Do Authors Comply with Mandates for Open Access?

Home , Our Research

COMMENT GOVERNANCE Make more ART Pre-Raphaelites LAB LIFE Memoir of PUBLISHING Engage more voices use of the patenting system interpreted discoveries neuroscientist and equality in the debate over Europe’s to regulate gene editing p.486 of a fecund age p.490 advocate Ben Barres p.492 open-access plan p.494 ILLUSTRATION BY SÉBASTIEN THIBAULT SÉBASTIEN BY ILLUSTRATION

Do authors comply with mandates for open access? The first large-scale analysis of compliance with open-access rules reveals that rates vary greatly by funder, report Vincent Larivière and Cassidy R. Sugimoto.

ast month, European research funders is open access1–4. Here, we report the first They highlight the importance to open access collectively called for research publica- large-scale analysis of compliance, focusing of enforcement, timeliness and infrastructure. tions to be made free, fully and immedi- on 12 selected funding agencies. Bibliometric And they underline the need to establish sus- Lately; so far, 14 funders have signed up. Before data are fraught with idiosyncrasies (see tainable and equitable systems as the financial that, at least 50 funders and 700 research insti- ‘Analysis methods’), but the trends are clear. burdens for science publishing shift from tutions worldwide had already mandated Of the more than 1.3 million papers we research libraries to authors’ research funds. some form of open access for the work they identified as subject to the selected funders’ support. Federally funded agencies and insti- open-access mandates, we found that some FREE FOR ALL tutions argue that taxpayers should be able two-thirds were indeed freely available to read. Funders with open-access mandates have to read publicly funded research, and that Rates varied greatly, from around 90% for varying incentives, opt-out mechanisms, broader accessibility will allow researchers work funded by the US National Institutes of copyright protections, deposit guidelines whose institutions do not subscribe to a par- Health (NIH) and UK biomedical funder the and other associated infrastructures and ticular journal to build on existing research. Wellcome Trust, to 23% for work supported by requirements. These affect when, how and However, few empirical analyses have the Social Sciences and Humanities Research how much work is made open. Our analysis examined whether work supported by fund- Council of Canada (see ‘Mandates matter’). did not assess licensing and instead counted ing agencies with such mandates actually Our findings have policy implications. articles found to be freely available to

COMMENT

MANDATES MATTER About two-thirds of papers under open-access (OA) mandates are free to read*, either from repositories (green OA) or journal websites (gold OA), with US funders favouring repositories. Of open papers, about half are available by both routes.

VERY VARIED ACCESS GREEN AND GOLD Rates of compliance vary greatly by funder, although they mostly trend Since 2009, the proportion of papers available to read* as both gold and green upwards. Dips in 2017 are due to embargoes (which delay access for has soared, even as the proportion of green-only access has stayed relatively xed periods after publication). constant, and gold-only has dropped.

Before mandate adoption Gold only Green and gold Green only No open access identi ed

UNITED KINGDOM AND EUROPEAN UNION 19% in 2016 16% 21% 12% (69% in 2009) (5%) (51%) (10%) 100 Wellcome Trust 80 MRC EPSRC ESRC BBSRC 60 United United EPSRC Kingdom Kingdom 40 ESRC ERC 20 Finch report published Other UK 25% 40% 19% 48% (20%) (6%) (22%) (17%) Other EU 0 Papers under OA mandates (%) under OA Papers 10% 4% 8% 3% (33%) (18%) (21%) (11%) UNITED STATES AND CANADA & C.R.S. V.L. ANALYSIS: WEB OF SCIENCE/ROARMAP. ANALYTICS UNPAYWALL/CLARIVATE SOURCES: DATA

100 NIH Wellcome Gates MRC 80 Trust United CIHR United Many papers under embargo Kingdom 60 NSF Kingdom NSERC 40 Other Canadian 10% 76% 9% 80% (9%) (40%) (11%) (57%) 20 Other US SSHRC 0

Papers under OA mandates (%) under OA Papers 53% 6% 12% 2% 2009 2010 2011 2012 2013 2014 2015 2016 2017 (53%) (8%) (10%) (3%) FUNDER EFFECT Even within the same discipline, access varies greatly by funder. NSF 17% NIH Of chemistry papers supported by the NIH, 81% were open United (17%) United access; 24% of NSF’s chemistry papers were. States States

Papers available (%) 24% 20 50 100 0 25 50 75 100 46% 40% (22%) (42%) (45%)

17% 9% 27% 7% (27%) (13%) (54%) (8%)

Biomedical research Clinical medicine Health Mathematics Earth and space Psychology Physics Biology Professional elds Social sciences Chemistry Engineering and technology All disciplines Gates ERC Wellcome Trust Multicountry Multicountry NIH MRC Gates 11% 63% 29% 37% (17%) (43%) (19%) (19%) BBSRC ESRC 79% 7% 45% 12% ERC (75%) (7%) (46%) (23%) CIHR 7% EPSRC (7%)

