The co-word analysis technique was first proposed by Callon et al. (1986). Since then, academic researchers have used co-word analysis to map the bibliometric structure of different fields, including creativity (Zhang, et al., 2015), environmental responsibility (Dai & Zhang, 2020; Yang et al., 2021), auditing (Uyar et al., 2020), and IC (Bamel et al., 2022; Quintero-Quintero et al., 2021).

Co-word analysis has been considered an effective method for content analysis and text mining (Feng et al., 2017; Zupic & Čater, 2015). One of its key advantages was that it revealed the conceptual structure of a discipline without the need to consult the full text (Romo-Fernández et al., 2013). Co-word analysis was based on the assumption that the co-occurrence of two or more keywords in a document indicated the correlation between them, and the higher the co-occurrence frequency, the stronger their relationship (An & Wu, 2011; Callon et al., 1986; Hu & Zhang, 2015; Ravikumar et al., 2015; Whittaker, 1989). Another assumption was that keywords were carefully selected by the authors and accurately represented the document's content (Feng et al., 2017). Co-word analysis can be used to quantify the links between research themes in a scientific discipline (Ding et al., 2001; Khasseh et al., 2017; Ravikumar et al., 2015; Sedighi, 2016), identify domains, subdomains, and hot topics (Dai & Zhang, 2020; et al., 2012; Zhang et al., 2015), and predict future trends (Uyar et al., 2020).

Social network analysis (SNA) was used for exploring the latent content of scientific texts. It was introduced in the 1960s by the renowned sociologist Harrison White. Social networks were defined as a network of relationships or interactions, where the nodes were people or actors, and the edges represented the relationships or interactions between them (Abbasi et al., 2011). The main element in a social network was the actor or keyword (Köseoglu et al., 2019). The relationship between these actors (or keywords) constituted ties or links (Yang et al., 2012), the sum of which formed the graphical networks in SNA or the conceptual map and the knowledge network that reflected the current state of a specific subject area (Uyar et al., 2020).

SNA has been increasingly employed by researchers in various fields such as information science (Otte & Rousseau, 2002), economic geography (Ter Wal & Boschma, 2009), communities of practice, and natural resource management (Cross et al., 2006; Prell et al., 2009), water pollution management (Cantner & Graf, 2006; Ruzol et al., 2017), creativity (Zhang et al., 2015), environmental responsibility (Dai & Zhang, 2020; Yang et al., 2021), medicine (Xie et al., 2020), auditing (Uyar et al., 2020), knowledge transfer (Marchiori & Franco, 2020) and IC (Bamel et al., 2022), among others.

VOSviewer has been used as a software tool for constructing and visualizing bibliometric networks, including networks of journals, researchers, or articles based on citation, bibliographic coupling, co-citation, or co-authorship relationships. VOSviewer also has text mining functionality that can be used to construct and visualize co-occurrence networks of keywords extracted from a body of scientific literature.

This study used the Clarivate Analytics WOS database to retrieve data and VOSviewer software to construct social networks. WOS has been widely used as a reliable source for the systematic review of texts (Benavides-Velasco et al., 2013; Khan & Wood, 2015; Köseoglu et al., 2019; Kumar & Jan, 2013; Uyar et al., 2020; Yan et al., 2015; Zupic & Čater, 2015). The period between 1975 and 2020 was covered since prior to 1975, scientific journals seldom required keywords, and the content was mostly unavailable online.

After an in-depth review of the texts, 19 keywords1 were selected as the most extensive representatives of IC research which were used with the WOS database on November 22, 2021. This application resulted in 22,613 documents, of which 10,303 belonged to unrelated disciplines (e.g., philosophy, nursing, history) and were removed. The final statistical population, covered the disciplines of economics, management, business, finance, social science, interdisciplinary, information science, library science, operations research, management science, and public administration, and consisted of 12,310 documents. Of these, 8,413 were articles2, 824 were books3, and 3,073 were proceedings papers. For a more detailed analysis, the population was divided into two periods, i.e., 1975 to 2000 and 2001 to 2020, the latter of which represented over 90% of the documents. Sample statistics are reported in Table 1. Fig. 1 also demonstrates the data collection framework.

