Research streams in corporate social responsibility literature: a bibliometric analysis

Corporate social responsibility (CSR) research is heterogeneous and still fragmented. In its interdisciplinary setting, researchers focus on different CSR aspects, secondary concepts and themes. The lack of a unifying paradigm indicates that the CSR literature should be summarized and classified. This study’s systematic overview of CSR research provides such a classification. Previous conceptualizations of CSR research mapped the literature from individual authors’ perspective, rendering different and partly inconsistent classifications. Using bibliometric methods, this paper offers an objective overview. We analyze the references of 1902 CSR journal articles by bibliometric techniques as (co-)citation, core/periphery, factor, and network analyses. By doing that, we provide an overview of the CSR research core, identify different research streams, describe their main publications’ topics and recent developments, and make suggestions to inspire future research in and across research streams. Our results show the increased relevance of formerly niche research streams, such as employee-oriented CSR research, or research on consumer skepticism. Among others, process-oriented and micro-level research, critical approaches, and mergers between themes from various research fields offer a wide scope for further research.


Introduction
Corporate social responsibility (CSR) issues have become a core part of corporate management practice (Crane et al. 2015), with multinational companies such as Coca-Cola, Pfizer, Disney and Wells Fargo releasing annual CSR reports that outline their efforts and achievements. At the same time, scientific interest in CSR and its literature's breadth and depth continues to grow and is turning out to be quite widespread (e.g. Crane et al. 2018). However, the heterogeneous and fragmented research field (Aguinis and Glavas 2012) still lacks a unifying theoretical lens, shared construct clarity, and applied methodology (Gond et al. 2020;Mitnick 2019;Wood and Logsdon 2019).
The variety in concept definitions, approaches, and perspectives is already evident from a nomological perspective: The 'corporate' component stresses CSR's business-centered perspective, the term 'social' addresses its societal perspective, and 'responsibility' emphasizes its ethical side. The diverse CSR research field can be grouped into instrumental, descriptive, and normative research (Mitnick 2019;Wood and Logsdon 2019). Contesting perspectives such as "instrumental/economic CSR" and "injunctive/social CSR" (Mitnick et al. 2021, p. 625) are based on contrasting paradigms. Depending on the applied research paradigm, this results in different implications regarding CSR motives, stakeholders to be favored and aspired outcomes. For example, taking a business-centric perspective, social activities are viewed as more desirable if shareholders react positively to them (Wang et al. 2016), but can be problematic from an ethical point of view.
Fragmentation and broadening the research field have led researchers away from the objective of identifying a unifying paradigm (Gond et al. 2020). Recent research studies and reviews cover single CSR topics, specific theoretical backgrounds or secondary concepts and subfields (Gond et al. 2020;Ji et al. 2021). Examples are reviews dedicated to CSR communication (e.g. Crane and Glozer 2016;Verk et al. 2021), CSR knowledge in communication literature (Ji et al. 2021), political CSR (e.g. Frynas and Stephens 2015) and its micro-foundations (e.g. Gond et al. 2017). Recent research also links CSR to related concepts such as sustainability (e.g. Ketprapakorn 2019), and this is sometimes distinguished from CSR (e.g. Bansal and Song 2017) but is also used as a synonym (e.g. Cantele and Zardini 2018).
Due to this heterogeneity, researchers characterize the literature as a low-paradigm research field (Gond et al. 2020;Mitnick 2019;Wood and Logsdon 2019) and as "'essentially contested'" (Mitnick et al. 2021, p. 624). They view CSR as an umbrella term (Frynas and Yamahaki 2016;Scherer and Palazzo 2007) and accept the heterogeneity and controversies. We also value research heterogeneity, but argue that there should be attempts to better integrate the various research streams (Ji et al. 2021) to facilitate dialogue and cross-fertilization (Mitnick et al. 2021).
Against this background, we strive to enrich the CSR research field by a broad, systematic review that provides a neutral perspective on the discipline as well as its development over time. Thereby this study complements narrative reviews that are inherently more restricted in their scope as well as more subjective in nature. We aim to identify the CSR literature's overall research structure and connections across subfields (Ji et al. 2021). We describe research streams based on bibliometric criteria that are not clearly discernable (Ma et al. 2012), making the "invisible college network" (Ji et al. 2021) visible. Going beyond qualitative literature reviews, we employ quantitative empirical bibliometric methods to ensure a more objective review process (Kuntner and Teichert 2016;Zupic and Čater 2015). By conducting citation, co-citation, core/periphery, factor, and network analyses, we contribute to the development of the research field from a methodological point of view. We also analyze how the research streams evolved over time and identify recent trends. Finally, we make suggestions that could inspire future research in and across CSR research streams.

