a changing access-effect trajectory across publication cohorts
Derived from the full evaluation — not a separate score.
Strengths
The analysis covers 11,013 articles across 11 biological and medical journals and reports both journal-specific and pooled estimates, including a sensitivity analysis excluding PNAS.
↳ Methods, pp. 5–6; Tables 1–2
The paper discusses skeptical observational studies and a prior randomized trial that found no citation advantage, using them to frame the access-versus-selection problem.
↳ Introduction, p. 3; Limitations, pp. 10–11
Table 3 translates estimated citation differences and author fees into journal-level cost-per-citation figures, creating a visible decision context for authors and institutions.
↳ Results, p. 7; Table 3
Limitations
Paid open-access status is author-selected, and the paper acknowledges that novelty and expected impact are unobserved. The models therefore cannot determine whether access or article selection produced the adjusted association.
↳ Methods, p. 6; Limitations, pp. 10–11
Despite acknowledging unresolved confounding, the Discussion attributes the advantage to “self-selection – not access” and infers that subscription access barriers are small and diminishing.
↳ Discussion, pp. 8–9; Limitations, pp. 10–11
The Abstract and Results report two significant positive full-model journal effects, but Table 2 displays three: Bioinformatics, Physiological Genomics, and PNAS.
↳ Abstract; Results, p. 7; Table 2
The paper’s main contribution is its common analysis of 11 author-choice programs, including pooled estimates, a PNAS sensitivity analysis, and a practical cost-per-citation calculation. The observable covariate adjustments are useful descriptively, but the Methods and Limitations acknowledge that novelty and expected impact remain uncontrolled. That limitation weighs heavily because the Discussion nevertheless attributes the association to self-selection rather than access and describes subscription barriers as diminishing. The temporal-decline claim is also less secure than its wording suggests because citations were retrieved once for cohorts of different ages. Finally, the count of significant journal-level effects conflicts with Table 2, reducing confidence in the per-journal summary without overturning the pooled estimate.
Nabu’s assessment, alongside the field’s view.
Are you an author of this paper?
Limited2.7
Confidence highThe study extends the open-access citation debate through a common analysis of 11 journals, pooled sensitivity estimates, a temporal pattern, and a cost-per-citation framing. The advance is meaningful but bounded by its observational identification.
“Analyzing all journals together, we estimate a 17% citation advantage”
The regression models control article age and several observable characteristics, with pooled analyses reported with and without PNAS. Author choice remains confounded by unobserved novelty and expected impact, yet the Discussion attributes the association to self-selection and diminishing access barriers.
“this method is unlikely to adequately deal with article characteristics that are unobservable to the researcher”
The paper follows a coherent progression from the access-versus-selection question through regression results, interpretation, costs, and limitations. Reporting is reduced by the claim that two journals are significant when Table 2 shows three, alongside several overlapping percentage estimates requiring careful model tracking.
“only two of the eleven journals show positive and significant open access effects”
The discussion engages supportive and skeptical observational studies and explicitly compares the findings with a prior randomized trial. Its limitation analysis is substantive, but the causal self-selection and access-barrier conclusions exceed the acknowledged inability to address unobservable article differences.
“self-selection – not access – is the explanation for the citation advantage”
Lower confidence on Positioning — domain match limited.
Caveats4 of 4 checks
The central aggregate estimate is internally traceable, but the reported count of significant journal-level effects conflicts with Table 2. Table 3 also describes its cost calculation imprecisely, although the displayed values reveal the intended calculation.
The study is a non-interventional bibliometric analysis using a named proprietary citation source, and no ethics or participant-consent requirement is apparent. Funding is acknowledged, with no specific conduct concern identified in the supplied text.
Flags: 0 declared / 5 total
18 of 22 checkable references verified
26 references in manuscript 4 are books, websites or datasets — counted, but not index-checkable
No retraction notice found in Retraction Watch.
Sources: Retraction Watch ✓
Where this paper’s evidence sits on the path from initial observation to real-world use.
The analysis uses operating-journal citation and fee data, making it more directly relevant than a simulated assessment. Readiness remains limited because the estimates are observational and the paper states that later cohorts would have different costs.
“the cost per citation for articles published in 2007 would be much higher”
AI-generated, human-governed. Something look off? Contact us to request a review.