| title | author | date | output | bibliography | csl | ||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
A Study of Institutional Spending on Open Access Publication Fees in Germany |
|
7 April 2016 |
|
literatur.bib |
frontiers.csl |
The rise of open access journals matches the increasing relevance of publication fees in academic publishing [@Davis_2011; @Laakso_2012; @Pinfield_Review_2015]. To cover these fees, also referred to as article-processing charges (APC), authors tend to make use of funding that grant agencies or academic institutions provide [@Suber_2012]. Yet, how and to what extent these research support activities are effective in terms of the number of supported articles and associated costs remains under debate.
The study of institutional spending on open access journal articles has been limited for several reasons. One is that payment on these charges is fragmented across the budgets of grant agencies, research institutions, and libraries, or is covered by personal budgets. Asking 9,645 authors from various disciplines how they financed publication fees, a comprehensive survey in 2010 revealed that the majority of the respondents had access to research funding or institutional support to cover these charges. By contrast, 12 % paid publication fees individually [@soap]. These findings are consistent with that of other studies, adding that sources of funding mostly exists in higher income countries, mainly to support research articles in the bio- and physical sciences [@Solomon_2011]. Personal budgets, on the other hand, are likely used to cover low price publication fees [@Solomon_2011; @Bj_rk_2015].
Another key problem in this regard is that funding for open access journals using publication fees lacks transparency because the parties involved - authors, universities, funders, publishers - neither release information on who pays for what nor the costs of publishing [@Bj_rk_learned_2014], a situation similar to the lack of transparency regarding journal subscriptions [@Lawson_Meghreblian_2015]. Empirical studies examining publication fees so far obtained price estimates by surveying authors [@soap], or from journal websites. Using the latter method, two studies investigating journals across a broad range of disciplines calculated similar price averages that ranged between 904 $ [@Solomon_2012] and 923 $ [@Walters_2011], as well as considerable price variances across journals and publishers. Accordingly, @Solomon_2012 suggested to cluster fully open access journal using publication fees into several groups. In descending order, these are high-impact journals, followed by biomedicine journals from commercial publishers, large multi-disciplinary journals, and mid-price journals from commercial publishers covering a large spectrum of disciplines. Lower priced journals were published by academic societies and by publishers from low-income countries.
Nevertheless, it remains unclear which factors contribute to pricing in academic publishing. Generally, these might include article processing, impact, rejection rates, management and investment, and profit margins [@Van_Noorden_2013]. While fixed prices for individual articles are common, agreements between publishers and institutions can lead to discounts, and publishers sometimes waive publication fees for authors from low-income countries [@BJ_RK_2012; @Lawson_waiver_2015]. Other factors leading to variable pricing schemes include submission or page charges [@BJ_RK_2012].
Hybrid journals substantially add to this complexity of open access funding [@Bj_rk_learned_2014; @Kingsley_2014; @Pinfield_2015]. These journals, allowing articles to be published immediately as open access after a charge was paid, rely both on subscriptions and publication fees as revenue sources. Although the uptake of open access through hybrid journals was described as lower and more expensive compared to that of fully open access journals [@Solomon_2012; @Bj_rk_learned_2014], this model has gained increasing attention because of recent open access science policies, notably from the UK [@Pinfield_Review_2015].
To address these problems of fragmented spending on publication fees and of in-transparency about what was being paid, some European research funders and research performing institutions have recently begun to disclose their payment for publication as open data. To our knowledge, the first research funders providing such data were the Wellcome Trust [@wellcome_apc_12] and the Austrian Science Fund FWF [@fwf_apc_13]. The not-for-profit company Jisc followed by collecting data from UK universities [@Lawson_data_2015]. Disclosed as publicly available spreadsheets, these data-sets self-report expenditure along with bibliographic information, including title, journal and publisher, and a persistent identifier to the publisher's version. Curatorial efforts focused on the disambiguation of publisher and journal titles as well as on detecting duplicates and persistent identifiers to the full text including the Digital Object Identifier (DOI) [@Neylon_2014; @woodward_2014]. Parts of Jisc's cost data was examined by @Pinfield_2015. Although the average spending on publication fees remained stable across the universities, they found large price variances, as well as a varying number of articles UK universities supported between 2007 - 2014, being consistent with earlier studies that collected price information from journal websites [@Solomon_2012].
