The academic, economic and societal impacts of Open Access: an evidence-based review

All authors contributed equally to the writing of this manuscript using the Overleaf collaborative writing platform.

Competing interests: JPT currently blogs for the PLOS Paleo Community, and works for ScienceOpen. CHJH is a Center for Open Science ambassador. DCJ and FW are members of the Open Access Working Group of EURODOC. PM is a Research Data Alliance member. LBC works for the University of Pittsburgh, which has an Open Access library publishing department. All views presented here are strictly personal.

This is an open access article distributed under the terms of the Creative Commons Attribution Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Version Changes

Revised. Amendments from Version 2

The final version of this manuscript includes minor edits reflected in the last review by Peter Suber, as well as stylistic edits noted by other commenters. We hope that this paper will continue to be shared and discussed, and that it helps with future developments in Open Access.

Peer Review Summary

Abstract

Ongoing debates surrounding Open Access to the scholarly literature are multifaceted and complicated by disparate and often polarised viewpoints from engaged stakeholders. At the current stage, Open Access has become such a global issue that it is critical for all involved in scholarly publishing, including policymakers, publishers, research funders, governments, learned societies, librarians, and academic communities, to be well-informed on the history, benefits, and pitfalls of Open Access. In spite of this, there is a general lack of consensus regarding the potential pros and cons of Open Access at multiple levels. This review aims to be a resource for current knowledge on the impacts of Open Access by synthesizing important research in three major areas: academic, economic and societal. While there is clearly much scope for additional research, several key trends are identified, including a broad citation advantage for researchers who publish openly, as well as additional benefits to the non-academic dissemination of their work. The economic impact of Open Access is less well-understood, although it is clear that access to the research literature is key for innovative enterprises, and a range of governmental and non-governmental services. Furthermore, Open Access has the potential to save both publishers and research funders considerable amounts of financial resources, and can provide some economic benefits to traditionally subscription-based journals. The societal impact of Open Access is strong, in particular for advancing citizen science initiatives, and leveling the playing field for researchers in developing countries. Open Access supersedes all potential alternative modes of access to the scholarly literature through enabling unrestricted re-use, and long-term stability independent of financial constraints of traditional publishers that impede knowledge sharing. However, Open Access has the potential to become unsustainable for research communities if high-cost options are allowed to continue to prevail in a widely unregulated scholarly publishing market. Open Access remains only one of the multiple challenges that the scholarly publishing system is currently facing. Yet, it provides one foundation for increasing engagement with researchers regarding ethical standards of publishing and the broader implications of 'Open Research'.

Keywords: Open Citation Advantage, Altmetrics, Open Access, Publishing, Copyright, Incentive System, Open Science, Text and Data Mining

Introduction

Open Access (OA) refers to the removal of major obstacles to accessing, sharing and re-using the outputs of scholarly research. The rationale is that the research process is facilitated by ensuring rapid and widespread access to research findings such that all communities have the opportunity to build upon them and participate in scholarly conversations. As such, the major drivers behind OA relate to within- and between-community equality ( Veletsianos & Kimmons, 2012), as well as bridging the global North-South research divide ( Adcock & Fottrell, 2008). Reflecting this ambition, there are currently over 700 OA policies and mandates recorded worldwide from a range of research institutes and funding bodies ( roarmap.eprints.org). OA pertains to documents made available via two main pathways: the Gold route and the Green route ( Harnad et al., 2008). The Gold route refers to freely accessible research articles at the point of publication. This route is often, although not always, accompanied by article processing charges (APCs). The Green route refers to author self-archiving, in which peer-reviewed articles and/or not peer-reviewed pre-prints are posted online to an institutional and/or subject repository, or to a personal website. This route is often dependent on journal or publisher policies on self-archiving ( sherpa.ac.uk/romeo). Some publishers require an embargo period before deposition in public repositories is allowed. These embargoes are applied in order to avoid putative reductions in subscription income due to such self-archiving, although there is little evidence to support the existence of such embargoes ( Berners-Lee et al., 2005; Bernius et al., 2013; Henneken et al., 2006; Houghton & Oppenheim, 2010; Swan & Brown, 2005). The Green route is also enabled through author rights retention, in which authors pre-emptively grant non-exclusive rights to their institutions before publishing any works. The institution then has the ability to make articles by these authors OA without seeking permission from the publishers (e.g., this is the case of the Dutch Taverne amendment that has declared self-archival of research after ‘a reasonable period of time’ a legal right ( Open Access NL, 2015)). Through these dual pathways, almost 25% of all scholarly documents archived on the Web are now obtainable via OA somewhere on the Internet ( Khabsa & Giles, 2014).

