Corporate social responsibility in international business literature: results from text data mining of the Journal of International Business Studies

Corporate social responsibility has been an important theme in management at least since the 1960s. International business became a recognized subfield in management around the same time. Logically, there might have been much dialogue about corporate social responsibility in international business research and publication, yet previous reviews of the literature indicate relatively little such research. This study complements previous literature reviews by employing text data mining to analyze a sample of 1188 articles published from 2000 to 2018 in the Journal of International Business Studies (JIBS). Results show that from 2000 to 2018 only 35 CSR focused articles appeared. CSR research has increased over time, highly influenced by editorial specification of special issues. These documents can be grouped into seven CSR topics, with corruption and embeddedness being the most salient. Strategies are suggested for increasing research on CSR in international business.


Introduction
The purpose of this research is to assess the inclusion of corporate social responsibility (CSR) or social responsibility (SR) in articles published from 2000 to 2018 in the Journal of International Business Studies (JIBS). JIBS is the official publication of the Academy of International Business (AIB) has ISSN: 0047-2506 (print) and 1478-6990 (electronic). Secondarily, this study assesses whether CSR focused articles have increased over time, and the geographic bases of authors represented by articles having at least 10% CSR topic weight. Many definitions of CSR and ways of measuring it have been proposed. The definition used to provide a framework in this study is the one put forth by the International Standards Organization (ISO) in ISO 26000. Moratis (2016: 41) summarizes the dimensions included in this standard as the following: Environment, Social, Stakeholder, Societal obligation, Quality of life; Ethical conduct; Law abiding, Human rights, Transparency. The ISO uses Social Responsibility (SR) as its referent term. Following Moratis (2016), the term CSR is used in this study, because it is historically the term often used, and from a research methodology viewpoint the term CSR subsumes SR. Thus, this study's methodology identifies documents in using either SR or CSR in their text.
Most previous literature reviews have asserted a neglect of CSR in the international business literature (Bhattacharyya, 2020;Doh et al., 2010;Kolk, 2016;Egri & Ralston, 2008;Pisani et al., 2017). Defining international business and enumerating all its representative scholarly outlets is beyond the scope of this analysis, but those articles appearing in the Journal of International Business Studies (JIBS) are deemed to be a logical representation of the field's interests. However, this is not to suggest that there are no other meritorious journals in international business exist. Indeed, many other journals are also representative of the field, but JIBS is the first journal sponsored by the Academy of International Business and the only specifically international business journal listed in the Financial Times Index, widely used by universities as an indication of high-quality academic work (Saunders et al., 2011), and is widely recognized as the leading journal in international business (Tüselmann et al., 2016).
Looking at a broad range of journals in international business and related fields going back to 1985, Pisani et al. (2017: 591) conclude, "CSR research is far from being global and still emerging in 'mainstream' IB. " On the other hand, Buckley et al. (2017), present topics related to CSR as important for the field of international business, indeed, as one of the "grand challenges" for maintaining relevance and developing the field in the future. This review seeks to answer the question of whether CSR has been a significant research topic in the international business literature, or whether it is a relatively neglected area. This review differs from others in several ways. First, previous reviews have generally used a more qualitative methodology, identifying themes and concepts the researchers considered to be associated with CSR. For example, in the case of Pisani et al. (2017), topics include items that might be associated with CSR, but also might have little to do with CSR, e.g., expatriate management. The study reported here uses a more quantitative methodology, one that allows terms and topics to emerge from textual analysis itself rather than to be derived from existing concepts and paradigms. The advantage of using this approach is that it facilitates a relatively objective analysis compared to content analysis or keyword analysis since it enables results to be obtained somewhat free from preexisting, subjective paradigms. This technique has been used in analyzing the content of other academic journals (Guan et al., 2019;Kulkarni et al., 2014) as well as corporate reports (Parra et al., 2017).
However, text data mining analysis is subject to bias of its own. For example, the frequency of the word "environment" might suggest high interest in sustainability and other topics related to "green" issues. In the case of JIBS, this term generally applied to the business environment, e.g., competition, government regulation, etc.
Another contrast is that this study focuses on only one journal. Focusing on one leading journal allows us to see whether CSR has been represented at the center of the field of international business scholarship. An argument could be made that this focus is too exclusive, that an analysis should be as broad and deep as possible. However, many different topics can be identified in a wide and deep analysis including many meritorious journals, and filtering occurs insofar as certain topics attract leading researchers and become central to a field, which would naturally be the topics studied in its leading journal. Inclusion of a wide range of journals could amplify the noise surrounding any important topic of central interest to researchers, and might obscure the apparent coverage of topics like CSR which may or may not have attracted top researchers' attention.
Another aspect of this study distinguishes it from others. Previous studies have gone back many decades in their analyses. One could argue that certain classic articles will provide a foundation for thought on CSR for decades to come, e.g., Friedman, 1970, Carroll, 1991 However, any scholarly field that survives evolves and develops with new scholarship, especially a field where data sources have expanded as rapidly as is the case with CSR. Inclusion of early time periods may be interesting from a sociology of knowledge perspective (Kot, 2014), but may dilute the significance of the CSR topic in the international business literature of most interest to scholars during recent decades. Thus, this study is restricted to 2000-2018.The structure of this paper starts with a review of international business literature reviews which focus on the inclusion of SR and CSR. Next comes a description of the methodology (text data mining-latent semantic analysis, or TDM-LSA) along with a discussion of the more technical details of this approach. Then results are presented. The final section of the paper presents limitations of the study along with implications of the findings for the inclusion of CSR in international business research, along with strategies and resources that may be useful for bringing CSR more into the scholarship of international business.

