Introduction

Open Access (OA) dissemination makes research outputs freely available on the public Internet, allowing users to read, download, copy, distribute, print, search, or link to the full text without any financial, legal, or technical barriers, in accordance with an open copyright licence [1]. One of the milestones of the widespread awareness of OA was 2001’s Budapest Open Access Initiative (BOAI), which established the first clear distinction between the two main types of OA: self-archiving (green OA) and a brand-new generation of journals that would allow scholarly material to be distributed in its full form, freely on the publisher’s website (gold OA). Gold open access journals may be associated with article processing charges, originally intended to cover publication costs. However, the origins of OA go back to the 1970s, when computer scientists were already sharing their papers with File Transfer Protocol (FTP) servers. The first OA journals appeared in the late 1980s with New Horizons in Adult Education, Psycholoquy (1989), and Public-Access Computer systems Review (1989). In the early 1990s, the movement had already picked up some steam with the creation of a dozen more electronic journals, but more importantly, the creation of the first mainstream self-archiving repository arXiv (1991) by Paul Ginsparg, which allowed physics and engineering researchers to deposit preprint version of their papers for free on the Internet. The year 1991 also marked the publication of Dr. Allen Bromley’s, then advisor to President George H. W. Bush, Policy Statements on Data Management for Global Change Research. The statement pleaded for global open access, open data, and preservation of this data in open repositories [2]. In 2000, BioMed Central, the first for-profit OA publisher, was founded.

Nearly half a century since the first use of FTP servers to share scientific papers, the OA movement is still gaining momentum, with the implementation of several OA policies by funders and universities worldwide and the development of new business models by for-profit and not-for-profit publishers. According to several studies, nearly half of research articles could be available online at no cost [3–5]. Previous studies have analyzed the availability of articles [3–5], their citation advantage [4, 6, 7], and compliance with OA mandates [8]. It has often been argued that OA could improve global participation in science, especially for developing countries. However, few articles [9, 10] have compared how different countries, especially developing ones, use OA literature. This paper provides a contemporary portrait of the adoption of OA worldwide, distinguishing between publishing in OA and citing OA publications and between the two main types of OA (green and gold).

Background

Methods

We collected all articles and reviews in the Web of Science (WoS) published between 2015 and 2019 (N = 8,590,184). Science being increasingly collaborative, publications are generally authored by multiple individuals, often from different institutions and countries, among whom the work has typically been unevenly distributed (Larivière et al. 2015). We argue that choices in terms of cited literature and publication venues are predominantly made by authors who played a leading role in the research. Thus, when assigning a publication to a country, we only use the institutional affiliations of the first and corresponding authors. When the first and corresponding authors are affiliated with multiple countries, the publication is fully counted for all the countries. We divided countries into four income level groups based on the 2021–2022 World Bank’s country classification (https://blogs.worldbank.org/opendata/new-world-bank-country-classifications-income-level-2021-2022).

We used the Unpaywall (http://unpaywall.org) database to determine the OA availability and OA type of a paper or reference. Because we use the DOI to link the WoS and Unpaywall data, the 381,101 WoS publications that do not include a DOI, and the 71,408 dois that were not found in Unpaywall were dropped, leaving us with a dataset of 8,137,675 publications. This data indicates whether the paper is published in OA or not, and the OA category it belongs to (gold, green, hybrid or bronze). Some of our analyses include all OA categories while others were limited to green and gold to stay true to the original BOAI definition. These two categories are not mutually exclusive and papers that are published in a gold OA journal and found in a repository were counted in both categories.

The percentage of OA publications varies drastically across disciplines [31] and the distribution of publications across fields varies by country. To account for this, we calculate the percentage of gold and green OA publications for each country and field, and then calculate the field-weighted average z-score for each country. The field-weighted average z-score is calculated as follows: Where:

• x is the share of OA publications or references of country c in field f,
• μ is the average share of OA publications or references for all countries in field f, and
• σ is the standard deviation of the share of OA publications or references for all countries in field f.

This makes values from the different indicators comparable and recenters them around the average (z = 0), negative values indicated lower than average attention and positive values indicating higher than average attention. These z-scores are used for all the country-level analyses below.

