Corporate Social Responsibility: Where Does It Come from, and Where Does It Go? Evolution of the Conceptual Structure from 1975 to 2021

Abstract

From the first discussions on Corporate Social Responsibility (CSR) in the 1950s to today, there is evidence of a ramification of concepts associated with CSR that produce a web of relationships that evolves in different directions. This paper analyses the conceptual structure of the CSR field, contributing to understanding its development and evolution between 1975 and 2021, opening a discussion of what these concept relationships might show in the understanding, development, and future application of CSR in business and society. For this purpose, networks of concepts were identified by using the SciMAT software on 6861 papers obtained from the WOS database in the Business and Management categories in the field of study. The results show that the field still appears to be far from understood and is very focused on the interests of companies. We offer and opportunity to rethink its purpose from a perspective that integrates other dimensions that concentrate on the society–business relationship. Based on these results, this study presents new research directions to explore both the drivers and results of the application of CSR.

1. Introduction

The increase in scientific production in the field of Corporate Social Responsibility (CSR), and how it associates with most areas of management, reflects its transversality and explains the interest in its study [1]. This interest is metaphorically presented by Ferramosca and Verona [2] as a “melting pot” or a “kaleidoscope”: a fertile field that tends to diverge and expand.

Studying CSR evolution, understanding where it is going, and analyzing whether it has taken a clear and specific direction seems to be a pending task for the field of study. Likewise, examining the concepts it has been associated with over time and identifying emerging research topics that may contribute to its integration into corporate management, thus improving the relationship between business and society, seems like a challenge for academia and a motivation for this study to be carried out.

Some studies seek to understand the evolution of CSR from different approaches. For instance, Mishra [3] compares the evolution of CSR in the USA and in India, divided into four phases: conceptualization, introduction, growth, and consolidation. She explores how the concept evolves based on the way it is used in these countries, shifting from philanthropy and paternalism to the establishment of rules.

This paper refers to the evolution of the conceptual structure of CSR in an attempt to explore how it has linked to different management concepts over time. This is the evolutionary perspective from which we seek to understand how CSR is applied. The present study poses two questions:

-

Where does CSR come from? This is accomplished by identifying concepts with which it has been associated for the period under analysis;

-

Where is it going? This is achieved by projecting future perspectives and exploring new angles of study and new uses from a management approach.

This paper conducts a conceptual analysis to explore the evolution and relationship of CSR concepts in the literature from 1975, which was the first year data became available in the Web of Science (WoS). The SciMAT Software was used for the analysis.

The conceptual structure analysis completes the latest analysis of the intellectual structure [1], as well as the bibliographic reviews in the field [4,5,6]. Although there are scientific literature reviews on CSR [4,5], this method of analysis reveals a number of biases [7]. Moreover, bibliometric and scientometric studies of intellectual structure establish patterns in the intellectual relations of a field [8] but not in the evolution of the interactions of concepts over time, nor do they identify which of these patters can be drivers in the field of study [9].

The conceptual structure of a field has been less studied than its intellectual structure [9]; it refers to the analysis of the co-occurrence of keywords, authors, references, or sources in scientific documents. Based on these relationships, the structure establishes networks of concepts according to statistical proximity, provides an evolutionary perspective, and proves the strength of the co-occurrence between concepts, thus generating semantic maps that help understand the knowledge behind this field of study [10,11].

The bibliometric analysis carried out by De Bakker et al. [12] contributes to the understanding of the field up to 2005. The authors take an evolutionary look at the concept of CSR with new concepts that have been associated with it over time. This is an indicator that the field of CSR “is a vibrant field”, where progress in the literature on corporate social responsibility “is obscured or hindered”, and where new constructs continually emerge in an attempt to understand it.

This paper deals with the challenge of contributing to the understanding of how CSR evolved between 1975 and 2021 by analyzing the management concepts associated with it from a broad perspective and by minimizing the restrictions involved in the study of the database in order to provide the purest results possible.

The SciMAT Software was selected for this analysis as it is capable of developing the science mapping approach [13]. This was applied to the database obtained from the Science Citation Index-Expanded (SCI-E) and the Social Science Citation Index (SSCI) of WOS spanning from 1975 to 2021, covering most of the important international journals in the area [14]. In order to broadly study the field without any bias, all the papers containing the concept of Corporate Social Responsibility in their titles, abstracts, and/or keywords were selected. Papers in English were considered and, given the focus of study, only the Business and Management categories were examined.

The database was obtained on 31 March 2022, with a total of 6861 papers. It then underwent data cleaning and re-processing that concluded on 21 August 2022. During this last stage, and considering the volume of publications per year, the period was divided into seven sub-periods [9,13].

This paper is divided into four sections: the first section provides a brief review of the literature, the second section describes the methodology used for the research process, the third section shows the results of the methodology and the software, and the final section offers a discussion and a conclusion with potential scope for future research.

2. Literature Review

To complete the conceptual structure analysis provided in this paper and following the same seven-period criteria used when applying SciMAT to the 6861-paper database, the twelve most cited academic papers were selected for each period of analysis up to 31 March 2022.

The aim of this review is to identify the research topics that aroused most interest in the scientific world for each period of study so they can be used in the subsequent analysis of the conceptual structure.

Between 1975 and 1991, a debate evolved about what CSR was. As early as 1975, Bowman and Haire [15] identified CSR as a problem for corporate strategy, as it is in conflict with the interests of investors and poses a cost–benefit dilemma. Along these lines, Jones [16] suggested that CSR is a “doctrine” that must be questioned.

