Introduction

The open access movement endeavours to achieve universal access to research findings and to realise the expectations to handle the journal crisis, (i.e., the high and ever-growing costs of journal subscription as opposed to decreasing budgets of libraries). Since scientific publications, whether openly accessible or not, are costly and investment-intensive, various open access models have been devised and followed to finance the costs. In the author-pays model, the article processing charges are paid by authors. This model is mainly seen in fully gold journals, where all articles are openly published in exchange for compulsory payment of the article processing charges, and in hybrid journals in which non-open access articles co-exist with the open access ones voluntarily paid by their authors (Pulverer, 2018).

The article processing charges are provided by authors or their sponsors in the form of grants or additional funds in research projects. The sustainability of this model entails, therefore, the tendency of authors and funding agencies towards it because authors have such priorities as journal quality and publication speed over the open accessibility issue (Solomon and Björk, 2012a). Meanwhile, the expensive charges which vary from $300 to $5,000, from publisher to publisher, and from journal to journal have increasingly concerned scientists (Ellers et al., 2017).

Furthermore, open access mandates that require grantees to make their research outcomes open (e.g., via self-archiving or submitting to open access journals) have not been widely effective (Björk, 2016; Siler et al., 2018). Therefore, for balanced growth and expansion, the model requires strong incentives on the part of writers and their sponsors. Increasing the influence of researchers in their scientific community can motivate them to support open access goals.

According to research studies, article-processing-charge-funded papers published in hybrid journals achieve a citation advantage compared to the self-archived and non-open access articles published in the same journals (Eysenbach, 2006; Gaule and Maystre, 2011; Harnad and Brody, 2004; Laakso and Björk, 2016; Mueller-Langer and Watt, 2018; Sabharwal et al., 2014; Sotudeh et al., 2015; Wang et al., 2015). They also outperform those article-processing-charge-funded papers published in gold journals (Chaudhuri and Thohira, 2010).

The open access citation advantage is attributed to either the higher visibility of open access articles or the self-selectivity of the authors for their high-quality papers to be made openly accessible (Björk and Solomon, 2015; McCabe et al., 2013; Swan, 2010). Whatever the underlying reason is, the increase in the open access citation advantage may imply an acceleration in knowledge circulation and dissemination which is its ultimate ideal.

However, open access contributing countries are not equal in their citation performances (Hajjem et al., 2006; Sotudeh and Horri, 2008, 2009) transmitting the Matthew effect to the open access realm. The phenomenon refers to the inequalities in science reward systems caused by the contributors' status in science. Accordingly, the systems recognize the scientists of high reputation for particular scientific contributions, while withholding the recognition from less famous ones (Merton, 1968).

At the country level, the phenomenon is rooted partially in the reality and partially in the cognitive, psychological and social biases in citation behaviour in general and against less-developed countries in particular. The countries' science systems are subject to inherent shortcomings, including the lack of institutionalised research spirit, the absence of scientific communities, inadequate interaction among universities and industries, isolation of research and its negligible role in decision making, technological and instrumental underdevelopment, inefficiency of reward and reviewing system, the lack and the misallocation of research funding (Sotudeh and Horri, 2008).

The research underdevelopment exacerbates disbelief and cognitive biases against the quality of science in these countries. A study conducted on Elsevier and Springer's article-processing-charge-funded papers showed that countries at different levels of scientific development were interested in contributing to the open access model in hybrid journals, although the model owed its bulk, either in terms of impact or annual growth, to the developed countries (Sotudeh and Ghasempour, 2018). Therefore, it is not clear if the open access citation advantage observed for the article-processing-charge-funded papers published in these journals (Sotudeh et al., 2015) is proportionately distributed among the contributing countries, or deprives less developed countries due to their relatively low-quality outputs, whether actual, perceived or biased.

Besides, affordability of this model is not the same for researchers in the countries with different income levels. Research evidence has shown that low-income researchers prefer less costly journals with lower impact factors (Bonaccorso et al., 2014; Ellers et al., 2017; West et al., 2014). Consequently, they pay the price of high-tiered journals where authors from the developed countries publish their papers. This is because open access is not a viable model for highly prestigious journals with expensive publication charges and high rates of rejection; their publication costs are inevitably provided by cross-subsidizing by low-tiered mega-journals (Ellers et al., 2017). It is, therefore, important to know whether the publication of article-processing-charge-funded papers would equally bring about cost-effectiveness for all countries, in terms of the citations earned for the money paid.

