Scholars like any other? An inquiry into productivity and citations in the field of contemporary China studies

Abstract

Social scientists have extensively studied the relationships between co-authorship, scholarly productivity, and reputation, but few studies have examined these phenomena in the field of contemporary China studies. This article aims to fill this gap by investigating the impact of co-author networks on productivity and citations. It draws its conclusions by analyzing 2784 articles published between 2011 and 2021 in nine major international journals dedicated to contemporary China. The findings reveal unique trends among specialists on contemporary China, characterized by low average productivity and a broad publication distribution across disciplinary journals. The study also demonstrates a positive correlation between co-authorship, network connections, productivity, and citations, emphasizing the role of mentoring strategies and extra-mural collaborations in enhancing productivity. Furthermore, the formation of “invisible colleges” within the main networks is observed, influencing the literature and citation landscape. However, these invisible colleges are not monopolizing publishing opportunities if compared with other fields of knowledge.

Keywords

Research collaboration productivity networks China reputation social sciences

Introduction

The idea that collaboration between scholars enhances the number of publications of a given scholar might sound neither new nor counterintuitive. Indeed, in the 1960s, years before the publish or perish system already at work in the United States in the 1930s became a more global phenomenon, several scholars already demonstrated that “the most prolific man is also by far the most collaborating” (De Solla Price and Beaver, 1966: 1014). Since then, countless papers have studied the relationship between research collaboration and the number of articles published in different domains of the so-called hard sciences (De Solla Price and Beaver, 1966; Kumar, 2015; Lee and Bozeman, 2005; Pravdic and Oluic-Vukovic, 1991; Zuckerman, 1967) and, to a lesser extent, in the fields of humanities and social sciences (Akbaritabar et al., 2018; Wanner et al., 1981; Wilder and Walters, 2019).

Regarding the impact of co-authoring on citation counts, the literature is also very substantial (Abbasi et al., 2011; Katz and Hicks, 1997; Li et al., 2013; Wallace et al., 2012), and it generally supports the idea that co-authoring, especially when it involves international collaboration, is beneficial. Using different metrics (degree centrality, betweenness centrality, and eigen centrality), specialists in social network analysis have widely documented how the position in a co-authorship network is a strong predictor of future citations (Biscaro and Giupponi, 2014; Guan et al., 2017; Ji et al., 2022; Kretschmer, 2004; Newman, 2001a). This phenomenon was found to be particularly powerful and influential in science, technology, engineering, and mathematics (STEM) fields, as early publications with an already prominent and well-connected scholar provided newcomers with a higher chance of belonging to the top 5% of most productive scholars by their mid-careers (Li et al., 2013).

However, the current literature also indicates that the influence of research collaboration and networks on scholars’ productivity is quite different from one field of science to another (Yair et al., 2022), and from one socio-historical context (Ma et al., 2014) or period (Moody, 2004) to another.

Given the impossibility of drawing a universal law of scientific productivity (Kretschmer and Rousseau, 2001; Murphy, 1973), each field and subfield of knowledge is worthy of investigation. Moreover, the growing amount of such research makes it possible to compare different fields of knowledge to each other over time (Henriksen, 2016). In other words, it is possible to evaluate whether a field of study is led by certain scholars or cliques (i.e. a small-world structure) or offers more room for less connected scholars to emerge and contribute.

Following this perspective, the present article inquiries into the influence of co-author network ties on the productivity of contemporary China specialists publishing in English in the major journals of this field. More specifically, this research mainly draws on metadata harvested from the platform Scopus, focusing on 2784 articles published between January 2011 and December 2021 in nine journals (China Information, China Journal, China Perspectives, China Quarterly, China Review, China: An International Journal, Chinese Journal of Sociology, Chinese Sociological Review, and Journal of Contemporary China). We selected these journals after an exploration of the main bibliographic references and co-citation networks, which informed us about their centrality to the subject of knowledge and their multi-disciplinary character (which was not the case for the Chinese Economic Review, for instance). Moreover, most of these journals are WOS–SSCI (Web of Science–Social Sciences Citation Index) indexed. Therefore, they receive many contributions from authors based in the United States, the United Kingdom, and China, where scholars need to publish in such journals to obtain tenure and promotion.

As further explained in the data and method section, these data enabled us to: (1) investigate the number of articles to which each of the 2918 authors contributed; (2) build co-author networks: and (3) investigate the influence of network ties on productivity (the number of published articles in the field) and reputation (citations to previously published works). To investigate progressively these different aspects, we initially retrieved and analyzed some key information about the networks of the co-authors (see the data and methodology section).

Next, the first part of the results section focuses on the number of articles published by scholars over 10 years in the selected journals. This section demonstrates that co-authoring leads to higher productivity (at a rate that is quite standard in social sciences), while solely single-authoring scholars can still be found among the most prolific (top 10% and top 5%) authors, which is quite a rare phenomenon. In addition, we point out that the main network key players (i.e. those who are in a position to control the information or to diffuse it from one part of the network to another) are over-represented among the most prolific scholars.

The second section focuses on the 144 most prolific authors (approximately the top 5%) and investigates how network ties increase the number of citations to previously published work. Our negative binomial logistic regression models show that scholars who do not belong to the largest co-author networks have much less chance of seeing their previous work cited. Moreover, inclusion in the main co-author networks appears as one of the most influential factors predicting the citation of previous work, together with seniority.

These findings ultimately lead us to the conclusion that co-authorship ties matter, while the field of contemporary China studies remains comparatively open. While we do not deny that well-connected scholars, who share expertise and often lead large teams of younger researchers, benefit from their invisible colleges (Crane, 1972), we argue that they are not responsible for a disproportionate amount of articles and do not excessively receive citations from their co-authors, especially when compared with other fields of knowledge.

Literature review

Since Lotka's (1926) seminal research on scientists’ productivity, an astonishing amount of research has investigated with whom and how scholars are collaborating, why scholars are seeking collaborators, and how co-authors and co-author networks benefit in terms of their productivity and reputation.