Literature review on the inclusion of CSR as topic in international business
Several recent reviews have indicated little attention being paid to CSR in international business (Doh et al., 2010;Kolk, 2016;Egri & Ralston, 2008;Pisani et al., 2017). Inkpen and Beamish (1994) conducted an analysis of the first 25 years of research published in JIBS, and categorized the corpus into 24 subdisciplines, none containing the words "social responsibility" or CSR. The most relevant term they identified was business-government interaction, finding it the 11th most frequent of 24 categorizations. A total of 53 articles were identified in this business-government interaction category, out of a total of 1220 articles in all categories. Buckley (2002) did a comprehensive analysis of the grand questions of the international business field in its formative years as a discipline, and no topic relating to CSR appears in the existing list or in the recommended topics for future research. Griffith et al. (2008) examined scholarly work in six leading international business journals from 1996 to 2006. They identified multinational enterprise (MNE) citizenship, ethical issues in international business, public policy issues, and environmental issues as emerging themes, however, these themes together constituted only about 10% of the total emerging themes. Kolk and Van Tulder, looking at CSR and sustainable development in the international business literature, observed: "Sustainable development and CSR have only slowly been taken up as relevant concepts of study in mainstream management journals in general, and in International Business publications in particular… it is still remarkable that, except for specialized journals on business ethics and business and society where CSR has received most attention, more than 97% of the articles in main management journals over the past two decades have not referred to either CSR or sustainable development at all" (Kolk & Van Tulder, 2010: 120).
Kolk examined the international business literature inclusion of social responsibility issues over a 50-year period, identifying three subthemes: "the (green) environment; ethics, rights and responsibilities; poverty and (sustainable) development" (Kolk, 2016: 23). However, the three themes were not highly represented in the literature examined. Buckley, Doh, and Benische identified as a "grand challenge" the topic of "understanding how MNEs respond to greater pressures for social responsibility and sustainability in the global operations" (Buckley et al., 2017(Buckley et al., : 1046. It is interesting to compare the field of international business with international marketing, because a contrast can be seen. For example, Eteokleous et al. (2016), using LSA methods like those used here, found considerable attention paid to CSR themes in the international marketing literature, and Fatma and Rahman (2015) identified several CSR themes in the literature on consumers. It is also interesting to consider CSR coverage in business ethics specific journals, where many international topics have been studied, but with different and distinct theoretical approaches from international business. Some studies in business ethics or business and society journals compare CSR attitudes or practices in different countries, e.g., Chen and Bouvain (2009). Others look at international codes such as the U.S. Global Compact (Sethi & Schepers, 2014), or international monitoring such as that done by Transparency International (Wang & Rosenau, 2001). Sustainability has emerged as a unifying theme (Chabowski et al., 2011). Clearly, there has been considerable interest in CSR in areas related to international business.