The value of citation (in dollars) which was estimated by Diamond (1985) for the first time in the traditional publication model was used by Harnad (2005) to examine the economic importance of the open access model. Aiming at identifying high-tiered journals with lower expenses, the study of West et al., (2014) was the only instance of using the value of citation in dollars to calculate the cost-effectiveness of article-processing-charge-funded journals through dividing the articles citations by their publication charges. Given the lack of studies on the cost-effectiveness of such papers for different wealth classes of authors, the present study attempted to examine it at the country aggregation level. To do so, it endeavoured to investigate the citation performance of the countries at different scientific development levels in their open access and non-open access articles published in a collection of hybrid journals. Following West et al. (2014), the cost-effectiveness of the articles was estimated by calculating the proportion of the citations to article processing charges.

Literature review

Contrary to what was expected, electronic publication did not result in a reduction of journals' subscription fees. Commercial publishers' offers, like big deals, led to the imposition of huge costs on libraries and, thereby, to the serial crisis (Albert, 2006; Björk, 2004; Guédon, 2004; Hedlund et al., 2004; King, 2007; Knezo et al., 2005; Young, 2009). Journal subscription offers a profit of $3 billion to the publishers, with Elsevier alone earning $0.76 billion (Tsikliras and Stergiou, 2013).

The open access movement emerged on the part of the science and library communities to address the serial crisis. In light of this movement, the enthusiasts of science all over the world can freely and without any payment read, download and copy the documents which they need. Although open access resources are accessible to the readers free of charge, the publication of the articles involves charges. According to Tsikliras and Stergiou (2013), the global publication cost of open access articles is estimated to be around $1.5 billion a year. The movement is believed to reduce the imposition of the publishers' exorbitant costs on the knowledge customers (Tsikliras and Stergiou, 2013) and thereby to save vast amounts of their budgets (Walters, 2007).

However, the establishment of the open access models did not eradicate the libraries' financial challenges. For instance, the green model, which allows the authors to self-archive the pre- or post-print version of their articles free of charge through their websites or institutional repositories, requires the libraries to provide financial and human resources and electronic tools for storing, maintaining and preserving self-archived articles (Bailey, 2006; Björk, 2004; Collins, 2005). Also, the delayed open access model, as a subset of the green model, does not unburden the libraries from journal subscription because researchers may be in urgent need of access to the articles during the embargo period. Besides, publishers accept the delayed model provided that they receive subscription fees during the embargo. On the other hand, the commercial publishers did not widely welcome this model because of the risk of falling revenues (Getz, 2005; Laakso and Björk, 2013; Morris, 2004; Wellen, 2004).

The costs of the gold model are provided either directly (through article processing charges paid by authors or their universities and sponsors) or indirectly (through advertisement, subsidies, charities, and donations of governmental or private sponsors) (Bailey, 2006; Joseph and Jha, 2013; Knezo et al., 2005; Mabe, 2004). While the authors and their sponsors play a key role in sustaining the former, no fees are paid by authors or their readers for the latter (Anderson, 2004; Collins, 2005; Kingsley, 2014; Prosser, 2003; Tenopir et al., 2017). The article processing charges range from $500 to $2,500 (Schroter and Tite, 2006; Regazzi, 2004).

The article-processing-charge-based model has not been widely welcomed by authors, libraries, universities and their sponsoring institutions due to its high cost as well as the emergence of some predatory publishers who abuse the model and undermine the quality and the authenticity of the scientific outputs, especially those from developing countries. Moreover, some researchers believe that the final costs of the model will be higher than those of the subscription-based model in the long term. Some even warn against the risk of compromising the quality of articles for the sake of profits (Chan and Costa, 2005; Hagenhoff et al., 2008; Jeon and Rochet, 2010; Walters, 2007). The imposition of exorbitant costs on universities and libraries is among the other factors inhibiting the gold model (Bird, 2010; Chesler, 2004; Guterman, 2004; Young, 2009).