Regarding the first aspect, Katz and Martin (1997) classified research collaboration based on levels (individual, group, department, institution, sector, and nation) and whether the collaboration is set up within or outside a given level. Martín-Sempere et al. (2002) adopted a similar approach, distinguishing between intra-mural and extra-mural collaboration, as geographic proximity is considered an important factor fostering the development of research collaboration. According to Beaver (2001), the most common form of collaboration is between a principal investigator and the students they supervise. While Etzkowitz's (1984) research echoed Beaver's statement and confirmed that scientists are becoming entrepreneurs who manage short-term resources and lead large teams of junior scholars and students, later studies have indicated that mentor–student relations may not always be the major reason for the collaboration.

According to Melin (2000), the supervisor–student relationship is far from being the major reason for collaboration. Instead, co-authoring researchers primarily seek co-authors with special competences (41%), or special data and equipment (20%), rather than seeking to establish supervisor–student relations (14%). While we may assume that the respondents in Melin's study may have partly responded in this way due to reasons of political correctness or conformism,¹ it remains the case that collaboration and co-authoring can arise for various practical, technical, and instrumental reasons (Bozeman et al., 2013; Bozeman and Corley, 2004; De Solla Price and Beaver, 1966). Moreover, these reasons vary from one field of science to another, as access to critical instruments and technical pieces of equipment certainly play a more important part in STEM fields.

Whatever the reasons for why scholars collaborate and co-author, countless studies have demonstrated the positive influence of collaborative habits on scholars’ productivity and reputation. Concerning the influence of the number of collaborators on the number of papers written over a certain period, the correlation appeared to be particularly strong in Newman's (2001b) study. In this study, which covered biomedical research, astrophysics, high-energy physics, theoretical physics, and computer science, the correlation between the number of co-authors and the number of published papers followed “a power law” with “a clear exponential cut-off” (Newman, 2001b: 407). Subsequently, well-known studies, such as that of Lee and Bozeman (2005), have verified this correlation while refining it. More precisely, it has been shown that collaborations correlate with productivity when a given scientist's total number of publications measures publishing productivity. However, “collaboration and publishing productivity are not significantly related” (Lee and Bozeman, 2005: 693) when productivity is measured by fractional count and once other control variables (e.g. age, citizenship, job satisfaction, discrimination, collaboration strategies, and field) are included in the models. Interestingly, in economics, the correlation between co-authoring and productivity was negative once contribution was fractionalized (Ductor, 2015; Hollis, 2001).

The comparison between these different studies highlights the existence of important nuances. Indeed, in each field of science, different cultures of teamwork develop and evolve depending on how science is administrated, not to say empowered. As Bourdieu explained, each scientific discipline, and sometimes each subfield in a given discipline, exists as a relatively autonomous world that is regulated by slightly different norms, practices, and expectations (Bourdieu, 1988). For instance, in the United States, historians usually write more books during their careers compared to economists and physicians (Yair et al., 2022). At the worldwide level, Parish et al.'s (2018) study demonstrated that social scientists are far more likely to be involved in large-scale collaborative projects. Unsurprisingly, social scientists are also relatively less productive in terms of peer reviewed articles compared to scholars in the hard sciences. For instance, in Ruiz-Castillo and Costas's (2014) study using Thomson Reuters data and covering the period 2003–2011, scholars in sociology and anthropology published an average of 2.2 articles over the period, and 75.3% of the authors published only one article. In economics and business, the average productivity was quite similar (2.3), but only 65.8% of the scholars published only one article. This article additionally showed that the proportion of scholars with “uninterrupted, continuous presence” (i.e. publishing every year) differed tremendously in the hard sciences, social sciences, and humanities.

While social scientists have generally fewer collaborations and are therefore less productive, several studies have demonstrated an increasing proportion of collaborative papers. This trend is particularly obvious for Dutch sociologists (De Haan, 1997). A quite similar but more pronounced trend exists in American sociology (Stoltz, 2023). Conversely, in French sociology, the percentage of co-authored papers is dramatically lower. In Cahiers Internationaux de Sociologie, only 12.5% of the articles between 1960 and 1995 were co-authored. This proportion reached 66.7% for the American Sociological Review during the same period (Pontille, 2003).

Given the diversities of national and disciplinary traditions, networks of co-authors play a more or less important role in knowledge production and diffusion. For instance, in Italian sociology, the network of co-authors is not highly cohesive (Akbaritabar et al., 2020). This means that most authors work independently or in small communities isolated from others. Given these structural differences, some fields are more or less prone to experience the small world of citation analyzed by Wallace et al. (2012).

Since each field and subfield are particular, in this study we measure how these phenomena are at work in the field of contemporary China studies. While several studies have already examined how collaborations are developing among Chinese scientists and social scientists (Hong and Zhao, 2016; Jonkers and Tijssen, 2008; Liu and Xia, 2015), no one has analyzed contemporary China studies as a field of knowledge composed of scholars from different countries. This paper addresses this gap in the literature. Ultimately, it invites colleagues to nuance their perceptions of the field, which is often seen as being largely dominated by a few scholars.

Data and methods

Unlike many scholars researching co-authorship networks or co-citation networks (Feng and Kirkley, 2020), we considered using Web of Science (WOS) data and Scopus data. Both databases offer many references, but they unequally cover different geographic areas, fields of science, and periods. For instance, WOS includes the Chinese Science Citation Database, which covers engineering journals published in the People's Republic of China. WOS also contains the largest amount of research published before 1900. However, for this research, which focuses on literature published by contemporary China specialists between 2011 and 2021, Scopus offered more data on key journals in the field.

We generated our raw data by searching each journal's International Standard Serial Number, filtering for the period 2011–2021 and keeping research articles only. We repeated this procedure for each journal and collected information on 2784 articles. The information for each article included a list of authors, a list of each author's Scopus identification number, the title of the article, the abstract, the affiliations of the authors, the keywords, the journal, the journal's issue, and volume numbers, and, more importantly for this research, a list of citations made by the authors of each article.

Based on this raw data, we followed the procedure elaborated by Goutsmedt and Truc to create different datasets (Goutsmedt and Truc, 2022). This procedure enables R users to prepare bibliographic data for analysis by using a limited number of packages (dplyr, tidyverse, janitor, and in this study bib2df and networkflow) with advanced formulations or regular expressions for string-search coding. Simplifying the description of this procedure, it excavated from the data a first dataset containing a list associating each article (identified by a newly created variable named “citing_id”) with one of its authors, the Scopus identifier of this author, their country, and their affiliation. This resulted in a first dataset with 4578 observations, as the majority of the authors (82.3%) had at least one co-authored paper. Regarding the second dataset, each reference in the bibliography corresponded to one row in the dataset. Each row contained information about the title of the book, article, or book chapter that was cited, the name of the cited authors, and, most importantly, the origin of the citation (variable citing_id).