Looking at Fortune-500 companies, Holmes [17] explored their adherence to CSR, Abbot and Monsen [18] examined the evolution of CSR actions over time, and Rosen et al. [19] conducted a survey to study investors’ behavior in relation to CSR. They concluded that younger and more educated investors are more willing to invest in CSR-oriented companies, although they are not willing to sacrifice financial returns.

In this first period, Carroll [20] refined the concept of Corporate Performance (CP); Wood [21] introduced the concept of Corporate Social Performance (CSP); and Druker [22] studied the meaning of CSR. The interest in understanding the relationship between CSR and Financial Performance (FP) began with Cochran and Wood [23] and McGuire et al. [24]. Aupperle et al. [25] empirically examined the relationship between CSR and profitability, but no relationship was found. Epstein [26] contributed with an interesting approach, seeking to understand CSR from the culture and human progress perspective.

Between 1992 and 1996 (second period of analysis), Swanson [27] continued to study the relationship between CSP and CSR, and suggested a model to explore CSR processes and their results on corporate behavior. Klassen and McLaughlin [28] continued the study of the impact of CSR on FP, based on the study by Cochran and Wood [23]. Sharfman [29] contributed to measure CSP, and Pava and Krausz [30] showed investment funds that influence their decisions towards investments with social performance.

During this period there was an interest in understanding what drives CSR. Ibrahim and Angelidis [31] found greater willingness to apply CSR in boards with external directors, as external directors show a greater concern for the discretionary component of CSR and a weaker orientation towards pure economic performance. Wang and Coffey [32] studied how the composition of boards of directors influences the charitable decisions of the company.

Burke and Logsdon [33] argued that CSR promoters think that CSR programmes are “worth” implementing because of the strategic benefits it brings to companies and the value created; Hammond and Slocum [34] showed that company reputation has more to do with financial performance than with any other type of behavior, such as social behavior; and Fryxell and Wang [35] showed that it is financial objectives that are relevant to reputation.

In different fields, Robertson [36] associates CSR with ethical behavior, Drumwright [37] uses a green economy approach to study socially responsible purchasing in firms, and Greening and Gray [38] show how organizational structures can be responsive to social and political demands by suggesting a model to respond to such demands.

From 1997 to 2001 (third period of analysis), there was an interest in understanding the impact of CSR in different business fields. Waddock and Graves [39] continued to explore the relationship with FP; Miles and Covin [40] investigated the impact of CSR on corporate reputation; Sen and Bhattacharya [41] examined how it influences consumer behavior; and Epstein and Roy [42] showed the importance of formulating a CSR strategy to impact CSP and FP.

Brown and Dacin [43], Mohr et al. [44], and Maignan [45] examined how market perception of companies’ CSR practices impacts consumer decisions, while Albinger and Freeman [46] studied how CSR actions are a factor in attracting employees. Additionally, Murray and Vogel [47] described how companies’ goodwill actions can be evaluated from an affect approach.

This period shows an interest in understanding what drives CSR. McWilliams and Siegel [48] analyzed the external pressure on companies to apply it, and, in 2001, the same authors [49] outlined a model of supply and demand for CSR, which depends on a number of endogenous and exogenous variables. They concluded that there is an “ideal” level of CSR, which managers can reach through cost–benefit analyses. Baron [50] explored CSR from a strategic approach.

Between 2002 and 2006 (fourth period), there was debate regarding what CSR is. Van Marrewijk [51] presented CSR concepts and definitions; Garriga and Melé [52] classified CSR into four groups of theories; Porter and Kramer [53] defined CSR as philanthropic actions, as a source of competitive advantage, although the same authors then argue that companies should stop thinking in terms of “corporate social responsibility” and instead think in terms of “corporate social integration” in order to build shared value instead of using CSR to carry out public relations campaigns [54].

Luo and Bhattacharya [55] established a relationship between CSR and company market value; Bhattachrya and Sen [56] further studied how and why consumers respond to CSR initiatives; and Sen [57] studied the returns of CSR by means of a field experiment.

Maignan and Ferrel [58] studied the relationship between CSR and marketing as a way of obtaining better stakeholder support; while Becker-Olsen et al. [59] further explored consumer behavior when perceiving CSR actions. They argued that these actions often seek to manipulate consumers.

Lichtenstein et al. [60] studied the effects of CSR on donations in non-profit organizations supported by businesses; and Branco and Rodrigues [61] analyzed how CSR actions bring benefits to companies, both by generating intangible resources and by improving their reputation.

Between 2007 and 2011 (fifth period of analysis), there was an interest in defining the factors that lead to apply CSR. Campbell [62] proved that CSR encourages responsible behavior in a clear regulatory environment; Godfrey et. al. [63] concluded that it can act as an insurance policy, generating shareholder value. Carroll and Shabana [64] identified four reasons for participating in CSR: to reduce costs and risks; strengthen legitimacy and reputation; build a competitive advantage; and create win-win situations. Aguilera et al. [65] presented a model to explain participation in CSR.

There is also interest in better understanding CSR, seeking to define it and explore the impact of its initiatives. In this framework, Dahlsrud [66] analyzed definitions of CSR and introduced five dimensions, while Matten and Moon [67] studied CSR initiatives in different countries.

The study of the relationship between CSP and PF continued during this period. Du et al. [68] explored the role of communications in maximising the return of CSR, while Surroca et al. [69] studied CSR and PF by looking at intangible resources as mediating variables.