To overcome some of the limitations of the green and the gold models, the hybrid model was introduced in which journals could publish open access as well as non-open access articles, depending on their authors' decisions to support open access by paying the article processing charges. Springer and Elsevier are among the first major publishers who converted to the model. The article processing charges vary among publishers and journals. For instance, the charges imposed by Elsevier range from $800 to $5,000 and average approximately $3,000 (Bernius et al., 2009; Chesler, 2004; Moskovkin, 2008). The average value also holds for other top publishers including Springer, Oxford University Press, Nature, and Wiley-Blackwell (Chakravarty and Sharma, 2012; Hansen, 2012; Mccabe et al., 2013).

The advantage of this model lies in its voluntary nature compared to the gold one. Authors and researchers do not have to pay for the publication of their manuscripts. Besides, this choice allows them to identify the most appropriate hybrid journals; this is believed to make publishers reduce the publication charges to compete for gaining high-quality research papers (West et al., 2014). On the other hand, publishers would not face financial problems for their support of open access because if the number of article-processing-charge-funded papers is low, the publisher would compensate the costs through subscription fees (Prosser, 2003).

Various studies have shown that article-processing-charge-funded papers in hybrid journals have a citation advantage over not only the self-archived and non-open access articles published in the same journals (Davis, 2009; Eysenbach, 2006; Gaule and Maystre, 2011; Harnad and Brody, 2004; Laakso and Björk, 2016; Mueller-Langer and Watt, 2018; Sabharwal et al., 2014; Sotudeh et al., 2015; Wang et al., 2015), but also those published in gold journals (Chaudhuri and Thohira, 2010). This fact may reinforce the incentive to support open access to promote the position of the authors and their universities in the scientific community. They not only enjoy higher visibility, but also are more likely to publish their high-quality articles through the hybrid model (Björk and Solomon, 2015; McCabe et al., 2013; Swan, 2010). Thus, the role of these journals would become more intense in advancing knowledge, by increasing their uptake of article-processing-charge-funded papers.

However, the hybrid model has led to some concerns for authors. According to Björk and Solomon (2014), the average charge in hybrid journals is $2,727, while articles published in the subscription-based and the gold models cost much less ($2,097, and $1,418, respectively). This may discourage authors and their investors, especially in developing countries (Chakravarty and Sharma, 2012; Gillies, 2014; Ho, 2011; Jeon and Rochet, 2010; Solomon and Björk, 2012b; Weber, 2009). Some countries, even in the developed block (e.g., Poland and Denmark), face problems in paying for the publication charges (Björk et al., 2014; Burchardt, 2014). The second concern is the double dipping phenomenon referring to the publishers' profits from both the subscription and the publication charges (Mittermaier, 2015; Mittermaier and Jülich, 2015). Therefore, the model, directly or indirectly, imposes financial burdens on library and research budgets.

Therefore, the hybrid model had not been widely welcomed when introduced. According to Dallmeier-Tiessen et al., (2010), from the total 2000 journals published by twelve major highly-renowned publishers, just 2% had adopted the model. Hence, to reduce the concerns, some measures have been taken, including delaying the payment until the paper's acceptance, and considering exemptions or discounts (Weber, 2009), especially for the authors in low-income countries (Lawson, 2015; Waugh, 2012; Young, 2009). Moreover, to eliminate the double dipping problem, the competitive market of the hybrid model has proposed some solutions including refunding the charges at list prices, value-based funding of the charges, and fixing the share of the funders (Björk and Solomon, 2014; Ware and Mabe, 2015). Some publishers, like Elsevier, have reduced the publication charges to compensate for the costs imposed on the libraries and the research institutes due to the double dipping phenomenon (Pinfield et al., 2016).

Academia has been found to pay or fund the article processing charges increasingly or constantly (Pinfield and Middleton, 2016; Pinfield et al., 2016). Although the hybrid journals are still more expensive than the gold ones (e.g., more than $2750 for big publishers including Elsevier, Springer, Wiley, and Taylor and Francis) (Vrana, 2016), there is a positive relationship between the charges and the quality of the journals regarding their citation ranks which may justify the higher amount of money requested by the prestigious journals (Pinfield et al., 2016, 2017). Furthermore, according to Sotudeh and Ghasempour (2018) publishing article-processing-charge-funded papers has been progressively growing in all countries, though the largest share has belonged to the developed countries.