NSF SSHRC CIHR NSERC Canada Canada SSHRC 7% All funded papers (11%) 9% 34%

MRC, Medical Research Council (UK); BBSRC, Biotechnology and Biological Sciences Research Council (7%) (24%) (UK); EPSRC, Engineering and Physical Sciences Research Council (UK); ESRC, Economic and Social Research Council (UK); ERC, European Research Council; NIH, US National Institutes of Health; Gates, Bill & Melinda Gates Foundation; NSF, US National Science Foundation; CIHR, Canadian Institutes of *Our analysis counted papers freely available to read on publishers’ websites as gold and those in Health Research; NSERC, Natural Sciences and Engineering Research Council (Canada); SSHRC, Social repositories as green. It did not consider conditions of reuse or whether free access happened at the Sciences and Humanities Research Council (Canada). same time as publication. Shown over time in Supplementary Information Figure S3.

COMMENT

read. We assessed whether access was gold (available on a journal website) or green ANALYSIS METHODS (available in a repository, such as PubMed Central, the preprint server arXiv or else- How we mined data on open-access compliance where, sometimes with a delay or ‘embargo’ of up to a year after publication). About half We first identified the funding sources as on Sci-Hub). As of April 2018, Unpaywall (47.5%) of open papers are both green and of papers using the published provided the open-access status of nearly gold (see ‘Green and gold’). acknowledgements (mandated by most 96 million scholarly documents. Both the NIH and the Wellcome Trust state funders). These have been indexed by Of the 12,495,074 journal articles we that they will withhold or suspend payments the Web of Science (WoS) since 2008 for matched with Unpaywall using DOIs, if articles are not made open access, although science and medicine, and since 2015 for 1,352,918 acknowledged funding from it is unclear whether they have done so. These social-sciences articles. There is no uniform 1 of the 12 funders we identified. agencies provide convenient repositories for format, so we looked for variations of To determine rates of compliance, we depositing articles: PubMedCentral for NIH- agency names (such as ‘NSF’ and ‘National matched Unpaywall data to our set of WoS funded work, and Europe PubMed Central in Science Foundation’) and aggregated these. articles and analysed them by funder and the case of Wellcome. Their policies encour- Next, we used Unpaywall, a platform discipline. WoS includes papers published age compliance and allow authors to publish that helps researchers to find open-access in about 12,500 journals annually, so some in journals that do not permit articles to be articles. It identifies the population of funded work is in journals not covered available immediately without a subscription. scholarly papers using the list of unique by our analysis, especially in the social Although articles must be in a repository at digital object identifiers (DOIs) registered by sciences and humanities. Our ability to the time of publication, free access can occur Crossref, a non-profit indexing organization. assign funders to papers is imperfect, later. For example, a paper with a 12-month Unpaywall mines all journal websites listed in given the various ways in which funder embargo published in the March 2016 issue the Directory of Open Access Journals, along names appear and because authors do not of a journal would become freely available in with databases such as PubMed Central always provide funding information. Rates the repository in March 2017. and 50,000 other journal websites and of estimated compliance are likely to be Funders that allow authors to deposit repositories. It intentionally excludes papers conservative; there might be funded papers papers after publication see lower rates of that are available on social-networking sites that are freely available online but which compliance, presumably because authors (such as ResearchGate) or illegally (such could not be found by Unpaywall. V.L. & C.R.S. lose track of this obligation. The Canadian Institutes of Health Research (CIHR) mandated deposit on publication from 2008 to between 2009 and 2016; the Engineering and NEXT STEPS 2015, but dropped this requirement when Physical Sciences Research Council went If funding agencies have their own data on the three main Canadian research councils up by 50 percentage points. This follows the compliance, the information should be adopted a joint, harmonized policy. Com- publication of the Finch report in 2012 (see openly published so that it can be used in pliance for CIHR-funded studies went from go.nature.com/2yojrkc) by a working group assessments of the march of open access, such around 60% in 2014 to around 40% in 2017. of academics, funders and publishers that as ours. That would also allow comparisons Other funders that have lower rates of was established in 2011 by David Willetts, to be drawn. Future research on compliance compliance than the NIH and the Wellcome then the UK science minister. It strongly with open-access mandates should evalu- Trust provide less enforcement and infra- recommended increasing access to research ate the utility of other data sources, such as structure. For example, the US National through article-processing charges and gold Scopus, 1findr, Kopernio and Dimensions Science Foundation (NSF) called for “volun- open access rather than by archiving papers (run by Digital Science, a firm operated by tary compliance” with open-access mandates in repositories. For the next assessment of the Holtzbrinck Publishing Group, which until early 2016 — its rate was around 47% in research institutions in England in 2021, has a share in Nature’s publisher. Nature is 2016. Its repository, which uses infrastruc- major UK funders have now decided to con- editorially independent of its publisher). We ture developed by the US Department of sider only open-access publications. must also create stronger reporting systems so Energy, has less visibility and fewer articles that these data are more readily available for than PubMedCentral does. However, that FIELD CULTURE analysis. This involves collaboration between might soon change, because deposition in We find variations by discipline, with nearly funders, publishers and indexers. Reporting this public repository is now mandatory for full compliance in biomedicine, clinical medi- should allow for analyses at the level of funded papers arising from NSF funding awarded cine and health research. The social sciences, projects, which would provide information after January 2016. Compliance at the CIHR chemistry and engineering all show lower on the time between funding and open dis- is hampered by similar barriers. Unlike rates (see ‘Funder effect’). Within the same semination. On a broader level, more research PubMed Central in the United States, Pub- discipline, compliance varies drastically by is needed to understand what makes scientists Med Central Canada was never the dedicated funding agency. For example, in chemistry comply with funder mandates and why. infrastructure for Canadian medical papers. research, 81% of work funded by the NIH Ultimately, open access needs a sustain- The Canadian repository faltered and then is publicly available, whereas that is true of able financing model. Libraries and other closed in February this year, and no strong only around one-quarter of chemistry stud- organizations have historically borne the environment of enforcement has arisen. ies supported by the NSF and CIHR. Differ- cost of publishing through subscription Factors include lower funding in Canada ent funders support different types of work, fees. Gold open access displaces those costs compared to the United States, which makes but the variations we found also remain con- on to authors (who often need to allocate it harder for authors to allocate funds for sistent within sub-disciplines (see Supple- funds from their research budgets to cover article-processing fees. mentary Information, Figure S5). Although publishing), even as libraries continue to The United Kingdom has seen a steep researchers cite norms and needs within dis- shell out for subscription fees. The cost of rise in open-access compliance across all ciplines as a reason not to comply with open- publishing in open-access journals ranges agencies (see ‘Very varied access’). Rates at access mandates, we believe that the funding from less than US$100 to more than $5,000 all four of the UK research councils stud- agency is a stronger driver of open access than per article, with dominant publishers such ied went up by at least 20 percentage points is the culture of any particular discipline. as Elsevier averaging $2,612 per paper in