A special focus is placed on the stakeholders’ perspective. Parmar et al. [70] studied the use and adaptations of the Stakeholder Theory, while Barnea and Rubin [71] classified the 3000 largest socially-responsible US corporations, and showed internal pressures from managers, shareholders, and experts to over-invest in CSR actions without bearing the costs. Barnett [72] tried to explain the effects of CSR actions on corporate finances as a consequence of the company’s relationship with its stakeholders.

Bear et al. [73] revisited the study of boards of directors and their effect on the promotion of CSR actions, concluding that gender diversity is particularly favorable in promoting CSR actions.

Between 2012 and 2016 (sixth period), Rao and Tilt [74] studied the relationship between CSR actions and gender diversity on boards; Jain and Jamali [75] examined the relationship of Corporate Governance (CG) with CSR initiatives; and Frynas and Yamahaki [76] reviewed the theories related to external and internal drivers of CSR.

In terms of the impact and performance of CSR, Saeidi et al. [77] explored the question “How does CSR contribute to company FP?”; Ioannou and Serafeim [78] provided a CSP index of social and environmental metrics, by using a sample of companies from 42 countries over seven years; Flammer [79] studied the conditions under which CSR brings positive benefits to companies, and she looked at the environmental action line of CSR to conclude that companies that report environmentally responsible behavior see their actions valued [80].

Aguinis and Glavas [81] linked the application of CSR to business outcomes, although they claimed knowledge gaps and the need for methodological approaches to better understand the impact of CSR.

Servaes and Tamayo [82] concluded that CSR is positively related to company value; Cheng et al. [83] found that it reduces restrictions on access to capital; and Kang et al. [84] suggested that it can sometimes be a repair mechanism for past irresponsibility. Finally, Hawn and Ioannou [85] argued that it contributes to a better market value.

Between 2017 al 2021 (seventh period), Beji et al. [86] concluded that gender diversity, age, and external directors in boards increase CSR activities; Gond et al. [87] revealed how environmental conditions can drive CSR actions; Iglesias et al. [88] proved that, in a more transparent and digitized environment, customers push for genuine CSR practices.

El Ghoul et al. [89] explored the relationship between CSR and the quality of institutions, and He and Harris [90] showed how the pandemic has acted as a driver for genuine CSR solutions.

During this period, there is also interest in studying the relationship between CSR and FP. Platonova et al. [91], and Vishwanathan et al. [92] have proven a positive relationship between CSR and FP actions; Cui et al. [93] and Kim [94] found an improvement in the reputation of the company; and Farooq et al. [95] studied the mechanisms through which CSR influences employee identification, and through employees, Firm Performance (FP).

Finally, Aguinis and Glavas [96] studied the relationship between CSR and job performance; Albuquerque et al. [97] concluded that systematic risk decreases and the value of the company increases when using CSR; and Kim [94] concluded that CSR impacts consumer confidence.

3. Materials and Methods

This brief literature review presents a research study that shows an evolutionary path of concepts associated with CSR. However, the study is not able to identify the networks of concepts related to CSR, as it cannot include thousands of research studies in the field. A bibliographic approach will always present biases [7], and the conceptual structure analysis seeks to reduce these biases. [10,11].

In order to avoid biases in the results in this field of study, Zupic and Čater [11] suggested several stages in the bibliometric study of management-related disciplines.

Stage 1: Research Design

For the scope, limits, and objective of this research study, the conceptual structure of the CSR-related field of study was identified, along with the evolution of concepts considered in the related academic production, and future challenges to be explored. The keywords from the academic production were used as the unit of analysis, through the co-occurrence method [98].

Stage 2: Compilation of Bibliometric Data

Papers from journals indexed in the Social Science Citation Index (SSCI) or Science Citation Index Expanded (SCI-Expanded) were selected from the main collection of the WoS database. The WoS database is a data source commonly used for bibliometric studies [11,99], as its indices cover the most important international journals in the areas of pure, applied, and medical sciences, as well as Social Sciences [14,100].

From this database, the study extracted all papers, reviews, and letters published in the journals between 1975 and 2021 containing the concept of Corporate Social Responsibility in their titles, abstracts, and/or keywords. Papers in all languages were considered and, given the focus of study, only the Business and Management categories were examined, resulting in 6861 papers.

Although this study considered the possibility of incorporating other concepts as search criteria, from the review of the literature before applying the method, the preliminary finding was the concept of “Corporate Social Responsibility”. CSR was indeed the main connector of the different ways of understanding the field. Other relevant concepts or disciplines (such as corporate sustainability, corporate citizenship, business ethics, etc.) were not always—or in most cases—present; incorporating these concepts into the search criteria could bias the analysis, thus forcing results from the CSR conceptual structure towards arbitrarily chosen dimensions.

Stage 3: Analysis

Once the data for the analysis have been obtained, it needs to be cleaned before processing. The appropriate techniques and bibliometric software are applied for the scope of the analysis in order to characterize the conceptual structure of the field of study [11].

The SciMAT software was selected to characterize the conceptual structure due to the fact that it can illustrate the relationship of keywords in the different concepts in terms of their time evolution and perform science mapping analysis [101]. Although there are other software for bibliometric and scientometric analysis, such as Bibliometrix [102] or VOSViewer [103], they tend to address the intellectual structure of the field of study [104], which has already been covered by other studies [1].