Methodology
Text Data Mining (TDM) is built on the basic notion of the word count. Latent Semantic Analysis (LSA) is a mathematical method for inductively identifying themes based on word frequencies and word associations within articles in a whole corpus (all articles in the journal for the time selected). First, all words in all articles (often termed "documents") are downloaded. Then words are excluded that have no special significance e.g., "the", "thus. " Words appearing infrequently are also excluded, and terms containing a common stem are combined. The result is a matrix of words of interest with their frequency of appearance. Through filtering, terms are weighted by normalizing frequencies. This allows an assessment to be made of the relative importance of each term in relation to other terms and to identify connections between terms. For example, if the term "CSR" appears often, but mostly in connection with "human rights", this connection will be shown.
Weights are assigned to identify associations between terms (i.e., topics) or between documents. LSA is used to obtain Singular Vector Decompositions (SVDs) from the resulting matrix composed of documents and their associated weighted terms. The SVDs show the importance of terms in each document (Manning et al., 2008). These SVDs can then be rotated to model the behavior of the data and to facilitate interpretation as well as labeling (Evangelopoulos & Visinescu, 2012;. Finally, post-LSA analysis may include comparing and classifying documents using either cosine similarity, or by using hierarchical or non-hierarchical clustering methods.  make some recommendations on LSA extensions and argue that researchers should use non-hierarchical clustering techniques such as K-means (Hartigan & Wong, 1979;Jain, 2010) or the expectation-maximization algorithm (Do & Batzoglou, 2008) for document summarization. However, for purposes of obtaining mutually exclusive groupings, when the number of groupings is not known, hierarchical clustering is preferred (Wang et al., 2009). To summarize, the frequency of terms, and the relationship (or clustering) of terms into topics emerges from the data, rather than from assignment to categories determined by human coding. Figure 1 illustrates this methodology. The first step is Unsupervised Topic Discovery (UTD), which entails focusing on SVDs that include the term eigenvectors, placing special attention on factor loadings obtained from transposing matrices for terms, which represent term clusters/groupings (or topics). However, a single document may have high factor loadings in more than one SVD. In other words, one document may allude to more than one topic. Indeed, any scholarly article will Page 4 of 14 Paul and Parra Int J Corporate Soc Responsibility (2021) 6:12 probably contain several topics. The next question is how important this topic is for defining a document as a whole, and what other topics are alluded to in the document. Consequently, documents that discuss CSR topics are selected by calculating the total sum of the factor loadings per document (or total document weight) as well as the relative contribution made to that total by the CSR SVD. This is the second step shown in Fig. 1. The third step is to then rank all documents in the collection and obtain a sub-collection of CSR documents including each article for which the CSR SVD contributed at least 10% to total document weight. Using a 10% cutoff conformed to standards within the acceptable range of other studies using LSA, as well as to results obtained in previous reviews of CSR treatment in international business literature (Inkpen & Beamish, 1994). The fourth step is to take this sub-collection and apply UTD once again to obtain CSR topics, and to identify CSR sub-categories. These steps are illustrated in Fig. 1. The final step, illustrated in Fig. 2, is to apply divisive hierarchical clustering to obtain a taxonomy of CSR topics treated from 2000 to 2018 (Wang et al., 2009). This taxonomical representation reveals connections between topics included the analysis. For example, corruption and embeddedness (or institutional logic) are split out into a branch that represents a separate cluster, while all other topics are remain in a different branch .

Results
The application of TDM analysis on the 1188 JIBS articles (dating from 2000 to 2018) yields a total of 35 SVDs or topics. Table 1 provides a list of representative words for all topics. CSR appears as Topic 18. This listing is not a ranking, but rather an enumeration of topics appearing in JIBS, which are comprised of term clusters. The summary term may be viewed as a term of convenience and carries no special significance other than to enable communication that differentiates between themes typically treated in JIBS. By focusing on what the analysis has labeled Topic 18 one can see which articles have fallen within this category, what sub-themes have appeared (again identified by term clusters within Topic 18). Table 1 is a list of topics identified by TDM.
The next step is to identify the articles in the collection of 35 articles in the CSR topic cluster (Topic 18) that have at least a 10% weight on CSR SVD, to eliminate items such as book reviews and cases where, for example, the term "environment" refers to the general business environment rather than to some environmental issue. Also, two relevant articles that came corrupted from the publisher were manually included. Finally, the resulting articles are analyzed according to the article's CSR SVD contribution using 10% as the basis for inclusion of the article in the sub-collection on which divisive hierarchical clustering is used. articles identified, arranged by subject, showing authors with the geographic location of their universities. The next consideration is the frequency of CSR related materials over time. Figure 3 provides time series data. Two tendencies merit notice. First, there is an increasing incidence of CSR related topics over time. This is consistent with the more general analysis provided by Pisani et al. (2017).
However, further examination shows that this incidence is heavily influenced by editorially determined special issues of JIBS. Note three CSR peaks in Fig. 3. The 2006 peak corresponds to a special issue on corruption. The 2012 peak is associated with a special issue on cultural distance. Finally, in 2016 and 2017 there were special issues on internationalization in the information age and on international corporate governance.

