arXiv and REF – together at last?

New draft REF2021 guidance was released for consultation on Monday morning. Buried half-way through this daunting 139 page document was an update to the REF Open Access policy.

This revised policy comes on the back of Research England’s report Monitoring sector progress towards compliance with funder open access policies which was released in June, and on which we have already commented.

From an Open Access perspective, additional flexibility for preprint servers has been added to the policy:

The funding bodies recognise that many researchers derive value from sharing early versions of papers using a pre-print service. Institutions may submit pre-prints as eligible outputs to REF 2021 (see Annex K). Only outputs which have been ‘accepted for publication’ (such as a journal article or conference contribution with an ISSN) are within scope of the REF 2021 open access policy. To take into account that the policy intent for ‘open access’ is met where a pre-print version is the same as the author accepted manuscript, we have introduced additional flexibility into the open access requirement: if the ‘accepted for publication’ text, or near final version, is available on the pre-print service, and the output upload date of the pre-print is prior to the date of output publication, this will be considered as compliant with the open access criteria (deposit, discovery, and access).

That’s a significant adjustment to previous advice and will be of considerable relief to many researchers who routinely publish their research in this way. Indeed, we have lobbied behind the scenes on this policy issue for more than three years.

But what does this actually mean and what should institutions and authors take from this?

Repositories, preprint servers – what’s the difference?

Firstly, this policy legitimises preprint servers (like arXiv, bioRxiv, SocArXiv and many more) and allows authors to use these systems without needing to worry about technical requirements.

This is in stark contrast to the way institutional and subject repositories are treated by the policy.  These repositories must meet all the requirements of the REF Open Access policy to be considered compliant, which is fine for most institutions because meeting the policy requirements is vital, but subject repositories are usually left in the lurch:

Individuals depositing their outputs in a subject repository are advised to ensure that their chosen repository meets the requirements set out at paragraphs 224 to 241 in this policy. REF 2021 guidance will not certify the repositories which fulfil policy requirements.

We’re still not sure if Europe PMC is compliant, for example.

Don’t just sit there!

However, just because preprint servers are okay, doesn’t mean that authors using preprint servers should assume they don’t need to do anything. There are two significant caveats to take note of:

  1. the manuscript deposited in the preprint server must be the “‘accepted for publication’ text”; and
  2. the manuscript must be uploaded prior to first publication.

Determining the deposit time is usually straightforward, so institutions will be able to monitor this aspect of the policy with some level of automation (especially for arXiv which is harvested by a range of publication systems).

However, the key challenge will be determining the manuscript version. We’ve previously described the work we do as manuscript detectives, so some level of checking with authors will still need to take place.

We are working internally at Cambridge on what our workflow will be to capture these outputs and we will be talking to our researchers on what they need to do or not once this is determined. We still encourage all of our researchers to upload manuscripts when accepted for publication until we indicate otherwise.

Regardless,

If there is one key recommendation we would make to all users of preprint repositories – annotate or label the records to clearly indicate the manuscript version (e.g. submitted, accepted, published).

It will help us, and you, in the long run.

Published 25 July 2018
Written by Dr Arthur Smith
Creative Commons License

‘No free labor’ – we agree.

[NOTE: The introductory sentence to this blog was changed on 27 June to provide clarification]

Last week members of the University of California* released a Call to Action to ‘Champion change in journal negotiations’ which references the April 2018 Declaration of Rights and Principles to Transform Scholarly Communication.  This states as one of the 18 principles:

No free labor. Publishers shall provide our Institution with data on peer review and editorial contributions by our authors in support of journals, and such contributions shall be taken into account when determining the cost of our subscriptions or OA fees for our authors.”

Well, this is interesting. At Cambridge we have been trying to look at this specific issue since late last year.

The project

Our goal was to have a better understanding of the interaction between publisher and researcher. The (not very imaginatively named) Data Gathering Project is a project to support the decision making of the Journal Coordination Scheme in relation to subscription to, and use of, academic journal literature across Cambridge.

What we have initially found is that the data is remarkably difficult to put together. Cambridge University does not use bibliometrics as a means of measuring our researchers, so we do not subscribe to SciVal, but we have access to Scopus. But Scopus does not pick up Arts and Humanities publications particularly well, so it will always be a subset of the whole.

