Mapping corporate social responsibility practices at the international level: systematic review and content analysis approach

Purpose – It has been more than 20years since the idea of binding multinational corporations directly to international law was abandoned. Since then, concerned actors have sought to manage corporate conduct through voluntary regulation. However, little is known about the instruments produced in this regard. This study aims to understand the properties of the instruments that govern or regulate corporate social responsibility at theinternationallevel. Design/methodology/approach – Systematic literature review and content analysis methods were combined to compile a list of 229 international corporate social responsibility instruments (ICSRIs) produced by intergovernmental (IGOs) and international nongovernmental (INGOs) organizations. These instruments were categorized accordingtoan adapted classi ﬁ cationframework. Findings – The majority of instruments from our sample are produced by INGOs, focus on management activities and are applicable to speci ﬁ c industries. The most common issues addressed by the instruments are related to worker protection, human rights, governance and the environment. A limited number of instruments specify stakeholders ’ involvement or feature an external orientation. Instruments rarely address issues related to product quality and safety, economic contribution or social performance. Practical implications – Without a comprehensive overview, it has been dif ﬁ cult to develop broad-based understandings about voluntary regulation as a mechanism for controlling corporate conduct internationally. This study ’ s ﬁ ndingsoffer valuable insights, allowing policymakersand industry practitioners to understand theeffectivenessof,andmake appropriate enhancements to,ICSRIs. © Osman


Introduction
The new millennium marked a dramatic shift in the discussion about corporate social responsibility (CSR). The idea that corporations had obligations toward society at large had already taken root (Agudelo et al., 2019). In the global north, the public's demand for corporate accountability had since been reflected by legislation and government policy (Knudsen, 2017). In the global south, however, corporate misconduct continued to be a cause for concern (Omoteso and Yusuf, 2017). Driven by the attraction of cheaper production opportunities, corporations increasingly turned their attentions toward developing countries (Bae et al., 2021;Nolan, 2017;Ruggie, 2017). It was not long before reports of exploitive and harmful business practices shocked consumers the world over (Hackett and Moffett, 2016;Ullah et al., 2021). Unfortunately, attempts by the international community to control the behavior of these multinational corporations (MNCs) through international law proved unenforceable, and repeated attempts to homogenize obligations across all nations were unfeasible (Bair, 2015). At this point, the concerned parties adopted the perspective that the only effective way to manage corporate conduct was through market and social pressures (Berger-Walliser and Scott, 2018). Consequently, the idea of binding corporations to the law at an international level faded, and questions regarding voluntary regulation began to surface (Bartley, 2022;Eller, 2017).
Many MNCs created individual codes of conduct for themselves, to appease their customers' demands for formal CSR commitments (Berger-Walliser and Scott, 2018;Tamvada, 2020). However, commentators soon began to question the implementation and overall impact of this self-regulation (Bae et al., 2021;Hackett and Moffett, 2016). While some suspected that these codes were nothing but lip service (Lebaron et al., 2017), others were concerned with the applicability of code elements (Tamvada, 2020). Civil society organizations, nonprofits, industry associations, intergovernmental bodies and even private companies all realized that they could achieve their objectives by designing or participating in the formulation of instruments that could be applied to a wide range of corporations regardless of geographic location or reach (Berger-Walliser and Scott, 2018;Camilleri, 2017;Choudhury, 2018). The analysis of these instruments is the subject of this study.
Much of the existing research is limited to in-depth investigations of individual instruments (Brown et al., 2018) or comparative investigations of a small number of instruments (Fransen et al., 2019;Pope and Lim, 2020). Few research has concentrated on exploring all the instruments in use within a particular area (Raynolds et al., 2014;Knudsen, 2017), and fewer studies have attempted to investigate all the instruments in use at a global level (Perez et al., 2019). Without a comprehensive overview of the range and characteristics of these international CSR instruments (ICSRIs), thorough understanding and appreciation of the voluntary regulation phenomenon remains limited.
This research paper will address the question: SAMPJ 13,4 Q1. What are properties of the instruments regulating corporate social responsibility at the international level?
