All posts by Arthur Smith

How open is Cambridge? 2017 edition

Welcome to Open Access Week 2017. The Office of Scholarly Communication at Cambridge is celebrating with a series of blog posts, announcements and events. In today’s blog post we revisit the question about the openness of Cambridge. 

For Open Access week last year I looked at how open Cambridge was using the extremely useful Lantern tool, developed by Cottage Labs, and which is the basis of the Wellcome Trust’s compliance tool. If you haven’t used it before, Lantern takes a list of DOIs, PMIDs, or PMCIDs and runs these through a variety of sources to try and determine the Open Access status of the publication. I found that, for publications in 2015, 51.8% of all of Cambridge’s research publications were available in at least one ‘Open Access’ source. How did Cambridge’s 2016 publications fair? Read on to find out.

Using the same method as last year, I first obtained a list of DOIs from Web of Science (n=9416) and Scopus (n=9124) for articles, proceedings papers and reviews published in 2016. Combining and deduplicating these lists returned 10,674 unique DOIs (~29 publications/day). I also refreshed the 2015 publication data using the latest Web of Science and Scopus information, which returned 10,090 unique DOIs. Year-on-year, this represents a 5.8% increase in the total number of publications attributable to Cambridge – more than inflation!

The deduplicated DOI lists for 2015 and 2016 (20,764 DOIs in total) were fed into Lantern and analysed in combination with information from Web of Science and the University’s institutional repository Apollo.

Figure 1. Distribution of papers, published in 2015 and 2016 which have a DOI, according to the Open Access sources they can be found in. 57.5% of 2016’s articles appear in at least one Open Access source, which represents a 4% increase over 2015. One third of all papers published in 2016 are available in Apollo.

Very pleasingly the percentage of publications available in at least one Open Access source increased to 57.5% in 2016 compared to only 53.4% for 2015 publications. Given that the total number of publications also increased during this period this result is doubly exciting. In raw numbers, this means that while 5384 publications were Open Access in 2015, an impressive 6135 publications were made Open Access in 2016.

Most of this increase can be attributed to the much larger share of publications that appear in Apollo, which is now the largest source of Open Access material for the University of Cambridge. An additional 822 publications were deposited in Apollo in 2016 compared to 2015, which is a 30% increase in one year alone.

You can now find more of the University’s research outputs in Apollo than in any other Open Access source. And because we operate an extremely popular Request a Copy service, potentially all of the publications held in Apollo, even those that are restricted and under embargo, are available to anyone in the world. You just need to ask.

Published 23 October 2017
Written by Dr Arthur Smith
Creative Commons License

Open Access policy, procedure & process at Cambridge

First up, HEFCE’s Open Access policy:

At the outset, let’s be clear: the HEFCE Open Access policy applies to all researchers working at all UK HEIs. If an HEI wants to submit a journal article for consideration in REF 2021 the article must appear in an Open Access repository (although there is a long list of exceptions). Keen observers will note that in the above flowchart HEFCE’s policy is enforced based on deposit within three months of acceptance. This requirement has caused significant consternation amongst researchers and administrators alike; however, during the first two years of the policy (i.e. until 31 March 2018) publications deposited within three months of publication will still be eligible for the REF. At Cambridge, we have been recording manuscript deposits that meet this criterion as exceptions to the policy[1].

Next up, the RCUK Open Access policy. This policy is straightforward to implement, the only complication being payment of APCs, which is contingent on sufficient block grant funding. Otherwise, the choice for authors is usually quite obvious: does the journal have a compliant embargo? No? Then pay for immediate open access.

One extra feature of the RCUK Open Access policy not captured here is the Europe PMC deposit requirement for MRC and BBSRC funded papers. Helpfully, the policy document makes no mention of this requirement; rather, this feature of the policy appears in the accompanying FAQs. I’m not expert, but this seems like the wrong way to write policies.

Finally, we have the COAF policy, possibly the single most complicated OA policy to enforce anywhere in the world. The most challenging part of the COAF policy is the Europe PMC deposit requirement. It is often difficult to know whether a journal will indeed deposit the paper in Europe PMC, and if, for whatever reason, the publisher doesn’t immediately deposit the paper, it can take months of back-and-forth with editors, journal managers and publishing assistants to complete the deposit. This is an extremely burdensome process, though the blame should be laid squarely at the publishers. How hard is it to update a PMC record? Does it really take two months to update the Creative Commons licence?

This leads us to one of the more unusual parts of the COAF policy: publications are considered journals if they are indexed in Medline. That means we will occasionally receive book chapters that need to meet the journal OA policy. Most publishers are unwilling to make such publications OA in line with COAF’s journal requirements so they are usually non-compliant.

What happens if you should be foolish enough to try to combine these policies into one process? Well, as you might expect, you get something very complicated:

This flowchart, despite its length, still doesn’t capture every possible policy outcome and is missing several nuances related to the payment of APCs, but nonetheless, it gives an idea of the enormous complexity that underlies the decision making process behind every article deposited in Apollo and in other repositories across the UK.

[1] Within the University’s CRIS, Symplectic Elements, only one date range is possible so we have chosen to monitor compliance from the acceptance date. Publications deposited within the ‘transitional’ three months from publication window receive an ‘Other’ exception within Elements that contains a short note to this effect.

Published 18 September 2017
Written by Dr Arthur Smith
Creative Commons License

How open is Cambridge?