As the workflow explains, the first dataset authors–articles gave birth to two new datasets. The first was named author–productivity and indicated the number of articles published by each author, the number of collaborative articles, the number of self-authored articles, and whether or not they belonged to the most prolific (top 10%, top 5%, and top 2.5%) authors. The second newly created dataset solely contained information about co-authored papers and their authors (see Figure 1).

While creating these two datasets, we had to be extremely careful about disambiguation issues. As Moody (2004) noted, authors’ identification is made difficult because authors’ names are inconsistent over time. Fortunately for us, Chinese women rarely change their patronym after marriage, and individuals could be matched thanks to their Scopus author identifier. Nonetheless, since we doubted the reliability of the Scopus author identifier, we also empirically disambiguated the 2918 authors. Like Moody, we found that American scholars’ middle names were not always recorded. For Chinese authors, this empirical disambiguation was also extremely instructive. When disambiguating the family names and first names (in Pinyin), we found 19 perfect homonyms and countless authors with the same family name and same first letter of their first name (e.g. Chen Zhang and Chunyu Zhang, corresponding to the same shortened identifier ZHANG, C.). In some rare cases, the family name and first name were inverted from one publication to another.

After this long preparatory work, we were able to construct the first basic representation of the co-author networks by using the packages biblionetwork, tidyverse, ggraph, and tidygraph in R.

This first basic graphical representation already proved insightful for understanding the shape and structure of co-authors’ networks and communities (see Figure 2). As can be easily seen, most of the collaborations took the form of a dyad (314) or triad (102), which were not connected to any other groups and are called small islands in the social network analysis literature. Therefore, large and well-connected communities of authors were rare. The space of co-authorship looked more like a constellation of small and isolated groups than a highly cohesive community.

The graph also enabled us to visualize four larger networks of co-authors—or big communities—that could themselves be divided into smaller communities by using clustering methods (Leiden in our case) (Traag et al., 2019). However, even if we enlarged our selection to the nine largest co-author networks, this subset only included 367 authors, representing 12.5% of the authors in the entire sample and 18.4% of the authors with a co-authorship. To offer a first comparison, in the field of biomedical research, Newman (2004) remarked that 92.6% of the authors belonged to the giant component (the largest connected subgraph (or component) within a network). Even in multi-disciplinary fields, such as scientists researching the evolution of cooperation, the proportion of authors in the giant component reached 33% (Liu and Xia, 2015).

The co-author networks of contemporary China specialists were also less cohesive than most national communities in the field of sociology. For instance, in Moody's (2004) research on sociologists publishing in journals belonging to the Sociological Abstracts database, 48% of the co-authoring authors were included in the giant component for the decade 1989–1999. Even in a country such as Italy, where co-authorship is less common and where sociologists exist as “a community of disconnected groups” of co-authors, 25.92% of authors belonged to the giant component (Akbaritabar et al., 2020).

This lack of connectedness among co-authoring groups is not surprising given the internalized and multidisciplinary nature of the field. The authors were affiliated with universities located in 49 countries or regions. In addition, their fields of expertise covered many different disciplines in social sciences (including but not limited to sociology, political science, international relations, and law), economics, public administration, and humanities (history and media studies).

Finally, the analysis of a non-anonymous version of the graph suggested that the communities were further disconnected by the tendencies of scholars to co-author with people working on a similar range of topics within a field (see Figure 3).

For instance, we found three large communities in the nine largest networks specializing in family and social stratification (in yellow–brown), social stratification and inequality (in light blue, in the upper part), and economic development (disconnected in red on the right). As a regular reader of the journals can sense, these different communities are generally publishing their papers in different journals.

Despite the low cohesion of the co-authors’ networks, a further investigation of the networks indicated that some well-connected scholars occupied an advantageous position. This position could confer on them some advantages in terms of productivity and citations. Indeed, these authors were key players who made up the largest network holding. In other words, if we retrenched them, there would be no more possible paths from one part of the network to another or, more importantly, between the co-authors. Indeed, 48.7% of the authors belonging to one of the nine largest networks had a betweenness centrality equal to zero, while 21 authors had a betweenness centrality above 1000. Moreover, the distribution of the betweenness centrality among the 367 scholars almost followed the Pareto principle, as 20% of the scholars with the highest betweenness centrality accounted for 76% of the total.

To sum up, the primary analysis of the network generated a series of insightful information points not only about the scarcity of relations between communities but also about the presence of key players, which was worth integrating into the author–productivity dataset. This step was easily realized by matching the different data, using the disambiguated names of the authors and their Scopus author identifiers. As shown in the first part of the results section, these data were later used to demonstrate to what extent collaborating and, more importantly, well-connected scholars benefited from their network ties.

The next step consisted of measuring the influence of network ties on the citations to previous productions. This required focusing on the most prolific (top 5%) scholars, integrating the data on self-citations and citations, and then enriching the data with relevant demographic information.

We were obliged to focus on the most prolific (top 5%) authors for two practical reasons. First, the citation count and the self-citation count extracted through our loop were inaccurate because the authors’ names in the cited references were stored in the family name–first letter of first name format. As we have already explained, many different Chinese authors could correspond to similar combinations of this type, and our dataset contained perfect homonyms. Moreover, it was impossible for us to proceed to a full disambiguation of the references because the list contained 138,890 direct citations. The second reason for our focus on the top 5% was the impossibility of finding enough relevant biographic data for most of the authors. Indeed, many people who had authored one paper over the decade had left academia without leaving behind many numerical traces of their careers. Most of the authors had never created a Google or ORCID profile through which we could find information about when they defended their PhD thesis or how they identified their gender. Finally, the shortage of data for the authors who were not the most productive was not randomly distributed. We were rarely able to retrieve up-to-date and exhaustive information for the scholars who were in the main networks but who had low betweenness. The same phenomenon existed for the scholars who were not co-authoring at all. Besides, given the extreme dispersion in terms of citations received from previous work among the most productive scholars (minimum = 0, maximum = 774, standard deviation = 109.34, and mean = 63.18), it was impossible to assume the representativity of a sample of 1% of the scholars.