SciMAT was applied based on the science mapping approach suggested by Cobo et al. [13]. It adds value to the analysis by preprocessing and cleansing data [101], thus adjusting the keyword clustering and co-occurrence algorithm.

Before performing the algorithms, data were subjected to pre-processing to clean the base and avoid bad data (outliers) or papers in the WoS collection associated with 2022 publications (early access). Then, keywords provided by the authors of the papers and those offered by the journals in which they were published were grouped together, avoiding redundancy problems (same concepts presented differently). Not grouping them could reduce their importance by understanding them as different keywords.

An example of redundancy is what happens with the keywords grouped under the concept of “FINANCIAL-PERFORMANCE”. In this case, independent keywords in the literature are grouped as: FINANCIAL-PERFORMANCE; FINANCIAL-PERFORMANCES; FIRM-FINANCIAL-PERFORMANCE; FINANCIAL-PERFORMANCE-(FP); and FIRM-FINANCIAL-PERFORMANCE-(FFP); among others.

There were also keywords that were grouped because they point directly to a level of detail that is not considered relevant for the purpose of the study, such as those for the concept of “STAKEHOLDER-THEORY”, listed as: STAKEHOLDER-THEORY; STAKEHOLDER-APPROACH; STAKEHOLDER-PRESSURES; STAKEHOLDER-RELATIONS—STAKEHOLDERS-INFLUENCE.

The same happened with the concepts of Corporate-Social-Responsibility; CSR; CSR-PREFERENCES; CORPORATE-SOCIAL-RESPONSIBILITY-STRATEGIES; and 20 other keywords that point to the same concept, treated as a single concept to improve the analysis of co-occurrence. This allowed us to move from 14,371 concepts to 12613, which is what the study worked with.

The analysis of the co-occurrence of keywords identified the networks that emerge from this relationship in the papers, linking them with keywords that tend to appear together. These connections are presented as a semantic map that configures the conceptual structure of the field [9,11], by simplifying the relevant elements of the papers and facilitating their understanding. To normalize the frequency of co-occurrence, the Equivalence Index Algorithm [13,103] was used, while the Simple Centres Algorithm was selected as a clustering method (to label the clusters) [13].

In order to have an evolutionary perspective of the concept networks, seven analysis periods were defined according to the number of relevant papers and the academic evidence of scientometric analyses [9,105]. The analyses seek to find trends and understand the conceptual structure of the last five years, as well as the state of the art of the field of study [9,11]. The analyses considered periods of time, except for the first period (1975–1991), due to the low academic production of papers on CSR in the WoS collection.

Table 1. Analyzed periods and number of papers.

Period	Years Covered	Number of Papers
1	1975–1991	47
2	1992–1996	27
3	1997–2001	53
4	2002–2006	202
5	2007–2011	1167
6	2012–2016	1915
7	2017–2021	3450

Source: Compiled by the authors, based on the data processed with SciMAT.

shows the time periods and the number of papers in each, according to the search criteria described.

Following Marchiori’s study [9], simulations were carried out on the database, with the aim to observe the behavior with different parameterizations [13]. These simulations help to verify that no clusters are generated by restricting the first periods and given the low volume of publications. On the contrary, not restricting the periods of high production volume generates so many clusters that make it impossible to analyze the resulting information. After six processes of adjustment and analysis, researchers selected the parameterization that produces a better analysis of the results [11].

Thus, for periods 5, 6, and 7, only keywords that were repeated at least 2, 3, and 4 times. respectively, were considered as candidates for the networks. Likewise, for a link between two concepts to be a candidate for a co-occurrence network, it must be present at least 3 times for periods 5 and 6 and co-occur and at least 4 times to be part of a network in the last period. For the other periods, no restrictions on occurrence or co-occurrence were considered.

Stage 4: Visualization

Once specifications were made, the study applied the SciMAT algorithm to obtain the characterization of the conceptual structure of the field of study. Based on the guidelines provided by Cobo et al. [13], the software graphic results were as follows.

• Keywords’ Evolution

The evolution of keywords shows the number of keywords in the papers for each period—represented by each circle—as well as the number of keywords that were already considered in the previous period (input arrow) and keywords that are no longer considered in the following period (output arrow). The figure also shows the weighting of keywords that are maintained from one period to the next.

• Strategic Diagram

The papers published in high-impact journals in the last 5 years of the period analyzed are considered as the state of the art in the field of study which define the guidelines for future research [9,11]. This is why the strategic diagram is needed. Not only does it highlight the networks of the period, but it also arranges them according to their contribution in a two-dimensional space. The first variable is Callon’s Centrality (horizontal axis), which represents the degree of interaction of the network with other networks in the same period, while Density (vertical axis) represents the internal cohesion of the network and the concepts included [101,106].

In order to highlight the impact of each network in each period, the size of the node representing the cluster is generated, based on the average number of paper citations in the cluster.

• Thematic Networks

They represent the interactions of the different clusters and are displayed for the driving clusters. The size of each node represents the importance of each concept within the cluster and the name of the network is given by the central concept, which is the most representative concept within the network, as it is the one that coincides the most with the rest of the concepts [105].

• Thematic Evolution

The clustering algorithm arranges keywords in specific networks which contain these concepts and label them based on the one that determines the network. Considering the analysis divided by periods, the algorithm facilitates visualising networks over time and links them based on their relationship with networks of the next period. The solid lines represent networks that carry the same label or one of the words contained in the other network. The dotted lines represent networks that share elements but not the determining concept of the cluster.

