Universities in the UK are funded for research under the Haldane Principle1 of dual support. Core funding is provided by the Funding Councils to provide for a ‘well founded laboratory’ and to enable universities to undertake ‘internally’ funded research as they see fit. Additional project-specific funding is provided by other funders such as the Research Councils (RCs)2, charities, companies, governments and other bodies like the European Commission (for example with their recently launched Horizon 20203 programme).
Since the late 1980s, the core funding for UK universities has been informed by the results of research assessment exercises, the most recent of which, the Research Excellence Framework (REF20144), is currently being undertaken. Panels of academic and user experts are now assessing research and research related information submitted last November and the results will be published in December 2014. The results of these assessments will directly drive this core funding (known as ‘quality related’, or ‘QR’ funding) part of the dual support system for the next six years from 2015/16 onwards. To give an indication of the importance of this exercise to UK universities, the mainstream QR funding for 2013/14 (informed by the results of the previous exercise) from the Higher Education Funding Council for England (HEFCE) alone, i.e. for England, was £1.018 billion5.
Finally, it should be noted that most project funders do not pay the full economic cost (fEC6) of the research that they support (for example most Research Council projects are funded at about 80% of fEC). The balance is provided by the universities themselves; so availability of QR funding also impacts on the ability of universities to attract and resource externally funded projects.
Research Excellence Framework
The full rules for REF2014 can be found on the REF website7; a very brief summary is provided here.
“… most project funders do not pay the full economic cost …”
Each UK (for the sake of brevity) university is invited to submit research outputs, research impacts and research environment statements from the 36 specified subject areas. For the purposes of this article, we will focus on the research outputs (normally publications), which count for 65% of the overall result of the assessment. For each member of staff selected for submission, up to (and normally) four outputs are submitted for assessment. For the 2014 exercise there is no requirement for any types of outputs to be open access (OA), just that they were first in the public domain (although confidential reports are allowable) between 1 January 2008 and 31 December 2013. Each output is assessed on a five-point excellence scale (four star, three star, two star and one star, down to unclassified) for originality, significance and rigour. Whilst the funding formula for the 2014 results has not yet been decided, in the 2008 exercise, a top-rated ‘world-leading’ (four star) output was, in some science areas, worth over £20,000 per annum in QR8. Bearing in mind that this may have been a journal article from 2001, and it will still be bringing in £20k in QR funding in the academic year 2014/15, it is clear that the decisions made for REF2020 will have far-reaching financial consequences. For a fuller explanation of the funding formula see the HEFCE website9; other national Funding Councils formulae are similar.
Proposals for OA requirements for REF2020
In Feb 2013, HEFCE issued a letter on open access and submissions to the REF post-201410, which was in effect a pre-consultation document affecting the wording in the following full consultation11 from July to October 2013. A short discussion is provided by Sabaratnam and Kirby in their blogpost12. The final consultation questions certainly indicate the direction of travel, and in early 2014 we are likely to know what the final open access requirements for REF2020 will be. As a consequence, some of the discussion that follows may be outdated by the time you read this, but it is hoped that the conclusions will still be relevant.
“… a ‘one size fits all’ rule may not be appropriate.”
The first thing to make clear is that any requirements for REF2020 will be restricted to research outputs that are either journal articles or conference proceedings. They exclude monographs, book chapters – although there are indications that these will be included in the following assessment exercise, REF2026 – and other forms of research outputs such as patents and installations. Journal articles and conference proceedings are, however, still a large majority of the outputs. In RAE200813 77% (of the 214,287) outputs submitted fell into these two categories. Given that this was an increased percentage from the 73% in RAE200114 and that the overall submission size of REF2014 is about the same as RAE2008, then the number of outputs of these types for REF2020 is likely to be in the order of 150,000 (although, as we shall see later, not all will need to be OA). It should also be noted that the distribution of output types varies quite dramatically by subject area: in RAE2008 ranging from 99% in chemistry down to 22% in art and design. So a ‘one size fits all’ rule may not be appropriate.
One main proposed criterion for REF2020 is that an article (or conference proceeding – hereinafter we will use ‘article’ to mean both types) that was not originally open access and subsequently made so would not be eligible; indeed the intention is that ‘the primary objective of this proposal is to stimulate immediate open-access publication’. This means that universities cannot wait until near the submission in 2019 and make just those articles that they wish to submit open access. A much more strategic view must be taken. Given the limited funding for ‘gold’ OA, universities must carefully manage15 what funds they have if they are not to prejudice future REF submissions. However, as the proposals are ‘colour blind’, it does provide a useful lever for universities to further their ambitions for ‘green’ OA and in particular to promote and develop the use of institutional repositories (IRs). Specifically, the proposal indicates that articles must be available through the university's (or a shared) IR. One major headache for research managers and administrators (RMAs, and we include library professionals in this group) is determining and ensuring eligibility of an article against this timing issue of acceptance/publication, and what scope for leeway there might be. It has not yet been decided which date (acceptance or publication) will be used; there are pros and cons for both. Perhaps the biggest issue for RMAs will be showing/proving that the article was OA at that time. If some administrative delay/error meant that for whatever reason it was not made (say) gold OA until the day after, would that really mean that the article would be ineligible? Surely the intention is to promote the spirit of OA, but what leeway might be appropriate? A month might be reasonable, particularly given the unproven nature of current OA and article processing charge-(APC-)related processes, let alone these future REF issues.