This paper focuses on how much German universities and research organisations spent on open access publication fees. In Germany, the Deutsche Forschungsgemeinschaft (DFG), the largest German research funder, has strongly influenced how universities manage institutional support for these charges. Before the DFG started to pay for centrally funded publication fees on a pro rata basis through its "Open-Access Publishing" program in 2011, only a few central funds existed [@eppelin_2012]. This is similar to the situation described in Canada [@Hampson_2014] or the UK [@PINFIELD_2012]. The DFG has enforced a set of criteria grantees have to comply with [@Fournier_2013], leading to related implementations for open access supporting across German universities. These criteria exclude sponsoring of articles in hybrid journals, and the funding of articles whose publication fee exceeds 2,000 €1. Grantees agree not only to pay for APCs, but also to find ways to improve the handling of those financial transactions. This includes central invoicing schemes and memberships, among others, agreed between university libraries and publishers that often lead to a discount on publication fees [@Fournier_2013].
Non-university research organisations, i.e. institutes organised in the Fraunhofer-Gesellschaft, Helmholtz-Gemeinschaft, Leibniz-Gemeinschaft, and Max-Planck-Gesellschaft, are not eligible for this DFG funding program. But in response, some have adopted similar processes to support authors. The Max-Planck-Gesellschaft operates their long-lasting open access activities, including handling spending and publisher agreements centrally, through the Max Planck Digital Library (MPDL) [@Schimmer_2013; @Sikora_2015], while the Leibniz-Gemeinschaft set up a dedicated open access fund in 2016.
The evolving institutional support to cover open access publication fees has led to calls for an unified approach towards open access funding in Germany. The MPDL called for re-allocating subscriptions in favour of open access journals in 2015 [@Schimmer_Geschuhn_Vogler]. At the same time, the Allianz der Wissenschaftsorganisationen2, a science policy board representing all major research organisations in Germany, marked price transparency as one way to sustain an "adequate open access publication system" [@allianz]. Reflecting Austrian and UK initiatives to share institutional spending on open access publication fees as open data, as well as professional discussions on open access publishing, Bielefeld University Library began to openly share its payment of publication fees in May 2014. After engaging with the working group "Electronic Publishing" of the Deutsche Initiative für Netzwerkinformation (DINI)3, other German institutions joined under the umbrella of the Open APC initiative soon after [@openapc].
The aim of the study was to examine how much German universities and research organisations spent on open access publication fees until 2015. Using self-reported cost data from the Open APC initiative, the analysis focused on the amount that was being spent on publication fees, and compared these expenditure with data from related Austrian and UK initiatives, both in terms of size and the proportion of articles being published in fully and hybrid open access journals. We also asked how thoroughly self-reported articles were indexed in Crossref, a DOI minting agency for scholarly literature, and analysed how the institutional spending was distributed over publishers and journal titles.
We analysed self-reported cost data released by the Open APC initiative on May 13, 2016,4 to assess institutional spending on open access publication fees in Germany. In addition to administrative data about the amount paid per article including value added tax, the reporting institution, and the year of invoicing, we used information about whether an article was published in a fully open access journal or in a hybrid journal, as well as the DOI reported in the data-set.
We fetched bibliographic metadata for each article from Crossref on May 19, 2016, on the basis of the reported DOIs. Although the Open APC initiative gathered metadata representing publishers and journals from Crossref as well, this information was retrieved at the time when the participating institutions submitted their expenditure. The Open APC initiative kept track of the date when these data-sets were submitted with Git, a version control system, increasingly used for enabling reproducible research [@Ram_2013], and made this information available via GitHub to be transparent over time. However, during these data collection activities Crossref had regularly updated metadata to represent ongoing mergers of publishing houses. A prominent example in this regard was the merger of the two large publishing houses Springer Business + Media and Nature Publishing Group announced on May 6, 2015, that operated as Springer Nature at the time of our study. To reflect these dynamics in academic publishing, we decided to retrieve updated bibliographic metadata from Crossref, and to merge these records with the administrative information rather than re-using the historical publisher and journal information contained in the Open APC data-set.