Fig. 1.

Data Collection Framework.

(0.72MB).

Co-occurrence refers to the presence, frequency, and proximity of similar keywords across articles and can reveal hot research topics. It includes thematically identical keywords, but not exactly the same. A total of 6,516 keywords were used in the documents between 1975 and 2020. Following prior research (Dai & Zhang, 2020; Zhang et al., 2015), a threshold was set for keyword frequency. The research period was divided into three parts: a threshold of 50 for the entire period (1975–2020); a threshold of 5 for the period 1975–2000; and a threshold of 50 for the period 2001–2020. Consequently, 159 keywords in the total time period, 55 in the period 1975–2000, and 156 in the period 2001–2020 met the specified thresholds. Table 2 shows the top 50 frequent keywords in these time periods.

Fig. 3 illustrates the co-occurrence network. The most frequent keywords were divided into 6 clusters with six different colors. Keywords that were similar in content were grouped in a cluster. For example, the keywords "Performance" and "Networks" were in the blue cluster and the keywords "Knowledge" and "Strategy" were in the green cluster. The size of the circles indicated keyword frequency, and the thickness of the lines indicated the strength of co-occurrence within and between clusters. As the figure shows, all the clusters were interconnected, and there were strong relationships between the 6 clusters. This indicated the high interdependence of different areas of IC research.

Fig. 3.

Co-occurrence network of keywords (1975–2020).

(0.72MB).

Overall, there were 159 keywords, 6 clusters, and 7,671 links in this network, with a total link strength of 44,502. Cluster 1 (red) with 56 keywords was the largest cluster and was represented by "Growth", Cluster 2 (green) had 34 keywords and was represented by "Knowledge", cluster 3 (blue) had 27 keywords and was represented by "Performance", cluster 4 (yellow) had 20 keywords and was represented by "Innovation", cluster 5 (purple) had 17 keywords and was represented by "Impact", and cluster 6 (turquoise) had five keywords and was represented by "Success".

Mapping keywords in terms of density can also be informative. In a density map, the closer a keyword is to the red areas, the higher is the keyword frequency. As shown in Fig. 4, keywords such as "Innovation", "Knowledge", "Networks", "Governance", "Business", "Company", "Performance", "Trust", "Market", "Model", "Management", "Effect", "Information", "Success", "Growth", and "Education" received a great deal of attention from researchers in recent years and were practically saturated. In contrast, keywords such as "Mediating Role", "Efficiency", "Gender", and "Firm Market Value" received relatively little attention. This finding could inform researchers about the most effective areas to focus on.

Fig. 4.

Keyword density map (1975–2020).

(0.57MB).

Figs. 5 and 6 show the keyword density map for the period 1975–2000 (with a co-occurrence threshold of 5) and 2001–2020 (with a co-occurrence threshold of 50), respectively. According to these figures, keywords such as "Growth", "Economic Growth", "Model" and "Earnings" that featured prominently between 1975 and 2000 (keywords in the red and orange sections) faded between 2001 and 2020. Moreover, keywords such as "Performance" and "Management" that were not prominent between 1975 and 2000 became very common in the period 2001 to 2020 and were in the red areas of the map. In addition, new areas such as "Mediating Role", "Entrepreneurial Orientation", "Competitive Advantage", and "Knowledge Management" became more prominent between 2001 and 2020.

Fig. 5.

Keyword density map (1975–2000).

(0.43MB).

Fig. 6.

Keyword density map (2001–2020).

(0.53MB).

The word clouds for the four dimensions of IC (human capital, structural capital, relational capital, and social capital) are illustrated in Fig. 7 using VOYANT software. Word clouds are a weighted list for visualizing text or language data and have become increasingly popular in recent years (Jin, 2017). It must be noted that a larger font size indicates higher keyword frequency.

Fig. 7.

Word clouds for the four dimensions of intellectual capital.

(0.87MB).