A core issue remains: universal or even marginal access to approximately 75% of articles is not directly possible unless one either is in a privileged position to work at an institute that has subscription access to these articles, or has enough money to pay on a per-article basis (given that journals provide this feature; some do not). Subscriptions to all peer-reviewed journals is not affordable for any single individual, research institute or university ( Odlyzko, 2006; Suber, 2012). Consequently, the potential impact of research articles is never fully realized, impeding scientific progress by a lack of use, while simultaneously negatively affecting the recognition of individual researchers ( Hitchcock, 2013) and the funders who support their work.

Because of these issues, free and unrestricted access to primary research literature has become a global goal of the OA movement. The steady increase in OA over the past two decades has required careful negotiations between a range of stakeholders (e.g., librarians, funders, academics). Much of the driving force behind this global change has been through a combination of direct, grassroots advocacy initiatives and policy reforms from universities, funders and governments. The debates regarding the benefits of OA over subscription-based access often hinge on the increased value to academics. However, increased access has broader benefits to research through enhanced visibility, facilitating innovation by businesses and decreasing financial pressure on academic/research libraries (known more broadly as the ‘serials crisis’ ( McGuigan & Russel, 2008)). Additionally, increased access to scholarly outputs might help foster a culture of greater scientific education and literacy, which in turn could have a direct impact on public policy ( European Commission, 2012; Zuccala, 2010), particularly in domains such as climate change and global health, as well as increasing public engagement in scientific research ( Stodden, 2010). OA also includes a moral aspect, where access to scientific knowledge and information is regarded as a fundamental feature of global human equality. For example, Article 27 of the United Nations Declaration of Human Rights states that " Everyone has the right to freely participate in the cultural life of the community, to enjoy the arts and to share in scientific advancement and its benefits." ( United Nations, 1948).

This review aims to provide information on the various impacts of OA to scholarly research. We consider the impact of OA from the academic, economic, and societal perspective. In addition, we shortly consider the broader implications of OA on Open Data, a closely related aspect united under a general theme of Open Research or Open Science. By aggregating evidence from a range of primary sources, this review should be useful to those broadly interested in the impact of open scholarly research, as well as policymakers and others involved in implementing OA policies and strategies. We refrain from making predictions about the future of OA publishing or policy recommendations, as these are both beyond the scope of this work.

A brief history of Open Access

The OA movement is intrinsically tied to the development of the Internet and how it redefined communication and publishing ( Laakso et al., 2011). With increased availability of Internet bandwidth, print articles have become virtually redundant, and sharing of information has never been cheaper. As a consequence, the costs per research article should have potentially decreased as a result of not investing material resources in publications printing and distribution. Therefore, widespread dissatisfaction with the expensive traditional publishing model has increased, resulting in the OA movement and concomitant innovations in scholarly publishing. A comprehensive timeline of the OA movement is provided as part of the Open Access Directory ( oad.simmons.edu/oadwiki/Timeline).

Interest in using the Internet for facilitating access to scientific research coalesced throughout the 1990s, culminating with the 2001 conference on "Free Online Scholarship" by the Open Society Institute in Budapest. The result of this conference was the release of the Budapest Open Access Initiative (BOAI), which is recognized as one of the defining points of the OA movement. The BOAI was the first initiative to use the term "Open Access" and articulated the following definition:

By "open access" to [peer-reviewed research literature], we mean its free availability on the public internet, permitting any users to read, download, copy, distribute, print, search, or link to the full texts of these articles, crawl them for indexing, pass them as data to software, or use them for any other lawful purpose, without financial, legal, or technical barriers other than those inseparable from gaining access to the internet itself. The only constraint on reproduction and distribution, and the only role for copyright in this domain, should be to give authors control over the integrity of their work and the right to be properly acknowledged and cited.

This definition is broadly equivalent to the Creative Commons Attribution license (CC-BY), which is widely considered to be a standard for OA ( creativecommons.org/licenses/). One result of the growing OA movement is the rise of OA-only publishers, who publish exclusively digital content and have demonstrated that such a business model is financially feasible (but does not necessarily sustain the current journal ecosystem). Some of these publishers are for-profit and some are non-profit. For example, pioneer OA publishers BioMed Central (for-profit) and the Public Library of Science (PLOS) (non-profit) were founded in the early 2000s and remain successful OA publishing businesses to date. More recently, OA publishing has gained increasing momentum among researchers, funders, and governments. This has led to a proliferation of innovative approaches to publishing (e.g., PeerJ, peerj.com; F1000Research, f1000research.com; Open Library of Humanities, openlibhums.org) and a range of different policies from research funders and institutes mandating OA. All of these different policies and new business models, combined with traditional publishers launching their own OA titles and programs, have made the overall OA ecosystem quite complex.