Categorizations of CSR literature
Publications that describe the field more holistically have identified different perspectives on CSR. Table 1 provides an overview of these diverse categories. Looking at the categorizations, it becomes evident that there is a lack of a unifying paradigm (Gond et al. 2020), as the research field is mapped in different ways that are based on different underlying views on CSR. The heterogeneity starts with an interpretation of what should be included in CSR research. For example, Dahlsrud (2008) extends the three CSR components of corporate, social, and responsible to five dimensions, namely the environmental, social, economic, stakeholder, and voluntariness dimensions.
The CSR categories differ largely by the categorization bases used to classify CSR research. They are grouped according to organizations' social roles (Klonoski 1991), key approaches (Windsor 2006), theoretical groups (e.g. Garriga and Melé 2004;Secchi 2007), theories and concepts (Frynas and Yamahaki 2016;Ji et al. 2021) or paradigmatic frameworks (Gond and Matten 2007;Scherer and Palazzo 2007). Even classifications applying the same categorization basis come to different conclusions. For example, Garriga and Melé (2004) as well as Secchi (2007) categorize CSR research by grouping theories behind CSR research. While the former research team divides CSR research into four groups that refer to instrumental, political, integrative, or ethical theories (Garriga andMelé 2004), Secchi (2007) differentiates between CSR research based on utilitarian, managerial, and relational theories.
Other inconsistencies result from allocating research streams differently into the various categories. This is exemplified by different classifications of the stakeholder theory: The early work of Garriga and Melé (2004) categorizes stakeholder management as an integrative theory, which separates this CSR research stream from works related to corporate citizenship and social contract theory (the latter being part of their political group). In contrast, Secchi (2007) assigns stakeholder theory to the relational theories along with corporate global citizenship and social contract theory. Furthermore, the granularity and contents of identified categories differ. Whereas Gond and Matten (2007) distinguish four paradigms in the CSR research field based on Burrell and Morgan (1979), Scherer and Palazzo (2007) summarize positivist and  (1) Category (2) Category (3) Category (4) Category (5) Category (6) Category (7) Klonoski ( Research streams in corporate social responsibility… post-positivist schools of thought and propose a politicized CSR conceptualization based on Habermasian philosophy. Previous categorizations often did not base on existing publications, but on subjectively desirable research streams, such as the yet unexplored category 'Habermas 2 CSR' (Scherer and Palazzo 2007). The exceptions are previous bibliometric, publication, and citation analyses (De Bakker et al. 2005;Lockett et al. 2006), which predominantly used only counting methods from the multitude of quantitative bibliometric methods and therefore require a methodological extension. Previous bibliometric studies that used sophisticated quantitative methods do not refer to individual research streams of CSR literature, but to the entire business ethics literature (Calabretta et al. 2011;Ma 2009;Ma et al. 2012). Some of these studies focus on concepts and theories in a specific subdiscipline, such as CSR research in communication literature (Ji et al. 2021), while we analyze all research streams of the CSR literature. This is particularly important, as we aim to describe the entire CSR research structure across these subfields (Ji et al. 2021), attempting to uncover the invisible network (Ji et al. 2021;Ma et al. 2012). Targeting the entire CSR research field enables us to provide insights about the actual roles that specific research streams play in the overall CSR research and to reveal additional research streams that may deserve more attention.

Data generation
Our analysis included several research steps. First, we generated data by searching for CSR-related articles in the Social Sciences Citation Index (SSCI). This database is most commonly and frequently used for bibliometric analyses in management and organizational studies (Zupic and Čater 2015) and recommended for its comprehensiveness, especially for our subject field (e.g. Ma et al. 2012). We had access to the index since 1994, therefore we searched for publications that appeared between 1994 and 2020. We chose 'corporate social responsibility' as a search term and added 'AND organization*' to establish the connection to organizational research, and looked for these terms in titles, abstracts, and/or author keywords. Without this organizational focus, for example, topics from technology science (e.g. Mills and De Paoli 2018) would have been included. By using truncation, we included terms from an employee perspective, such as organizational justice, organizational citizenship behavior (Rupp et al. 2013), or organizational commitment (Brammer et al. 2007).
To ensure a high scientific quality, we focused on double-blind reviewed articles written in English, as is common practice in bibliometric studies (e.g. Wörfel 2019). Our data included 1902 articles citing 146,220 references. While we restricted the search to reviewed articles, their references included numerous other types of publications such as books (e.g. Bowen 1953;Freeman 1984) or Friedman's important New York Times Magazine article published in 1970, as we will show in the results of the co-citation analysis.