The top ten most frequently cited articles are listed in Table 3. The results show that the article titled "The Benefits of Facebook Friends: Social Capital and College Students' Use of Online Social Network Sites" was the most frequently cited article in the field of IC with 3,519 citations. This information can help researchers and readers identify the most relevant research studies.

Co-citations between documents were also analyzed. A co-citation network can visualize the evolution of a scientific field. Co-citation refers to the frequency with which two documents are cited together in a third document (Small, 1973). The results indicated 268,369 co-citations across the documents. Fig. 8 shows the co-citation network of documents with at least 80 co-citations. It must be noted that 246 documents met this threshold. Overall, 246 documents, 4 clusters, and 19,455 links are shown in Fig. 8. In addition, the total link strength is 223,107. Cluster 1 (red) was the largest cluster with 73 documents. It was based on the article by Nahapiet & Ghoshal (1998), with 1,445 citations, 228 links, and a total link strength of 15,012. The second largest cluster in green had 65 documents. It was based on the article by Barney (1991) with 700 citations, 228 links, and a total link strength of 7,375. The third largest cluster was blue with 63 items. It was based on the article by Edvinsson (1997) with 897 citations, 217 links, and a total link strength of 8123. The fourth and last cluster was yellow with 45 items. It was based on the article by Lucas (1988) with 586 citations, 168 links, and a total link strength of 2,398.

Fig. 8.

Co-citation network (1975–2020).

(0.5MB).

Table 4 lists the top ten authors in terms of the number of documents published. According to these statistics, "Jeffrey Chen" is the top author in the field of IC with 32 documents. This list can help readers identify the most prominent researchers in this field.

The co-authorship network is also illustrated in Fig. 9. Co-authorship networks have been used in various studies to understand the structure of a research field (Andrikopoulos & Kostaris, 2017; Chan et al., 2009; Faraji et al., 2020; Kılıç et al., 2019). There were a total of 20,226 authors in the field of IC, but only those who had written at least eight documents and received ten citations were illustrated in Fig. 9. 65 authors met this threshold. Overall, there were 65 authors, 45 clusters, and 31 links. In addition, the total link strength was 120. Authors in a network have research collaborators. For example, "Nick Bontis" has collaborated with "Muhammad Khalique" and "Francesca Sgrò" on a number of research projects. Of course, "Nick Bontis" has also collaborated with "John Dumay", but these collaborations have not been extensive enough to form a co-authorship network.

Fig. 9.

Co-authorship network (1975–2020).

(0.25MB).

Table 5 shows the ten most prolific countries. Most articles on IC have been published in the United States (about 27% of articles). China (about 12% of articles) and the United Kingdom (about 8% of articles) were in second and third place, respectively. It must be noted that in terms of publications, there was a big gap between the United States and the next ranked countries.

The co-authorship network of countries was also illustrated in Fig. 10. A total of 139 countries were involved in writing and publishing works, as suggested by the organizational affiliation of the authors, with only 68 countries meeting the threshold of at least 12 documents and 100 citations.

Fig. 10.

Co-authorship network of countries (1975–2020).

(0.43MB).

Overall, there were 68 countries, 6 clusters, and 736 links in the network, with a total link strength of 3,574. The size of the circles indicated the number of documents published by each country through international collaboration. The larger the circle, the more active the country was in international research. The lines between the two countries indicated the frequency of collaboration, and the thicker the line, the more extensive the collaboration and the closer the relationship. For example, the United States had extensive collaboration with many countries, including China, Canada, South Korea, Australia, Taiwan, Singapore, and the United Kingdom.

Table 6 lists the top ten publications by the number of articles (Panel A) and the number of citations (Panel B). This list can help researchers identify the most prominent journals involved with IC research to publish their findings. A total of 3,188 journals were active in this field. The “Journal of Intellectual Capital” was the top journal in terms of the number of articles with 208 documents, and the “Academy of Management Journal” was the most frequently cited journal with 11,914 citations. In Panel A, three of the top ten journals were conference journals.