We used the R package rcrossref [@rcrossref], developed and maintained by the rOpenSci initiative5, to access Crossref's REST API6. We requested the XML-based format application/vnd.crossref.unixsd+xml in which full and abbreviated journal titles as well as the ISSN media types, the International Standard Serial Number used to identify journals, were distinguished. It also contained normalised publisher information, thus avoiding confusion about naming of publishing houses other studies were faced with when working with self-reported data [@woodward_2014]. In cases where no bibliographic information could be obtained, we used the Open APC values. Because Crossref was not the only registration agency for DOIs, but also the agencies DataCite and Medra minted DOIs for scholarly work, we furthermore obtained the DOI agency for each article with the help of the rcrossref client.
Data collection also involved obtaining cost data from related open data initiatives. To compare self-reported spending on open access journal articles by Germany universities and research organisations with that of other initiatives, we consulted the openly available data-sets from the the Austrian Science Fund FWF [@fwf_apc_14; @fwf_apc_15], Jisc [@jisc_14; @jisc_15] and the Wellcome Trust [@wellcome_apc_13_14; @wellcome_apc_14_15]. For analysis, we obtained the overall publication fee spending on both fully and hybrid open access journal articles. In the case of FWF, we gathered the cost information from the accompanying spending reports. We used the spreadsheet data to calculate Wellcome Trust's and Jisc's spending, and converted the prices from GBP to Euro in accordance with the average foreign exchange reference rates provided by the European Central Bank. Our comparison of the open data initiatives focussed on the last two years 2014 and 2015. Because Wellcome Trust's spending was reported for the fiscal periods 2013 - 2014 and 2014 - 2015, we referred to the average exchange rates of the full two-year period as we could not determine the actual invoicing dates from the data. We excluded articles from the analysis with missing information about the cost or the journal type. In the case of Jisc's 2014 data [@jisc_14], for instance, we excluded spending on 2,812 publications that amounted to 4.861.772 € from the analysis because no publication type was given in the data-set.
Data curation activities of the Open APC initiative and those of the other initiatives differed in some respects. For instance, whereas the DOI was a mandatory element in the Open APC data template that the participating institutions were required to report, in case of the Wellcome Trust spending data, DOIs were additionally identified by automated compliance checks. Our first screening of the data-sets revealed that some articles published in Crossref-indexed journals lacked a DOI. Because of these different methods to curate the cost data, and as our main focus was institutional funding for publication fees in Germany, we decided to only compare German spending with that reported by other initiatives. We, therefore, did not analyse the distribution of spending over publishers and journal titles, or the indexing coverage in Crossref for the Austrian and UK spending data.
After excluding payments being made for non-journal articles as well as articles invoiced in 2016, we retrieved 7,417 open access journal articles 30 German universities and research institutions financially supported between 2005 and 2015. As illustrated in Figure 1, payments made for open access journal articles increased over the years. While one institution supported 5 articles in 2005, most institutions included in our study shared their expenditure from 2013 onwards. With 1,999 articles, the year 2015 was best represented. However, at the time of analysis, not all, but 27 institutions contributed cost information, suggesting a time lag between payments made and reporting these spending to the Open APC initiative.
Figure 1: Growth of Open APC Initiative
Among all articles, fees amounted to 9,627,537 € including value added tax, the average payment was 1,298 € (Median = 1,231 €) . Figure 2 presents the distribution of institutional spending on publications. We observed that 6,996 (94 %) of the publication fees were paid in accordance with the DFG price cap of 2,000 €. Most payment on publication ranged between 1,000 - 1,250 €.