Even with this growing prevalence of publishers that facilitate OA to the scholarly literature, OA is still hardly ubiquitous. Bjork et al. (2009) estimated that the total number of published articles in 2006 was approximately 1,350,000. Of these, 4.6% became immediately accessible and an additional 3.5% became accessible after an embargo period of typically one year. Furthermore, usable copies of 11.3% could be found in repositories or on the author’s home pages. Since the U.S. National Institutes of Health (NIH) mandated archival of articles in the public PubMed Central repository in 2008, the cumulative number of OA articles in PMC has increased more than the number of non-OA articles (see Figure 1 ). In 2013, the total percentage of OA articles available was estimated at 24% of English-language scholarly documents accessible on the Web ( Khabsa & Giles, 2014).

Since 2004, the growth rate of OA articles is significantly higher than that of non-OA articles.

Although these estimates show OA is on the rise, the full potential of OA is far from achieved. Björk et al. (2014) pointed out that 62% of journals (from the top 100 journal publishers indexed by SCOPUS) endorse immediate Green OA self-archiving by their authors, 4% impose a 6-month embargo, and 13% impose a 12-month embargo. As such, 79% of articles published in any recent year could already be OA within 12 months after publication via Green OA, 62% of them immediately if authors were actually self-archiving properly ( Gargouri et al., 2010; Gargouri et al., 2012). The disconnect between practice and what is allowed has three potential explanations: (i) researchers are unsure whether they have the legal right to self-archive, (ii) they fear that it might put their article’s acceptance for publication at risk, and (iii) they believe that self-archiving may be a lot of work ( Harnad, 2006). Research funders and institutions worldwide are now beginning to realize that they need to alter their conditions to make OA mandatory ( Vincent-Lamarre et al., 2016) in order to counteract these misconceptions of self-archiving ( Carr et al., 2007; Swan & Brown, 2005). Swan & Brown (2005) have indicated that the vast majority of researchers (81%) would comply with mandatory OA if it were a condition of funding. On the other hand, it is worth mentioning that ensuring compliance with OA policies set by research institutions is rather difficult. Some tools, such as the Open Access Monitor ( http://symplectic.co.uk/elements-updates/introducing-open-access-monitor), help institutions to track compliance with their OA policy.

Table 1 shows a non-exhaustive summary of the developments in the advancement of scholarly publishing and the OA movement. Included are the founding of major institutions in the movement as well as policy and legal developments. Several controversial moments are included, because they have spurred action or generated awareness for the movement. One of them is the suicide of Aaron Swartz, who was arrested for downloading JSTOR articles on the grounds that he allegedly intended to make these publicly available. Another ongoing controversy is scholarly piracy; this includes the Sci-Hub and LibGen projects, which have created an online repository of pirated scholarly papers (around 50 million at the time of writing). Both projects gained increased attention after becoming the target of a lawsuit by the publisher Elsevier. There have been mixed responses to these kinds of activities, polarising the view that illegal acts regress or weaken the case for OA, while some hail the development as the ‘Napster moment’ (i.e., a change inducing disruption; Rosenwald, 2016) for the OA movement, which will force the established industry to change. Regardless of the legality of it, Sci-Hub is used by a large number of people from all over the world to access research articles ( Bohannon, 2016; Elbakyan & Bohannon, 2016).

Table 1.

The effect of Open Access upon academia

The two main ways in which OA affects academia are (i) through association with a higher documented impact of scholarly articles, as a result of availability and re-use, and (ii) through the possibility of non-restrictively allowing researchers to use automated tools to mine the scholarly literature. For the former, major arguments in favor of OA include the evidence that work that is openly available generates more academic citations, but also has more societal impact. In addition, appropriately-licensed OA works play a major role in academic education, including re-use in classes and for dissertations. The latter major argument involves non-restrictive access to the scholarly literature through appropriate licensing, making it possible to use automated tools to collect and analyze the entire body of scholarly literature in a legally sound framework and irrespective of copyright laws. The following sections cover these two effects of OA.

The potential impact advantage

Academic impact. Academic impact is frequently measured through citation counts, and these remain fundamental as the ‘currency units’ for researchers, research groups, institutes and universities. Lawrence (2001) was the first to propose that OA would have a citation advantage. The utility and consistency of the citation advantage across different research fields has been intensively debated because its magnitude substantially varies depending on the discipline ( Table 2 ). However, the general tendency identified by studies to date indicates that there is at least some association between OA publishing and increased citation counts across most disciplines ( Hajjem et al., 2006; Antelman, 2004) ( Figure 2 and Table 2 ). A comprehensive and annotated bibliography of studies documenting potential citation impacts was created by Steve Hitchcock ( eprints.soton.ac.uk/354006/1/oacitation-biblio-snapshot0613.html) and has been managed by SPARC Europe since 2013 ( sparceurope.org/oaca/).

The majority concluded that there is a significant citation advantage for Open Access articles. Source: Data from The Open Access Citation Advantage Service, SPARC Europe, accessed March 2016.

Table 2.

Main scientific papers that have investigated and quantified the citation advantage as well as its origin.

