Gold Open Access: Counting the Costs

Research Councils UK (RCUK) have recently announced a significant amendment to their open access (OA)  policy which requires all research papers that result from research partly or wholly funded by RCUK to be made open access [1]. To comply with this policy, researchers must either; a) publish in an open access journal, termed Gold OA, which often incurs an article processing charge (APC); or, b) ensure that a copy of the post-print is deposited in an appropriate repository, also known as Green OA.

A subsequent clarification from RCUK stated that Gold OA is the preferred mechanism of choice to realise open access for outputs that they have funded and have announced the award of block grants to eligible institutions to achieve this aim [2]. Where a Gold OA option is unavailable, Green OA is also acceptable; however, RCUK have indicated that the decision will be ultimately left up to institutions as to which route to take [3].

Since RCUK are the major funder of research in the United Kingdom, this new policy will not only have a major impact on how researchers publish their work, but also huge implications for their budgets. Many research institutions funded by RCUK are currently investigating how they will implement this policy and are looking at the costs for open access publication, and how they can support the adoption of open access within their organisation. The ball is very much in the court of institutions to decide how to play the open access game.

One of the key factors that will affect institutions is the cost that publishers will set for their APCs. So far RCUK have steered clear of suggesting an appropriate fee, leaving individual publishers to determine the market level of the APCs as per the current situation. Meanwhile there seems to be a huge variability in costs. There is a general expectation that over time APCs will settle to a reasonable rate and similarly journal subscriptions will lower to reflect the gradual change in business model from subscription fees to APCs. Most publishers have not yet been upfront about what impact they will have on journal subscriptions, if any, and it is hard to access and assess real-life data. RSC Publishing is one notable exception since it has introduced a system of waiving a proportion of APC fees based on institutional subscription costs.

Much of this transition period to full open access will have to be navigated through uncharted territory, where no one has a clear handle on the costs involved. The rationale of this article is to present data on article processing charges gathered over the past five years, report on trends seen within this data, to suggest some approaches and to generally contribute to and inform the policy discussion.

The Problem

To put some rough-and-ready figures on the table, the University of Edinburgh publishes in the region of 4,000-4,500 peer-reviewed journal articles per year; this figure does not include other publication types like working papers not affected by the RCUK policy. Assuming an average Article Processing Charge (APC) of £1500 [4], the total publication costs to make all of these outputs Gold would be in the region of £6m. It is clear that even with guaranteed funding from HEFCE, and other funders of research, large research-intensive universities will not be able to pay for all of their research to be published under Gold OA. How to allocate funding to researchers will be a difficult choice that many institutions are currently asking themselves - will it be on a first-come-first-served basis, funder-specific, or will REF-submitted material take priority?

Equally problematic are the difficulties we face in fully assessing an institution’s total spend on open access. Whilst it is possible to find out through aggregate sources like Web of Science how many articles are published in fully open access journals. It is virtually impossible to find out the number of open access articles published in hybrid journals as there is currently no flag in the metadata which indicates the open status of the paper. A hybrid journal is a traditional subscription journal that offers open access to individual articles upon payment of an APC. Of course it is possible to find hybrid open access content through EuropePMC.org; however this will only give a snapshot for the biomedical and life sciences. With current systems and processes it is virtually impossible to gauge this spend accurately.

Cost Data

Unfortunately, financial data about open access publishing is scarce. The University of Edinburgh (UoE) has recently implemented account codes to allow the finance systems to track this spend going forwards; however, finding out costs retrospectively remains problematic. Furthermore, institutions are not typically in the habit of publishing this data with others. The institutions that have shared data show a degree of variability. In 2010, the foremost initial supporter and enabler of Gold Open Access publishing in the UK, the Wellcome Trust, found that the average cost of publication under the author-pays model was $2,367 (approximately £1,500) [4]. RCUK in their recent press release on block grants for open access estimate the average APC as £1,727 plus VAT [2], whilst, based on figures in the Finch Report, the University of Nottingham paid on average £1,216 [5].

All these figures are useful as they give a ballpark figure upon which further estimates can be based. The precise cost of individual APCs levied by publishers is generally unavailable in a form which easily enables further analysis. Typically this information is available from publisher’s Web sites; however, aggregating the data is cumbersome as there is no consistent way to interrogate the Web sites and APCs commonly vary from title to title in the publishers’ portfolio. There have been some commendable attempts to gather this information, for example the SHERPA RoMEO listing of Publishers with Paid Options for Open Access [7]. Here about 100 publishers have been surveyed and their APCs are listed. A large cost variance exists for some publishers’ records as individual journals often have different APCs, and also institutional subscriptions/memberships can reduce costs in a non-uniform way. It takes a lot of effort to gather these data and keep them it up to date. Other approaches have tried to crowd-source this activity, for example Ross Mounce’s survey of open access publishers, publications, licences and fees. Here approximately 130 publishers’ web sites were surveyed to find out what licences are being used on the open access content; the cost being a secondary focus of the survey. Analysis of these data shows less than 5% of publishers claiming 'open access' are fully compliant with the Budapest Declaration on Open Access [7].