Figure 2: Distribution of institutional spending on publication fees by German research organisations
Figure 3: Institutional spending on publication fees by German research organisations per year
Figure 3 presents institutional spending per article and year. Large price variations could be observed. Publication fees that were paid by the German universities and research organisations ranged between 40 € and 7,419 € (SD = 486). However, the average price paid varied somewhat during the period 2011 and 2015 (1239 - 1423 €).
The number of APC payments per institutions varied considerably (see Table 1). With 2,856 reported articles, the Max Planck Society contributed 39 % of the overall article volume. By contrast, we observed a lower number of supported open access journal articles for several universities that only recently begun to set up centrally managed open access funds to cover publication fees.
Table 1: Institutional spending on open access publications (in €)
Table 2 compares Open APC spending data with that of the Austrian FWF, as well as with Jisc's and Wellcome Trust's expenditure. Prices were converted according to the average Euro exchange rate of the examined periods, and gathered for both fully and hybrid open access journals. The comparison revealed that the Open APC initiative lacked cost information about hybrid journals, whereas the related Austrian and UK open data initiatives reported a large share of spending on these journals between 2014 and 2015. Over the years 2005 – 2015, 3 out of 30 German universities and research institutions reported 60 hybrid journal articles to the Open APC initiative, representing 0.81 % of all articles included in the data-set. In terms of the number of supported articles and the amount being spent on publication fees, by contrast, the Open APC data-set provided the most comprehensive price information for fully open access journals compared to what the Austrian and UK initiatives had reported.
Comparison of average prices revealed that publishing in hybrid journals was, on average, more expensive than in fully open access journals. Price differences between these two categories were also reported earlier, indicating that prices for fully open access journals were on average lower [@Solomon_2012; @Pinfield_2015]. In 2014 and 2015, the mean price for fully open access journals calculated from all data-sets was below the DFG price cap of 2,000 €.
To identify publication fee spending on the article-level, as well as to gather bibliographic metadata, DOIs were a mandatory part of the Open APC initiative's data collection activities. The participating institutions reported DOIs for 7,373 out of 7,417 articles. Using these DOIs, we retrieved additional metadata from Crossref for 7,346 publications, representing 99 % of the total article volume. Articles for which no metadata could be obtained, were registered with the DOI agency DataCite (10 articles) or Medra (two articles). For eight articles, our parser could not gather the XML resource, although these publications were registered with Crossref at the time of our study. Seven DOIs reported to the Open APC initiative did not resolve.
We used the DOI to automatically fetch publisher and journal names for each article from the Crossref REST API. Table 2 shows the top ten publishers in terms of the number of financially supported articles. These publishers represented 92 % of all articles include din our data-set. In total, payments were made to 139 publishing houses. In comparison with data from the UK, we observed a greater share on total spending of some open access publishers. @Pinfield_2015, for instance, reported remarkably lower numbers for the open access publishers MPDI, Copernicus, and Hindawi.
Table 2: Publication fees paid per publisher (in €)
Most of publication fee spending in Germany was on articles published in Springer Nature journals, likely profiting from the merge with the open access publisher BioMed Central in 2008, and that between the well-established publishers Springer Science + Business Media and Nature Publishing Group in 2015. Using the Crossref-Member-ID instead of the publisher name, we were able to differentiate between journals formely published by Springer Science + Business Media and Nature Publishing Group. In terms of articles, the majority of payments made were for publications in journals formerly associated with Springer Science + Business Media. Springer Science + Business Media journals accounted for 2,027 articles, representing 94 % of the overall Springer Nature article volume recorded by the Open APC initiative, and 92 % in terms of the amount being spent. Median publication fee spending slightly differed between Springer Science + Business Media (1,355 €) and Nature Publishing (1,386 €). However, price variation was higher for Nature Publishing journals (SD = 848€) compared with that of the former Springer Science + Business Media titles (SD = 313 €).