The main objectives of the paper are to: identify a dependable and extensive body of work, build a comprehensive list of instruments and classify these instruments. Answering the question contributes to theory and practice. Regarding theory, our study serves as a schematic for the systematic identification and the quantitative analysis of CSR instruments. Furthermore, the findings of our study will extend those from previous conceptual and descriptive research, offering insight into scale to complement existing knowledge of depth. With respect to practice, the findings of our study will offer new understanding of existing ICSRIs. Concerned actors will be able to appreciate the scope of these instruments and identify the issues and activities that they typically neglect. This will enable policymakers, civil society activists and industry influencers to improve or develop instruments that balance out the shortcomings of existing ones. The findings of this study can have an impact on the improvement of voluntary CSR regulation at a global level leading MNCs to fulfill their social obligations more effectively. This will, in turn, contribute to the implementation of the Sustainable Development Goals.
2. Literature review 2.1 Theoretical background Carroll (1979) stated that CSR "encompasses the economic, legal, ethical, and discretionary expectations that society has of organizations at a given point in time." Just over a decade later, Carroll (1991) arranged the components of his definition into the well-known pyramid structure that has epitomized his work. Economic expectations were placed at the base of the pyramid, and discretionary expectations were placed at the top. He also introduced a new concept into his definition that would become a core theme in many of his later publications (Carroll, 1999(Carroll, , 2016Carroll and Shabana, 2010;Schwartz and Carroll, 2003). He stated that organizations should manage their social expectations in relation to their stakeholders. In this way, a connection was made between his CSR pyramid and stakeholder theory. Stakeholder theory can be traced back to the pioneering work of Freeman (2010), who proposed that businesses were dependent on their stakeholders for their ultimate survival, and therefore had a responsibility to create value for them. He defined stakeholders as "any group or individual who can affect or is affected by the achievement of a firm's objectives" including, but not limited to, customers, employees, investors, suppliers and local communities (Freeman, 2010, p. 25). This idea gained widespread academic popularity after the new millennium as public perceptions of corporations became increasingly negative in the wake of several scandals (Parmar et al., 2010). Although it has been suggested that stakeholder theory and CSR are complimentary frameworks, rather than co-dependent concepts, the idea that businesses have responsibilities to their stakeholders has become a fundamental principle of modern CSR understanding (Dmytriyev et al., 2021).
The social license to operate (SLO) concept is a useful framework to understand how industry approaches CSR (Gehman et al., 2017). The SLO was examined by Joyce and Thomson (2000), who explained that, with respect to mining companies, it referred to "the approval [and] the broad acceptance of society to conduct its activities." While the concept was originally developed by mining industry consultants between 2000 and 2013, it has since been applied to a wider range of industries (Gehman et al., 2017). Although it seems quite similar to stakeholder theory, the SLO concept is distinct in two ways. First, the focus has typically been on "local communities [. . .] and those stakeholders who can affect Corporate social responsibility practices profitability," rather than all stakeholders (Voyer and van-Leeuwen, 2019). Second, it concerns the reputation of an organization, rather than its responsibilities (Hurst et al., 2020). Institutional theory is a useful lens to understand how policymakers approach CSR (Contrafatto et al., 2019). Meyer and Rowan (1977) stated that organizations could only become legitimate, and therefore survive, if they aligned themselves with the norms and practices of their environment, a process they labeled isomorphism. They mentioned that these norms and practices were set by social, economic and cultural institutions. The link between institutional theory and CSR was first made by DiMaggio and Powell (1983), who discussed how coercive, mimetic and normative isomorphism influenced how organizations responded to social expectations. Coercive isomorphism is the pressure exerted by both formal and informal authoritative entities that an organization depends on. Mimetic isomorphism is an organization's desire to mimic other organizations, who they perceive as more advanced or competent. Normative isomorphism is an organization's desire to standardize the conditions and methods of their industry. A critical implication of this theory is that organizations, within a particular environment (Scott, 2008), and even across different environments (Matten and Moon, 2008), will eventually homogenize.