As part of Open Access Week 2016, the Office of Scholarly Communication is publishing a series of blog posts on open access and open research. In this final OAWeek post Dr Arthur Smith analyses how much Cambridge research is openly available.

For us in the Office of Scholarly Communication it’s important that, as much possible, the University’s research is made Open Access. While we can guarantee that research deposited in the University repository Apollo will be made available in one way or another, it’s not clear how other sources of Open Access contribute to this goal. This blog is an attempt to quantify the amount of Cambridge research that is openly available.

In mid-August I used Cottage Labs’ Lantern service in anLantern_Oct2016_Graphic attempt to quantify just how open the University’s research really is. Lantern uses DOIs, PMIDs or PMCIDs to match publications in a variety of sources such as CORE and Europe PMC, to determine the Open Access status of a publication – it will even try to look at a publisher’s website to determine an article’s Open Access status. This process isn’t infallible, and it relies heavily on DOI matching, but it provides a good insight into the possible sources of Open Access material.

To determine the base list of publications against which the analysis could be run,  I queried Web of Science (WoS) and Scopus to obtain a list of publications attributed to Cambridge authors. In 2015, the University published 9069 articles, reviews and conference papers according to Web of Science. Scopus returned a slightly lower figure of 7983 publications. Combining these two publication lists, and filtering to only include records with a DOI, produced one master list of 9714 unique publications (that’s ~26 publications/day!).

In 2015 the Open Access team processed 2746 HEFCE eligible submissions, so naïvely speaking, the University achieved a 28.3% HEFCE compliance rate. That’s not bad, especially because the HEFCE policy had not yet come into force, but what about other Open Access sources? We know that other universities in the UK are also depositing papers in their repositories, and some researchers make their work ‘gold’ Open Access without going through the Open Access team, so the total amount of Open Access content must be higher.

In addition to the Lantern analysis, I also exported all available DOIs from Apollo and matched these to the DOIs obtained from WoS/Scopus. WoS also classifies some publications as being Open Access, and I included these figures too. If a publication was found in at least one potentially Open Access source I classified it as Open Access. Here are the results:

Lantern_Oct2016_Figure1
Figure 1. Of 9714 DOIs analysed by Lantern, 51.8% appear in at least one open access source.

It is pleasing that our naïve estimate of 28.3% HEFCE compliance closely matches the number of records found in Apollo (26.2%). The discrepancy is likely due to a number of factors, including publications received by the Open Access Team that were actually published in 2014 or 2016, but submitted in 2015, and Apollo records that don’t have a publisher DOI to match against. However, the most important point to note is the overall open access figure – in 2015 more than 50% of the University’s scholarly publications with a DOI were available in at least one “open access” source.

Let’s dig a little deeper into the analysis. Using everyone’s favourite metric, the journal impact factor (JIF), the average JIF of articles in Apollo was 5.74 compared to 4.33 for articles that were not OA. Other repositories and Europe PMC achieved even higher average JIFs. On average, Open Access publications by Cambridge authors have a higher JIF (6.04) than articles that are not OA, which suggests that researchers are making value judgements on what to make Open Access based on journal reputation. If a paper appears in a low(er) impact journal, it’s less likely to be made Open Access. Anecdotally this is something we have experienced at Cambridge.

Lantern_Oct2016_Figure2
Figure 2. Average 2015 JIF of papers classified according to their open access status.

The WoS and Scopus exports contain citation information at the article level, so we can also look at direct citations received by these publications (up to 16 August 2016)  rather than relying on the JIF. I found that Open Access articles, on average, received 1.5 to 2 more citations than articles that are not Open Access. However, is this because authors are making their higher impact articles Open Access (which one might expect to receive more citations anyway) and are not bothering with the rest? Or this is effect due entirely to the greater accessibility offered by Open Access publication? Could the differences arise because of different researcher behaviour across different disciplines?

My feeling is that we have reached a turning point – the increased citation rates of Open Access material is not caused by the article being Open Access as these articles would have naturally received more citations anyway. Instead of looking at formal literature citations, the benefits of Open Access need to be measured outside of academia in areas that would not contribute to an articles citations.

Lantern_Oct2016_Figure5
Figure 3. Average citations received by papers according to their open access source.

Breaking it down by the source of Open Access reveals that articles that appear in other repositories receive significantly more citations than any other source. This potentially reveals that collaborative papers between researchers at different institutions are likely to have greater impact than papers conducted solely at one institution (Cambridge), however, a more thorough analysis that looks at author affiliations would be needed to confirm this.

If we focus on the WoS citation distribution the difference in average citations becomes clearer. Of 8348 WoS articles, not only are there fewer Open Access articles with no citations (14% vs 17%), but Open Access articles also receive more citations in general.

Lantern_Oct2016_Figure4
Figure 4. Citation distribution of papers found in WoS depending on their open access status.

What can we take away from this analysis? Firstly, Lantern is a valuable tool for discovering other sources of Open Access content. It identified over a thousand articles by Cambridge researchers in other institutional repositories that we did not know existed. When it comes time for the next REF, these other repositories may prove a vital lifeline in determining whether a paper is HEFCE compliant.

Secondly, more than 50% of the University’s 2015 research publications are potentially Open Access. Hopefully a similar analysis of 2016’s papers will show that even more of the University’s research is Open Access this year. And finally, although Open Access articles receive more citations than articles that are not Open Access, it is no longer clear whether this is caused by the article being Open Access, disciplinary differences, or if authors are more likely to make their best work Open Access.

Published 28 October 2016
Written by Dr Arthur Smith

Creative Commons License