Month: September 2016

Imperial College 2015-2016 open access compliance report to RCUK

Earlier today Imperial College submitted its annual report on compliance with the Research Council’s open access policy to RCUK. The RCUK OA policy envisages a five year journey after which 100% of RCUK funded scholarly papers should be available as open access in 2018. To support the transition to open access, RCUK have set up a block grant that makes funds available to institutions to cover the cost for article processing charges (APCs) and other OA-related expenses. Funds are awarded in relation to RCUK research funding for institutions, and Imperial College has the second largest allocation, just behind Cambridge and followed by UCL. The annual reports to RCUK give an overview over institutional spend and on compliance.

The headline figures for the 2015/2016 College report are:

  • £1,051,130 block grant spend from April 2015 to March 2016
  • 89% overall compliance, split in 31% via the gold and 58% via the green route
  • 570 article processing charges paid at an average cost of ~£1,800
  • The top five publishers are: Elsevier, Wiley, Nature, ACS and OUP

Like every year when discussing the RCUK report figures I think it is important to highlight that compliance rates between universities cannot meaningfully be compared without understanding the data sources and methods used. Just to give one example: the College could also have reported 81% green and 8% gold from the same data.

Why do I caution against directly comparing the numbers? For starters, research-intensive universities find it difficult to establish what 100% is. With hundreds, or in the case of Imperial College many thousand papers published every year we rely on academics to manually notify us for each paper who the funder is. Even though the College has made much progress improving its processes and data over the past few years we have to acknowledge that data collected through such a process will never be complete or fully accurate. For the College report we decided, like in the previous years, to base our analysis on outputs we know to have been RCUK-funded. For this year the size of the sample was 1,923 papers (compared to 1,326 in 2014). With a different sample the numbers would have been different, and other universities may have taken a different approach to analysing the data.

Sadly, it is currently not easy to establish whether an output was made available open access. Publishers do not usually add licensing information to metadata, and searching for manuscripts deposited in external repositories is possible but not necessarily accurate. The process we used for analysis was:

  • Cross-reference the sample with the journal list from the Directory of Open Access Journals; class every article published in a full OA journal as compliant ‘gold’.
  • Take the remaining articles and cross-reference with the list of articles for which the College Library has paid an APC; class all those articles as compliant ‘gold’.
  • Take the remaining articles and cross-reference with the outputs from ResearchFish that show a CC BY license; class all those articles as compliant ‘gold’.
  • Take the remaining articles and cross-reference with list of outputs deposited in the College repository Spiral; class all those articles as compliant ‘green’.
  • Take the remaining articles and cross-reference with list of outputs that have a Europe PubMed Central ID; class all those articles as compliant ‘green’.

As in previous years we also put remaining outputs through Cottage Labs Lantern tool but this showed no additional open access outputs. The main reason for that, I suspect, is the high compliance via the green route: some 81% of outputs in the sample had been deposited to the College repository Spiral or to Europe PMC. As the College prefers green over hybrid gold it would have been in line with our policy to report them as green, but as the RCUK prefers gold OA we decided to report all outputs know as gold as such, like in previous years.

I could write more about reporting issues around open access, but as I have done that on a few other occasions I refer those who haven’t suffered enough to my previous posts.

One other caveat should be raised for those planning to analyse the APC spend in comparison with previous years: The APC article level data is based on APCs paid during the reporting period. This differs from the APC data reported in the previous period which was based on APC applications published. There are, therefore, a small number of records duplicated from the previous year. These have been identified in the notes column.

If you want the data from the full report, please visit the College repository:

I would like to thank everyone involved in putting the information together, especially colleagues in the Central Library’s open access team.