David among Goliaths: Open access publishing in scientific (semi‐)periphery

This case study analyses data on papers of Croatian authors published in 2017 from four Web of Science Core Collection citation indexes (SCI‐EXP, SSCI, AHCI, and ESCI). The primary dataset (5,176 articles and reviews) was divided into two subsets, the open access (OA) subset (2,964 papers) and non‐OA subset (2,212 papers). We also used the primary dataset to create a subset of papers published in Croatian journals (1,588) as opposed to foreign ones. All were screened for full‐text OA status, journal JCR quartile ranking, journal dominant discipline, and language of publication. OA papers prevailed with 74.4%. Most were available at publisher websites. The percentage of OA papers in Croatian journals was 99.8%. The share of OA papers was the highest in the humanities and social sciences, which also saw the highest share of papers in the Croatian language.


INTRODUCTION
The countries of the European Union (EU) produce one quarter of the world's scientific papers (Noorden & Butler, 2019), and according to the European Commission (EC), 40.4% of them were freely accessible in 2017 (EC, 2020).The EC recommends (Com-leading EU countries in the number of OA papers (EC, 2020).
According to its development indicators, Croatia belongs to the group of semi-peripheral countries.Semi-periphery countries are positioned between the periphery, characterized by low R&D investment, poor research infrastructure, etc. and the core, scientifically and technologically advanced countries.They play a major role in mediating economic, political, and social activities linking core and peripheral areas (Sorinel, 2010).Domazet and Marinovi c Jerolimov (2014) describe the European semi-periphery, especially its post-socialist countries, as 'the countries with a high level of development, widespread educational and healthcare attainments, close to European cultural sphere'.Croatia is still going through transitional changes in almost all of its economic, educational, social, and political life.By joining the EU in 2013, Croatia has become its youngest member state.Its investment in research and development in 2017 was 0.85% of GDP, which is far below the EU average of 2.3% (The World Bank, 2019).However, the number of papers published by Croatian authors has steadily been growing.
According to the total number of internationally visible papers between 2013 and 2017, Croatia was in the group of seven lastranked EU member states.But research intensity measured by the number of papers per 10,000 population shows that Croatia, with 11 papers per 10,000, ranks 18th among EU countries, higher than France, Greece, Slovakia, Hungary, Bulgaria, etc. (Noorden & Butler, 2019).Between 2013 and 2017, the Croatian production was 31,853 documents in journals covered by the Web of Science Core Collection (WoSCC).Of these, 13,023 were available through OA.And while the increase in overall 'output' in this 5-year period was 21%, the increase in the overall number of OA papers was as high as 58.7%.These increases were partly influenced by the changes in the WoSCC indexing policies and the introduction of Emerging Sources Citation Index (ESCI) (Testa, 2009; Web of Science: Emerging Sources Citation Index: Search and discover a trusted set of journals with comprehensive coverage by subject and region, 2020).
It has been a while since Guédon (2008) noted that OA journals would probably be crossing the divide between peripheral science, where they started, and 'core' or 'mainstream' science.He argued that international scientific competition had gradually led to a two-tier communication system, the international acting as the main quality arbiter and national reflecting national politics and priorities.He emphasized that Garfield's Science Citation Index had 'a strategic role in the present power structure of world science', reflecting the dominance of mainstream journals.However, peripheral journals improved their position with the launch of the Emerging Sources Citation Index (ESCI), which ensured important research was visible in WoSCC, 'even if it is not yet internationally recognized' (What is the Emerging Sources Citation Index?, 2020).For example, ESCI covers more than 50% of all Croatian journals indexed in WoSCC.
Kie n c (2017) reported that authors from scientific periphery published in OA journals more often than authors from core countries, 46% versus 20% of papers, respectively, to be more exact.Archambault et al. (2013) found that publishing in gold OA journals was much more common in eastern Europe and hypothesized that it was because gold journals more frequently allow publishing in languages other than English.Croatia, however, does not fit that profile most Croatian scientific journals, especially in the STEM fields, are published in English (Pulišeli c & Petrak, 2006).
About 180 scientific and technical journals in Croatia were supported by public funds in 2017.Their publishers are mainly notfor-profit organizations (universities, research institutes, learned societies) under the patronage of the Croatian Ministry of Science & Education (MSE).Among the MSE's most important criteria for subsidizing a journal is that it is indexed by at least one international bibliographic database and that it provides OA to the entire journal's content on the Hrc ˇak repository portal (Ministarstvo znanosti i obrazovanja, 2018).Hrc ˇak was launched in 2006 as an outlet for journals not having infrastructure for e-publishing and an interface for free access to all journal content (Stojanovski, Petrak, & Macan, 2009).As a platform, Hrc ˇak gave Croatian scientific journals better international visibility, which eventually increased the number of Croatian journals indexed in relevant international bibliographic databases.In 2007, Web of Science databases indexed only 16 Croatian scientific journals; in 2011, they indexed 61 (Macan, Piki c, & Mayer, 2012), and their current number is 124 (Web of Science Group, a Clarivate Analytics Company, 2020).Meanwhile, again mainly because of Hrc ˇak requirements, Croatian journals upgraded their editorial practices, which was the other main reason why so many are indexed in international bibliographic databases.Piwowar et al. (2018) suggested further investigation to better understand what drives OA.The aim of this study was therefore to explore in depth the OA production of Croatian authors indexed by the WoSCC and published in 2017.We focused on OA papers, their disciplinary and quartile distributions, and the characteristics of Croatian OA journal publishing.