Results

Tables 1 and 2 present an overview of our dataset by discipline and type of OA in terms of publications and references, respectively. In Table 1, we see that 42.9% of the articles published during the 2015–2019 period were Open Access at the time of our data collection. The proportion of publications in OA varies by fields, ranging from 21.2% in the Humanities to 50% in the Medical and Health Sciences. The very small proportion of papers that are in gold OA only (0.5%–6.1%) shows the large overlap between green and gold OA.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Proportion of OA publications by type and by fields (2015–2019).

https://doi.org/10.1371/journal.pone.0272730.t001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Proportion references to OA by type and by fields.

https://doi.org/10.1371/journal.pone.0272730.t002

In Table 2, we see that the share of references to OA papers also varies by fields with Medical and Health Sciences citing more OA papers. However, we can observe that the proportion of references to gold OA articles (7.2%) is generally lower than the proportion of gold OA publications presented in Table 1 (18.1%). Overall, the proportion of OA references (39.6%) is lower than the proportion of OA publications (42.9%).

Figs 1 and 2 reveal that Sub-Saharan African countries tend to publish and use OA more than the rest of the world. In North America, the United States publishes and cites OA more than the world average, however, Canada publishes in OA less often but cites OA more often. South American countries are generally slightly over the average, with western countries such as Ecuador, Peru, and Chile generally publishing and citing OA more, while Brazil is at the world average. In Europe, Western European countries mostly publish and cite OA, while the trend is opposite for the Eastern European countries. In Asia, North Africa and the Middle East, most countries publish or cite OA less often.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Weighted z-score and normalized map of the number of OA publications by country.

Red indicates that a country is above the world average, blue indicates it is below the world average. White represents the world average. Contains information from OpenStreetMap and OpenStreetMap Foundation, which is made available under the Open Database License.

https://doi.org/10.1371/journal.pone.0272730.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Weighted z-score and normalized map of references made to OA publications by country.

Red indicates that a country above the world average, blue indicates it is below the world average. White represents the world average. Contains information from OpenStreetMap and OpenStreetMap Foundation, which is made available under the Open Database License.

https://doi.org/10.1371/journal.pone.0272730.g002

Fig 3 illustrates the relationship between the weighted and normalized OA publications and the weighted and normalized references. The Pearson correlation shows a moderate relationship between the two indicators (r = 0.4805; p<0.001), indicating that the more a country publishes in OA, the more it is likely to make references to OA papers.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Correlation between normalized OA publications and normalized cited references by country.

https://doi.org/10.1371/journal.pone.0272730.g003

Fig 4 shows plots of countries according to their OA publications and references compared to the field-normalized world average. Overall, as Fig 1 suggests, the two indicators correlate. The correlation is, however, much stronger for low-income (r = 0.884; p<0.001) and lower-middle-income countries (r = 0,827; p<0.001) than for upper-middle-income countries (r = 0.524; p<0.001) and high-income countries (r = 0.630; p<0.001). On average, the different groups are positioned in different quadrants: the high-income countries are mostly positioned in the first and third quadrants, while upper-middle-income countries are in the third quadrant. On the other hand, low-income and lower-middle-income countries are generally located in the first quadrant. This shows that overall, lower-income countries tend to both publish and cite OA more than the rest of the world, while the upper-middle-income countries are generally publishing and citing OA under the world average. Open access practices tend to vary a lot from a country to another in high-income countries, with about half being over the average and the other half being under.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. OA publication vs. references per country income (red dots; average for all the countries of the same income category).

https://doi.org/10.1371/journal.pone.0272730.g004

Fig 5 shows the average weighted z-score for green and gold OA publications and references to green and gold OA publications by income categories. It shows that while low-income countries tend to publish and cite both green and gold OA, high-income countries mostly use green OA while considerably underusing gold OA. Looking at middle-income countries, lower-middle-income countries tend to use gold OA more than green OA, while upper-middle-income countries underuse both types of OA. S1 Appendix shows detailed world maps of gold and green OA publications and reference patterns.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Average weighted z-score for green and gold OA publications (left) and references to green and gold OA publications (right) by income category.