The data we present here is an attempt to enrich the data available to interested parties and make them available in a reusable format for further analysis. It comprises articles funded by the Wellcome Trust at the University of Edinburgh between 2007 and 2012. In total there are 260 articles published in a mixture of open access journals and traditional subscription journals with an open access option (sometimes known as hybrid). All of the journals charged an article processing fee. Overall, the total cost incurred was £452,713.40. The mean article processing charge was £1,741.21, with the median value £1,644.22. The full data can be accessed online at the Edinburgh DataShare repository [8].

APC Cost through Time

As shown in Table 1 the number of articles supported by the Wellcome Trust open access grant has increased each year:

Year	No.of articles	Total cost	Mean APC
2007/08	19	£32,712.85	£1,721.73
2008/09	38	£58,648.88	£1,543.39
2009/10	49	£90,304.78	£1,842.95
2010/11	67	£124,359.79	£1,856.11
2011/12	87	£146,687.10	£1,686.06
Total	260	£452,713.40	£1,741.21

Table 1: High-level details of the Wellcome Trust open access expenditure at UoE

Over the past five years the costs have been fairly stable – the yearly variation we see here is likely to be a function of the small sample size. It is interesting to note that costs have not appreciably risen in line with inflation.

The jump in number of articles from 2009/10 onwards may be directly attributed to the hardening of the Wellcome Trust’s open access policy and the introduction of sanctions - backdated to 2009. Any non-compliant papers identified in submitted end-of-grant reports will now impose 10% of the budget to be withheld [9]. It appears that Open Access policies require rigorous compliance monitoring to be successful, and seem to be more effective when punitive sanctions are imposed.

APC and Impact Factor

Next we decided to look at the relationship between APC cost and the prestige of a journal. Bearing in mind the well known and documented limitations (for example as discussed by Kurmis [10]), we used the well-established metric journal impact factor as a proxy for prestige. The impact factor for the host journal of each article was calculated using the 5 year average as published in the Thomson Reuters Journal Citation Reports [11]. The five-year average value was chosen to normalise impact factors across the data census period.

The figure below shows the journal impact factor plotted against journal article processing fee. The graph shows a positive correlation between APC and impact factor; as demonstrated by the simple linear regression line (trendline).

Figure 1: Graph showing cost of article processing fee plotted against journal impact factor

At a first glance it appears that some journals with high impact factors are applying significantly higher article processing charges. Looking specifically at the APCs over £3,000, we find the mean journal impact factor is 11.30. To put this in context with the whole dataset, the mean APC cost and journal impact factor is £1,741.21 and 5.92 respectively.

APC and Journal Type

The first point to note is that during the five-year sample period, publication in hybrid journals (n=185) was significantly more popular than publishing in full OA journals (n=75). This may be due to the fact that there are more hybrid journals to publish in. Secondly, Figure 1 clearly shows that there is a significant difference in price between hybrid journals (red squares) and full OA journals (green triangles). The full OA journals plot completely under the trend line; whilst all of the APC fees over £2,000 were incurred by hybrid journals only. We found the average APC cost for hybrid journals was £1,989.79 compared to £1,128.02 for full OA journals – a difference of £861.77.

Limitations

This study has a number of limitations that should be noted; the biggest of which is the relatively small data size. We propose and strongly encourage any other institutions which hold similar information to share their data to create a larger dataset to see if these results still hold true.

For some journals there is a large variation in cost (shown in Figure 1 by a linear vertical spread of costs for the same journal) due to unseen factors such as foreign currency conversion, or institutional membership schemes reducing costs. This artificial noise detracts from the signal we are trying to interpret.

Discussion

The average APC we have seen is higher than both the previously reported 2010 average by the Wellcome Trust and the 2010-11 figures from the University of Nottingham; however, it is broadly in line with the latest APC figures published in the Finch Report which suggests this is fairly representative.

The relationship between impact factor and cost is also interesting to note. Research- intensive institutions are likely to be hit twice; since they publish more articles and more frequently in higher-impact journals, their share of Gold OA bills is likely to be disproportionally larger.

The causes of significantly higher APC costs for high impact factor and hybrid journals are hard to identify and the suggestions made here are purely speculative:

1. Higher rejection rates: It is often reported anecdotally that journals with high impact factors generally have higher rates of rejection for submitted manuscripts compared to other journals. If this is the case, then these journals will incur higher overheads in order to administer their peer-review process. There is a fundamental lack of transparency in the cost of peer review which prevents any detailed analysis. Do extra peer review costs for one paper merit a price increase £2,000 over and above the average APC costs seen for some other journals?
2. Reprints: various publishers have commented that they maximise their income streams by selling commercial reprints [12]. A fully open licence (for example Creative Commons Attribution CC-BY) would remove this as users are free to distribute and reuse without further payment. It is precisely for this reason that Nature Publishing Group has recently imposed a premium of ~£200 on their CC-BY licences [13]. In this case, higher APCs are being used to offset revenue loss. It may be the case that other publishers are including this factor in their costs but are not as up front about it.
3. Brand: Scarcity is commonly used by retailers as a mechanism to stimulate demand or prestige for their brand. In this sense, publishers are no different. One way of artificially creating scarcity is to increase retail costs; consequently there is a valid business case for publishers to increase APCs for their higher impact journals.
4. Value: Related to the issue of brand, there is a commonly held view that having high costs for publishing articles in high impact journals is justified as this is a valued service for which researchers are willing to pay a premium.