Key points
• Three quarters of papers published by Croatian authors are available in open access (OA).
• Most papers are in English, even in Croatian journals; Croatian prevailed only in the humanities.
• Croatian journals are mostly non-commercial, and many receive government subsidies, provided they are OA; they generally do not charge for article processing.
• The prevalence of OA over non-OA papers is the highest in the humanities, followed by the social sciences.
• Open access publishing might increase international visibility of journals from periphery or semi-periphery countries.

METHODS
The of the authors' affiliation addresses, so our results were not restricted to the corresponding authors.
The results were then filtered by document type to include only 'article' and 'review' papers.What remained was our primary dataset of 5,176 papers, which we used as the basis for all further analysis.Clarivate Analytics InCites was used for disciplinary distribution analysis, as well as for quartile distribution (Q) of papers according to the journal's impact factors (JIF).Hrc ˇak and the journals' homepages were used for the identification of the Croatian journals, and DOAJ was used for the additional checking of their OA status.
Our primary dataset was divided into three subsets according to the goals of our analysis: • OA subset comprised 2,964 papers.It combined two subsets: • a subset of 2,574 papers obtained by refining the primary dataset by the WoS open access filter (all OA types according to the WoS typology were included) and • a subset of 390 papers published in Croatian journals that were not marked in WoS as OA papers but were identified manually by journal name search in Hrc ˇak and DOAJ.
• Non-OA subset included the remaining 2,212 papers.
• A random sample was created in order to check if there was any 'hidden' OA paper within the non-OA subset.
Randomization was based on the expected OA proportion of 20%, 95% confidence interval, and type 1 error of 5%.The OA status of thus obtained 212 papers was checked manually using the article title as a Google Scholar search string.All retrieved papers were inspected to find respective OA full text.
• Croatian journals subset comprised 1,588 papers published by Croatian journals.

DATA ANALYSIS
All papers were analysed to determine OA to full text, disciplinary orientation and JCR quartile distribution of journals, journal's country of origin, and language of publication (with focus on Croatia/Croatian).
Clarivate Analytics defines OA papers as 'freely accessible peer reviewed versions of an article from either a publisher's website or repository' (Clarivate Analytics, 2018).In WoSCC, OA papers are grouped into five categories: DOAJ Gold, Other Gold, Bronze, Green Published, and Green Accepted.For the purpose of this study, we decided to use only two broad OA types: papers available at publisher websites (irrespective of their specific publishing model as gold, bronze, hybrid, delayed OA, etc.) and those available on repositories, personal websites, and academic social networks (green OA).We did not verify the validity of WoS data.
Regarding the country of publication, the papers were divided into two groups: published in Croatian journals and published in foreign journals.There were three language groups: papers in English, papers in Croatian, and papers in other languages.
The records from WoSCC were transferred to the Clarivate Analytics InCites platform to determine discipline and JCR ranking.For disciplinary distribution of the papers, we used the first level of OECD

Primary dataset
Our search of the four WoSCC citation indexes resulted in 5,176 papers with at least one Croatian affiliation.They were published in 117 Croatian and 2,018 foreign journals, most of them (87.2%) in the English language.
A total of 3,768 (72.8%) of all papers in the primary dataset were published in JCR-ranked journals.Among them, 31.7% were published in Q1 journals, 25.7% in Q2, 18.8% in Q3, and 23.6% in Q4.Most papers in the Q1 category belong to natural sciences (63%).The distribution, however, varies greatly between Croatian and foreign journals (Table 1).
Papers published in journals covered by ESCI (23.7%) could not be distributed by the quartiles as impact factors are not calculated for these journals.The same is true for the AHCI journals, which published 2.9% of papers.Journals covered by ESCI published 58.4% of all social sciences and 63.3% of all humanities papers (Table 1).