Estimates for the open citation advantage range from +36% (Biology) to +600% (Agricultural Sciences) ( Swan, 2010; Wagner, 2010). In a longitudinal study, Eysenbach (2006) compared the bibliometric impact of a cohort of articles from a multi-disciplinary journal ( Proceedings of the National Academy of Sciences) that offers both OA and non-OA publishing options. After adjusting for potentially confounding variables, the results indicated that non-OA papers were twice as likely to remain uncited six months after publication when compared to OA articles. Additionally, the average number of citations for OA articles was more than double than that of the non-OA articles. The study also differentiated the type of OA article, namely the self-archived (i.e., Green OA) and the publisher version of record (VOR) that is freely available (i.e., Gold OA). Gold OA was found to have a higher overall academic impact than Green OA.

Despite strong evidence for a citation advantage, the magnitude of this advantage remains variable. The substantial heterogeneity in observed citation advantages can be due to different academic cultures or could simply be spurious. For example, self-archiving prior to publication is a community standard in fields such as high energy physics or mathematics, but has yet to be widely adopted among the life sciences. Such ‘pre-prints’ have also been associated with an overall increase in the average number of citations, the total number of citations, and the speed of citation accumulation ( Aman, 2014; Gentil-Beccot et al., 2010). Other studies could only replicate immense citation advantages (+600%) if relevant predictors were omitted ( McCabe & Snyder, 2014), which indicates a potential spurious effect. When taking into account these relevant predictors, the citation advantage became much smaller (+8%). When the citation advantage is low or non-existent, this could suggest that in those research fields there is a sufficient level of access to the literature such that OA confers no localised access advantage, or that adoption of OA has not yet reached a level where any such advantage has become statistically evident.

One alternative explanation for the existence of citation advantages could be that researchers choose to publish OA when a finding is more impactful, but empirical evidence contradicts this selection effect. Gargouri et al. (2010) compared citation counts for articles which were self-selected as OA or mandated as OA (e.g., by funders). The study concluded that both were cited significantly more than non-OA articles and showed no differences in citation rates. As such, these findings rule out a selection bias from authors as the cause for the citation advantage ( Gargouri et al., 2010). However, research that is selected to merit funding by funding agencies may, in itself, be perceived to be more impactful than research that is not funded. Additionally, as no single OA mandate is ever 100% effective, it might be the simple case that authors are more likely to comply with a mandate for the research they perceive to be of higher impact. In a study of articles in the field of psychology, Anderson (2013a) found that publications with funding sources reported in the text were found to be more highly cited and connected to other highly-cited publications (this type of publication is called "generative" in the study) than publications with no reported funding sources. Furthermore, research that was privately funded was found to be more generative than publicly funded research. In a similar study in the Library and Information Sciences field done by Zhao (2010), the citation counts for grant-funded publications were "substantially higher" than publications without grant funding. Although these studies indicate that grant funding is correlated with increased citation rates, the openness of articles was not addressed in either study. Future research will be required to demarcate the potential causality and to determine the conditions under which we could see whether or not OA has an effect on citation counts. For example, this could be conducted through a randomised controlled trial in which research articles from a particular funder are randomly assigned to OA and non-OA routes, with the citation counts assessed after a certain time.

In sum, evidence indicates that OA is broadly related to increased academic impact in terms of citations ( Figure 2 ; see also McKiernan et al. (2016)), but given the large variability in results, further research should aim to synthesize these findings in a meta-analysis and try to explain the cause of this variability.

Broader societal impact. Scholarly articles also have a societal impact, such as when they are covered in news media or are discussed in social media channels; alternative metrics, or altmetrics, can be used as a guide to measure this mode of impact ( Liang et al., 2014). Information such as social media usage, Mendeley readership, and media attention ( Piwowar, 2013) can be tracked by various altmetrics providers (e.g. ImpactStory, Plum Analytics, and Altmetric.com). As such, when an article generates discussions outside of the academic literature, altmetrics are capable of tracking this. Despite limitations (such as academics discussing their own research on platforms like Twitter), altmetrics provide a general view of the wider societal impact of research articles. Considering the increased pressure on researchers and research institutes to communicate research findings to the public, altmetrics can provide additional insight into which research drives public interest. A working group established by NISO is investigating the future role of altmetrics in research communication and assessment ( www.niso.org/topics/tl/altmetrics_initiative/).

OA articles would be expected to have an altmetrics advantage compared to the non-OA literature; if an article has fewer restrictions for journalists, citizens, businesses, and policy-makers, it seems logical that this would enable the research to be publicly re-used. Furthermore, those parties may be more likely to promote articles which are publicly accessible into different communication channels. In other words, increased access removes barriers to widespread societal engagement, whereas a relative lack of article access discourages engagement.