Having investigated seventy-eight publishers in Sherpa-Romeo, Björk (2012) found that the number of hybrid journals increased from 2000 in 2009 to 4300 in 2012. Furthermore, the number of open access papers published in these journals has increased from 8,000 to 12,000. Springer and Elsevier published most of the hybrid journals during this period. Kozak and Hartley's (2013) investigation of 9,000 open access journals in Directory of Open Access Journals showed that 28% of the journals followed the gold model; medicine-related journals gained the highest rank (47%), and sociology-related journals gained the lowest rank (4%). By reviewing the hybrid journals of the five big publishers, Laakso and Björk (2016) found out that the number of charge-funded papers in these journals increased from 666 in 2007 to 13994 in 2013. Furthermore, similar to Pinfield et al. (2017) and Pinfield and Middleton (2016), they showed that such papers were mostly published in medicine-related fields. Vrana (2016) indicated that 17.79% of the articles in Directory of Open Access Journals adhered to the model, suggesting a probable decline in the tendency towards the model, when compared to the twenty-six percent previously reported by Morrison (2015).

Based on the research evidence, it seems that this model has been stabilizing, though steadily and slowly. However, it is necessary to guarantee its balanced development through the world and the proportionately distributed rewards to ensure its survival.

Objectives

The main purpose of this study was to determine the relative cost-effectiveness of article-processing-charge-funded papers among different countries and scientific blocks. The following questions were, therefore, raised:

1) Do the article-processing-charge-funded papers have citation advantages for all contributing countries and scientific blocks?

2) Are such papers relatively cost-effective?

3) Is there any significant difference between the relative cost-effectiveness of such papers across the countries in different scientific blocks?

Method

Applying a comparative citation analysis method, this research investigated a sample of papers published in forty-seven Elsevier hybrid journals during 2007-2015. The sample was selected since the collection indicated an open access citation advantage (Sotudeh et al., 2015) that has been sustained in recent years (Sotudeh and Estakhr, 2018). Besides, Elsevier publishes the highest number of hybrid journals (Morrison, 2017). Furthermore, during 2007-2015, the model got established and used by many publishers (Sotudeh and Ghasempour, 2018; Sotudeh et al., 2015). An acceptable citation time window was also observed by concentrating on the publication time span.

Data collection procedures

To collect data, the researchers first rechecked the forty-seven hybrid journals, identified by Sotudeh et al. (2015) and Sotudeh and Ghasempour (2018) through the Science Direct database to ensure their adherence to the hybrid model. The bibliographic and bibliometric data including publication years, sources, citations, DOI, affiliations, correspondence addresses, document types, and subjects were collected from Scopus. During 2007-2015, the journals published 160,168 articles, including 5,135 papers funded by article processing charges. The countries of the corresponding, first and last authors were retrieved through the affiliations and the correspondence addresses fields. The three positions were analysed because there were no definite rules to detect the authors responsible for paying the charges (Davis, 2009) and these positions were the most probable ones for the main authors (Kennedy, 2003; Wren et al., 2007) who were probably responsible for the payment. The rationale underlying the choice of Scopus and Science Direct was a wider journal coverage by the former compared to other prestigious citation databases like WoS, and a clear tagging system for identifying charge-funded papers applied by the latter.

The identified countries were then categorised into four scientific blocks following Wagner et al. (2001). Based on nations' science and technology investment and capacity, they provided four categories of countries, including twenty-two 'scientifically advanced countries' (with a science and technology capacity well above the international mean), twenty-four 'scientifically proficient countries' (with a positive standing in scientific capacity when compared to the rest of the world), twenty-three 'scientifically developing countries' (with some features of scientific capacity, and a positive trend in spending for science and technology but a scientific capacity below the international mean), and finally eighty-one 'scientifically lagging nations' (with little data indicating scientific capacity).

To detect the publication charges between 2007 and 2015, at the outset, the researchers referred to the Elsevier database. However, this database only showed the charges of the current year. Therefore, using the Internet Archive, the researchers tried to retrieve and analyse the contents of Elsevier's archived web pages in the past years. The pages announcing the charges had been tagged as 'sponsored articles'

Annual changes in the article processing charges in Elsevier journals

According to the results of the content analyses of the archived pages, the charges requested by the studied journals were $3,000 and were stable from 2007 to 2012. This is in line with previous literature estimating Elsevier's publication charges to be $3,000 till 2012 (Björk, 2011; 2012). However, our analyses showed that it had been fluctuating from $500 to $5,000 in 2013 and 2014. Figure 1 shows the average publication charge in the studied journals in 2010 to 2015. As seen, although the average values had been constant in 2010 to 2012, the trend curve was slightly declining on a significant exponential basis (R2=0.7677, F=13.78, Sig.=0.021) in the following years. This signifies a reduction, though negligible, in Elsevier's costs of publishing open access articles in the hybrid journals as a response to prevent the double dipping phenomenon and to support the Research Council UK's open access policy (Mittermaier, 2015), or a result of cross-subsidizing the costs of its highly-prestigious, and thus expensive, journals by its lower-tiered journals (Ellers et al., 2017).