In all likelihood the cost of APCs will be influenced by all these factors and more. Commercial publishers may seek to set the APCs at a price point which they think the market can bear. In theory, researchers can choose exactly where to publish and are free to publish elsewhere if they don't like the prices. Unfortunately this thinking could be seen as somewhat naïve. It has been widely mentioned, for example by David Prosser of Research Libraries UK [14], that market forces are unlikely to force down costs so institutions will have to be very careful about how they set up and manage their central open access funds.

We share this concern that with an inelastic market - researchers are unlikely to shop around - and where the costs are sheltered - central funds mean that researchers are not exposed directly to costs - the APCs would remain high because normal market forces would not drive costs down. We believe one very effective course of action is to be as maximally transparent with Gold APC costs as possible. Unlike the current situation we find ourselves in, where traditional journal subscription costs are kept hidden from readers by confidentiality clauses in the contracts, all institutions, including universities and RCUK, should strive to publish a detailed analysis of their spend on Gold OA.

Conclusion

Bearing in mind the study’s limitations, the data we present here suggest some important points:

1. Hybrid journals seem to be more popular venues for Open Access  publication, and
2. Hybrid journals generally charge more than full OA journals independent of journal impact factor, and
3. There is a positive correlation between APC cost and impact factor for both hybrid and full OA journals.

Some more reflective points that this study raises are:

1. It appears that Open Access policies require rigorous compliance monitoring to be successful, and seem to be more effective when punitive sanctions are imposed.
2. Research-intensive institutions are likely to be hit by a cost ‘double whammy’; they not only publish more articles, but they also publish them more frequently in high-impact-factor journals.
3. Institutions need to be more open about costs, and publish the data in a format that allows reuse.

The dataset upon which this paper is based is available to download from Edinburgh DataShare [8].

Acknowledgements

Many thanks to the Wellcome Trust for allowing us to reuse and publish part of its data.

References

1. Research Councils UK Policy on Access to Research Outputs
   http://www.rcuk.ac.uk/documents/documents/RCUK%20_Policy_on_Access_to_Research_Outputs.pdf
2. RCUK announces block grants for universities to aid drives to open access to research outputs http://www.rcuk.ac.uk/media/news/2012news/Pages/121108.aspx
3. RCUK Open Access Policy – When to go Green and When to go Gold
   http://blogs.rcuk.ac.uk/2012/09/28/rcuk-open-access-policy-when-to-go-green-and-when-to-go-gold/
4. Wellcome Trust and the author-pays model
   http://blog.europepmc.org/2011/02/wellcome-trust-and-author-pays-model.html
5. Accessibility, sustainability, excellence: how to expand access to research publications: Report of the Working Group on Expanding Access to Published Research Findings. Section 7.5 http://www.researchinfonet.org/wp-content/uploads/2012/06/Finch-Group-report-FINAL-VERSION.pdf
6. SHERPA RoMEO: Publishers with Paid Options for Open Access
   http://www.sherpa.ac.uk/romeo/PaidOA.html
7. A survey of 'Open Access' publisher licenses
   https://sites.google.com/site/rossmounce/misc/a-survey-of-open-access-publisher-licenses
8. Edinburgh DataShare http://datashare.is.ed.ac.uk/handle/10283/250
9. Wellcome Trust Authors’ FAQs http://www.wellcome.ac.uk/About-us/Policy/Spotlight-issues/Open-access/Guides/WTD018855.htm#ten
10. Kurmis, A.P. Understanding the limitations of the journal impact factor. J Bone Joint Surg Am. December 2003; 85-A(12):2449-54 http://www.ncbi.nlm.nih.gov/pubmed/14668520
11. Thomson Reuters Journal Citation Reports
    http://thomsonreuters.com/products_services/science/science_products/a-z/journal_citation_reports/
12. NPG position statement on open access publishing and subscription business models. http://www.nature.com/press_releases/statement.html
13. NPG expands Creative Commons Attribution license options
    http://www.nature.com/press_releases/cc-licenses.html
14. David Prosser. "Cash alone will not cure the research market". The Impact of Social Sciences blog, 18 September 2012
    http://blogs.lse.ac.uk/impactofsocialsciences/2012/09/18/cash-alone-will-not-cure-the-research-market/

Author Details

Theo Andrew
Open Scholarship Development Officer
University of Edinburgh

Email: theo.andrew(at)ed.ac.uk
Web site: http://about.me/theo_andrew

Theo Andrew is the Open Scholarship development officer based at the Digital Library Section at the University of Edinburgh. Professional interests include: enhancing scholarly communication using new technologies; promoting the open agenda within academia; research analytics and altmetrics; and research information/data management.