We analyzed the citations as well as the co-citations of cited references to derive information about the CSR literature's intellectual structure (White and Griffith 1981). To overcome the 'long tail end' of the citation frequencies' distribution, we first limited our analysis of referenced documents to the most prominent articles. We selected references cited at least 30 times, which resulted in 291 key references used for co-citation analysis. Later, we also analyzed all citing articles up to 2020. This combination of analyses allowed us to overcome the inherent problem of co-citation analysis that recent papers have a lower chance of being co-cited and to link current research to identified long-term research streams and to reveal their current trends.

Data analysis
Following a descriptive analysis of citing papers, our bibliometric study bases on a multistage analysis of the cited documents as their intellectual basis. A co-citation analysis quantifies how often publications are cited together (Pilkington and Teichert 2006). While a high number of citations are associated with a publication's importance in the research field, a high number of co-citations of two publications indicate that they share similar ideas and belong to an overarching research stream (Small 1977). Co-citations can therefore identify key publications as well as distinct streams. We build a co-citation matrix of cited references and analyze them in multiple steps. First, we analyze the entire co-citation network, differentiating core publications from publications in the periphery. We then describe the relative contribution of individual publications by statistical measures on a single node level. Research streams are been identified by means of factor analysis and their temporal development is analyzed.
To conduct a core/periphery analysis based on the co-citation matrix, we used the software UCINET 6.665 (Borgatti et al. 2002). The core forms a cohesive structure with publications that are central to the topic and strongly related to each other. Publications in the periphery are only loosely connected and they relate to specific research themes (Borgatti and Everett 2000). Furthermore, we calculated statistical measures to indicate the research core's overarching structure, called degree centrality, eigenvector centrality, and betweenness centrality. We used Freeman's degree to measure the total number of a publication's co-citations in the database (Freeman 1978). The eigenvector centrality represents an extension of the degree centrality by considering the importance of articles to which a publication is related (Bihari and Pandia 2015). The eigenvector score, which also underlies Google's page ranking (Langville and Meyer 2006), reflects various publications' centrality in the entire research field better than unweighted measures do. Finally, the betweenness centrality reveals the extent to which a publication acts as a 'bridge' between otherwise unconnected publications in a network.
Factor analyses were conducted with SPSS to identify different research streams in the core as well as in the periphery (Kuntner and Teichert 2016). Statistical measures infer single publications' relevance in each research stream. The factor loadings indicate the publications' fit to the research streams' context, while the factor scores reversely indicate how much the publication influences the respective research stream (Teichert and Shehu 2010). Finally, all citing articles up to 2020 are linked to the research streams based on their citation patterns. The number of citations an individual research stream receives per year (publication year of the citing articles) is set in relation to the total number of citations (Kuntner and Teichert 2016). This final analysis reveals the evolution of the research streams and helps identify recent trends in CSR research. Table 2 provides a descriptive overview of our initial database of 1902 citing articles. It lists the top 10 journals as well as the top 10 most prominent articles (measured by their received number of citations).

Descriptive overview of the most prominent citing articles
While the 1902 articles were published in over 400 journals, most appear in a few key journals. Nearly half of all identified articles were published in the top 10 A view of the 10 most often referenced publications provides a first glimpse of important topics in CSR research. These articles include overviews on CSR drivers (Aguilera et al. 2007) and on CSR's link to sustainability (Van Marrewijk 2003). Other articles focus on different stakeholder groups such as customers (e.g. Lichtenstein et al. 2004;Sen and Bhattacharya 2001) or employees (e.g. Brammer et al. 2007), as well as on marketing (e.g. Maignan and Ferrell 2004). Links between CSR, financial performance, or strategic benefits are also analyzed (e.g. Burke and Logsdon 1996;Surroca et al. 2010), highlighting the relevance of strategic approaches to CSR. Institutional CSR research is been investigated by a study of CSR in different institutional contexts (Doh and Guay 2006) as well as by an exploration of institutional CSR drivers (Campbell 2007). In sum, various topics are been identified by looking at top journals and single top publications. Yet it is unclear whether these topics are representative of all CSR research. The following analyses therefore broaden the scope of investigation and assess schools of thought in the research field by a complementary co-citation analysis of cited references.

Core articles
A core/periphery analysis identified the shared basis of cited references in the research field. It separated 43 publications in the core from 248 publications in the periphery. First, we zoomed into the analysis of core works to give an overview of the foundation of CSR research. Table 3 lists these key publications and provides a summary of derived statistics, showing their degree centrality, eigenvector centrality, and betweenness centrality.