Data analysis

To calculate the open access citation advantage, the percentage of open access articles' citation surplus was estimated proportional to the citation average of non-open access articles for each contributing country using the following formula:

(citation average of open access articles - citation average of non-open access articles)/citation average of non-open access articles * 100

Inspired by Bonitz et al. (1997; 1999), the countries were then categorised into three categories, including the right world (or open access citation winners), the middle world (equal open access and non-open access citation averages) and the left world (or open access citation losers). The groups' open access citation advantages were higher than, equal to or lower than zero, respectively.

To calculate the cost-effectiveness of the charge-funded papers, the cost-effectiveness of all (open access and non-open access) articles published by a nation was used as a benchmark. It is obvious that the authors or their sponsors paid only for the open access articles. However, they could have paid the charges for their non-open access articles if they had chosen to. Therefore, the non-open access papers had the same, though not paid, publication charges as their charged peers published in the same journals. In other words, the publication charge was de facto for the open access articles, while it was potential for the non-open access articles. Therefore, the citation cost-effectiveness of a group of papers could be calculated based on the papers' citations ratio to their publication costs (whether actual or potential) as follows:

citation cost effectiveness=citations/article processing charges

The relative citation cost-effectiveness for a given country in its open access contributions could be, then, calculated based on the ratio of the citation cost-effectiveness yielded for its charge-funded papers to that of the total papers published in the journal set. If the result of the fraction approximated one, the citation cost-effectiveness of the charge-funded papers was estimated to be as much as expected (i.e., at the same level as the overall citation performance). If it was more than one, the cost-effectiveness was higher, and if it was less than one, it was lower than the expected level.

Findings

As mentioned above, the contributing countries were extracted from author by-lines at three positions of the corresponding, the first and the last authors. Given the similarities of the results for these different authorship roles, here are just the findings related to the countries to which the corresponding authors were affiliated.

Comparing the citation performance of open access and non-open access articles among the contributing countries

To compare the citation performances of open access and non-open access models across the countries, first of all, the citation means in different countries were compared considering the publication years. Due to the non-normality of the data, the U-Mann-Whitney test was used. As shown in Table 1, except for 2008, the Z value is larger than the critical value in all years (Z=1.96). Thus, in the years 2007 and 2009-2015, the null hypothesis is rejected; there exist significant differences between the citation means of the models (α=0.05).

Open access citation advantage in the countries of the world

The open access citation advantage was calculated through the proportion of the citation mean of the charge-funded papers to that of the non-open access articles for each contributing country in different scientific blocks. It is illustrated in Appendices 1 to 4. In the figures, the positive values indicate the countries' open-access citation advantage, signifying that they received more citations for their open access articles compared to their non-open-access papers. The negative values show their citation disadvantage for their open access contributions relative to their non-open access ones.

The verification of the results showed that the citation advantage of some countries was seemingly an artefact of their low number of open access articles or affected by their fluctuations over the years. These included the Russian Federation (in the scientifically developed block), Cuba (in the scientifically proficient block), Egypt and Venezuela (in the scientifically developing block), and Lebanon, Laos and Nigeria (in the scientifically lagging block) that are in the extreme positions in their group in terms of open access citation advantage. For example, none of the forty-two articles published by Venezuela in 2007-2011 and 2013 was charge-funded. The papers received 739 citations. The country gained twenty citations through publishing four charge-funded papers in 2012, and 2014-15, while it received sixty-five citations for eighteen non-open access articles in these years. This yielded open access citation advantages of -0.25, 0.83 and 11 for 2012, 2014 and 2015 respectively, averaging to 3.86 of open access citation advantage. As a result, its open access citation advantage increased dramatically.

The status of the countries in various scientific blocks in terms of their average open access citation advantage is summarised in Table 2.