Managing corporate social responsibility across borders
MNCs have challenged the way that governments across the globe think about national and international regulation (Crane et al., 2019). MNCs typically take one of two forms, namely, actor-based or network-based (Ruggie, 2017). In the actor-based model, the corporation is headquartered in one country, but is directly linked to its subsidiaries that are based in other countries. In the network-based approach, the corporation is headquartered in one country, but is indirectly linked to a vast web of suppliers across multiple countries, which are often, themselves, subsidiaries of other companies that are headquartered in yet other countries. Therefore, the "parent" company is legally separated from its subsidiaries and suppliers, as they are in different regulatory jurisdictions (Nolan, 2017). In this regard, there are two main criticisms. First, MNCs can protect themselves from liabilities in institutionally strong countries (Ruggie, 2017). Second, MNCs can exploit vulnerabilities in institutionally weak countries (Giuliani, 2019;Ullah et al., 2021).
With respect to exploiting institutionally weak countries, there are two main issues. First, MNCs will pursue a direct policy of shifting their operations to the countries with the weakest regulations to take advantage of cheaper production opportunities and reduced liability (Bae et al., 2021). In turn, these countries will be prompted into competing with one another vis-à-vis deregulation and tax enforcement (Kocher, 2021). Second, for MNCs that pursue a policy of outsourcing, suppliers will tend to outsource labor-intensive and lowvalue requirements to other suppliers as well (Davies and Ollus, 2019). In turn, these secondary suppliers will then outsource these requirements to yet other subcontractors, and so on (Crane et al., 2019). Ultimately, production will take place in factories that are hidden from view, where exploitive labor practices and human rights violations are core elements of the business model (Khan, 2018).
In this context, international treaties and conventions that place obligations on states, rather than MNCs, are ineffective in controlling conduct (Cutler and Lark, 2020;Ruggie, 2017). In this regard, Bair (2015)  obligations of MNCs toward host countries, and, after revision, the obligations of host countries toward MNCs. However, 18 years after drafting had begun in 1974, disagreements between member states, as well as changing economic circumstances, led to the abandonment of the Code of Conduct. As Mantilla (2009) discusses, the Draft Norms were a second attempt at creating a legally binding code for MNCs. Although, in contrast to the Code of Conduct, the Draft Norms were presented as a summary of existing international laws with mechanisms for monitoring, verification and enforcement. The idea was that MNCs could be held directly accountable for their conduct without the need for a new international treaty. However, as Bair (2015) states, the negative reaction from governments and the business community led the United Nations to reconsider the legal requirement of the instrument. Ultimately, after seven years since drafting began in 1999, the Draft Norms were declared to be entirely voluntary.
Consequently, when it comes to governing CSR at an international level, soft law and private regulation are important options (Cutler and Lark, 2020;Eller, 2017;Omoteso and Yusuf, 2017;Tamvada, 2020). In contrast to hard law, which concerns legal and enforceable obligations and rights, soft law refers to statements of expected behavior without enforceability, typically in the form of norms, principles and standards (Choudhury, 2018). In contrast, private regulation refers to rules that companies submit themselves to voluntarily (Eller, 2017). Internationally, these regulations are monitored by external agencies and are seldom legally enforceable (Berger-Walliser and Scott, 2018).  (Choudhury, 2018), private regulation instruments can be produced by either IGOs or international nongovernmental organizations (INGOs) (Berger-Walliser and Scott, 2018). There are also ICSRIs that have been coproduced by both IGOs and INGOs (Henry et al., 2019).