In addition, Fig. 11 illustrates the co-citation network of journals, which can help identify the most important journals in the field of IC. The results indicate 102,326 co-citations among journals. This network included journals with at least 290 citations. It must be noted that 199 journals met this threshold. Fig. 11 had 4 clusters, 199 journals, and 18,975 links with a total link strength of 4,301,902. Cluster 1 (red) with 87 journals was the largest cluster. It was based on the “American Economic Review” with 8,232 citations, 198 links, and a total link strength of 231,316. The second-largest cluster was cluster number 2 (green), with 50 journals. This cluster was based on the “Strategic Management Journal” with 9,491 citations, 198 links, and a total link strength of 487,632. The third-largest cluster was cluster number 3 (blue), with 43 journals. It was based on the “Journal of Intellectual Capital” with 11,542 citations, 197 links, and a total link strength of 294,930. The fourth and last cluster was cluster number 4 (yellow), with 19 journals. It was based on the “Research Policy” journal with 4,691 citations, 198 links, and a total link strength of 184,437.

Fig. 11.

Co-citation network of journals (1975–2020).

(0.71MB).

Table 7 lists the top ten universities by the number of articles. This list can help interested readers identify universities active in IC research for potential research collaborations. The results showed that a total of 5,920 universities and research institutes were involved in publishing research papers in the field of IC, and the “National Bureau of Economic Research (NBER)” was the world's leading research institute with 84 research papers. It must be noted that 7 of the top 10 universities were located in the United States. This suggests the substantial investment in IC research and the commitment of US universities to this important research area.

The co-authorship network between universities is illustrated in Fig. 12. A total of 5,920 universities and research institutes were involved in conducting and publishing IC research, as suggested by the organizational affiliation of the authors, with 97 universities meeting a threshold of at least 25 documents and 100 citations. Overall, there were 97 universities, 8 clusters, and 580 links, with a total link strength of 845. The size of the circles indicated the number of documents published through inter-university collaborations within and between countries. The larger the circle, the more active the university was in inter-university collaborations. The lines between the two universities indicated the frequency of collaboration, and the thicker the line, the more extensive the collaboration and the closer the relationship. For example, Harvard University had extensive collaborations with the University of Columbia, University of Pennsylvania, UC Berkeley, University of Melbourne, University of Manchester, and University of Massachusetts among others.

Fig. 12.

Co-authorship network between universities (1975–2020).

(0.57MB).

In order to examine regional (continental) differences in IC research, the co-occurrence of keywords in Asia, Europe, North America, South America, Oceania, and Africa are highlighted in Table 8. The results showed that except for ​​South America, where the most frequently used keyword was “Innovation”, “Performance” was the most frequently used keyword in all the other continents. Moreover, to clarify the co-occurrence networks, the co-occurrence of the keywords was visualized for each geographical region (Figs. 13–18).

Fig. 13.

Co-occurrence network in Asia (1975–2020).

(0.72MB).

Fig. 14.

Co-occurrence network in Europe (1975–2020).

(0.84MB).

Fig. 15.

Co-occurrence network in North America (1975–2020).

(0.74MB).

Fig. 16.

Co-occurrence network in South America (1975–2020).

(0.56MB).

Fig. 17.

Co-occurrence network in Oceania (1975–2020).

(0.79MB).

Fig. 18.

Co-occurrence network in Africa (1975–2020).

(0.66MB).

The interpretation of these regional co-occurrence networks was similar to Fig. 3. Each network was made up of several clusters, and keywords that were similar in content were grouped together. The size of the circles indicated keyword frequency, and the thickness of the lines indicated the strength of co-occurrence within and between clusters. For example, Fig. 13 consisted of 99 keywords, 5 clusters, and 2,791 links with a total link strength of 10,841. The red cluster with 30 keywords was the largest cluster represented by “Innovation”, the green cluster with 24 keywords was the second largest cluster represented by “Management”, the blue cluster with 17 keywords was the third-largest cluster represented by “Performance”, and the yellow and purple clusters with 14 keywords each were the fourth and fifth clusters represented by “Impact” and “Trust”, respectively. Moreover, the high interconnectedness of different clusters indicated strong relationships between different keywords. The interpretation of other regional co-occurrence networks (Figs. 14–18) was the same.