This was not the case for the most productive scholars, who were often more established (32 associate professors and 83 professors out of 144 scholars). For them, we were able to acquire almost exhaustive data, as we only lacked the year when the PhD was obtained for six scholars. The descriptive table presenting the data is presented in the online Appendix.

Once we had finally gathered these data, we considered different modeling approaches to investigate how inclusion in networks was beneficial to citations. Various methods have been used to predict citation counts in previous research, including: ordinary least squares (OLS) linear regression on citations per publication (Aksnes et al., 2013); OLS regression based on a logarithmic transformation of the citations (to which one unit is added) (Stewart, 2023; Tang, 2013); distribution-free regression methods (Peters and van Raan, 1994); and, more commonly, negative binomial regression (Thelwall and Wilson, 2014). While Thelwall and Wilson (2014) argued that taking the logarithm of the citation counts and adding to it one unit before running linear or multiple linear regression is a better strategy, we considered that the properties of the data should determine the selection of the method.

In our case, the citation count almost followed a discrete lognormal distribution, and the large gap between the mean and the variance of the random variable immediately ruled out the possibility of performing Poisson regression. Given the distribution of our dependent variable, we attempted both OLS regression with the logarithm of the citations with one unit added and negative binomial regression. Some OLS regression models were relevant, with R-squared values as high as 0.6. However, others that included additional predictors (e.g. geographic locations of the scholars) did not match the assumption of linearity (tested using the R package vglma). Conversely, the assumptions of the negative binomial regression were not violated when these control variables were added. We preferred the negative binomial regression models because the graphical and numerical interpretations were more straightforward and in line with the results obtained with the reliable OLS models.

Results

Unequal productivity, co-authoring patterns, and network ties

Before presenting the correlations between co-authoring, network ties, and productivity, a general overview of productivity accompanied by some comparisons of productivity distributions in other fields of knowledge is necessary to avoid overstatement regarding the beneficial nature of network ties.

Because authors in the field of contemporary China studies have also published their papers in other disciplinary fields (economics, sociology, political science, etc.), the average productivity in the field (i.e. the nine journals we selected) and the proportion of articles authored by the most prolific authors remained relatively low.

Indeed, 75.70% of the authors only published one article in the nine journals over the 10 years, while 14.15% published two articles. In other words, the top 10% of most productive scholars only published three articles or more over the decade. This phenomenon can be explained by four main factors. First, some scholars may have left the field after publishing a paper during their PhD while not pursuing an academic career afterwards. Second, It is also likely that some Chinese authors, who accounted for 33% of the total number of contributions, were also publishing articles in journals published in Chinese. Third, many scholars may also have considered publishing in journals of economics, sociology, or other disciplines. Finally, among the authors, we found scholars such as a French demographer who was not a specialist in China but offered technical assistance to contemporary China specialists while conducting their research on other societies.

As a result, the production in the field was comparatively low and dependent on a limited number of productive scholars. In Table 1, we simulate the number of scholars who would have authored two or three papers (and so forth) if Lotka's law was observed. Table 1 shows that many more authors should have produced a larger number of articles. The average productivity of contemporary China specialists in their field was even lower than the average productivity of the sociologists appearing in Moody's (2004) study for the decade 1989–1999. Indeed, in Moody's study, only 67.71% of the authors published one paper, while 16.7% published three articles or more. In our data, the proportion of authors with only one paper was 75.7%, while only 11.2% of the authors had three or more articles.

Table 1.

Distribution of researchers’ productivity.

Number of articles	Frequency	Lotka's theory	Percentage
1	2209		75.70%
2	413	552.3	14.15%
3	152	245.4	5.21%
4	72	138.1	2.47%
5	31	88.4	1.06%
6	17	61.4	0.58%
7	14	45.1	0.48%
8	4	34.5	0.14%
9	1	27.3	0.03%
11	1	18.3	0.03%
14	1	11.3	0.03%
17	1	7.6	0.03%
19	1	6.1	0.03%
20	1	5.5	0.03%

In De Solla Price and Beaver's (1966: 1013) study, the most productive (top 5%) of authors contributed a quarter of the authorships, while only 56% of the authors had only one authorship. For contemporary China specialists, the top 5% were responsible for 21% of the authorships, while a far larger proportion of authors only published one paper over the 10 years. Table 2 summarizes the proportion of papers published by the most prolific (top 2.5%, 5%, and 10%) authors.

Table 2.

Contribution of the most prolific scholars to the articles published in the field.

Most prolific authors	Corresponding to	Number of articles	Percentage
Top 2.5%	5 or more	423	15.2%
Top 5%	4 or more	598	21.5%
Top 10%	3 or more	1012	36.4%

To conclude this first step, a general tendency in science is verified within the field of contemporary China studies (Ruiz-Castillo and Costas, 2014). Nonetheless, the field remains relatively open since it welcomes many contributors from diverse horizons, while the proportion of the papers published by the most prolific scholars remains reasonable compared to other fields of knowledge. While the winners do not take all, they still have a noticeable part of the pie.

Although the most prolific authors did not monopolize the publication space, co-authoring was still beneficial and almost necessary to belong to this group, as shown in Table 3.

Table 3.

Number of articles published over the decade depending on co-authoring status.

Published articles		1	2	3	4	5	6	7	8 or more
Co-authoring	Frequency	1425	307	130	61	30	15	14	10
	Row percentage	71.5%	15.4%	6.5%	3.1%	1.5%	0.8%	0.7%	0.5%
	Column percentage	64.5%	74.3%	85.5%	84.7%	96.8%	88.2%	100.0%	100.0%
Not co-authoring	Frequency	784	106	22	11	1	2	0	0
	Row percentage	84.7%	11.4%	2.4%	1.2%	0.1%	0.2%	0.0%	0.0%
	Column percentage	35.5%	25.7%	14.5%	15.3%	3.2%	11.8%	0.0%	0.0%

Notes: Cramér's V = 0.0956; Chi-square p-value < 0.001.