Unsurprisingly, foundational conceptual models, meta-analyses and reviews of CSR assume focal positions in this list of core publications. They are cited most often, rendering the highest degree centrality as well as eigenvector centrality values. The publications by McWilliams and Siegel (2001) and Orlitzky et al. (2003) are cited 5773 and 5254 times respectively, and therefore stand out in terms of centrality. The latter study is a meta-analysis of 52 studies showing many ties to articles throughout the core CSR literature network. The stakeholder framework by Clarkson (1995) received fewer citations. A comparably low eigenvector centrality (ranked 25) characterizes the article, indicating that it had a more focused impact.
The table's findings also illustrate the importance of the CSR business case in the core literature. Articles that analyze the relation between CSR or rather corporate social performance and corporate financial performance (e.g. Margolis and Walsh Research streams in corporate social responsibility… The core contains overviews of definitions (Carroll 1999), theories (e.g. Garriga and Melé 2004), and drivers of socially responsible behavior (e.g. Aguilera et al. 2007;Aguinis and Glavas 2012;Campbell 2007). It also includes fundamental corporate social performance models (Carroll 1979(Carroll , 1991Wood 1991). According to these models, corporations do not only have economic and legal responsibilities, but are also responsible in an ethical and discretionary sense. As shown by the most relevant citing articles, the descriptive statistics of the core references confirm the relevance of stakeholder theory in CSR research (e.g. Donaldson and Preston 1995;Freeman 1984). Unsurprisingly, the Freeman (1984) book on stakeholder management is among the publications with the highest centrality values. The core also includes a few fundamental publications on institutional CSR perspectives (e.g. Campbell 2007;DiMaggio and Powell 1983;Matten and Moon 2008). We therefore confirm the descriptive results of the citing articles, as well as earlier findings that demonstrate the importance of stakeholder and institutional approaches in CSR theorization (Frynas and Yamahaki 2016). Over time, scientists have concluded that it would be fruitful to analyze how CSR affects specific stakeholder groups (Peterson 2004). Our findings also reflect this. The core includes many CSR publications connected to employees (e.g. Brammer et al. 2007;Carmeli et al. 2007;Kim et al. 2010;Rupp et al. 2006Rupp et al. , 2013) and a few articles associated with consumers (e.g. Brown and Dacin 1997;Luo and Bhattacharya 2006). It should be noted that, compared to general fundamental CSR conceptualizations (e.g. Carroll 1979Carroll , 1991 and reviews (e.g. Carroll 1999), more specific and recent studies on CSR and employees (e.g. Rupp et al. 2013) exhibit low values of betweenness centrality. Therefore they do not act as a 'bridge' between otherwise unconnected articles.

Research streams
The single articles' descriptive statistics provided a first impression of the CSR research. In another step, we executed factor analyses to investigate the underlying patterns of heterogeneity and to gain insights about different research streams. A factor analysis of publications of the periphery complemented one of core publications, allowing a cross-validation of the findings. In the core, we identified five factors that explain 55.75% of the variance in the co-citation patterns after eliminating two outlier publications based on low communality values. As expected, the factor analysis of the periphery showed a higher heterogeneity, with several articles excluded due to low communality values and seven factors explaining 42.35% of the variance. A content analysis confirmed that the periphery's topics relate to the core's five research streams. These were jointly named 'CSR and employees,' the 'CSR  ' 'CSR and marketing,' and 'CSR and institutional theory.' Two further research streams in the periphery, 'CSR and consumer skepticism' and 'environmental CSR as source of competitive advantage,' do not appear in the core. To provide an overview of all research streams, Table 4 presents the top five most representative and influential publications in each research stream.
The research stream with the highest explained variance covers the topic 'CSR and employees,' highlighting the prominence of this stakeholder perspective in CSR research. The publication of Brammer et al. (2007), also among the top citing articles, is recognized as the most representative article (FL = 0.88) as well as the most influential one (FS = 1.97) in this factor. In this and related articles, the authors use social identity theory as an approach to investigate the perceived CSR influence on employees' organizational commitment (Brammer et al. 2007;Kim et al. 2010;Turker 2009b). Another important aspect covered in this research is the combination of employee-oriented CSR and organizational justice (Rupp et al. 2006(Rupp et al. , 2013. To summarize, this research stream addresses the outcomes of CSR practices related to employees, which include CSR's influence on concepts like organizational identification (e.g. Jones 2010; Kim et al. 2010) or organizational citizenship behavior (e.g. Rupp et al. 2013). Articles in the periphery also address CSR's micro-foundation in combining employees' perceived CSR judgments. This includes CSR's influence on commitment (e.g. Farooq et al. 2014;Glavas and Kelley 2014), organizational identification (e.g. De Roeck et al. 2014), organizational justice (Rupp and Mallory 2015), job satisfaction (Glavas and Kelley 2014), or CSR and job performance (Vlachos et al. 2014). More interpretative qualitative studies are needed for a better understanding of employees' views on CSR. It appears that this research stream has paid limited attention to employees' influence on CSR implementation.