COMMENT

article-processing charges and Springer To move the conversation forward, we Vincent Larivière is associate professor at the Nature (which publishes Nature) averaging need a greater sense of the implications University of Montreal and associate scientific $1,913 (see go.nature.com/2cn3zuy). The of open access on the scientific system’s director of the Observatory of Science and system as a whole risks charging multiple financial structure. We must study how Technology, Montreal, Canada. Cassidy R. actors for the same product, and could price certain publishing models will put pres- Sugimoto is associate professor of informatics some places and people out of publishing. sure on some parts of the system while at Indiana University Bloomington, USA. Advocacy must be balanced with evidence alleviating it from other areas, or even e-mail: [email protected] in the open-access debate. Our research enriching them. We need to ensure that 1. Nature 508, 161 (2014). demonstrates that funders can clearly shape the mandates are sensitive to financial 2. De Groote, S. L., Shultz, M. & Smalheiser, N. R. compliance through their mandates, and inequity across countries, disciplines, PLoS ONE 10, e0139951 (2015). 3. Gargouri, Y., Larivière, V., Gingras, Y., Carr, L. that this compliance needs to be monitored. institutions and researchers. & Harnad, S. Preprint at https://arxiv.org/ Real barriers — such as infrastructure and Universities, industry and funding abs/1206.3664 (2012). funding — must be overcome to make agencies should think collectively about 4. Vincent-Lamarre, P., Boivin, J., Gargouri, Y., Larivière, V. & Harnad, S. J. Assoc. Inform. Sci. mandates efficient. However, the rhetoric robust and scalable models. Cooperation Technol. 67, 2815–2828 (2016). surrounding disciplinary barriers might and foresight are the only ways to ensure that be more a myth than a reality: when the everyone has open access to research — both Supplementary Information accompanies this article: see go.nature.com/2yitfgn. C.R.S. did this proper structure and incentives are in place, for readers who want to consume it, and for work while at the US National Science Foundation, researchers comply. authors who wish to publish it. ■ funded by an independent research programme. IMAGINECHINA/REX/SHUTTERSTOCK

A gene-edited ‘micropig’ was developed in 2015 by the BGI genomics institute in Shenzhen, China. Use the patent system to regulate gene editing Governments should use patents to shape the deployment of CRISPR–Cas9 as they have done for past technologies, argues Shobita Parthasarathy.

ext month, researchers, policy Washington DC nearly three years ago, ethical, economic and environmental makers, ethicists and social researchers have continued to apply concerns raised by manipulating plant and scientists will meet in Hong Kong the versatile gene-editing technology animal genomes, including our own. But, so Nfor the second International Summit on CRISPR–Cas9 to diverse domains — from far, governments have struggled to develop Human Gene Editing. crop enhancement and pest eradication viable approaches to regulation. Since the first summit, held in to human disease. Many have flagged the A crucial part of the arsenal for shaping