Cramér's V indicated a relatively weak correlation between the two variables due to the scarcity of authors publishing five articles or more. Nevertheless, authors who were co-authoring were more likely to belong to the top 10% (publishing three articles or more) and the top 5% (publishing four articles or more) of productive authors; conversely, they had less chance of having only one contribution. Moreover, only 15.3% of the authors who were solely single authoring exceeded the average productivity, and only 0.3% of them succeeded in belonging to the top 2.5% of most prolific authors. As is the case in other fields of knowledge, and despite the relatively low average productivity of contemporary China specialists in their field, the most prolific authors were those who were collaborating more.

Additionally, the authors’ productivity also depended on their positions in the main networks of co-authors.

Table 4 classifies the authors into four groups by combining two variables. The first group corresponds to the authors included in one of the main networks of co-authors, and the authors in this group had a betweenness above the median. Therefore, these authors had a more prominent influence since they bound communities together and could potentially bring information from one part of the network to another (or control the flux of valuable information). Interestingly, 43.3% of the authors in this group belonged to the top 5% of prolific authors. For the authors who did not hold such a key position but were still in the networks, the proportion shrank to 5.8%. Concerning the authors who had some co-authorships but were more likely to belong to small islands, they seemed almost equally likely to belong to the top 5% of prolific authors.

Table 4.

Chances of belonging to the top 5% of most productive scholars depending on authorship and network status.

		Belonging to the top 5% of productive scholars
Network status		No	Yes	Total
In the main networks— high betweenness	Frequency	51	39	90
	Expected frequency	86	4	90
	Row percentage	56.7%	43.3%	100%
	Pem	−12.6%	25%
In the main networks— low betweenness	Frequency	261	16	277
	Expected frequency	263	14	277
	Pem	−0.8%	1.5%
	Row percentage	94.2%	5.8%	100%
Not in the main networks— co-authoring	Frequency	1550	75	1625
	Expected frequency	1545	80	1625
	Row percentage	95.4%	4.6%	100%
	Pem	0.4%	−3.5%
Not in the main networks— not co-authoring	Frequency	912	14	926
	Expected frequency	880	46	926
	Row percentage	98.5%	1.5%	100%
	Pem	1.7%	−22%
Total	Frequency	2774	144	2918
	Expected frequency	2774	144	2918
	Row percentage	95.1%	4.9%	100%

Notes: χ² = 306.759; df = 3; Cramér's V = 0.324; Fisher's p = 0.000.

The additional interpretation of the percentage of maximum deviation from independence (PEM) (Cibois, 2009) confirmed that the main networks of key players were significantly more likely to belong to the top 5% of most prolific scholars, while the likelihood for the authors who were solely single authoring was significantly lower. In other cases (i.e. “In the main networks—low betweenness” and “Not in the main networks—co-authoring”), the likelihood was not significantly higher or lower since the local PEM was neither above 5% nor below 5%.

To conclude, a strong correlation existed between co-authoring and productivity. In addition, the key players in the main networks exhibited higher productivity, as 63.3% of them belonged to at least the top 10% of productive authors.

Network ties as one of the most important predictors of citations of previous work

Collaboration and by extension network ties mattered not just in terms of productivity. The negative binomial regression models all indicated that network ties tremendously increased the number of citations to previous work received by the 144 most productive scholars.

The incidence rate ratio for network status left no doubt about the importance of network ties. Scholars in the main networks were expected to have a higher rate of citation compared to those not in the networks, increasing from 104% (Model 2) to 159% (Model 1).

More interestingly, network status was more influential than other demographic characteristics (e.g. gender and job position) that were determinants in other studies (Wu, 2023). Indeed, the p-value for gender only appeared to be slightly significant for Model 2 (p = 0.058) and was insignificant once the variable academic field was added (Model 3). To quote Xie and Shauman (1998: 864), it remains possible that “women and men scientists are located in different academic structures with different access to valuable resources”. Therefore, once these parameters were controlled, the pure effect of gender was not statistically significant.

Regarding job position, this was not significant because seniority was included in the model. Including these two variables in Models 2, 3, and 4 showed that citation counts increased thanks to reputation and know-how that were accumulated over years, rather than because of a job position that might signal expert status (see Table 5). Put differently, older scholars received more citations because they had become highly recognized over the years, not because they held a specific title.

Table 5.

Negative binomial logistic regression models on citation count.

	Model 1	Model 2	Model 3	Model 4
(Intercept)	−0.746 (0.456)	0.551 (0.538)	0.139 (0.573)	0.184 (0.621)
Network status
Not in the main networks
In the main networks	0.950 *** (0.151)	0.712 *** (0.155)	0.779 *** (0.160)	0.738 *** (0.161)
Gender
Female (baseline)
Male	0.283 (0.156)	0.281 (0.149)	0.219 (0.148)	0.220 (0.147)
Productivity in the sub-field
Log (article in the sample)	1.009 *** (0.229)	0.466 (0.257)	0.359 (0.256)	0.380 (0.257)
Seniority
Log (years since PhD)	0.846 *** (0.120)	0.534 *** (0.148)	0.593 *** (0.148)	0.579 *** (0.155)
Self-citations
Self-citations		0.036 *** (0.007)	0.040 *** (0.007)	0.038 *** (0.008)
Academic position
Not professor (baseline)
Professor		0.267 (0.175)	0.174 (0.176)	0.130 (0.182)
Academic field
Economy and law (baseline)
Other humanities			0.297 (0.383)	0.213 (0.391)
Political science			0.591 ** (0.226)	0.532 * (0.231)
Sociology			0.401 (0.222)	0.366 (0.226)
Geographic location
Hong Kong–Macao–Tawain
Chinese mainland				0.014 (0.189)
Other location in Asia				−0.163 (0.256)
Other Global North countries				0.137 (0.220)
United States or Canada				0.267 (0.195)
n	138	138	138	138
Akaike information criterion	1317.194	1305.260	1305.238	1309.930
Bayesian information criterion	1334.758	1328.678	1337.438	1353.839
Pseudo R²	0.561	0.608	0.625	0.634

Notes: *** p < 0.001; ** p < 0.01; * p < 0.05.

Concerning seniority in the field, this aspect also largely accounted for the discrepancies between scholars. The variable years after PhD, which was log-transformed in Models 2, 3, and 4, indicated that a 1% change unit in the year after the PhD was associated with a 53.4–57.9% change in expected citations. Consequently, seniority in association with inclusion in one of the main networks could lead to tremendous gaps in terms of citations.