Categorizing international corporate social responsibility instruments
ICSRIs can be distinguished according to the way that commercial organizations are able to participate in or use them (Berger-Walliser and Scott, 2018). Certain ICSRIs may require certification (Raynolds et al., 2014;Schönherr et al., 2022). This is a verification of compliance, often resulting in the enterprise being allowed to use a particular label on their products (Lebaron et al., 2017). Bourguignon et al. (2020) discuss the International Framework Agreement (IFA) as an international CSR instrument. These are agreements between international unions and individual companies. Some CSR instruments may be membership based (Raynolds et al., 2014;Choudhury, 2018). These are instruments produced by organizations to set out the behavioral expectations of their own members (Perez et al., 2019). In addition, instruments may be open, such that compliance is completely discretionary with no obligations for the producing organization or the enterprise (Mähönen, 2020). Dankova et al. (2015) distinguish ICSRIs based on the type of activity involved in their application. They differentiate between management and reporting instruments. Management instruments focus on the day-to-day activities, operations and practices of a company (Abbas, 2020). Reporting instruments instruct companies on how they should report the impacts of their activities to the public (Camilleri, 2015;Pope and Lim, 2020;Siew, 2015). Henry et al. (2019) mention marketing and communication instruments as another distinguishing activity. These instruct companies on the ethical ways to promote their products or brand. Siew (2015) draws attention to assessment instruments as another distinguishing activity. These instruments are designed to allow companies to measure and evaluate their own impacts.
Corporate social responsibility practices Dankova et al. (2015) and Liubachyna et al. (2017) use stakeholder involvement as a distinguishing characteristic in their frameworks. Similarly, the external orientation of an instrument could be a distinguishing characteristic as well. Certain ICSRIs are designed with the intention of encouraging corporations to address wider social, environmental or economic problems, beyond only managing their own impacts or internal operations (Brown et al., 2018).
Both Dankova et al. (2015) and Raynolds et al. (2014) categorize instruments according to the specific CSR issues they address. Both make distinctions for human rights, labor practices, social performance and environmental performance; Slavery and gender equality are common topics of discussion when it comes to human rights issues (Ullah et al., 2021). Labor practices refer to the core rights set out in the International Labour Organization's Declaration on Fundamental Principles and Rights at Work, covering themes such as wages and worker safety (Khan, 2018). Social performance refers to development and value creation criteria aimed at local communities, including themes such as support for community organizations and infrastructure development (Hermans and Reyes, 2020). Environmental performance refers to impacts on natural spaces and resources (D'Souza et al., 2020), and covers themes such as conservation and pollution (Raynolds et al., 2014). Dankova et al. (2015) include economic performance and product responsibility criteria in their framework. Economic performance is the level of contribution to local or national economic objectives, covering themes such as job creation and knowledge transfer (Agudelo et al., 2019). Product responsibility is a combination of various product related attributes including themes such as safety and quality (Zeisel, 2020). Given that issues such as transparency and corruption are commonplace in CSR programs (Kaymak and Bektas, 2017), governance can also be understood as a distinguishing issue.
ICSRIs can be distinguished according to the industry sectors to which they are applicable (Fransen et al., 2019;Knudsen, 2017;Perez et al., 2019). These can be broad, such as the extractive and garment industries (Knudsen, 2017), or specific, such as the tea and flower industries (Fransen et al., 2019). Moreover, some schemes may be applicable to multiple, and even all, industry sectors (Perez et al., 2019).

Research design
To achieve the objectives of our study, we first had to define both the type of material we wanted to collect and the type of instruments we wanted to extract from that material. Regarding the type of material, we set the collection boundary to articles from high-impact journals written in English. No restrictions were placed on publication dates. Concerning the type of instruments, we defined CSR instruments as tools of governance (Steurer et al., 2012) that promote cultural and managerial transformation, reinforce accountability, delineate rules of behavior and establish systems for controlling activities (Zinenko et al., 2015), with respect to society's expectations (Carroll, 2016).
These instruments had to have been produced by international organizations. In keeping with Pease's (2016) work, international organizations were defined as organizations with members or activities in more than one country, including those comprising multiple governments, as well as non-and for-profit private organizations that were active in multiple countries. Furthermore, the instruments had to target commercial enterprises, and should not have been created for the sole purpose of managing the producing organizations' own CSR obligations.
For the first objective, we used a systematic literature review design, based on the recommendations of Snyder (2019). The literature review used the Preferred Reporting Items SAMPJ 13,4 for Systematic Reviews and Meta-Analyses (PRISMA) framework outlined by Page et al. (2021). This produced a collection of journal articles that contained references to ICSRIs. A brief statistical analysis of the articles was performed using several of the publicationmetrics proposed by Donthu et al. (2021). For the second objective, Neuendorf's (2017) content analysis method was used to extract the ICSRIs from the articles and compile them into a list. To achieve the third objective, we adapted a framework from Dankova et al. (2015) to classify the listed ICSRIs.