Some information that we thought would be helpful simply isn’t. We do have an institutional Altmetric account, so we were able to pull a report from Altmetric of every paper with a Cambridge author held in that database.  But Altmetric does not give a publisher view – we would have to extract this using doi prefixes or some other system. 

Cambridge uses Symplectic Elements to record publications from which, for very complicated reasons, we are unable to obtain a list of publishers with whom we publish. As part of the subscription we have access to the new analysing product, Dimensions. However, as far as we have managed to see, Dimensions does not break down by publisher (it works at the more granular level of journal), and seems to consider anything that is in the open domain (regardless of licence) to be ‘open access’. So figures generated here come with a heavy caveat.

We are also able to access the COUNTER usage statistics for our journals with the help of  the Library eresources team. However these include downloads for backfiles and for open access articles, so the numbers are slightly inflated, making a ‘cost per download’ analysis of value against subscription cost inaccurate.

We know how much we spend on subscriptions (spoiler alert: a lot). We need to take into consideration our offsetting arrangements with some publishers – something we are taking an active look at currently anyway.

Reaching out to the publishing community

So to supplement the aggregated information we have to hand, we have reached out to those publishers our researchers publish with in significant quantities to ask them for the following data on Cambridge authors: Peer Reviewing, Publishing, Citing, Editing, and Downloading.

This is exactly what the University of California is demanding. One of the reasons we need to ask publishers for peer review information is because it is basically hidden work. Aggregating systems like Publons do help a bit, although the Cambridge count of reviewers in the system is only 492 which is only a small percentage of the whole. Publons was bought out by Clarivate Analytics (which was Thompson Reuters before this and ISI before that) a year ago. We did approach Clarivate Analytics for some data about our peer reviewing, but declined to pay the eye watering quoted fee.

What have we received?

Contrary to our assumptions, many of the publishers responded saying that this information is difficult to compile because it is held on different systems and that multiple people would need to be contacted. Sometimes this is because publishers are responsible for the publication of learned society journals so information is not stored centrally.  They also fed back that much of the data is not readily available in a digestible format. 

Some publishers have responded with data on Cambridge peer reviewers and editors, usage statistics, and citation information. A big thank you to Emerald, SAGE, Wiley, the Royal Society and eLife. We are in active correspondence with Hindawi and PLOS. [STOP PRESS: SpringerNature provided their data 30 minutes after this blog went live, so thanks to them as well].

However, a number of publishers have not responded to our requests and one in particular would like to have a meeting with us before releasing any information.

Findings so far

The brief for the project was to ‘understand how our researchers interact with the literature’.  While we wrote the brief ourselves, we have come to realise it is actually very vague. We have tried to gather any data we can to start answering this question.

What the data we have so far is helping us understand is how much is being spent on APCs outside the central management of the Office of Scholarly Communication (OSC). The OSC manages the block grants from the RCUK (now UKRI) and the Charities Open Access Fund, but does not look after payments for open access for research funded by, say the Bill and Melinda Gates Foundation or the NIH. This means that there is a not insignificant amount of extra expenditure on top of that  coordinated by the OSC. These amounts are extremely difficult to ascertain as observed in 2014.

We already collect and report on how much the Office of Scholarly Communication has spent on APCs since 2013. However some prepayment deals makes the data difficult to analyse because of the way the information is presented to us. For example, Cambridge began using the Wiley Dashboard in the middle of the year with the first claim against it on 6 July 2016, so information after that date is fuzzy.

The other issue with comparing how much a publisher has received in APCs and how much the OSC has paid (to determine the difference) is dates. We have already talked at length about date problems in this space. But here the issue is publisher provided numbers are based on calendar years. Our reporting years differ – RCUK reports from April to March and COAF from October to September, so pulling this information together is difficult.

Our current approach to understanding the complete expenditure on APCs, apart from analysing the data being provided by (some) publishers, is to establish all of the suppliers to whom the OSC has paid an APC and obtain the supplier number. This list of supplier numbers can then be run against the whole University to identify payments outside the OSC.

This project is far from straightforward. Every dataset we have will require some enhancement. We have published a short sister post on what we have learned so far about organising data for analysis. But we are hoping over the next couple of months to start getting a much clearer idea of what Cambridge is contributing into the system – in terms of papers, peer review and editorial work in addition to our subscriptions and APCs. We need more evidence based decision making for negotiation.