The second research stream deals with the 'CSR business case.' McWilliams and Siegel's (2001) theory of the firm perspective and Donaldson and Preston's (1995) publication on stakeholder theory provide important theoretical underpinnings for this research. Owing to inconsistent findings on the relationship between corporate social and financial performance, many researchers have analyzed this relationship (e.g. McWilliams and Siegel 2000;Waddock and Graves 1997). Meta-analyses mainly find a positive link (e.g. Orlitzky et al. 2003). Specific empirical works locate in the courses' periphery: Hillman and Keim (2001) as well as Berman et al. (1999) analyzed the influence of concepts like stakeholder management and social issue participation on shareholder value. Other influential articles investigated the economic benefits of CSR activities from a risk management perspective (Godfrey et al. 2009). To summarize, this research stream's core provides the theoretical basis, whereas its periphery comprises empirical studies on the link between social performance and financial performance measures (e.g. Barnett and Salomon 2006;Griffin and Mahon 1997). To look more deeply and from a new perspective at the effects of CSR, this research stream would benefit by process studies and alternative research approaches.
The third research stream, 'CSR conceptualizations,' covers conceptual research on the role of business in society. The two most representative publications written by Carroll (1979Carroll ( , 1991 describe the different CSR dimensions by conceptual models such as the CSR pyramid (Carroll 1991). In such frameworks, researchers

3
Research streams in corporate social responsibility… summarized and refined corporations' different responsibilities and define the corporate social performance concept (Carroll 1979(Carroll , 1991Wood 1991). The research stream's core includes reviews of CSR definitions (Carroll 1999) and a classification of diverse theories in CSR research (Garriga and Melé 2004). The research stream in the periphery contains publications on the political role organizations play Palazzo 2007, 2011), according to which corporations are active in administering citizenship rights (Matten and Crane 2005). Altogether, works in this discourse highlight the contemporary importance of political theories for CSR research, as with "Habermas's theory of democracy" Palazzo 2007, p. 1096). Publications on political CSR in this research stream are however mainly theory-driven and conceptual (e.g. Palazzo 2007, 2011). Relatively recent political CSR conceptualizations therefore offer much potential for empirical research.
The fourth research stream refers to 'CSR and marketing.' It contains empirical articles that quantitatively investigate market responses and the conditions for successful CSR initiatives (Sen and Bhattacharya 2001). Highly influential and representative articles use quantitative methods to investigate concepts like customers' product evaluations (Brown and Dacin 1997), their product purchase intentions and company evaluations (Sen and Bhattacharya 2001), customer satisfaction, and their influence on firms' market value (Luo and Bhattacharya 2006). Publications in the periphery further investigate moderating and mediating effects by zooming into concepts such as customer-corporate identification (Bhattacharya and Sen 2003;Lichtenstein et al. 2004) or trust in explaining how CSR initiatives influence consumer reactions (Morgan and Hunt 1994;Vlachos et al. 2009). Further research investigates consumers' assessment of corporates' motives for practicing CSR (Sen et al. 2006;Vlachos et al. 2009). The findings highlight the positive effects of valuedriven reasons for implementing CSR in marketing campaigns (Vlachos et al. 2009). This topic is predominantly investigated from an uncritical perspective, assuming a positive relationship between CSR initiatives and customer reactions. Critical analyses might enrich this research stream by providing new perspectives.
The fifth research stream deals with 'CSR and institutional theory.' The most representative (FL = 0.71) article analyzes why and how CSR practices vary between the USA and Europe, highlighting the need for corporations to adapt to environmental changes (Matten and Moon 2008). Whereas Campbell (2007) focuses on institutional drivers to behave in a socially responsible way, Aguilera et al. (2007) show multi-level factors that put pressure on organizations to implement CSR practices. Reflections on institutional isomorphic processes (DiMaggio and Powell 1983) serve as a theoretical framework of this debate.
The periphery also contains basic research publications on institutional processes, for example to gain legitimacy (Meyer and Rowan 1977;Oliver 1991;Scott 1995). Organizations often face incompatible institutional logics. How they respond to contradictory institutional demands is one thematic aspect addressed in the periphery (e.g. Greenwood et al. 2011). Specifically, researchers examine the management of complexity in multinational corporations and their impacts on organizational legitimacy (e.g. Kostova and Zaheer 1999). Critical perspectives also enrich this research stream, such as the criticism that multinationals increasingly practice strategic forms of CSR at the expense of a more social and stakeholder-oriented understanding of CSR (Bondy et al. 2012). Apart from Asian exceptions (Chapple and Moon 2005), the focus is on developed economies, for example comparisons between the USA and Europe (e.g. Doh and Guay 2006;Matten and Moon 2008).