Figure 3 helps visualize this huge difference by showing the predicted citations depending on seniority and network status, based on Model 4.

For instance, approximately seven years after receiving a PhD, when a scholar is supposed to achieve tenure and apply for a promotion to associate professor (corresponding to a log of 2), the scholar was predicted to have received 52 citations if they belonged to one of the main networks of co-authors and 25 if this was not the case. Moreover, 20 years after PhD graduation (corresponding to a log of 3), a scholar not included in the main networks might receive fewer citations than a scholar included in a main network seven years after graduation.

Regarding self-citations, their noticeable influence needs to be both reported and nuanced. First, there is no doubt that some of the top 5% of most prolific scholars had more self-citations. For instance, the scholar who cited himself the most, self-cited 109 times. With these self-citations alone, he received more citations than 137 scholars in the sample. Nonetheless, once the self-citations were retrenched, this scholar remained the most cited scholar, as his self-citations represented only 14% of the citations he received. This was below the average self-citation rate of the most productive scholars (16.2%), which is relatively standard compared with other research (a variation between 10% and 36% depending on discipline) (Wallace et al., 2012). More generally, self-citations hardly changed the hierarchy between researchers. The Pearson coefficient of correlation between the ranking of researchers with self-citations excluded and the raw number of citations was 0.97 (for a maximum of 1).

Finally, self-citations remained quite low in the field, with an average of 1.73 citations to previous work in each newly published paper (see Figure 4). The magnitude of self-citations might therefore signal that those productive scholars built their new research questions based on their previous research, but they did not compulsively seek citations through self-citations. There might be some exceptions to this general rule, as 7.6% of the scholars (11 individuals) combined an average number of self-citations per article with a self-citation ratio twice as high as average. In other words, only a few scholars build their reputation through zealous self-citation. Interestingly, none of these 11 scholars was a key player in the main networks of co-authors. This suggests that zealous self-citation might be more a result of isolation, rather than a strategy employed by the most prominent scholars, who are also the most connected.

Discussion and conclusion

Without being abnormal specimens, specialists in the field of contemporary China studies show similarities and dissimilarities to scholars in other fields of the social sciences. First, similar patterns about the relation between productivity and collaboration have been found in other social science fields. As explained above, in social sciences, the average productivity of scholars is typically low, and a large proportion of authors only publish one article over an extended period. For instance, Moody's (2004) study of American sociologists showed that in the decade 1989–1999, two-thirds of authors only produced one article in a peer-reviewed journal indexed by the Sociological Abstracts database. Such a productivity rate is still observable worldwide in the twenty-first century, as Ruiz-Castillo and Costas's (2014) study showed that 75.3% of authors belonging to the fields of sociology and anthropology and authoring in journals indexed by Thomson Reuters published only one paper in the period 2003–2011. In light of this, the average productivity (1.459 articles) and the proportion of contemporary China specialists with only one contribution (75.7%) are not surprising at all.

However, the reasons why the productivity is skewed might be slightly different. As mentioned earlier, contemporary China specialists are often publishing in a wide range of disciplinary journals and in different languages. For instance, the most prolific scholar published 20 articles in the journals selected for this study, which only corresponded to 19.4% of the articles he published in peer-reviewed journals over the decade. He also published 21 articles in Chinese. Therefore, it is not surprising to find so many authors with few papers and such a limited gap between the most and least productive. Moreover, if we stipulate, following Bourdieu (1990), that social agents never accomplish disinterested actions, it would be more profitable for authors to publish in disciplinary journals with higher impact factors.

This phenomenon of engagement in multiple fields might explain why some solely single-authoring scholars belonged to the top 10% and 5% of most prolific scholars in contemporary China studies. Such an achievement would be barely thinkable in the hard sciences and also hardly achievable in American sociology. From this perspective, the possibility of becoming one of the most prolific scholars while not collaborating with others appears as the most salient characteristic of the field of contemporary China studies.

Nevertheless, as is generally the case in the hard sciences and social sciences (Melin, 2000), co-authoring and being a key player in the main networks of co-authors were both positively correlated with productivity. This phenomenon can be explained by two factors, which take their origins in the perspectives of scientific and technical human capital (Bozeman and Corley, 2004) and the social division of labor (Durkheim, 2013).

First, in line with Bozeman and Corley's (2004) study, the high productivity of key players in networks partly arises from their mentoring strategies. Indeed, among the contemporary China specialists, the key players were often well-established scholars leading large projects in which several of their PhD students, post-doctoral students, or former students were participating. In the networks of co-authors, some of the key players appeared to be linked to a few well-established scholars, but they were more widely linked to many co-authors who did not have any other ties. In these cases, the key players were likely to take part in many articles for which a substantial number of operations (e.g. data collection and analysis) were carried out by people they supervised. For instance, in one of the communities composed of 39 researchers (detected through the Leiden clustering method), we found 17 actual or former students of one key player. In this case, the high productivity of the key player partly corresponded to his capacity to collect research funding and skillfully supervise a network of younger collaborators, as well as pursuing collaborations with them once they graduated. This phenomenon, which is particularly pronounced for a few other key players, is in line with the findings of Fox and Mohapatra (2007) and Bozeman and Corley (2004). In China, sociologists give the name shimen (community with the same professor) to this community formed by one leading scholar and their present and former students.

Second, key players’ productivity may also partly arise from their capacity to establish collaborations outside of their institution with prominent scholars working on similar topics. Indeed, previous research demonstrated that the capacity to establish international collaborations led to higher productivity (Jonkers and Tijssen, 2008; Martín-Sempere et al., 2002). There are some reasons to believe that such a tendency is at work in the field of contemporary China studies. For instance, the most productive and important key player in the first cluster had 38 different co-authors, located in 12 different institutions in countries including not only the United States and China but also Belgium. Only four of these co-authors were present or former students, while the majority were senior scholars specializing in a similar topic. As explained by Bozeman and Corley (2004), such international collaborations might not only contribute to building invisible colleges but also provide opportunities to acquire and transmit technical skills and resources that ultimately benefit productivity and reputation.