Footnote

* There has been some discussion in listservs about who is behind the Call to Action and the Declaration. Thanks to Jeff MacKie-Mason, University Librarian and Professor, School of Information and Professor of Economics at UC Berkeley, we are happy to clarify:

  • The Declaration is by the faculty senate’s library committee – University Committee on Library and Scholarly Communication (UCOLASC)
  • The Call to Action is by the University of California’s Systemwide Library and Scholarly Information Advisory Committee, UCOLASC, and the UC Council of University Librarians, who: “seek to engage the entire UC academic community, and indeed all stakeholders in the scholarly communication enterprise, in this journey of transformation”.

Published 26 June 2018 (amended 27 June 2018)
Written by Dr Danny Kingsley & Katie Hughes
Creative Commons License

Observations on a data gathering project

The Office of Scholarly Communication provides information, advice and training on research data management.  So when faced with running a research project that involves a considerable amount of data, it is telling to see if we can practice what we preach.

This blog post is a short list of how we have approached managing data for analysis. Judging by our colleagues’ faces when we described some of the advice here, this is blindingly obvious to some people. But it was news to us, so we are sharing it in case it is helpful to others.

Organising and storing the data

As is good practice we have started with  a  Data Management Plan. Actually we ended up having to write two, one for the qualitative and one for the quantitative aspect of the project. 

We have also had to think through where the data is being stored and backed up. All of the collected data is currently being stored on a shared Cambridge University Google Drive where only invited users with a Cambridge University email address can view the data. This is because it can handle Level 2 confidential information and was accessible on and off campus. Some of the data is confidential and publishers have asked us to keep it private.

The data is also stored on a staff member’s laptop computer in her Documents folder (the laptop is password protected) that is backed up by the Library on a daily basis. There is a second storage place on the Office of Scholarly Communication’s (OSC) Shared Drive to ensure that there are two backups in two different locations.

One dataset has proven difficult to use as it is 48MB and Google Drive does not seem to be able to handle that file size well.

Each dataset was renamed with the file naming syntax that the OSC uses. This includes a three letter prefix at the beginning (e.g. RAW for raw data), a short description, then a version, and finally the date that the data was received. Underscores separate each section and there are no spaces. An example is MEM_JCSBlogData_V1_20180618.docx

To organise and summarise the metadata, we have created two spreadsheets. One is a logbook that records the name of the file, a description of the data, size of the file, if it is confidential, and what years it covers. The second spreadsheet records what information each dataset covered, i.e. Peer Review, Editing, Citing, APCs, and Usage. The spreadsheet also records correspondence with the publishers.

Assessing our data

At first glance, we were unsure whether we could do cross comparisons between publishers with the data that we had collected. Although most datasets were provided in Excel (with the exception of the Springer 2017 report on gold open access and eLife), they were formatted differently and covered different areas.

Dr Laurent Gatto, one of Cambridge’s Data Champions, very kindly agreed to meet with us and look over the data that we had collected so far. He suggested a number of ways that we could clean up the data so that we could do some cross comparison analysis. Somewhat to our surprise he was generally positive about the quality and analysability of the data that we had collected.

Cleaning up data for analysis

After an initial look at the data, Laurent gave us some excellent suggestions on managing and analysing the data. These are summarised below:

  • Have a separate folder where the original datasets will be saved. These files will remain untouched.
  • When doing any re-formatting, a new file will be created using the same naming convention, but updating the version. A record of any changes to the dataset will need to be recorded in a spreadsheet.
  • Ensure that all of the headers are uniform across the different spreadsheets, to allow analysis across datasets. Each header must be the same down to the last lowercase letter and cannot include any spaces
  • Dates must also be uniform using Year-Month-Day format
  • Only the first row of a spreadsheet can include the header. Having more than one row with header information will cause problems when you are starting to code.
  • Create a readme file where every header will be recorded with a short description.

Next steps

After speaking with Laurent we are more optimistic about the data that we have collected than we were before. We were concerned that there was not enough information to do analysis across publishers; however, we are more confident that this is not the case. As we start the analysis it will also give us a better understanding of what data is missing.

We will provide an update as we close in on our findings.

Published 26 June 2018
Written by Katie Hughes & Dr Danny Kingsley
Creative Commons License