Scientific journal publishing: yearly volume and open access availability

Introduction. We estimate the total yearly volume of peerreviewed scientific journal articles published world-wide as well as the share of these articles available openly on the Web either directly or as copies in e-print repositories. Method. We rely on data from two commercial databases (ISI and Ulrich's Periodicals Directory) supplemented by sampling and Google searches. Analysis. A central issue is the finding that ISI-indexed journals publish far more articles per year (111) than non ISI-indexed journals (26), which means that the total figure we obtain is much lower than many earlier estimates. Our method of analysing the number of repository copies (green open access) differs from several earlier studies which have studied the number of copies in identified repositories, since we start from a random sample of articles and then test if copies can be found by a Web search engine. Results. We estimate that in 2006 the total number of articles published was approximately 1,350,000. Of this number 4.6% became immediately openly available and an additional 3.5% after an embargo period of, typically, one year. Furthermore, usable copies of 11.3% could be found in subject-specific or institutional repositories or on the home pages of the authors. Conclusions. We believe our results are the most reliable so far published and, therefore, should be useful in the on-going debate about Open Access among both academics and science policy makers. The method is replicable and also lends itself to longitudinal studies in the future. Scientific journal publishing: yearly volume and open access availability file:///C|/Users/s982780/Desktop/docs/Scientific%20journal%20publishing%20%20yearly%20volume%20and%20open%20access%20availability.htm[5.5.2010 14:05:11] Introduction It is important to begin this paper with two definitions that are central to the entire discourse. By Scientific Journal Paper we mean a paper describing scientific research results, which has undergone some form of anonymous peer-review and which is published in a regularly appearing serial, usually by a third party publisher and not by the university of the author. Journals fall into the science, technology and medicine category as well as social science and the humanities. An alternative term often used is scholarly journals, but we have chosen the term scientific to cover all these subjects. For instance Tenopir and King (2000) sometimes speak of scientific scholarly journals in their influential book. Papers are typically 3,000 to 10,000 words in length and are written following longestablished conventions concerning style, referencing, tables of content etc. Other types of scientific publication include conference papers, book chapters, books and reports. Journal publishing is the most common form of dissemination of new research results, in particular in science and medicine. In some scientific domains, such as computer science, conference publishing is quite important and, in the humanities, book publishing is an important channel. Our analysis deals only with peer-reviewed papers published in journals. Compared with the other types of scientific publication, journal papers are comparatively easier to obtain, even years after publication, because of the large holdings of journals by university libraries. Today, the vast majority of recent journal papers is also available electronically. Most of the larger universities have licenses offering access to all the titles of major publishers (e.g. Science Direct) and many publishers also offer pay-on-demand services for the purchase of individual papers. Open Access means access to the full text of a scientific publication on the Web, with no other limitations than possibly a requirement to register, for statistical or other purposes. This implicitly means that open access material is easily indexed by general purpose search engines. There are several widely quoted definitions on the Web, for instance the Budapest Open Access Initiative (2002). For the scientific journal papers in particular, open access can be achieved using two complimentary strategies: gold open access means journals that are open access from the start, whereas green open access means that authors post copies of their manuscripts to open access sites on the Web (Harnad et al. 2004). As there are numerous different types of parties involved in the scientific publishing value chain (Björk 2007), such as publishers, libraries and authors, with sometimes conflicting interests, much of what is written about open access is strongly biased either towards promoting open access or describing the dangers of open access to the scientific publishing system. There has also been a discussion among open access advocates which of the two strategies (gold or green) is better. There is thus an urgent need for reliable figures concerning the yearly volumes of journal publishing and the share of the yearly volume, which is available as open access via different channels. In most of the earlier discussions about the economy of journal publishing the focus has been on the number of journals and costs (such as the subscription cost) have been mainly related to the individual title (e.g. European Commission 2006). This was natural because Scientific journal publishing: yearly volume and open access availability file:///C|/Users/s982780/Desktop/docs/Scientific%20journal%20publishing%20%20yearly%20volume%20and%20open%20access%20availability.htm[5.5.2010 14:05:11] of the easy availability of subscription information for individual titles and for the handling of paper copies in libraries all over the world. We argue that since the advent of the digital delivery for the contents and the electronic licensing of vast holdings of journal content (the big deal), the focus should be more on the individual papers as the basic molecule of the journal system and that any average costs should be related to the paper. We also think that the ratio of open access papers to the overall number of papers published is a much more important indicator of the growing importance of open access than the number of open access titles compared to the number of titles in general. Total number of papers published A central hypothesis in this calculation was that the journals indexed by Thomson Scientific's (ISI) three citation databases (Science Citation Index, Social Science Citation Index and Arts and Humanities Citation Index) on average tend to publish far more papers in a volume than the often more recently established journals not covered by the ISI and that this should explicitly be taken into account in the estimation method. We proceeded as follows. To estimate the total number of scientific peer-reviewed titles we used Ulrich's Periodicals Directory and conducted a search with the following parameters; Academic/Scholarly, Refereed and Active. In winter 2007, this yielded a total of 23,750 journals. For the case of the journals indexed by the ISI it was possible to extract the total number of papers published in the last completed year (2006) by conducting a search in the Web of Science (WoS). A general search was done covering all three