Mentoring and collaborations with other prominent scholars from other institutions may also create or strengthen invisible colleges, helping scholars to increase their impact in the field. Indeed, the communities in the main networks did not seem to be the only platforms for the cross-fertilization of ideas and the accumulation of scientific and technical human capital. They were invisible colleges as defined by Crane (1972) because they were groups in which particular references, methodologies, perspectives, and results were gathered to the point that scholars mutually influenced one another, trusted one another, and ultimately cited one another. In other words, these communities worked as groups, sharing the same mental universe and then influencing the literature and citation landscape. Indeed, our results showed without any ambiguity that co-authored papers often referred to the previous work of the most senior scholars involved in the collaboration. As noticed, the raw number of citations originating from such collaborations was sometimes substantial. However, we must insist on the idea that self-citations were not purposely and prevalently used to increase researchers’ H-index scores.

Aside from self-citation, inclusion in one of the main networks of co-authors had a tremendous net impact on citation counts. This suggests that well-connected authors’ past publications are taken seriously not only by their immediate collaborators but also by the acquaintances of their collaborators, following the friends of friends pattern, described by Boissevain (1974). In sum, our findings indicate that networks of co-authors in contemporary China studies were not only networks of collaboration but also networks of inter-knowledge and inter-acknowledgement. To reuse Nahapiet and Ghoshal's (1998: 252) terminology, co-author networks not only provided the insiders with access to valuable information and knowledge but also led to “reputational endorsement”.

Although networks provided benefits, we should bear in mind that invisible colleges might be less influential in the case of contemporary China specialists. As previously mentioned, the 396 scholars included in the nine largest co-author networks represented only 12.5% of contemporary China specialists. Besides, they only contributed to 11.4% of the articles published over the 10 years. Therefore, it would be completely inaccurate to claim that a few key scholars together with their students and acquaintances monopolized journals and systematically used them to ensure their self-promotion. At least, further research adopting an author-based approach and retrieving the entire co-authorships of all the scholars (i.e. co-authorships in and outside the field of contemporary China studies), would be necessary to support this claim.

This fact ultimately invites us to relativize the discourse about the influence of brothers with the same teachers (shixiong) and communities with the same professors (shimen). Contrary to some claims, these communities are not monopolistic. Ultimately, and as surprising as it might sound, the field of contemporary China studies is more open than other scientific fields. To this extent, the contemporary China specialists covered in this study were not exactly like scholars in any other field.

Footnotes

Authors’ contributions

Aurelien Boucher was involved in research question formulation, the collection of data, the construction of network analysis, the interpretation of the results, the exploration of the literature, and the writing of the article. Xiaoguang Fan was involved in research question formulation, the interpretation of the results, the exploration of the literature, and the writing of the article.

Ethical approval

The project “Understanding Chinese Social Sciences Through big data” has received the ethical approval of the afferent authority from The Chinese University of Hong Kong, Shenzhen (CUHK-Shenzhen).

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by HSS Grant( No.: G10120230297) of the Chinese University of Hong Kong, Shenzhen (CUHK-Shenzhen).

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

ORCID iDs

Aurélien Boucher

Xiaoguang Fan

Supplemental material

Supplemental material for this article is available online.

Notes

References

Abbasi

Altmann

Hossain

(2011) Identifying the effects of co-authorship networks on the performance of scholars: A correlation and regression analysis of performance measures and social network analysis measures. Journal of Informetrics 5(4): 594–607.

Akbaritabar

Casnici

Squazzoni

(2018) The conundrum of research productivity: A study on sociologists in Italy. Scientometrics 114(3): 859–882.

Akbaritabar

Traag

Caimo

, et al. (2020) Italian sociologists: A community of disconnected groups. Scientometrics 124(3): 2361–2382.

Aksnes

Rørstad

Piro

Sivertsen

(2013) Are mobile researchers more productive and cited than non-mobile researchers? A large-scale study of Norwegian scientists. Research Evaluation 22(4): 215–223.

Beaver

(2001) Reflections on scientific collaboration (and its study): Past, present, and future. Scientometrics 52(3): 365–377.

Biscaro

Giupponi

(2014) Co-authorship and bibliographic coupling network effects on citations. PLoS One 9(6): e99502.

Boissevain

(1974) Friends of Friends: Networks, Manipulators and Coalitions. New York: St. Martin Press.

Bourdieu

(1988) Homo Academicus. Stanford, CA: Stanford University Press.

Bourdieu

(1990) In Other Words: Essays Towards a Reflexive Sociology. Stanford, CA: Stanford University Press.

10.

Bozeman

Corley

(2004) Scientists’ collaboration strategies: Implications for scientific and technical human capital. Research Policy 33(4): 599–616.

11.

Bozeman

Fay

Slade

(2013) Research collaboration in universities and academic entrepreneurship: The-state-of-the-art. The Journal of Technology Transfer 38(1): 1–67.

12.

Cibois

(2009) Percentage of maximum deviation from independence (PEM): Comment on Lefèvre and Champely’s analyse d’un tableau de contingence. BMS: Bulletin of Sociological Methodology / Bulletin de Méthodologie Sociologique 103(1): 66–74.

13.

Crane

(1972) Invisible Colleges: Diffusion of Knowledge in Scientific Communities. Chicago, IL: University of Chicago Press.

14.

De Haan

(1997) Authorship patterns in Dutch sociology. Scientometrics 39(2): 197–208.

15.

De Solla Price

Beaver

(1966) Collaboration in an invisible college. American Psychologist 21(11): 1011–1018.

16.

Ductor

(2015) Does co-authorship lead to higher academic productivity? Oxford Bulletin of Economics and Statistics 77(3): 385–407.

17.

Durkheim

(2013) De la division sociale du travail [On the social division of labor]. Paris: Presses Universitaires de France (in French).

18.

Etzkowitz

(1984) Entrepreneurial scientists and entrepreneurial universities in American academic science. Minerva 21(2–3): 198–233.

19.

Feng

Kirkley

(2020) Mixing patterns in interdisciplinary co-authorship networks at multiple scales. Scientific Reports 10(1):7731.

20.

Fox

Mohapatra

(2007) Social-organizational characteristics of work and publication productivity among academic scientists in doctoral-granting departments. The Journal of Higher Education 78(5): 542–571.

21.

Goutsmedt

Truc

(2022) Networkflow: Functions for a workflow to manipulate networks. Available at: https://github.com/agoutsmedt/networkflow (accessed 14 January 2023).

22.