As shown in the table, from a total of the seventy countries publishing article-processing-charge-funded papers, 81.43% were open access citation advantageous, and the rest 18.57% lost their potential open access citations. None of the countries was categorised in the middle world (i.e., equally performing in the citations received by their open access and non-open access papers). In each of the scientific development blocks, the majority of the contributing countries were open access citation winners.

The comparison of the countries' open access citation advantage in various scientific blocks

Due to the non-normality of the data, Kruskal-Wallis test was used to compare the open access citation advantage of the countries in different scientific blocks.

As seen in Table 3, there is no significant difference across the scientific development blocks regarding their mean ranks of open access citation advantage (α=0.56 >0.05). As the citation advantage is the surplus of the open access citations of a given country to its own non-open access citations, one could conclude that the scientific blocks were equally able to improve their citation performance via open access in comparison with their non-open access publications.

The comparison of the citation cost-effectiveness of open access and non-open access articles

Due to the non-normality of the data, U-Mann-Whitney U-test was used to compare the citation cost-effectiveness of the open access and non-open access models. As shown in Table 4, the Z values are larger than the critical values in all years, except for 2008, (Z=1.96). Thus, concerning the years 2007 and 2009 to 2015, the null hypothesis is rejected. In other words, there is a significant difference between the models' citation cost-effectiveness in favour of the former (α < 0.05).

The relative citation cost-effectiveness of open access articles

As seen in Figure 2, the relative citation cost-effectiveness values of the open access model are more than one in all years, implying that over the years, the citation cost-effectiveness of the article-processing-charge-funded papers was higher than that of the total number of the articles published in the hybrid journals. It is at its lowest level in 2009 (1.22), and at its highest level in 2007 (1.52). Given the difference in the citation time windows of the articles published in different years, the citation cost-effectiveness of the articles published in a given year would be expected to be higher compared to those published in the preceding years. However, after a decrease from 2007 to 2012, it increased again since 2012. This may be due to the decrease in the charges since 2012 (Figure 1).

The open access relative citation cost-effectiveness across the contributing countries

As the publication charge for Elsevier was constant ($3,000) in 2007 to 2012 (Figure 1), the analysis of the contributing countries' relative citation cost-effectiveness in these years would reflect the same image as their citation advantages. Thus, this section deals only with the analysis of 2013-2015. The results plotted in Appendices 5 to 7 represent the mean values of the relative citation cost-effectiveness for the contributing countries.

It should be mentioned that the number of the open-access contributing countries reduced to 52 for the 2013 to 2015 publication time span. France (7.91), the Czech Republic (4.54), and Venezuela (4.19) gained the highest ranks, and Hong Kong (0.2), Slovenia (0.31), and Croatia (0.27) gained the lowest ranks in 2013, 2014, and 2015, respectively. The high value of the relative citation cost-effectiveness for some of these countries (e.g., the Czech Republic, Egypt, Venezuela, Thailand, and Lebanon) is due to publishing a low number of highly cited article-processing-charge-funded papers.

Table 5 illustrates the frequency of the open access contributing countries categorised into the winner and loser worlds based on their relative citation cost-effectiveness values.

As observed, of the fifty-two open-access contributing countries during 2013 to 2015, the majority (81.25%) belonged to the right world (winners), while the minority (18.75%) belonged to the left world (losers) in terms of their mean relative citation cost-effectiveness. No countries were situated in the middle world. In the scientifically developed block, 95.23% of the participating countries were ranked as winners, while just 4.77% were ranked as the losers. In the proficient scientific block, the image is slightly different, so that a smaller percentage of the total seventeen contributing countries (64.70%) belonged to the right world of winners, while a relatively higher percent (35.30%) were the losers. Moreover, all nine of the twenty-four scientifically developing countries that contributed to the open access model were the winners. Finally, from fourteen scientifically-lagging countries that contributed to the charge-funded papers, 71.43% were the winners and 28.57% were the losers in this regard.

The comparison of the open access relative citation cost-effectiveness across the scientific development blocks

Kruskal-Wallis test was used to compare the scientific development blocks in terms of their relative citation cost-effectiveness. The results illustrated in Table 6 reveal that there are no significant differences between the countries of the world in different scientific blocks (Sig.>0.05).