indexes (Science Citation Index Expanded, Social Sciences Citation Index and the Arts and Humanities Citation Index). The parameters were set as follows; Publication year = 2006, Language = All languages, Document type = Article. Since the system has a limitation in the number of items shown of 100,000 it was not possible to directly get the total number of indexed papers. The problem was solved by systematically going through the alphabet by setting the Source Title as A*, B*, C* etc. This worked well for all other letters, for which the total number was less than 100,000, except for A and J. For the letter A more detailed search on AA*, AB* etc was enough, for J we had to go down to the level of Journal of A*, Journal of B* etc. The total number of papers we arrived at in this way was 966,384. ISI, as a rule, only indexes peer-reviewed journals, but with at least one notable exception, the Lecture Notes in... series published by Springer, which publishes conference proceedings in computer science and mathematics in book form. By doing a search using the above as Source Title we got the number of papers published in this series, which was 20,484. Subtracting this number from the earlier total leads to a final number of 945,900 ISI papers. If we know the exact number of titles that the ISI tracked in the Web of Science in 2006, we can easily derive the average number of papers published in a year by each title. As we did not have access to exact figures from ISI we had to find a roundabout way to estimate this figure. One indication is given by the number of journals included in the Journal Citation Reports. When searched from Ulrich's and defining Journal Citation Reports as a further search criterion, the result is 6,877 titles. For one reason or another, the search directly from Journal Citation Reports for 2006 gives more journals: 6,166 titles indexed Scientific journal publishing: yearly volume and open access availability file:///C|/Users/s982780/Desktop/docs/Scientific%20journal%20publishing%20%20yearly%20volume%20and%20open%20access%20availability.htm[5.5.2010 14:05:11] in Science Citation Index and 1,768 in Social Science Citation Index. Arts and Humanities Citation Index journals are not included in the Journal Citation Reports. We can, however, estimate the number of titles by assuming that Arts and Humanities Citation Index journals on average publish as many papers a year as Social Science Citation Index journals (53.1) which would result in an additional 532 titles. Adding these together, we would get 8,466 titles. Using these numbers as a base, we are able to estimate the average number of papers published in journals indexed in Web of Science by ISI as 111.7 per title. This can, for instance, be compared to the figure of 123 papers per year for 6,771 US publishers reported by Tenopir and King (2000). The number of titles indexed in the Web of Science is probably slightly higher for two reasons. The main reason is a time lag between the inclusion in the indexes and the first journal citation report produced for a specific journal. According to ISI (Personal communication from David Horky, Thomson Scientific, 17th of January, 2008) the number of titles indexed in the citation databases at the end of the year 2007 was 9,190 journals. At the beginning of 2008, according to ISI's Web-pages, the number of journals had risen to 9,300. Assuming that the number of journals indexed rises steadily every year, this would indicate that the number would have been somewhere between our estimate and this information. However, we have chosen to use our earlier mentioned estimate (8,466) because the number of titles does not influence the number of ISI-papers we have obtained separately. It does affect our estimate of the number of non-ISI journals, since these are obtained by subtraction (see text below). As we have estimated these to have a much lower number of papers published in a year, the effects of a possible mistake in our number of ISI-titles of 1% would be only around 0.2% in the total number of papers. Taking as a starting point the total number of titles as 23,750 and the number of titles indexed by the ISI as 8,466 we arrive by subtraction at the number of titles not indexed by the ISI as 15,284. In order to arrive at a total number of papers we now need to estimate how many papers these journals publish yearly, on average. This was done using a statistical sample of journals. The basis was Ulrich's database from which a sample of 250 journals was taken. We set the search so that we chose only journals that have an online presence. This might statistically result in a slight bias, but was the only practical way we could study the publication volumes of the journals in the sample. Then we extracted the number of papers published in 2006 until we had data for 104 journals (journals in the original sample that were indexed by the ISI or for which the number of papers could not be found were discarded). In this group the average number of papers published was 26.2, which, as we had suspected, was considerably lower than for ISI-indexed journals. Five of the journals had published no papers and the journal with the highest output had published 225 papers. Multiplying 26.2 by 15,284 results in an estimate of papers published in 2006 of 400,440. Adding the figures for ISI brings the estimate of the total number of peer-reviewed papers to 1,346,000 (rounded off) with 70% covered by the ISI. In their answer to a UK House of Commons committee in 2004, Elsevier estimated that some 2,000 publishers in science, technology and medicine publish 1.2 million peerreviewed papers annually (U.K. Parliament... 2004). Taking into account publishing in the social sciences and the humanities our estimate seems to be well in line with these figures. Scientific journal publishing: yearly volume and open access availability file:///C|/Users/s982780/Desktop/docs/Scientific%20journal%20publishing%20%20yearly%20volume%20and%20open%20access%20availability.htm[5.5.2010 14:05:11] Table 1: Estimated total number of journal papers published in 2006. The figures in bold have been extracted from the two databases used (ISI and Ulrich's). The highlighted figure in the centre was determined based on counting papers for a statistical sample of non-ISI journals with tables of content on the Web. The figures in italics result automatically from the other parameters by simple arithmetic operations. Share of open access publishing In policy discussions concerning open access publishing a very important question is, 'What share of all scientific papers is available openly?'. For a given year (in our case 2006) this concerns both papers directly published as open access (the so-called gold route in open access jargon) and papers published in subscription based journals, but where the author has deposited a copy in a subject-based or institutional repository (the green route). It is easier to estimate the number of gold route papers. In the case of copies in repositories, the evidence is much more scattered and there is the additional difficulty of checking the nature of the copies (copy of manuscript submitted, personal copy of approved manuscript or replica of published article).