There is research showing evidence for an altmetrics advantage for OA articles, but this does not reflect itself in the most impactful articles. Wang et al. (2015) found evidence that OA articles receive more attention through social media. The authors compared social media attention (Twitter and Facebook) between OA and non-OA articles at Nature Communications and found that OA articles get 1.2–1.48 times as much social media attention as compared to non-OA articles (see also Adie, 2014). Nonetheless, of the top 100 articles of 2015 as presented by Altmetric.com, only 42 articles were OA ( www.altmetric.com/top100/2015/). This 42% is larger than the overall proportion of OA articles in the literature, which indicates that OA contributes relatively more impact per paper. However, it also indicates that the open impact advantage can be overshadowed by the intrinsic nature of the research published or by the traditionally prestigious journals with a larger and dedicated media apparatus (e.g., Nature, Science; Brembs et al., 2013).

Allen et al. (2013) found that a social media announcement of the release of a research article increases the number of users who view or download that article, but does not translate to increases in the citation count in the field of clinical pain research. Costas et al. (2015) found a relatively weak correlation between social media activity and citation counts for the articles in their sample (over 1.5 million article records), while Mohammadi et al. (2015) found that the number of Mendeley readers with a status of graduate student or faculty correlated with citation counts. When OA to the articles is factored into an analysis, there is a potential recursive relation between citation counts and altmetrics due to OA. Eysenbach (2011) indicated that there is a moderate correlation (0.42–0.72) between the tweets and citations of articles from an OA journal ( Journal of Medical Internet Research). Highly tweeted articles were eleven times more likely to be highly cited than less-tweeted articles, or vice versa (75% of highly tweeted articles were highly cited; 7% of less-tweeted articles were highly cited). However, it is difficult to assess causality in these cases: do research papers that have more academic impact make their way more frequently into societal discussions, or does online discourse increase their potential citation rates? Overall, this evidence implies that there is a general media advantage with OA (see also McKiernan et al. (2016)), which can be used as a proxy or pathway to indicate greater societal impact.

Altmetrics themselves should not be conflated with citations when it comes to assessing impact, even though some providers such as Altmetric.com supply a single score that can be used to rank an article in a similar way to a journal’s Impact Factor. Each measure of altmetrics tells a different story about the impact of research, and a careful understanding of the altmetrics landscape in conjunction with citation-based metrics can lead to a clearer picture of societal impact of scientific research.

Open Access and text- and data-mining

Traditionally, in order to publish a paper, researchers hand over their copyright via a Copyright Transfer Agreement. Copyright transfer as the default has far-reaching consequences on the ability of both the authors and others to re-use that published research, and many authors are not aware of the impact of these transfers on their ownership of the work. Academics frequently give the copyright to the publishers in exchange for the perceived prestige of publishing in one of their venues (e.g., Müller-Langer & Watt, 2010). In some cases, institutes adopt rights-retention OA policies that grant authors non-exclusive rights to their institutes before signing copyright agreements with publishers, which enables them to make articles OA without requiring permission from publishers ( cyber.law.harvard.edu/hoap/Good_practices_for_university_open-access_policies). Essentially, copyright is a pre-digital tool wielded by traditional publishers to maintain revenues rather than fostering creativity, innovation, or protecting authors ( Okerson, 1991; Willinsky, 2002). For example, the Author’s Guild sued Google Books for copyright infringement because they provided freely available digital copies; the court rejected this suit in 2016, stating that Google Books served the public interest and that copyright’s "primary intended beneficiary is the public" ( EFF, 2015). In the digital age, copying is essential to perform necessary research tasks. These activities range from viewing the article (i.e., downloading requires copying) to re-using figures from an article in a book. The interaction of OA and copyright is complex and deserves extended research in itself (e.g., Scheufen, 2015). We will highlight how OA views copyright and relate this to its effects on text- and data-mining (TDM).

The majority of ‘born OA’ journals and publishers do not request or receive copyright from authors. Instead, publishers are granted non-exclusive rights to publish, and copyright is retained by authors through a Creative Commons license (typically CC-BY). Importantly, this represents a power shift from publisher-owned to author-owned rights to research. This model of author-retained copyright appears to be favoured by the majority (71%) of the research community ( Hoorn & van der Graaf, 2006). Shifting copyright to stay with the author, combined with appropriate open licensing, allows for wider re-use, including TDM, and forms the basis for a robust scholarly ecosystem.

As such, copyright in OA publications is non-restrictive and also allows machines to freely access it. In traditional publishing, human reading and computer reading are seen as two separate things which require different agreements, whereas OA publishing views them both in the same, non-restrictive manner. In other words, in order to mine OA journals, one only needs the technical skills to do so. In order to mine closed access journals, one needs to sign or negotiate access conditions, even if legitimate access to the articles has already been bought ( Bloudoff-Indelicato, 2015).