In contrast to Springer Nature, other well-established publishing houses such as Elsevier and Wiley-Blackwell ranked lower in our analysis.
Table 3: Publication fees paid per journal (in €)
Prices also varied within journals. Based on the number of articles paid for, Table 3 illustrates the top ten out of 732 journals. We normalised PLOS journal titles, because the name changes from "PLoS" to "PLOS" was only partly represented in the Crossref metadata at the time of our study. Articles published in the top ten journals represent 45 % of the overall article volume. The multidisciplinary journal PLOS ONE ranked highest. In addition, the journals New Journal of Physics, Atmospheric Chemistry and Physics Discussions and Frontiers in Psychology, all of which publish contributions from all branches of their respective discipline, were also well represented in our study. In the case of Atmospheric Chemistry and Physics Discussions, the large price range can be explained by the fact that this journal charges per page and also takes the submission's file format into consideration.
In Germany, institutional spending on publication fees charged by open access journals has increased over the years. These findings are consistent with the general trend of publication fees as revenue source for open access publishing [@Davis_2011; @Laakso_2012; @Pinfield_Review_2015]. They also demonstrate the growing institutional support for researchers in Germany to cover these charges [@Fournier_2013]. Similar to the expenditure on publication fees at an institutional level in the UK [@Pinfield_2015], spending volume varies across the German universities and research organisations. With a proportion of 39 % of the total article volume, the Max Planck Society, a large non-university research organisation, supported most open access journal publications included in our study. A possible explanation is the centralised library support at the Max Planck Society where the Max Planck Digital Library has managed open access agreements with publishers over the last decade on behalf of most Max Planck institutes [@Schimmer_2013; @Sikora_2015]. This centralised approach presumably not only resulted in a large number of supported open access articles, but also in central access to cost data provided by publishers, as well as in advanced capabilities and skills needed to report on these expenditure on a regular basis. Many universities and research organisations, on the other hand, disclosed a remarkably lower number of supported articles.
To answer data curation issues while unifying UK spending data [@Neylon_2014; @woodward_2014], re-using DOIs to gather bibliographic metadata from Crossref is a promising approach. In our study, Crossref thoroughly indexed open access journal articles disclosed in the Open APC data-set, providing information about publisher and journal titles for 99 % of all articles included in the Open APC data-set. Making use of metadata from Crossref, therefore, reduces extensive validation work of bibliographic records as long as the DOIs are made available in the cost data. Beyond identifying publishers and journals, mandatory reporting of DOIs in the spending data can also increase the re-use of such data to study other aspects of open access journals using publication fees. For instance, impact analyses in the field of altmetrics make heavy use of DOIs as well [@Haustein_2016]. Drawing on Crossref has the potential to increase the comparability of cost data to prepare future negotiations with publishers on open access agreements, because Crossref's metadata represents current developments in academic publishing in terms of ongoing mergers of publishing houses. In addition to these practical consequences, future comparative studies of publication fee spending using data on the article-level can also benefit from such an approach.
This study is limited in some respects. One is that we cannot assess whether publishers and journals granted publication fee discounts. The Open APC initiative uses a minimal data scheme to encourage self-reporting, and, therefore, does not track this kind of information. However, large price ranges suggests that different pricing levels and varying pricing schemes are in place, an observation also made earlier [@Solomon_2012; @Pinfield_2015]. Adding to this complexity, it is likely that some institutions only paid parts of the publication fee. Take for instance the journal Nature Communication. Charges reported to the Open APC initiative ranged between 2000 €, the DFG price cap, and 4.403 €, a journal that can be grouped as pricy high-impact journal according to @Solomon_2012. Although such payments being made from several budgets are a proposed strategy to sustain publication funds at German universities [@Fournier_2013], these pro-rata payment were not made transparent in the Open APC data, leading to a possibly flawed representations of publication fee spending in Germany. In another case, one university included its contributions to the SCOAP^3 consortia and presumably divided the sum by the articles published by their authors in SCOAP^3-covered journals.^[26] This is very arbitrary, since averages for an institution can only be determined after the end of a full 3-year funding cycle. Other factors affecting price variations are exchange rates and differend tax rates for some organisations in Germany. For instance, the Max Planck Society has a limited input tax reduction. The refund of input value added tax for publication fees is 20 %. To increase the transparency of publication fee spending, @Pinfield_2015 generally suggested to disclose tax rates and the currency the payment was made in future cost data-sets. In addition, the Open APC data-set does not track funding sources so that we cannot determine which funders co-financed publication fees.