In the periphery, we identified two more research streams, one of which describes 'CSR and consumer skepticism.' Due to corporate misconduct, many consumers have become skeptical of whether CSR initiatives are sincere (Skarmeas and Leonidou 2013). The highly representative and influential articles show that CSR skepticism is determined by consumers' attribution of companies' motives for practicing CSR (e.g. Forehand and Grier 2003;Yoon et al. 2006). Doubting altruism, many consumers recognize opportunistic motives which they then evaluate negatively (Skarmeas and Leonidou 2013). Consequently, researchers highlight the importance of communication strategies that organizations use to mitigate negative consequences (e.g. Du et al. 2010;Morsing et al. 2008). Studies mainly base on attribution theory (Jones and Davis 1965;Kelley 1973). From a paradigmatic perspective, they are predominantly functionalist, with a focus on effectiveness (e.g. Yoon et al. 2006). In the future, alternative theories such as critical management studies might enrich this perspective.
We also identified a research stream called 'environmental CSR as a source of competitive advantage' in the periphery. Researchers validate relations, such as between environmental CSR, resources and capabilities, and firms' competitive benefits and performances (e.g. Russo and Fouts 1997;Sharma and Vredenburg 1998). To substantiate these links, they use the resource-based view (Barney 1991;Wernerfelt 1984) and the natural-resource-based view (Hart 1995) as their theoretical foundation. Interpreting CSR aspects from a resource perspective is especially important for small and medium-sized enterprises, as they should not treat CSR initiatives solely as costly external effects (Jenkins 2006). Future research might add dynamic and network perspectives to investigate how competitive advantages are been created through CSR initiatives.

Development of research streams over time
Next, we show how the discussion along the different research streams developed over the last 15 years. While the results in the previous section base on cited references, the following analysis bases on the citing articles assigned to the relevant research streams. The citing articles' greater actuality (compared to the cited articles) allowed us to show current trends in the research streams' development. Figure 1 visualizes the research streams' evolution in the core.
Analogously, Fig. 2 shows the trends of the research streams in the periphery. The figures reveal highly dynamic patterns in the evolution of CSR literature. As expected, there are bigger changes in the periphery than in the core.
Research works on 'CSR and employees' gained prominence and are currently by far the most cited publications. This is in line with previous calls for more microlevel CSR research (Aguinis and Glavas 2012;Frynas and Yamahaki 2016), which appears to have stimulated the increase (Gond et al. 2020). It also indicates a shift away from a business point of perspective on CSR to a broader, stakeholder-oriented perspective in CSR research.
References to the 'CSR business case' in research are particularly declining, especially in the periphery. Until 2008, CSR business case research was the most cited research stream in the periphery, but this prevalence decreased significantly in 2009. In the core, CSR business case literature almost always remains the second most referenced research stream until 2014 and again in 2020. Even so, this research stream lost prevalence while employee-oriented CSR research gained importance.
The research stream 'CSR conceptualizations' shows a similar declining trend over the entire period, hinting to a phase of saturation in derived conceptual  underpinnings. Until 2015, 'CSR conceptualizations' was the core's most referenced research stream. In 2016, references to employee-oriented CSR publications overtook CSR conceptualizations. However, the evolution of political CSR conceptualizations in the periphery differs: Citations of this political subdiscourse increased significantly in 2008. Further peaks occurred in 2011 and 2013 after the release of novel political CSR publications Palazzo 2007, 2011). Thereafter, its importance decreased. Notably, empirical political CSR research might again revitalize the discourse.
Another stream of declining relevance is 'CSR and marketing.' Starting in 2010, this stream became the least often referenced stream out of the five core research streams. In the periphery, marketing-oriented CSR research was more prevalent, peaking in 2008 and for a short period in between 2012 and 2014. Even it decreased afterwards, it has a higher prevalence in 2020 compared to 2006. Nevertheless, we can conclude that the broader discourse on the 'CSR business case' as well as the specific discourse on 'CSR and marketing' became less important over time, deemphasizing the corporate or business view in the CSR discourse.
The number of core publication citations on 'CSR and institutional theory' increased from 2007 to 2009. However, it plays only a moderate role in the CSR research core over time. In contrast, institutional CSR research has high prevalence in the periphery. Although these citations also decreased after 2010, the research stream almost always remained the most important until 2016.
Publications on 'CSR and consumer skepticism' in the periphery gained momentum, especially from 2016 to 2020. Due to rising corporate misconduct (Skarmeas and Leonidou 2013), research on consumer skepticism increasingly replaced the mainstream discourse on positive consumer reactions to CSR initiatives. The growing trend let us assume that this research stream might become more central in the entire research field.