Guan

Yan

Zhang

(2017) The impact of collaboration and knowledge networks on citations. Journal of Informetrics 11(2): 407–422.

23.

Henriksen

(2016) The rise in co-authorship in the social sciences (1980–2013). Scientometrics 107(2): 455–476.

24.

Hollis

(2001) Co-authorship and the output of academic economists. Labour Economics 8(4): 503–530.

25.

Hong

Zhao

(2016) How social networks affect scientific performance: Evidence from a national survey of Chinese scientists. Science, Technology, and Human Values 41(2): 243–273.

26.

Jin

, et al. (2022) Co-citation and co-authorship networks of statisticians. Journal of Business and Economic Statistics 40(2): 469–485.

27.

Jonkers

Tijssen

(2008) Chinese researchers returning home: Impacts of international mobility on research collaboration and scientific productivity. Scientometrics 77(2): 309–333.

28.

Katz

Hicks

(1997) How much is a collaboration worth? A calibrated bibliometric model. Scientometrics 40(3): 541–554.

29.

Katz

Martin

(1997) What is research collaboration? Research Policy 26(1): 1–18.

30.

Kretschmer

(2004) Author productivity and geodesic distance in bibliographic co-authorship networks, and visibility on the web. Scientometrics 60(3): 409–420.

31.

Kretschmer

Rousseau

(2001) Author inflation leads to a breakdown of Lotka’s law. Journal of the American Society for Information Science and Technology 52(8): 610–614.

32.

Kumar

(2015) Co-authorship networks: A review of the literature. Aslib Journal of Information Management 67(1): 55–73.

33.

Lee

Bozeman

(2005) The impact of research collaboration on scientific productivity. Social Studies of Science 35(5): 673–702.

34.

Liao

Yen

(2013) Co-authorship networks and research impact: A social capital perspective. Research Policy 42(9): 1515–1530.

35.

Liu

Xia

(2015) Structure and evolution of co-authorship network in an interdisciplinary research field. Scientometrics 103(1): 101–134.

36.

Lotka

(1926) The frequency distribution of scientific productivity. Journal of the Washington Academy of Sciences 16(12): 317–323.

37.

Chen

(2014) Study of the collaboration in the field of the Chinese humanities and social sciences. Scientometrics 100(2): 439–458.

38.

Martín-Sempere

Rey-Rocha

Garzón-García

(2002) The effect of team consolidation on research collaboration and performance of scientists: Case study of Spanish University researchers in Geology. Scientometrics 55(3): 377–394.

39.

Melin

(2000) Pragmatism and self-organization research collaboration on the individual level. Research Policy 29(1): 31–40.

40.

Moody

(2004) The structure of a social science collaboration network: Disciplinary cohesion from 1963 to 1999. American Sociological Review 69(2): 213–238.

41.

Murphy

(1973) Lotka’s law in the humanities? Journal of the American Society for Information Science 24(6): 461–462.

42.

Nahapiet

Ghoshal

(1998) Social capital, intellectual capital, and the organizational advantage. The Academy of Management Review 23(2): 242–266.

43.

Newman

MEJ

(2001a) Scientific collaboration networks. II. Shortest paths, weighted networks, and centrality. Physical Review E 64(1): 1–7.

44.

Newman

MEJ

(2001b) The structure of scientific collaboration networks. Proceedings of the National Academy of Sciences of the United States of America 98(2): 404–409.

45.

Newman

MEJ

(2004) Who is the best connected scientist? A study of scientific coauthorship networks. In: Ben-Naim

Frauenfelder

Toroczkai

(eds) Complex Networks (Vol. 650). Berlin, Heidelberg: Springer Berlin Heidelberg, pp.337–370.

46.

Parish

Boyack

Ioannidis

JPA

(2018) Dynamics of co-authorship and productivity across different fields of scientific research. PLoS One 13(1): e0189742.

47.

Peters

HPF

van Raan

AFJ

(1994) On determinants of citation scores: A case study in chemical engineering. Journal of the American Society for Information Science 45(1): 39–49.

48.

Pontille

(2003) Authorship practices and institutional contexts in sociology: Elements for a comparison of the United States and France. Science, Technology, and Human Values 28(2): 217–243.

49.

Pravdic

Oluic-Vukovic

(1991) Distribution of scientific productivity: Ambiguities in the assignment of author rank. Scientometrics 20(1): 131–144.

50.

Ruiz-Castillo

Costas

(2014) The skewness of scientific productivity. Journal of Informetrics 8(4): 917–934.

51.

Stewart

(2023) Achievement and ascriptive processes in the recognition of scientific articles. Social Forces 62(1): 166–189.

52.

Stoltz

(2023) Rising co-authorship in sociology, 1895–2022 (preprint). SocArXiv.

53.

Tang

(2013) Does “birds of a feather flock together” matter—evidence from a longitudinal study on US–China scientific collaboration. Journal of Informetrics 7(2): 330–344.

54.

Thelwall

Wilson

(2014) Regression for citation data: An evaluation of different methods. Journal of Informetrics 8(4): 963–971.

55.

Traag

Waltman

van Eck

(2019) From Louvain to Leiden: Guaranteeing well-connected communities. Scientific Reports 9(1):5233.

56.

Wallace

Larivière

Gingras

(2012) A small world of citations? The influence of collaboration networks on citation practices. PLoS One 7(3): e33339.

57.

Wanner

Lewis

Gregorio

(1981) Research productivity in academia: A comparative study of the sciences, social sciences and humanities. Sociology of Education 54(4): 238–253.

58.

Wilder

Walters

(2019) Quantifying scholarly output: Contribution studies and productivity studies in sociology since 1970. The American Sociologist 50(3): 430–436.

59.

(2023) The gender citation gap: Why and how it matters. Canadian Review of Sociology/Revue Canadienne de Sociologie 60(2): 188–211.

60.

Xie

Shauman

(1998) Sex differences in research productivity: New evidence about an old puzzle. American Sociological Review 63(6): 847–870.

61.

Yair

Goldstein

Rotem

, et al. (2022) The three cultures in American science: Publication productivity in physics, history and economics. Scientometrics 127(6): 2967–2980.

62.

Zuckerman

(1967) Nobel laureates in science: Patterns of productivity, collaboration, and authorship. American Sociological Review 32(3): 391–403.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.01 MB