Stage 5: Interpretation

Considering the interpretation of conceptual structure findings for other fields of study provided in the academic literature by de las Heras-Rosas and Herrera [100], Díaz-López et al. [105], and Marchiori et al. [9], the keyword evolution explains the stability or complexity of the field based on how the relevant concepts were incorporated over time. The thematic evolution helps to understand the origins and evolution of the networks highlighted in the strategic diagram of the last period of analysis, which shows the state of the art of the field of study. Finally, the thematic networks help us understand the interaction of concepts contained in each cluster.

4. Results

Graphical results are obtained from statistical and scientometric methods based on the evolutionary analysis of the keywords in the literature, their co-occurrence, the generation of clusters or networks of concepts that link them, and the variables of Centrality and Density of the networks for each period.

Firstly, it is interesting to see the magnitude and persistence of concepts considered as keywords for the field of study during the analyzed period. Here, Figure 1 shows the number of keywords contained in the papers in each period, from the first period (1975 to 1991) to the last one (2017 to 2021). The horizontal arrow shows the concepts that are maintained from one period to the next, together with the corresponding percentage for the concepts in the following period. The upper arrows show the number of “new” concepts incorporated into the field of study in that specific period, together with the concepts that are no longer considered.

From these results, it is worth noticing the growing and ongoing incorporation of new concepts over time, as well as the tendency to preserve concepts from previous periods. This could mean that academic approaches have not necessarily undergone radical changes, but they rather add relevant concepts for analysis in each period, maintaining the previous ones, going from 4 concepts in the first period to 1269 in the last, as shown in Figure 2, along with the number of papers from each period.

4.1. State of the Art of the Field of Study

Applying the clustering method on each period analyzed allows us to detach networks from concepts—or clusters—that are built from the links determined by the co-occurrence of keywords. Each network presents concepts that tend to co-occur to a greater or lesser extent. Concepts that link with all the other words, or with the literature, are considered as the name of the network, through co-occurrence.

In view of the above, a strategic diagram was designed for each period, showing the distribution of the algorithm’s defined networks based on the centrality and density of each of them. Figure 3 shows the strategic diagram for the last period of analysis, thus representing the current situation of the CSR field of study [9,11], hence providing information on trends and challenges in the field.

Based on the understanding of how the Density and Centrality variables behave, each quadrant represents networks that share common characteristics for the field of study. Therefore, the clusters are categorized according to those that are drivers in the field of study, basic and transversal, emerging or declining, and isolated and highly studied [105].

4.1.1. Networks Driving the Field of Study

Networks with greater density and centrality (upper right quadrant) can be considered essential for the field of study; they contain concepts that co-occur a lot with one another and with other networks. The driving networks in this quadrant are:

• Board of Directors: The network with the highest density and one with the highest centrality. Corporate Governance is the most studied concept in the network, followed by Disclosure, and Corporate Performance;
• Financial Performance: The network with the highest centrality among the driving networks. It contains concepts such as Stakeholder Theory, Management, Strategy, and Reputation. Additionally, this network links with elements of performance management, such as Corporate Social Performance, Environmental Performance, and Financial Performance;
• Organizational Citizenship Behavior: This network is mainly associated with people management concepts, such as Commitment, Leadership, Human Resource Management, Employees, and Satisfaction, among others;
• Brand: Concepts are closely related to marketing practices and their expected effects, such as Consumers, Loyalty, Trust, and Purchase Intention;
• Agency Theory: It is worth mentioning the research on the role of “manager”, with concepts such as Ownership Structure, Investor Protection, Institutional Investors, and CEO Power;
• Environmental Management: It associates keywords such as Sustainability, Institutional Pressures, and Competitive Advantage, and relates keywords associated with the supply chain, such as Supply Chain Management and Environmental Supply Chain;
• Sustainability Reports: Two types of concepts are present in this network. On the one hand, those concepts that drive or motivate reporting (Transparency, Legitimacy, Sustainability Assurance), and on the other, concepts associated to the characteristics of reporting (Global Reporting Initiative, Quality, Nonfinancial Information, Statement).

4.1.2. Basic and Transversal Networks

Networks with lower density and high centrality (lower right quadrant), as well as fewer co-occurrences than the essential networks for the period in question. However, they have an important link to other networks, so they tend to be present in papers on the field of study, but not so deep as a cluster. The driving networks in this quadrant are:

• Performance: One of the most centralized networks in the field, thus showing relevance to CSR. As a driving cluster, the Financial Performance network already considered a number of important performance elements such as Environmental Performance and Corporate Social Performance. Within this cluster, Performance is the most present concept in the literature, with a frequency of 836 times;
• Identity: It identifies concepts such as Brand Identity (Corporate Marketing, Corporate Brand, Image) and Person Identity (Citizenship Behavior, Person–Organization, Ethnocentrism, Citizenship);
• Corporate Philanthropic: It links the concept of philanthropy to features of a decision-maker (Upper Echelons, CEO, Manager), and tends to be associated with the size of the company (Emerging Markets, Company Size, Resources, and Sponsorship);
• Behavior: It is almost exclusively associated with ethical concepts such as Ethical Consumption, Business Ethics, Gender Differences, Values, Religiosity, and Ethics;
• Industry: Three specific industries are characterized for this network, which were present in the study of the field for this last period, i.e., the Banking Sector, Tourism and Hotels.