Automated extraction of information from scholarly research via TDM is a methodology that can be applied to investigate the scholarly literature at an enormous scale, creating new knowledge by combining individual findings. This has already proven to be useful for a large variety of applications (e.g., Glenisson et al., 2005; Martone et al., 2016; Swanson, 1987). Moreover, OA publishers facilitate TDM on a massive scale by allowing multiple options for collecting the literature needed. For example, PLOS is non-restrictive and allows users to scrape articles directly from the website or using its API. As a result, scraping tools can be used, such as rplos , an R package developed to search and download full-text scholarly papers ( Chamberlain et al., 2016).

TDM is not only a knowledge-generation tool; it also allows for automated screening for errors and automated literature searches that renew scientific discovery. With TDM it becomes possible to easily compare one’s results with those of the published literature, identify convergence of evidence and enable knowledge discovery ( Natarajan et al., 2006) or discover frequent tentative hypotheses that can be used for new research ( Malhotra et al., 2013). It has already been used to make major advances in fields such as biomedicine ( Gonzalez et al., 2016). TDM also allows for computer applications that can download all scholarly literature given certain search terms (e.g., ContentMine’s getpapers tool ; github.com/ContentMine/getpapers), simplifying and shortening the tedious literature search. TDM can also serve a screening purpose similar to plagiarism scanners, helping to detect statistical errors in the scholarly literature (e.g., Nuijten et al. (2015)). TDM can be used in various innovative ways and is an emerging and rapidly advancing field; non-restrictive licensing through OA certainly promotes its wider application.

Given the exponential increase in the number of scholarly publications, (semi-)automated methods to synthesize results have become increasingly important. TDM decreases the time dedicated to the search for relevant information in scholarly literature by categorizing information ( Leitner & Valencia, 2008), highlighting and annotating relevant results to specific users ( Shatkay et al., 2008), and profiling research ( Porter et al., 2002). Furthermore, TDM also prevents researchers and readers from wasting time on reinventing the wheel, simply because one can no longer keep up with the huge amount of published literature available ( Harmston et al., 2010).

Because of traditional copyright transfers, TDM has often been stymied by traditional, closed access publishers who frequently see it as a copyright infringement. Researchers using software that harvests data from online publications have been (threatened to be) cut off from accessing the articles. These researchers found themselves trapped in negotiations to resume their research (even though their universities had paid subscription fees for access (e.g., Bloudoff-Indelicato, 2015; Van Noorden, 2012)). Standard subscriptions do not permit systematic downloads because publishers fear that their content might be stolen and their revenue therefore lost ( Van Noorden, 2012). In 2014, Elsevier opened its papers for TDM via a proprietary API ( Van Noorden, 2014), but placed restrictions on the researchers using it; however, researchers are not legally required to comply with these restrictions in some countries (e.g., U.K., U.S.A., Handke et al., 2015).

To make the enormous corpus of closed access papers retrospectively available to the public might be possible through legal action at an institutional or governmental level. The Dutch Government, for example, has recognized OA as a right, with Dutch citizens capable to make their scientific publications free to access after a ‘reasonable period of time’ ( Open Access NL, 2015). Such steps are further supported by Shavell (2010) and Eger & Scheufen (2012) who ascertained that transition towards an OA model could not be smooth without first undertaking the necessary legislative steps. The position of institutes regarding copyright transfer remains generally unclear. While academics themselves may have little power in broader debates regarding copyright, institutes could claim ownership of the work by invoking their rights under the work made-for-hire doctrine ( Denicola, 2006). However, OA policies at universities generally use a system of non-exclusive rights, presupposing that faculty are the owners of their work and can grant non-exclusive rights to the university for use (for examples of approaches and language used when drafting OA policies, see ( Shieber & Suber, 2016)). Importantly, this means that OA through the ‘Green’ route does not always depend on permission from publishers, and increasingly is becoming dependent on rights retention by authors, through carefully-drafted and widely-supported university policies. While these are positive steps towards making research available for TDM, in light of the potential copyright problems for closed access articles and the fact that not all research is available through institutional Open Access policies, TDM will be easier and legally safer for OA journals. As a consequence, TDM is likely to be more readily applied to OA literature when compared to closed access literature.

The economic impact of Open Access

The effect on publishers

Any publisher has to cover operating costs, which are primarily made of (i) article processing charges (APCs), (ii) management and investment costs, and (iii) other costs. Article processing includes editing, proofreading and typesetting, among other things. Management and investment are instead the marginal costs needed to establish and keep the journal running. Other costs include promoting the journal, hosting and infrastructural services, sponsoring conferences, and other services that are extrinsic to research articles themselves. The average production cost for a single research article is estimated to be around $3500–$4000 ( Van Noorden, 2013) but these costs are highly depending on the publisher. For example, Philip Campbell (Editor-in-Chief of Nature) stated that his journal’s internal costs were at $20,000–$30,000 per paper ( Van Noorden, 2013), due in part to the high selectivity and rejection rate at Nature (i.e., this is an average cost per published paper, and not the production costs associated with publishing a single accepted paper). However, these are at the high end of the cost spectrum, with other journals, such as the Journal of Machine Learning Research (JMLR) costing between $6.50–$10 per article ( blogs.harvard.edu/pamphlet/2012/03/06/anefficient-journal/). Other publishers are completely transparent about their direct and indirect production costs, such as Ubiquity Press, which levies an APCs of $500 ( ubiquitypress.com/site/publish/). One possible reason for such variation between journals and publishers is that it is generally unclear whether proposed costs relate to those directly involved in article processing or those required in order for a publisher to ‘break even’ if they receive zero subscription income for an article made OA.