In contrast, the interest in the research stream 'environmental CSR as a source of competitive advantage' has declined. Since 2016, this research stream has received the lowest number of citations of all research streams in the periphery. A reason might be that important publications root back to the 1990s (e.g. Hart 1995;Sharma and Vredenburg 1998) such that a state of saturation was achieved. The discourse might regain attention in the future, considering the widespread diffusion of new consumer movements that link consumer skepticism with environmental aspects and climate change.

Discussion and conclusion
This paper provides an in-depth analysis of the research streams in the CSR literature. It offers guidance on the CSR literature to new scientists and inspiration for future research to experienced ones. Previous categorizations have led to divergent classifications of CSR research based on researchers' perspective. Our study provides a more objective overview of current CSR research streams by using sophisticated quantitative bibliometric methods (Kuntner and Teichert 2016). One objective was to make the invisible CSR knowledge network visible (Ji et al. 2021;Ma et al. Research streams in corporate social responsibility… 2012). The analyses reveal five research streams constituting the core of the research field as well as its periphery. These research streams are 'CSR and employees,' the 'CSR business case, ' 'CSR conceptualizations,' 'CSR and marketing,' and 'CSR and institutional theory.' In the periphery we find two more research streams dealing with 'CSR and consumer skepticism' and 'environmental CSR as source of competitive advantage.' A further aim was to show how the research streams evolved over time. We identified a high degree of dynamics. The research streams on 'CSR and employees,' and 'CSR and consumer skepticism' increased most over time. Comparing 2006 with 2020, the research stream 'CSR and institutional theory' also increased in the core. In the periphery, the latter research stream shows an increasing trend only from 2019 to 2020. Political CSR research also increased in the periphery, when comparing 2006 with 2020. In contrast, the relevance of 'CSR conceptualizations' decreased in the core, just as the core research stream 'CSR and marketing'. The importance of the research streams on the 'CSR business case,' and 'environmental CSR as a source of competitive advantage' also declined. Generally, our results show an overall decreasing significance of business-centered CSR research.
To summarize, Table 5 gives an overview of the different main foci and main methodologies of each research stream. According to our last research objective, it also provides inspirations for future research.
Although the employee-oriented research streams are already prominent, we found major gaps that can inspire future research. The most cited studies examine the positive link of perceived CSR on desirable attitudes and work behavior. However, in the context of corporate misconduct (Lee et al. 2018), it would be interesting to study irresponsible corporate behavior, its impact on employees, and organizational strategies to prevent such misconduct. While there are initial studies (e.g. Lee et al. 2013), this focus area has not yet established itself as a school of thought in CSR research and calls for further research.
In addition, instead of just focusing on how CSR influences employees, future research should also investigate how employees can influence the implementation of CSR practices. While this perspective has been taken sporadically (Bolton et al. 2011), it has not yet established itself as part of the employee-related research streams. A strategy-as-practice approach (Whittington 1996) might be used for such an investigation. Other fruitful sample theories might be the sensemaking theory (Weick et al. 2005) or a critical tension-centered approach (e.g. Trethewey and Ashcraft 2004) for analyzing how employees make sense of CSR and how they derive meaningfulness of work (Aguinis and Glavas 2019;Mitra and Buzzanell 2017).
In general, future research should pay more attention to the role of policies in the implementation of CSR practices. Method-wise, qualitative research methods would also enrich employee-related CSR research as most extant studies are quantitative with a one-sided view of the impact of perceived CSR practices on employees (e.g. Brammer et al. 2007).
Research on the 'CSR business case' can be also enriched by using a strategyas-practice approach (Whittington 1996). Khan's (2018) process-oriented study, based on sensemaking theory (Weick et al. 2005), might guide future research investigating the link between CSR strategy formulation and implementation, firm Research streams in corporate social responsibility… performance, and reputation with a focus on managers' cognitions. Additionally, in contrast to functionalist approaches that deny potential tensions between the ethical and the business case for CSR, more critical studies might focus on the trade-offs between social and economic goals. Paradox approaches would be suitable for this purpose. This is addressed by Hoffmann (2018), who shows how paradoxes of CSR are constituted or denied in research discourses. Contingency approaches might also enrich future CSR business case research that go beyond the convergence model to challenge the perfect alignment between ethical behavior and economic rewards. One such example is the recent published pluralism model by Lynn (2021), which abandons the search for a universal relationship between ethics and economics to instead focus on the situated social mechanisms to foster ethical actions. The core research stream 'CSR conceptualizations' could benefit from conceptualizations of political CSR, that we find increasingly in the periphery. Publications on the political role of organizations are often conceptual Palazzo 2007, 2011), even in the citing articles (e.g. Wickert 2016). Consequently, the CSR discourse would benefit from empirical investigations of the postulated relationships, such as between responsible leadership styles based on upper echelon theory and the micro-foundations of political CSR (Maak et al. 2016), or between firmlevel determinants and their influence of pursuing political actions in emerging contexts (Shirodkar et al. 2018). Until now, the resource dependence theory (Pfeffer and Salancik 1978) has been under researched in studies on political CSR and might be fruitful for future empirical research (Shirodkar et al. 2018).