Discussion and conclusion

In its endeavours to democratise science, the open access movement had to find solutions to finance open access resources. Among a variety of open access models, the hybrid model tries to solve the financial problems of the previously proposed models by covering the open access publication costs from those authors who would like to set free their scholarly outputs. No studies have explored the model's capacity to change the citation performance of the countries at different scientific development levels. This is in spite of the fact that science system has been revealed to suffer from inequalities in recognition and rewards, especially against those in less affluent and hence less scientifically-developed nations. The objective of the present study was to examine the cost-effectiveness of the article-processing-charge model in terms of the citations earned by the contributing countries across different scientific blocks.

The results reconfirmed the previous findings suggesting the citation advantage of the charge-funded papers.(Davis, 2009; Eysenbach, 2006; Gaule and Maystre, 2011; Harnad and Brody, 2004; Laakso and Björk, 2016; Mueller-Langer and Watt, 2018; Sabharwal et al., 2014; Sotudeh et al., 2015; Wang et al., 2015). Moreover, the majority of the countries contributing to the open access model enjoyed the article-processing-charge citation advantage. In addition, the open access citation winners were at different scientific development levels all over the world. This is in line with previous studies confirming the role of the open access gold model in improving the citation performance of different countries, including the less-developed ones (Sotudeh and Horri, 2009).

The present study also tried to estimate the citation value of the open access publications by comparing the open access and non-open access articles in this regard. The charge-funded papers showed to be citation cost-effective relative to the non-open access articles published by the same countries in the journal set. The majority of the countries enjoyed the relative citation cost-effectiveness. All scientific blocks reached a relative cost-effectiveness, and there was no significant difference between them in this regard. In other words, the citation advantage of the open access articles is sustained at different levels of scientific development of the contributing countries, even after controlling for the effect of the publication charges. Evidently, the publication charges of the hybrid journals varied widely across journals, ranging from $500 to $5,000 (Bernius et al., 2009; Chesler, 2004; Moskovkin, 2008). As a result, the costs of the citations were not equal. Depending on the journals' charges, the cost of each citation, whether paid by authors, sponsors or the publisher, would vary. On the other hand, depending on their wealth and scientific development levels, authors chose different hybrid journals with different charges, so that less affluent authors preferred low-impact journals with lower publication charges (Bonaccorso et al., 2014; Ellers et al., 2017; West et al., 2014).

Based on the findings of the present study, it seems that the article-processing-charge model could improve the relative citation performance of different countries of the world, including the less-developed ones. However, the improvement was relative to the citation performance of their own non-open access articles. Consequently, the findings should not be interpreted as a conclusion that publishing charge-funded papers would bridge the overall citation gap between the scientifically rich and poor countries. Further studies are required to shed more light on the model's effect in this regard. Moreover, the results were limited to the hybrid journals published by a single prestigious publisher and should be interpreted and generalized with caution. Because, the citation performance of journals depends on different factors, including the prestige, reputation and quality of their publishers. Consequently, the study should be replicated on different publishers to further validate and generalize the results.

Another important point is that the study limits itself to the hybrid model, since the model's sustainability depends highly on its affordability for authors and their willingness to pay. Obviously, this is not the case for those open access journals imposing no charges on authors and libraries (e.g. the platinum ones (Haschak, 2007), which may have their own financial challenges endangering their sustainability (Foxall and Nailor, 2016)). Furthermore, the results should not be considered as a credit to the hybrid model as an appropriate and advantageous open access model, given the existence of such serious, not-yet-resolved challenges as double dipping, high charges, and low visibility (Mittermaier, 2015).

To summarise, the findings of the present study showed that the charge-funded papers had a higher citation advantage than non-open access ones for different countries. Moreover, they had a higher relative cost-effectiveness encompassing the majority of countries in all scientific blocks.

About the authors

Zohreh Estakhr received her MA in Information Management from Shiraz University. She can be contacted at zohrehestakhr@gmail.com.
Hajar Sotudeh is a professor of Knowledge & Information Sciences at Shiraz University. Her primary research interests include open access, scientometrics, altmetrics, and natural-language-processing and citation-based information retrieval. She can be contacted at sotudeh@shirazu.ac.ir.
Javad Abbaspour is an assistant professor of Knowledge & Information Sciences at Shiraz University. His main research interests include recommender systems, information system design and evaluation. He can be contacted at javad.abbaspour@gmail.com.

How to cite this paper

Appendices