https://doi.org/10.1371/journal.pone.0272730.g005

Discussion

Our findings show that for the period between 2015 and 2019, Sub-Saharan Africa is publishing and citing OA at a higher rate than the rest of the world, while we found that the Middle East and Asia are the areas where the proportion of publications available in OA is lower, as is their use of OA. This could be explained by the fact that APCs are often waived for low-income countries, such as most Sub-Saharan countries (e.g. through the Research4Life program), but are not or only partially waived for middle-income countries such as the ones found in the Middle East and Asia. One other possible explanation is that the national and transnational OA initiatives such as Plan S for the European Union and PubMed Central in the United States are almost non-existent in these countries, and the Registry of Open Access Repositories shows that institutional repositories are not well developed. Looking at income levels, our findings also reveal that lower middle-income and low-income countries are those who publish and cite the most OA. Again, this may be partly explained by the fact that APCs are generally waived for these countries, making publishing in gold OA more accessible for them. Moreover, some commercially owned gold OA journals may have less strict publication criteria and given the lack of research infrastructure in these countries, they may be more likely to publish in these journals [32]. We also found that the upper-middle-income countries tend to behave similarly to the higher-income countries. However, the underlying mechanisms behind the use of OA potential may be different. For instance, upper-middle-income countries may lack the resources to pay for APCs on top of high subscription prices for closed journals (both don’t qualify for waivers), which is not necessarily the case for high-income countries. There may also be other factors, such as a lower reputation of OA journals in certain parts of the world. Another interesting finding is that while low-income countries use both green and gold OA, high-income countries generally tend to favour green OA. This may imply that even in the wealthiest countries on the planet, researchers may still struggle to pay APCs or be opposed to them. A possible limitation of the study is our use of the DOIs to identify the papers, given that the number of journals using DOIs might not be evenly distributed around the world, which could have an effect on our analyses of certain countries. Another potential limitation is the use of the Web of Science as the main data source. A study by Basson et al. [33] underlined the importance of data sources and their effects on the measurement of OA, especially when assessing the global South, where a large body of the published research may not be indexed in the most popular bibliometric data sources such as Scopus or Web of Science. Overall, our results seem to confirm previous studies [9, 10], which found that low-income countries tend to publish in OA more than other countries. Ultimately, while high-income countries have mandates and repositories, and low-income countries have waivers, our results highlight national differences in OA uptake and suggest that more OA initiatives at the institutional, national, and international levels are needed to support wider adoption of open scholarship.

Conclusion

Our study has provided a snapshot of the state of OA over the 2015–2019 period. While a significant share of the scientific output was available in open access at the time of this study, concerns have been raised about the preservation of open access literature and the integrity of the scholarly communication record [34]. Combining several bibliographic indexes such as Scopus, Ulrichsweb, and the Directory of Open Access Journals, the authors traced journals through the Internet Archive’s Wayback Machine. Their results showed that 174 OA journals from various disciplines and geographic regions had completely disappeared from the web between 2000 and 2019. While this number does not appear to be particularly high, the authors noted that their results were most likely a low estimation since the data availability limitations did not allow them to assess the full extent of the phenomenon. They also noticed some common traits between the vanished journals: they tended to be affiliated with academic institutions or scholarly societies located in North America or published social sciences and humanities research. Ultimately, their study underlined the importance of collaborative action to ensure continued access and prevent further scholarly knowledge loss. It remains uncertain how this may play a part in the general uptake and sustainability of OA. Further research on this topic is certainly needed.

The recent COVID-19 pandemic has underlined the importance of OA, which has proven essential for faster and more efficient dissemination and use of scholarly literature. At the journal level, several publishers such as Springer-Nature, AAAS (Science), Elsevier, and Massachusetts Medical Society (NEJM) have announced the free opening of their COVID-19-related papers. However, it remains unclear if these papers will remain OA. Elsevier, for instance, already mentioned that their Novel Coronavirus Information Center will only be free for the duration of the pandemic [35]. Despite the existence of geopolitical tensions [36–38] and initiatives such as China’s move to ensure that funded studies on COVID-19 will be published in Chinese journals over international journals [39], the majority of countries have increased their number of international collaborations and have generally increased their share of OA literature [38, 40]. However, it has been shown that countries’ previous OA dissemination practices had no significant effect on their OA publication practices during the ongoing global pandemic: the extent to which a country was affected by COVID-19 was the main factor for OA publication [38]. Another important finding from that study is that countries with a higher GDP may not engage as much as other countries in international collaboration and open access during the global crisis, most likely because they are less dependent on outside collaborators. These findings have several implications for our study. First, the pandemic may help middle-income countries to participate and alleviate the divide in the uptake of OA since they do not necessarily have the necessary funding and resources to produce national COVID-19 research. Second, with high-income countries being more reluctant to participate in global collaboration and publish their COVID-19 research in OA, it highlights the importance of the various OA mandates, policies, and repositories in place. It remains to be seen whether this recent trend created by the pandemic will persist over time. Ultimately, the current crisis shows the necessity for more sustainable OA dissemination channels: as long as the research community is dependent on private for-profit publishers, the “openness” of articles will depend on either the “good will” of those publishers or on the scientific community’s capacity and willingness to pay for it.

Supporting information

Acknowledgments

The authors would like to thank Vanessa Sandoval-Romero for her participation in the initial phases of the project. We would also like the thank Marion Maisonobe and Iryna Kuchma for their time and their invaluable comments and suggestions. This article is a revised version of an ISSI paper with updated data that was originally published in the Proceedings of the 18th International Conference on Scientometrics and Informetrics.