The CSR marketing research could be enriched by more critical analyses of the link between perceived CSR and consumer responses. One example might be the influence of corporate greenwashing on consumers (e.g. De Jong et al. 2018). Thus, the marketing discourse may gain from insights of the rising research stream on 'CSR and consumer skepticism.' Method-wise, critical discourse analyses might be useful to unveil the ideological functioning of CSR marketing practices, such as cause-related marketing campaigns, which can be uncovered as strategy to avert ethical corporate criticism and to consolidate business profits (Brei and Böhm 2014).
Our bibliometric analysis confirmed the relevance of institutional theory in CSR research (e.g. Frynas and Yamahaki 2016). Among the cited articles of this discourse, we hardly find any empirical studies of CSR in special institutional contexts, such as in developing, emerging, transitional or post-colonial countries. While we find some works among the more recent citing articles (e.g. Munro et al. 2018), they still rely on mainstream institutional approaches. In the future, critical theories such as post-colonial approaches, for example applied by Jammulamadaka (2020), or power relations approaches might enrich research on CSR issues in such special contexts (Blowfield and Frynas 2005). Method-wise, institutional ethnography could be used as a research practice for such investigations (e.g. Campbell and Kim 2018).
The current trends highlight the growing relevance of psychological CSR (Ji et al. 2021). Attribution theory (Jones and Davis 1965;Kelley 1973) is particularly referenced in the growing research stream on 'CSR and consumer skepticism' (Ji et al. 2021). These studies are predominantly quantitative (e.g. Skarmeas and Leonidou 2013) and paradigmatic functionalist by focusing on organizational benefits due to effective communication (e.g. Du et al. 2010;Yoon et al. 2006).
Mainstream management researchers focus on communication strategies that companies use extrinsically to convince that CSR practices are mainly driven by intrinsic motives (Hoffmann 2018). Alternative perspectives could further enhance these insights by stressing the communicative constitution of a paradox of conflictual, intermingled and dynamic extrinsic and intrinsic CSR motives (Hoffmann 2018). This research stream might also benefit from inspecting controversial industries that are particularly vulnerable to skepticism about CSR initiatives (Aqueveque et al. 2018). Future critical studies might use qualitative websites and documentary analyses as well as hermeneutics to examine how corporations of such industries communicate about their CSR practices to protect and favor their interests Petticrew et al. 2018).
In the future, 'environmental CSR as a source of competitive advantage' could be analyzed from a more dynamic perspective and with focus on social network relations. For instance, Zhao et al. (2019) consider the mediating effects of dynamic capabilities and social capital on the relation between CSR and competitive advantages In future research, qualitative in-depth studies might enrich this discourse such as Adamik and Nowicki's (2019) study on CSR problems and paradoxes of co-creation in generating competitive advantages.
The examples above provide inspirations for future research directions within each research stream. Future research might as well profit from integrating research across research streams. One possible promising research route may be merging employee-oriented CSR research with studies on consumer skepticism to advance research on employee skepticism (Chaudhary and Akhouri 2018). An integration between employee-oriented studies with political CSR research would also be promising to explore the micro-foundations of organizations' (new) political role (Maak et al. 2016).
Our study has limitations. It bases on peer-reviewed journal articles listed in the SSCI. Other databases such as Scopus or EBSCO Business Source Premier could have extended our initial database of citing articles. However, the analysis of cited references-and therefore the co-citation analysis-goes above and beyond this dataset. Search terms were rather broad and the inclusion of other wordings or terms for individual CSR dimensions could have led to more divergent findings. However, 1,902 citing articles including their references already constitute a database that is large enough to distinguish current schools of thought, identify trends, and infer future research needs.
An inherent problem of bibliometric methods is that recent papers have a smaller chance of being co-citated in other works. However, by also analyzing not only (co-)citations but their citing articles as well, we were able to identify research streams and to link them to their current developments. Our contribution highlights the pillars of CSR research, current literature trends, and inspires future research. We particularly hope to stimulate further cross-sectional research to combine the essence of individual research streams.
Funding Open Access funding enabled and organized by Projekt DEAL.