4.1.3. Emerging or Declining Networks

Networks with lower density and centrality (lower left quadrant), not deeply studied and less linked to other networks. The identified clusters can be emerging or in decline, which could be clarified by a longitudinal view of the field of study. The driving networks in this quadrant are the following:

• Entrepreneurship: The most representative cluster of the quadrant, which links keywords such as Social Enterprise, Co-Creation, and Developing Countries to the concept of Entrepreneurship;
• Networks: Unrelated to one other and only related to the Network keyword. It is built by grouping concepts such as Social Networks, Knowledge, Business Models, and Institutions;
• Government: This cluster shows coherence between the link of concepts that characterize a political dimension of CSR. It identifies concepts such as Public Policy, Political Connections, CSR-Political, State, and Firm Self-regulation;
• Customer Satisfaction: Based on the relationships of the following concepts Service Quality, Signaling Theory, Stock Market, and Behavioral Intention;
• Organizational Commitment: It only contains two concepts, which could characterize two dimensions of Commitment: Customer-Orientation, and Employee Values and Engagement;
• Collaboration: It is characterized by concepts such as Suppliers, Partnership, Alliances, and Cross Sector Collaboration.

4.1.4. Isolated and Highly Studied Networks

Networks with high density but low centrality (upper left quadrant). They are deeply studied, but with little link to other networks. The driving networks in this quadrant are the following:

• National Cultures: The network with the lowest centrality; that is, the most isolated network from the rest of the clusters. It links the concept to the dimensions of culture, as established by Hofstede [107], who highlighted the effects of national culture on behavior;
• Eco-Innovation: Within the networks for the period, this is one of the densest networks and it links to the keywords Environmental Innovation, Green Innovation, Technology, and Capabilities;
• Multinational-Corporations: Many concepts are associated with this network, where the relationship of co-occurrence with the concept of International Business stands out;
• Socially-Responsible-Investment: Within the networks of the quadrant, this is the one with the lowest density; that is, very close to the quadrant of emerging or declining clusters. The concepts of Investment and Mutual Funds stand out;
• Communication: A network of ambiguous interpretation, as it is very close to the centre of the diagram so that, should the parameters change, it could be considered as part of another quadrant. The effect of communication actions is presented in the cluster through keywords such as Public Engagement, Organizational Legitimacy, Dialogue, Credibility, and Corporate Reputation, as well as communication channels such as Media, Social Media, and Impression Management.

The driving clusters or networks of the last period in the strategic diagram in Figure 3 reveals the current situation of the field of study [9,11]; whereas Figure 4a,b shows the concepts associated with each network in each quadrant, so as to contribute to a better understanding of the field of study.

To complete Figure 4a,b,

Table 2. Basic cluster indicators for the 2017–2021 period.

Cluster	Category	Centrality Range	Density Range	Documents Count	h-Index	Average Citations	Total Citations
BOARD-OF-DIRECTORS	Driving	0.96	1	406	44	20.64	8381
FINANCIAL-PERFORMANCE	Driving	1	0.87	800	60	21.23	16,983
ORGANIZATIONAL-CITIZENSHIP-BEHAVIOR	Driving	0.83	0.83	127	29	25.71	3265
BRAND	Driving	0.74	0.78	224	32	17.56	3933
ENVIRONMENTAL-MANAGEMENT	Driving	0.87	0.65	246	36	21.26	5231
SUSTAINABILITY-REPORTS	Driving	0.7	0.61	111	26	22.5	2498
MULTINATIONAL-CORPORATIONS	Isolated and highly studied	0.43	0.7	76	20	21.59	1641
PERFORMANCE	Basic and transversal	0.91	0.22	358	37	16.41	5874
COMMUNICATION	Isolated and highly studied	0.48	0.52	80	22	16.66	1333
AGENCY-THEORY	Driving	0.78	0.74	74	22	20.03	1482
GOVERNMENT	Emerging or declining	0.39	0.35	51	16	16.82	858
BEHAVIOR	Basic and transversal	0.61	0.26	123	24	14.87	1829
IDENTITY	Basic and transversal	0.65	0.3	59	21	21.81	1287
CORPORATE-PHILANTHROPIC	Basic and transversal	0.52	0.17	32	15	15	480
INDUSTRY	Basic and transversal	0.57	0.04	37	17	20.59	762
ECO-INNOVATION	Isolated and highly studied	0.3	0.96	35	14	18.03	631
COLLABORATION	Emerging or declining	0.22	0.48	19	10	18.16	345
NETWORKS	Emerging or declining	0.26	0.09	17	9	13.71	233
CUSTOMER-SATISFACTION	Emerging or declining	0.35	0.39	27	13	16.7	451
ENTREPRENEURSHIP	Emerging or declining	0.17	0.13	12	7	10.92	131
NATIONAL-CULTURES	Isolated and highly studied	0.04	0.91	7	5	11.43	80
ORGANIZATIONAL-COMMITMENT	Emerging or declining	0.13	0.43	8	5	14.12	113
SOCIALLY-RESPONSIBLE-INVESTMENT	Isolated and highly studied	0.09	0.57	13	8	14.15	184

Source: Compiled by the authors, based on the data processed with SciMAT.

shows each of the clusters for the 2017 to 2021 period, together with their degrees of density and centrality, number of papers associated with the cluster, h-Index, average number of citations of papers belonging to the cluster, and the total number of citations accumulated by the published scientific papers associated with each Thematic Network.

Table 2 shows that the cluster with the highest number of publications is Financial Performance, followed by Board of Director, which are both driving clusters, followed by Performance, which is a basic and transversal network. The concepts in these three networks account for 53.2% of the papers published.