In order to cover those costs and make a profit, closed access publishers charge for access via subscriptions, whereas many OA publishers or journals charge to publish. Due to increased subscription costs, closed access publishing is becoming an increasingly unsustainable business model ( Odlyzko, 2013) with prices estimated to have outpaced inflation at 250% in the past thirty years ( www.eff.org/issues/open-access). This will slowly but surely diminish the scope of access to the scholarly literature as fewer organisations are able to pay such high costs. Only recently has any transparency into the detailed costs of subscriptions been gained by using Freedom of Information Requests to bypass non-disclosure agreements between libraries and publishers ( Bergstrom et al., 2014; Lawson & Meghreblian, 2015). These requests provide the basis for understanding the economics of scholarly communication. For example, Bergstrom et al. (2014) found that commercial publishers, including Emerald, Sage, and Taylor and Francis, have prices of ten times the amount of non-profit publishers per citation for PhD-granting institutions. Two potential ways to prevent future retention of an unsustainable model is through decreasing the subscription prices, thereby lowering publishers’ profit margins and the financial burden on subscribers, or through switching to new OA-oriented business models and creating new value. Either way, price transparency will be essential for future bargaining efforts between academic libraries and publishers, and will be of interest to those involved in public policy and scholarly publishing. The concept of transitioning from a subscription-based model to one driven by APCs will be financially appealing to journals that operate with minimal profits or at a loss, and can be a pathway to achieve financial security and long-term journal sustainability. As such, increasing revenues is a strong incentive for OA ( osc.hul.harvard.edu/programs/journal-flipping/public-consultation/4/6/, accessed 26/04/2016).

OA publishing has become associated with a ‘pay-to-publish’ model, whereas around 70% of peer-reviewed OA journals do not charge APCs, according to data from the Directory of Open Access Journals (DOAJ) (see blogs.harvard.edu/pamphlet/2009/05/29/what-percentage-of-open-access-journals-charge-publication-fees/ and citesandinsights.info/civ16i4.pdf). However, approximately 50% of all articles published in peer-reviewed OA journals are published in APC-based venues ( Crawford, 2015; Laakso & Björk, 2012; Walters & Linvill, 2011). Authors paying to publish can be viewed as a fundamental conflict of interest for researchers. Nonetheless, this payment model has proven itself to function properly when editorial decisions are separated from the business-side of the publisher (i.e., editorial independence), removing the problem of ‘publication-bribery’. Additionally, many journals have always levied charges for to cover the costs of publishing regardless of OA; for example, PNAS charges $1225 per regular research article (with an additional $1350 for OA; pnas.org/site/authors/fees.xhtml), and Cell charges $1000 for the first colour figure and $275 for each subsequent one ( cell.com/cell/authors; as of April 2016). Therefore, equating OA with ‘pay-to-publish’ is actually a bit of a misnomer, as several closed journals charge to publish and many open journals do not. Furthermore, many publishers (e.g., PLOS, PeerJ), as well as many learned societies, operate fee waiver schemes for researchers unable to obtain funds to cover publication fees.

For those OA publishers implementing a pay-to-publish model, around 68.8% of publishers offer fee waivers to low- and middle-income countries ( Lawson, 2015), while other journals offer fee discounts often given in lieu of total fee waivers. Solomon & Björk (2012) investigated the sources of funding used by authors for APCs, indicating that these are highly variable across academic disciplines. For example, while 45.5% of authors in Health Sciences, Biology and Life Sciences use grant or contract funding as source for APCs, only 10.4% use this in Business and Economics, with 45.8% coming from personal funds. Other sources include national funding bodies, and discretionary funds administered by institutions, as well as institutional funds specifically in place to support OA policies (see also Dallmeier-Tiessen et al., 2011). Sources for APCs are also highly variable depending on the per capital GNP of the authors’ country, as well as the size of the APC ( Solomon & Björk, 2012). According to MacKie-Mason (2016), one potential outcome of authors seeing the price of APCs and securing funding for them is that authors may begin to take the price of APCs into account (in addition to other factors such as prestige and topic) when selecting a journal for their research output, which may drive market competition and could as a consequence lower the price of APCs. However, a potential negative consequence of an increasingly APC-driven model of OA is that some researchers may struggle to procure funds in order to publish and conform to mandates at different levels. This might impact early-career researchers and those working in fields were research grants and publishing fees are more difficult to obtain.