OA subset
The OA subset accounted for 57.3% (n = 2,964) of all retrieved papers.Most full texts were available at publishers' websites (94.2%).Many of them also had their 'green' versions available in repositories, academic social networks, and similar platforms (38.5%).A total of 1,585 (53.5%) papers in this subset were published in Croatian journals, and all of them were available on publishers' websites.
The OA subset shows a disciplinary distribution similar to the primary dataset, with natural sciences having the greatest number of papers (30.2%) (Fig. 1).

Non-OA subset
The non-OA subset comprised 2,212 papers.Only three were published in Croatian journals.Almost half of the papers (47.8%) in the non-OA subset belonged to the field of natural sciences and only 4.5% to the humanities (Fig. 1).Altogether, 88.7% (n = 1,962) of papers in the non-OA subset were published in JCR IF journals (Fig. 2).Most were published in Q1 journals (30.9%), followed by those in Q2 journals (25.8%), while only 8.8% were published in ESCI journals.

Non-OA randomized sample
Our analysis of the randomized sample (n = 212) in the non-OA subset discovered that 85 papers were actually OA (40.1%).
Twenty had free full-text access on the publisher's web pages, while 65 were available in repositories (39 papers) or academic/ social networks (26 papers).If we apply the 40.1% to the entire non-OA subset, as many as 887 papers in the non-OA subset could actually be OA.If we add these 887 papers to the number from the OA subset, the overall share of OA in the primary dataset rises to 74.4% (Fig. 3).
Most of 'green-only' papers were published in Q1 and Q2 journals (45 out of 65), which suggests the importance of repositories, social networks, and similar platforms for papers published in the most prestigious paywall journals with high JIF.

Croatian journals subset
Of all primary dataset papers, 30.7% (n = 1,588) were published in 117 Croatian journals.Only two Croatian journals in the primary dataset had a paywall (were non-OA), and they published three papers that met our search criteria.The remaining 115 journals publish in OA, most without charging for article processing.This is why 99.8% of all papers in Croatian journals were OA (n = 1,585).They accounted for 53.5% of the papers in the OA subset and for 56.8% of those published on the publisher's web pages.
Further subset analysis showed that 68 of the 117 journals were indexed by DOAJ and that 91 used one of the Creative Commons licences (most frequently CC BY), but most of them did not show this information in the article (e.g. in headers or footers).Twelve more journals did not refer specifically to the CC licence but allowed use of full texts for non-profit purposes in the journal's policy section.Twenty-three journals did not assign Digital Object Identifiers (DOI) to the papers published in 2017.
The OA status of 390 papers published in 34 Croatian journals was not determined in WoS.The main reason may be that 33 of those journals were not registered in DOAJ in 2017, and many of them had unclear copyright and OA policies.
More than half (52.8%) of papers were published in Croatian journals covered by ESCI, followed by 30.7% published in Q4 journals.Only seven papers were published in the three Croatian journals ranked Q1 (Table 1).Disciplinary distribution showed that only 8.5% of all natural sciences papers and 15.6% of all medical and health sciences papers were published in Croatian journals.In contrast, Croatian journals published 57.1% of all social sciences and 73.5% of all humanities papers from the primary dataset.
Just over a third (36.5%) of papers in this subset were published in Croatian language (11.2% of the primary dataset), most of which in the humanities and social sciences.Croatian is a predominant language in humanities papers across all datasets: 82.2% in Croatian journals, 75.4% in the OA subset, and 60.4% in the primary dataset.In contrast, only three medical papers were published in Croatian (Fig. 4).