“Surely the intention is to promote the spirit of OA …”
The issues relating to the availability of articles through an IR are relatively unproblematic except perhaps for the definition of what a shared IR might be, or indeed whether multiple IRs would be allowable. Also, the definitions of the requirement for articles to be searchable, allow reuse and be text minable are ill defined. It might be assumed that the text in a diagram used in an article need not be searchable (i.e. it could be an image) for the article to be REF2020 OA compliant, but where would the limits lie, and how might this be checked? There is a related issue of allowable embargo periods for green OA. The proposal is that Research Councils UK (RCUK)-permitted embargo periods are used, but these differ: some allow up to six months and others 12, and there is not a one-to-one mapping of RCs to REF units of assessment (UOAs). Also, RCUK is currently allowing longer embargo periods (12 and 24 months) for a transitional period: what will happen during the time around the change in allowable embargo periods? The issue is further complicated by interdisciplinary research, as highlighted for example in the ARMA16 response to the consultation.
“… when is a journal article not a journal article?”
Unlike the RCUK policy, there is no requirement in the current REF2020 proposals to make the underlying data freely available, so this is not an issue for RMAs (at least for REF2020) at the moment. However, it seems likely that this will be a future requirement (possibly for REF2026, if there is to be one) and so probably the problems have only been deferred.
Given the international nature of research, there are a great many articles submitted for assessment that have international authors. Such papers are exempt from the OA requirement so there should be no barrier to attracting high quality international staff to the UK, who might have a number of articles published through non-OA routes prior to coming to the UK, which would otherwise be ineligible for REF2020. However, the current proposal is that only those articles that do not have a UK higher education institution (HEI) address would be exempt. This would mean, for example, that if (say) Dr Hix from the Unseen University (a non-UK HEI) has co-authored a paper with Dr Piercemuller from Poppleton University (in the UK) and the latter had not arranged for the article to be REF2020 OA compliant then, when Dr Hix comes to the UK (perhaps to a different university) his article would not be eligible. However, if he had not co-authored with Poppleton then the article would be eligible. This is perhaps an edge effect, but does point to problems with full compliance, which is anticipated by the proposal.
To take another example, when is a journal article not a journal article? Some RAE submissions, quite legitimately classified a combined output (such as a journal article and editorial; or two related journal articles) as type ‘other’ rather than ‘journal article’ as it better described the content. I am sure that no unscrupulous REF2020 submissions would include non-OA-compliant articles as type ‘other’ in order that they become eligible, but should the combined type described above need the journal article part(s) to be OA compliant in order for the research output to be eligible? How would this be checked, and by whom?
The HEFCE proposal states that there should be a two-year period of grace, with the requirement only coming into force from January 2016. This would allow for preparation time and for articles already in the publication/ review process not to be rendered ineligible. It should be noted that there are less than two years until January 2016 and that each passing day allows less time for the preparation of institutional processes and procedures, for universities as well as publishers and intermediaries, as well as general culture change.
“… there should be a two-year period of grace …”
To ameliorate a number of the issues identified above, the proposal is that 100% compliance for articles (with UK HEI authors, from 2016 onwards) will not be required. In reality this will mean either identifying specific exceptions or a blanket compliance level. From an administrative viewpoint the latter would be preferable; defining, justifying and evidencing exceptions would be very burdensome. A similar process, for ‘individual staff circumstances’ (to make the case for staff submitting fewer than four research outputs without penalty), was necessary for REF2014, and proved to be time consuming and difficult for both academics and RMAs.
It is hoped, therefore, that the requirement will be that a percentage of articles submitted must be OA compliant. Given the different maturity of OA in differing subjects, it seems prudent to have different percentage compliance levels depending on the subject area. This could be as fine-grained as by UOA, but perhaps (as is one of the suggestions in the proposal) a compliance level for each of the four main panels would be sufficient. Given the other differences between the main panels, this would not present an administrative overhead for RMAs compared to that for a single overall compliance level, and would allow for disciplinary differences. It has been suggested17 by Professor Adam Tickell, Provost and Vice-Principal of the University of Birmingham, that the proposed compliance levels would not be too difficult to achieve, based on the University's REF2014 submission, as the majority of journals allowed for OA. However, the question of affordability is still key.
“… defining, justifying and evidencing exceptions would be very burdensome.”
The proposed requirements for open access compliance for journal articles and conference proceedings for REF2020 provide a huge imperative for UK HEIs to embrace and support OA. Whilst there appears to be sufficient time for institutions to prepare and adapt their processes and procedures in the next few years, there are sure to be many hurdles and unexpected issues on the way. The sector must work together to try and address as many of these as possible. The support of Jisc working with other partners such as ARMA18, RLUK19, funders and publishers, utilizing their new co-design approach20, will be crucial.
There is a great potential to over-engineer the solution to the OA REF2020 compliance problem (or rather, opportunity) and to produce complex administrative systems. We must remember that this initiative gives us a chance to promote and support open access, where appropriate. It must not become a stifling, self-serving process dragging the valuable time of academic staff away from the important task of actually undertaking the research itself. Nevertheless, it should be recognized that, however light touch the process is, there will still be a large administrative burden; of the estimated 150,000 articles to be submitted, around 100,000 would be post-2016, and so, perhaps 70,000 would need to be OA or otherwise exempt (but accounted for). Research managers and administrators will be holding their breath in anticipation!
“… there are sure to be many hurdles and unexpected issues on the way.”