It must also be noted that reporting to the Open APC initiative is voluntary. Therefore, not all institutions in Germany that provide central funding for publication fees contribute cost data to this initiative. In a qualitative survey, asking why German institutions are reluctant to share their cost data through the Open APC initiative, one institution feared that an increase in transparency would allow publishers to adjust prices in their favour. Others pointed out that the workload to produce such a data-set could be too extensive [@deppe_2015]. As no reliable registry of institutional open access funds or related support structures in Germany exists, we cannot assess the number of non-participants in Germany.
Our analysis of how institutional spending for open access articles was distributed over publishers and journals indicates that open access publishing is heterogeneous and concentrated at the same time. While we were able to identify 139 individual publishing houses that were supported by the German universities and research organisations, the distribution is highly skewed. Ten publishers collected 92 % of open access publication fee spending, being consistent with an observed high concentration of few publishers in current academic publishing [@Larivi_re_2015]. However, our study could not confirm that publications in open access journals owned by traditional publishing houses account for most of the spending on publication fees as observed by @Pinfield_2015. Rather, open access publishers such as Public Library of Science (PloS), Copernicus, or MPDI rank higher in our study than in the analyses of cost data in the UK.
One possible explanation why traditional publishers are less well represented in our study is the lack of cost information about hybrid open access journals. 99 % of all articles German universities and research organisations financially supported were published in fully open access journals. This presumably reflects the DFG funding policy that opposes paying for open access articles published in hybrid journals. However, while reviewing self-reported cost data from Austria and the UK where hybrid open access journals are generally supported, we observe a much higher share of payments for articles in hybrid open access journals. Because publication fee spending is fragmented and often in-transparent, it remains open to speculation whether authors affiliated with German universities and research organisations avoid opting for open access when publishing in hybrid journals, or simply use other budgets that are not covered by the Open APC initiative.
Our study reveals the size and extent of spending on open access journals using publication fees in Germany. Drawing on self-reported cost data from the Open APC initiative, support from German universities and research institutions have grown over the years. Comparing these expenditure with those from Austria and the UK, German open access funding is mostly focussed on fully open access journals, raising important questions about hybrid open access journals as publication venue. Given our findings, and in light of the general discussion about funding policies addressing hybrid open access journals, questions about whether and to what extent science policy interventions, as well as the availability of institutional support, influence how researchers publish are of particular concern.
Using self-reported data and gathering publisher and journal information from Crossref, our study extends methods and improves data collection activities for researchers and practitioners alike, as well as contribute to a better understanding of factors affecting the analysis of publication fees in open access publishing. In this regard, our research highlights large variation in the distribution of spending that needs to be taken into consideration when studying payment on publication at the institutional level. We have also confirmed previous studies showing large price variations across publishers and open access journals that need to be better understood in future.
We thank Andrea Hacker and Ada-Charlotte Regelmann for valuable comments on the first draft of this paper. We also thank Christoph Broschinski, Vitali Peil, and Dirk Pieper, the members of the DINI working group "Electronic publishing", and all data contributors7 of the Open APC initiative.
Footnotes
-
Guidelines for the funding program can be found here: http://www.dfg.de/formulare/12_20/ ↩
-
https://github.com/OpenAPC/openapc-de/tree/v2.4.3 ↩
-
rOpenSci: https://ropensci.org/ ↩
-
https://github.com/CrossRef/rest-api-doc/blob/master/rest_api.md ↩
-
https://github.com/OpenAPC/openapc-de#contributors ↩