The same behavior is observed in the number of citations, where once again the papers associated with the clusters Financial Performance, Board of Director, and Performance accumulate the highest number of citations, where the concepts associated with these three clusters account for 53.9% of total number of citations.

In relation to the Hirsch index or h-Index, which generally favors long-standing authors who publish papers with a lasting and above-average impact [108]; it is worth noting that, in this case, it is led by authors who present concepts associated with publications in the Board of Directors, Financial Performance, and Performance clusters.

4.2. Longitudinal View of the Field of Study

Figure 5 shows the evolution diagram of the clusters through the study periods. The solid lines connect the networks that carry the same label or one of the words contained in another network. The dotted lines link networks that share elements but not the determining concept of the cluster.

Each column of Figure 5 shows the clusters for each period, which are linked over time. Thus, for the first period only one cluster is set up under the name of Business Ethics (BE), which groups four concepts (Figure 6). For the second period, three clusters are generated, two of them from concepts present in the first period cluster. The BE cluster is maintained for the second period, and the FP cluster stems from it. Then FP becomes independent, thus naming a new cluster. The third cluster of the second period is independently generated from the first period cluster.

The same happens in every period until the last, where 23 clusters are generated, all of which are arranged according to density and centrality in the Strategic Diagram (Figure 3).

From a longitudinal view, it is possible to understand the origins of driving clusters for the last period, along with the most representative networks of the other three quadrants. The analysis is shown below.

• Board of Directors: A driving network since the second period of analysis, after integrating the FP cluster. In the fourth period, it joins the cluster led by the concept of Governance, which is identified as a basic and transversal cluster. In the sixth period, the Governance network becomes a driving cluster for the field, and in the last period of analysis, the Board of Directors concept takes the lead in the network;
• Financial Performance: Although it was set up as a network in the second period, FP was already a relevant concept in the only cluster identified for the first period, as shown in Figure 6. Therefore, the concept has always been relevant to the field of study. For the first periods, this network is presented as the driver of the field. In the third and fourth periods, the number of clusters increases, and new concepts emerge, such as Stakeholders, Marketing, or Supply Chain Management, thus diluting the contribution of each concept to the field and characterizing the Financial Performance network as basic and transversal. From the sixth period onwards, it once again takes on a leading role, becoming the driver in the last period;
• Organizational Citizenship Behavior: This network becomes relevant in the last period. It is detached from the Attitudes network of the previous period, which is eventually absorbed by the Behavior cluster;
• Brand: In the fourth period, the concept is present as part of the Alliances cluster. For the following period, the concept becomes independent and forms a network named Brand. It is a low-density, medium-centrality cluster. In the sixth period, it becomes the driver of the field of study;
• Agency Theory: This concept is part of the Governance network in the fourth period. In the sixth period, it becomes part of the CEO Compensation cluster. In the last period, it becomes a driving cluster;
• Environmental Management: Identified as a driving cluster in the last period, it takes on the concepts of Sustainable Development and Supply Chain Management (both clusters for the previous period). Likewise, these two concepts detach from the Sustainability network in the fifth period. For the sixth period, the concept of Sustainable Development becomes independent and forms its own network, while the previous Sustainability cluster is renamed Supply Chain Management, even though Environmental Management is considered as one of the relevant concepts. For the last period of analysis, the cluster changes its name back to Environmental Management and brings back (absorbs) the cluster that was detached from Sustainable Development before the relationship;
• Sustainability Reports: It belongs to the Sustainable Development network in the sixth period. It emerges as a driving cluster in the last period of analysis;
• Performance: It is formed for the last period, breaking away from the Governance network (driver of the field in the previous period). It is a basic and transversal cluster, containing the two concepts that characterized networks in the same quadrant for the previous period, i.e., Market and Innovation;
• Entrepreneurship: A network for the sixth period that was present in the Sustainability network in the fifth period. When identified as a network, its centrality and density decrease. Fewer and fewer concepts are linked to it;
• National Cultures: It begins as part of the quadrant of isolated and highly studied networks and becomes a cluster only in the last period. It has no links to previous periods.

5. Discussion and Conclusions

From the literature review provided by this paper, three very clear focuses of research on CSR can be identified. The first focus is the permanent search to contribute to improve its understanding and to present new definitions [16,17,20,22,26,51,52,54,65,66]. This is shown from the first period up to the fifth period. In the sixth and seven periods, there are no papers seeking to explore definitions of CSR in the selected literature, which is not to say that it has not been addressed in other literature selections.

The second focus is the understanding of the financial, strategic, and reputational benefits that the application of CSR brings to companies [15,16,23,24,28,33,39,40,41,53,55,85,91,92,93,94,96,97]. This is presented in the literature in the seven periods, with a special emphasis on FP.

The third (but not the least significant) focus is the ongoing search to understand what drives companies to implement CSR [31,48,62,63,64,76,82,84,88,90]. This appears from the second period of analysis and is consistently maintained until the last period.

It is worth noting that the most cited literature in the field of study is that which focuses on exploring CSR from the interests of the company, making Epstein’s [26] study very exceptional; his particular approach set out to understand CSR from the culture and human progress perspective.

From this literature review, it is very complex and biased to conclude on how the CSR concepts evolve. [7]. The analysis of the conceptual structure allows for a rich discussion on how concepts evolve and what drives CSR. This paper presents some of the findings that researchers considered relevant to open the debate, which will be expanded with future analysis based on the results obtained from applying SciMAT. The parameters of the algorithms used in the software have followed the recommendations of the literature. However, this may generate limitations and biases in the analysis. Future research ought to apply different parameterization criteria to validate or refute the findings presented here.