DISCUSSION AND CONCLUSIONS
Our findings show that 74.4% of papers with at least one Croatian affiliation are OA, which ranks Croatia shoulder to shoulder with the leading OA nations -Brazil, Netherlands, Switzerland, and United Kingdomwith more than 65% of OA in their overall WoS-indexed publication output (Science-Metrix, 2018).For a country on the scientific semi-periphery with relatively scarce financial and infrastructural research resources, OA to research output is an effective tool for increasing international visibility and research collaboration on the international scientific scene dominated by more productive and influential nations.
In his survey comparing OA publishing by authors from core and periphery countries, Kie n c (2017) argues that most journals in periphery countries are owned by academic societies and universities and are run as non-profit initiatives, often on public money.For most of them, journal publishing is a part of their public mission.Publishing in OA is the best way to accomplish that mission.
This is also the case with the majority of Croatian journals that publish in OA.They operate on public money and can pursue OA policy without article processing charges.Namely, OA is mandated if the journal is to receive government support/subsidy.However, the government has not issued an OA policy mandating publishing publicly funded research in OA outlets.This has been left to individual research institutions to decide, and only a few have imposed an OA mandate so far (Macan, 2018).Many Croatian researchers signed the Croatian open access Declaration in 2012 'inviting the state administration, headed by the ministry responsible for science, as well as scientific and educational institutions, organisations, professional associations, and all the others involved in gathering and publishing scientific information to act decisively and in coordination in order to store all the Croatian scientific information in open access form' (Budin et al., 2012).Until now, Croatia has not produced a universal strategy that would encourage its researchers to publish all their research OA.
As Kie nc explains it, '…for authors from the global periphery OA is the least important factor in choosing a journal to publish work in, while journal impact factor and abstracting services are the most important and that these two factors favour 'mainstream' journals that are usually published in the core countries' This culture of 'publish in high-profile journals or perish' (Wang et al., 2004) is prevalent in Croatia too.This is especially true for the authors in the fields of science, engineering, and biomedicine who are more focussed on the international scientific community and are more inclined to publish in international journals (Table 1).
Publishing in prestigious international journals enhances the recognizability of the author or the institution (Joki c & Šuljok, 2009).
Besides, publishing in the ranked international journals, whether they are OA or not, has become the most important instrument of the academic promotion in Croatia.Our results show that 23.1% and 18.8% of all papers were published in Q1 and Q2 journals, respectively.However, the share of papers indexed by ESCI is the largest (23.7%), and this is where the prevalence of OA papers is the highest.OA also prevailed in papers from Q4 journals, and in both categories, the share of papers from Croatian journals was high (Table 1).show that 40% of 'hybrid OA authors' drew on funds from their research funder for cover APCs.For 'fully OA authors', the figures are even higher; 59% of them pay their APCs from grant money (Monaghan, Lucraft, & Allin, 2020).With Croatian journals, the choice is mandated by the journal OA policy, which stems from public subsidies rather than author preference.
The analysis of our randomized sample indicates the importance of national or institutional repositories and academic/social networks that constitute the so-called 'green' OA.Most of these 'green-only' papers were published in Q1 and Q2 journals The Croatian example confirms that publicly funded support for OA publishing in local scientific journals may improve dissemination and access of the scientific results, as well as the chance of their international visibility.SSH disciplines especially benefit from such a policy.Croatian authors from other disciplines, such as natural and biomedical sciences, more often publish in the prestigious international journals with high OA publishing fees, and their papers more often remain closed or 'green-only' in their institutional repositories.

LIMITATIONS
Our findings are limited by the study design to a single year and a single country.They may help, however, to understand OA in similar settings at scientific semi-periphery.

Figure 1
Figure1Disciplinary distribution of papers within primary dataset and its subsets.

Figure 4
Figure 4 Distribution of papers in Croatian journals across scientific fields by the publication language.

(
Kie n c, 2017).Similar was reported by a Taylor & Francis 2019 researcher survey (Taylor & Francis researcher survey 2019: Researchers' views on the present and future of scholarly communication, 2019)the availability of open options was less important to authors than the reputation or impact factor of the journal.
(n = 45; 71.4%), which implies the importance of the green route for papers published in the most prestigious paywall journals with high JIF.Until recently, the awareness of the need for OA digital repositories among Croatian research and high education institutions was rather low, and their implementation and maintenance was a big technical challenge.Only researchers working on international competitive projects (such as the ones funded by the EC through the Horizon 2020 framework) were mandated to archive their papers into an OA repository or a similar infrastructure.The initiative for the development of a common national digital repository infrastructure resulted in the establishment of Digital Academic Archives and Repositories -DABAR (https://dabar.srce.hr), a centrally implemented technological solution based on opensource software (Macan & Petrak, 2019).It complements Hrc ˇak by enabling authors affiliated to the Croatian academic institutions to archive their theses, journal articles, book chapters, conference papers, learning objects, datasets, etc. as green OA.It would further increase the OA share in the Croatian publication output in the years to come.
primary sources of research data were the Clarivate Analytics

Table 1
Distribution of all papers and papers published in Croatian authors by scientific field, JCR quartile (where available), and ESCI Directorate for Science, Technology and Industry, Committee for Scientific and Technology Policy, 2018).This classification consists of six major fields: natural sciences, engineering and technology, agricultural sciences, medical and health sciences, social sciences, and the humanities.This scheme is similar to the official Croatian classification of science (Nacionalno vije ce za znanost, 2009).
Croatian journals account for 30.7% of all papers published by Croatian authors and for 53.5% of all papers in the OA subset.They seem to be an important communication and visibility channel if they are covered by major international bibliographic databases.Matanovi c & Berbi c Kolar, 2018), crucial for the survival of the national culture and language (Zrinšc ˇak, 2011) and not only for the communication of the SSH research results.The prevalence of local topics in SSH papers is also supported by the number of papers published in the Croatian language: 82.2% of papers in the humanities (301 out of 366) and 43.6% in social sci- heritage.The prevalence of SSH papers in Croatian journals indexed in WoSCC may reflect the importance of local social, historical, and linguistic phenomena that have not yet been sufficiently explored.The SSH journals are considered the national identity factor (