Discussion
Results show the proportion of CSR related articles appearing in JIBS is less than 5%, with robust statistical significance (p = .01). Editorial direction using special issues appears to have been a significant influence on the amount of CSR focused material appearing in JIBS. In particular, four of the nine articles in the legitimation category were published in the 2016 special issue on internationalization in information age.
Furthermore, five of the six articles in the corruption category were published in the 2006 special issue on the topic. Foreignness and embeddedness each started as topics in 2012, coinciding with special issues on cultural distance and internationalization. The role of editorial direction and selection of appropriate topics for special issues has been critical for the inclusion of CSR material in this journal.
The geographical location of the CSR work that has appeared in JIBS during the period studied was also considered. Authors stating affiliations with universities from North America (the United States, Canada and Mexico) have contributed most CSR articles, with European authors contributing a substantial proportion, and a smaller proportion from Asia, mostly in co-authorships with North American or European authors. No authors publishing on CSR in JIBS during this period provided African, Middle Eastern, or South American university affiliations. This analysis indicates that JIBS has included relatively little CSR material during the period studied, 2000-2018. The articles which have featured CSR topics have tended to cluster around special issues where topics of specific CSR interest were solicited. This finding affirms previous studies but remains rather curious given that so many pressing CSR topics involve globalization, multinational companies, and governance issues of high international relevance.

Limitations and suggestions for further research
Although the journal used in this analysis is an obvious place to assess the significance of a topic in international business, many other journals could be considered. Perhaps CSR is a topic where interest develops at the margin of the field and flows from newcomers or outsiders to its central thought leaders, and from more marginal journals to the mainstream. Also, new topics may be in the pipeline, and even dissertations and conference presentations may include more CSR research than that which has passed through the entire review process and been published. However, the possibility that established academic networks and research teams may be resistant to new research topics and difficult for researchers interested in new topics to penetrate should be considered (Newburry & Gonzalez-Perez, 2015;Tilt, 2016). Another limitation is the use of TDM itself, a somewhat mechanical and arbitrary type of analysis that may miss critical, thoughtful, imaginative categorization or even subtleties of terminology. The main question that arises from this study is why so little CSR material would be contained in the leading journal in international business. Prior to the period studied here, one might suspect that theory and data on fsas, Rugman, Verbeke, home region, triad, regional, diversification, intra-regional Regional strategy      CSR were not very well established, and this would limit publication of CSR research in a leading journal which would naturally require rigorous quantitative analysis as well as substantive theoretical contributions. However, CSR is now well established in the management literature with many data sources. Either a lack of familiarity with available data or a pragmatic calculation that CSR is not a career enhancing field of study for international business researchers might have influenced scholars not to focus on CSR.

Conclusions
A continuation of intentional efforts on the part of international business gatekeepers such as journal editors and accreditation organizations may be necessary if CSR topics are to gain more attention in international business. In addition, since quantitative studies dominate articles published in JIBS and other prestigious management journals, it might be helpful to promote awareness of the databases available on CSR topics. Since corruption is already an established area in international business research, the various indices on bribery and perceptions of corruption produced by Transparency International are highly relevant. Other publicly available data sources for international companies on a variety of CSR dimensions are available through the Global Reporting Initiative, a source which also collaborates with more than 150 other standards setting organizations in various areas of CSR. The International Standards Organization's Standard 26,000 formed the basis for the definition of CSR that guided this study, and companies will be providing data structured around this standard (Zinenko et al., 2015), especially since it guides companies in compliance with the OECD Guidelines for Multinational Enterprises. Whereas early articles on CSR had very little data to use for evaluating company CSR performance in connection with financial performance, the area has become more developed now, with many sources available, e.g., the MSCI KLD 400 Social Index, the MSCI USA ESG Select Index, the Calvert Index, and others that focus on aspects of CSR such as human rights and animal rights. The Global Impact Investor Network publishes an annual report including data from hundreds of socially screened funds. Many reviews of databases and indices may be found in the information technology and accounting literature, as well as the literature on socially responsible investing, now known as impact investing. Consequently,