Subscription-based publishers still frequently produce print versions of their journals, which increases production costs, potentially to justify charging for readership or to satisfy a small demographic who prefers this mode of reading. After all, subscriptions to print journals make sense and, if large-scale printing is still in place, simply transferring this idea to the digital versions creates continuity. Print versions are accompanied by logistical costs to print and ship each issue, but these are partially offset with reprint orders, additional charges for colour figures, and print-based advertising. For some of the largest subscription-oriented publishers the annual net profit on investment reaches up to 40 percent, which makes academic journal publishing highly lucrative for investors ( Satyanarayana, 2013), further increases investment to sustain this type of publishing model, and allows maintenance of an oligopoly ( Larivière et al., 2015).

OA publishers only publish digitally and have opened up avenues for innovation. For example, PeerJ has introduced a wholly different OA business model, where readers pay nothing to access articles, but authors pay a membership fee once to publish for a lifetime. The Open Library of Humanities (OLH) is another innovative business model in which libraries pay a small fee to support OLH and scholars are able to publish for free (subscription for publishing rather than subscription for access); this support also enables the OLH to help journals transition from a subscription model to OA (for example, the recent case of Lingua; timeshighereducation.com/research-intelligence/open-library-humanities-aims-flip-journals-open-access). Library publishing has also developed in response to the OA movement; in this model, academic libraries begin publishing operations in the interest of providing added value to their patrons and contributing to the growth of knowledge ( librarypublishing.org). In terms of innovating in the publishing platform itself, eLife have introduced the Lens as a novel way of viewing research articles online ( lens.elifesciences.org/about/), and F1000Research has introduced so-called ‘living figures’ to enable researchers to interact with data underlying research findings (e.g., Colomb & Brembs, 2015).

With this innovation comes massive scope for reducing the costs associated with publishing through implementing more efficient procedures. In this case, costs are reduced by eliminating the need for type-setting and copy-editing, with web-hosting costing only $15/year, and a total operating cost of between $6.50–$10.50 per article. Other platforms such as ARPHA offer an end-to-end XML-based publishing service, utilised by publishers including Pensoft, with a more efficient and integrated publishing workflow, which should highlight and reduce the real costs of publishing. In addition, OA has the potential to increase the speed of publication, as seen in journals like eLife and PeerJ ( blog.dhimmel.com/plos-and-publishing-delays/), which combined with ‘pre-print’ servers like biorXiv and platforms that offer post-publication peer review like Research Ideas and Outcomes ( riojournal.com/), F1000 Research, and ScienceOpen ( www.scienceopen.com/), can exceptionally accelerate the speed of research communication. Such innovations add value to the research communication process (contrary to services such as paying to print colour figures) and represent just several cases of innovation across the publishing ecosystem. One can imagine that publishing costs in OA journals become dependent on the value added on a per-article basis, which can help reshape and improve scholarly communication. As such, making publication costs dependent on the value added aligns the interests of publishers with those of scholars, where improving the quality of the process of scholarly communication is the end goal. The motivation behind this could come from the currently available data, which suggest that hybrid publishing options offered by traditional publishers, while being of higher cost due to supposed prestige, provide a much lower overall quality publishing process ( blog.wellcome.ac.uk/2016/03/23/wellcome-trust-and-coaf-open-access-spend-2014-15/). It is noteworthy that in spite of the higher costs of hybrid publishing compared to ‘pure’ or ‘born’ OA publishing, some reports, such as the highly influential and somewhat controversial Finch Report in the UK ( www.researchinfonet.org/publish/finch/), favoured the former model and high-priced Gold OA over a Green model.

The effect on non-publishers

The implementation of OA models has implications beyond the publishing industry in terms of economics. Research funding comes from multiple sources, including national funding agencies and industries, as well as private funders. Much primary research actually takes place outside of academia, inside R&D departments; if R&D in the private sector can access more research findings, this will ultimately benefit the public interest as well. A report from 2004 by Arzberger and colleagues into the scientific, social and economic development of access to research results concluded that access should be promoted to the largest extent possible. According to this report, access to research results can only be responsibly restricted in the case of national security, privacy, or those involving IP rights of the authors ( Arzberger et al., 2004). A major principle underlying this is the ownership of research results: publicly funded research and data are public goods and because they have been produced in the public interest they should be considered and maintained as such. Indeed, such a principle has become one of the focal rallying points of the global OA movement. Appropriate licensing and accessibility can influence re-use through commercialization, and can empower citizens and industry to recognize great economic benefits. This apparently resonates with many organisations, as indicated by the increased numbers of OA policies on a global basis (see Figure 3 ).