The study of the conceptual structure of the CSR field reveals a strong and permanent increase in concepts and clusters associated with the field of study, from the 4 concepts in the first period and only 1 cluster to 1269 concepts and 23 clusters in the last period (Figure 1 and Figure 5). CSR seems to be a fertile, diverging [12], and expanding field [2].

The clusters found in the Strategic Diagram (Figure 3) show Board of Directors (BD) and Financial Performance (FP) as their two main drivers for the last period. It is evident from the Thematic Evolution (Figure 5) that Board of Directors (BD), and Corporate Governance (CG) are concepts that have been historically associated with the field of CSR. Their presence in the driving networks can be observed from the second period of analysis (1992–1996) as part of the FP cluster. CG is shown from the fourth period onwards and remained as such until the sixth period, becoming consolidated in the last period under the name of Board of Directors.

For the last period, the Board of Directors cluster is the main driver of the field (due to its high density and centrality), whereas CG is the concept with the highest number of papers and average citations in the cluster, as shown in the Thematic Network (Figure 4a,b). Zaman et al. [109] reviewed 27 years of research on the relationship between CSR and GC and concluded that although the relationship has been extensively studied, there is no consensus on its nature. For their part, Endrikat et al. [110] analyzed 82 empirical studies that explored the relationship between CSR and BD and concluded that the studies yielded inconsistent findings. This paper shows that the academic study of CSR is highly related to BD, and that there is an ongoing interest in understanding how board composition affects decisions to implement CSR initiatives [29,32,72,73,74,84].

The presence of FP as the second most important driving cluster found in the field in the last period reveals a strong interest in measuring the impact on company performance when applying CSR. From the Thematic Evolution (Figure 5), this interest has proven to be present throughout the analyzed period as FP is represented from the first period in the only cluster built by the clustering algorithm (as shown in Figure 6).

As for the Performance cluster, it appears separately from the FP cluster, showing high centrality and lower density (basic and transversal network) by incorporating other dimensions of performance that are different and less studied than the financial one, such as Perception, Innovation, Leadership, and Commitment Performance, among others. As a result, this paper can conclude that there is academic interest in understanding how CSR impacts the different dimensions of performance and particularly the financial one.

This is in line with the second focus of research interest identified in the literature on CSR, in addition to the constant need to seek explanations as to what drives CSR application (third focus of research interest in the literature).

The Stakeholder Theory (ST) has been extensively studied in the field of CSR. According to Johnson-Cramer et al. [111], the ST is adding new domains and continues to transform the way we think about business and society. Based on the results of their study, the ST only appears as a driving cluster for the third period (1997–2001), although it is a less studied concept than Performance within the same network and is diluted as a driving cluster in the following period. It reappears as part of the FP network, which is the driving cluster in the fifth period (2007–2011) and in the last period.

As a conclusion, the Stakeholders’ perspective is being incorporated into the study of CSR, although this study shows that it does not promote the field as expected. Its participation seems to be mainly associated with its interaction with FP. This brings the intentions of CSR strategies and their real impact on Stakeholders into question. The conceptual structure of the field seems to show that CSR has been developing from a perspective that focuses on the interests of the company (Financial Performance, Board of Directors, Corporate Governance, Brand, Agency Theory).

Based on the above, answering the question “Where does CSR come from?” seems simple. The results from the conceptual structure analysis, together with what is found in the literature, allows us to conclude that CSR has been associated with the interests of the company from the first up to the last period of analysis. CSR practices are associated with financial achievement rather than other objectives, such as the impact on certain stakeholders or the improvement in the relationship between business and society. Even the ST has been dominated by FP.

As for the question “Where is CSR going?”, the answer seems more complicated. The strategic diagram shows emerging spaces for the last period, so there is a challenge for scholars to explore new ways to boost the application of CSR. It would be interesting to focus the analysis beyond the interests of the company or shareholders, in order to respond to the needs identified in all the dimensions considered as important for the field of study.

The need to model the application and effects of CSR is justified by the characterization of the state of the art in the field of study, as illustrated in the four quadrants of the strategic diagram for the last period. This opens up new research directions. It is interesting not only to explore what currently drives the strong relationship between CSR and networks such as the Board of Directors and Financial Performance, but also to find out whether low-centrality and low-density networks are declining or emerging and what their future relationship with the research field will be.

The diversification of concepts shown by the evolution of keywords (Figure 1) adds to the reality of networks of critical concepts highly studied in the field but isolated from each other, which could be better articulated with national, cultural, innovation, and responsible investment factors. These areas are present in the literature but have been seldom studied (Figure 3). They could be instrumental to drive CSR in its future direction.

This paper shows a great focus of CSR on the interest of the company [41,42,43,46,56,57], leaving a space for academia to explain whether CSR has definitely become a utilitarian tool to be used for the needs of the company or if there is still room for it to evolve and be looked at from new perspectives.

Understanding the external environment in which the company develops its business could well be the starting point for the scientific community, faced with this challenge, to suggest new models that shift CSR from a highly company-centered perspective to a genuine focus on business–society interaction, i.e., a perspective less focused on the interests of business alone. This would lead to a systemic view, integrating new dimensions from an evolved economy that demands genuine and modern behaviors from companies in society [112].