Introduction

Scope

In this study, an analysis of 204,781 articles published in 2023, in 1,812 hybrid Springer Nature journals (in the imprints Palgrave Macmillan, Springer and Springer Nature), is used to make a comparison between gold OA, green OA, and non-OA articles.¹⁴ Hybrid journals enable a comparison of similar articles, while the size of Springer Nature’s portfolio allows the topic to be explored in depth and at a disciplinary level. Nature Portfolio hybrid journals were excluded from the analysis to maintain comparability across the wider hybrid journal portfolio.

Discipline analysis

The 204,781 articles in this study were associated with 33 different disciplines, which varied considerably in size from 14 articles to 39,501 articles. The 8 disciplines where more than 10,000 articles were identified were included in the disciplinary level analysis, along with an aggregated Humanities subject area and Social Sciences subject area.

The Humanities and Social Sciences subject areas were created as aggregations, as the disciplines within the subject areas were quite small. Nonetheless, it is useful to examine whether the main findings for Science, Technology and Medicine disciplines hold true for the Humanities and Social Sciences. The Humanities category was created by aggregating articles from Architecture / Design, Cultural and Media Studies, History, Linguistics, Literature, Philosophy, and Religious Studies. Social Sciences was created by aggregating publications from Business and Management, Economics, Education, Finance, Geography, Law, Political Science and International Relations, Psychology, and Social Sciences.

Sources of insight

Following the example of previous studies, the impact of articles is explored through the number of citations, and reach is explored through the number of downloads and the Altmetric Attention Score. The Altmetric score is a compound measure of the attention a publication has received from traditional media, patents, policy documents, social media, and other online sources.

The reach and impact of the articles in this data set were analysed according to the publishing models identified by Unpaywall and using the statistical programming language R. Citation data comes from Dimensions, usage is based on the number of downloads from Springer Nature’s servers according to COUNTER 5.1 standards, and the Altmetric score comes from Altmetric.

The download data enables an analysis of the time taken to accumulate downloads, with usage data calculated for 3 months, 6 months, and 12 months after publication date. The download data also enables an exploration of reach into countries with different levels of income, using the World Bank classification of countries.¹⁵

The data were captured in early 2026, providing 2 full years, 2024-2025, in addition to the year of publication, 2023, for the initial reach and impact to be made.

Modelling

Negative binomial regression models were developed to take into account other factors at the document level, the author level, and the journal level that can influence the impact and reach of research publications, besides their OA status. 3 negative binomial regression models were developed for making predictions about the number of citations, number of downloads, and the Altmetric score. Additional models were also built for predicting the downloads over time. Although the models control for a range of factors, this remains an observational study and the findings demonstrate associations between publishing route and the indicators measured.

This modelling approach is suitable as in each case the dependent variables are count data, with a variance exceeding the mean. The median is also included in the tables due to the skewed nature of citation data, with the highest cited articles receiving many times more citations than the typical document. For each of the 3 dependent variables (citations, downloads, and Altmetric score), 6 variables were used in the model:

• OA status
• Clarivate’s Journal Impact Factor (JIF) 2024
• Discipline of the journal
• Times Higher Education (THE) World University Ranking 2024 of corresponding authors’ institutions
• Country of the institution of the corresponding author
• Funder acknowledgement as a binary feature—either there was, or wasn’t, a research funder associated with the article.

To maximise the number of records included in the study, where there were no data available, suitable substitutions were made: JIF was set to 0; country of corresponding institution was set as unknown; and a ranking of 1,907 was given (the number of ranked institutions).

OA classification

The OA status of the different publishing models was gathered from Dimensions, which in turn makes use of Unpaywall’s OA typology.¹⁶ Articles were classified into 1 of 3 categories:

• Gold OA: The VOR is immediately accessible upon publication on the publisher’s website, typically under an open licence.
• Green OA: The version of record requires a subscription on the publisher's website, but a version has been made publicly available in an institutional or subject repository which may include accepted manuscript (AM) versions as well as preprints.
• Non-OA: All other subscription articles.

The data underlying this report has been made openly available on Figshare.¹⁷

Considerations

There have been considerable changes in the use and measurement of social media and other sources tracked by Altmetric over the past 5 years, with a fall in the Altmetric scores. Therefore, the Altmetric results are only used at the overall level where they may be considered indicative for the purpose of comparison with previous studies, rather than being conclusive given the substantial changes in tracked sources and platform behaviour.

We did not have access to download figures for green OA publications stored in institutional repositories. As a result, the observed advantage of green OA over non-OA articles in terms of downloads may be conservative.

Finally, the 2 aggregated subject areas, the Humanities and Social Sciences, should be understood as rough approximations. Many of the disciplines will be split across both the Humanities and the Social Sciences (e.g. Geography). However, the rough approximation enables insights to be drawn into areas that wouldn’t be possible otherwise, while not drawing conclusions about the Humanities based on primarily Social Science data.