3.2 Collecting the journal articles and performing the publication-metric analysis Microsoft Academic, a scholarly search engine with an extensive index and semantic search, was used to identify articles in the data set (Choong et al., 2014;Visser et al., 2021). Key word combinations were entered into the search engine, composed of the key word "CSR," one regulatory and governance key word, and one international-level key word, as exhibited in Table 1. The key word "CSR" was chosen over "corporate social responsibility" after a preliminary investigation on the search engine. Both terms produced identical results. Altogether, there were 680 key word combinations, an example of which is "CSR Accountability Africa." These produced record lists in the search engine, which we then filtered to display journal articles only. Details of individual records were extracted from these lists based on the Microsoft Academic tags, displayed underneath each record, and placed into a Microsoft Word document. Only the records with tags that matched all three key words were included. A total of 1,597 records were collected in this way, over five consecutive days in April 2021.
After removing 568 duplicates, we screened the titles and abstracts of the remaining 1,029 records to remove the articles that did not meet our quality and language requirements. We removed 266 articles published in journals that were not listed on the Notes: For Regulatory and Governance, "Administrate," "Authorize," "Controlling," "Convention," "Guideline," "Legislate" and "Legislating" were included; for International Level, "Intergovernmental organization" was included. However, these yielded no results Corporate social responsibility practices SCImago Journal and Country Rank, and 4 articles that were not written in English. Next, we screened the full-text articles of the remaining 759 records, which we accessed via the websites and databases of the relevant publishers (e.g. ScienceDirect, Emerald Insight, Wiley Online Library). This involved reading each article to identify references to ICSRIs that matched our criteria. Accordingly, 228 articles were removed for not containing references to CSR governance or regulatory instruments, 132 articles were removed because the referenced instruments were not produced by international organizations and 3 articles were removed because the referenced instruments were for internal use only. This process left us with 396 articles, each including at least one reference to a relevant ICSRI. Figure 1 illustrates the overall collection process. Before continuing, an Excel sheet was used to analyze the articles according to their publication metrics, such as date of publication, journal of publication and research design.

Compiling and classifying the list of international corporate social responsibility instruments
The 396 articles collected from the systematic literature review represented the units of sampling for the content analysis. The units of data collection, and subsequent units of analysis, were the in-text mentions of ICSRIs, relevant to the predetermined criteria, contained within those articles. To find these mentions, we read through the main body of each article. When a mention of a CSR instrument was detected, it was extracted and listed into a Microsoft Word document. As the objective was to find the actual instruments, we focused on the context of each mention. However, each mention needed to contain a referenceable instrument and organization. Where appropriate, the instrument was recorded at its highest designation. For example, mentions of "ISO 14001" or "14031" were recorded under the title "ISO 14000 Series." Overall, a list of 229 ICSRIs was compiled. Using Microsoft Excel, the listed ICSRIs were classified according to seven categories, namely, producing organization, participation, activity focus, stakeholder involvement, external orientation, CSR focus and industry application.

Publication metrics
As shown in Figure  The collected articles were published by 215 different journals. As displayed in Figure 3, there were only nine journals that were represented by five or more articles each. These accounted for 115 (29%) articles. The three most represented journals were the Journal of Business Ethics, Corporate Governance (Emerald Publishing) and Sustainability (MDPI). Table 2 shows the research design and methodologies of the articles. The most popular research design was qualitative (n = 265, 67%), followed by quantitative (n = 102, 26%), mixed methods (n = 23, 5%) and book review (n = 6, 2%). The three most popular methodologies were conceptual (n = 213, 54%), followed by secondary data analysis (n = 60, 15%) and content analysis (n = 31, 8%).

Content Analysis
In total, the titles of 229 ICSRIs were collected. Figure 4 displays each of the 20 ICSRIs that were mentioned in 22 or more articles. The remaining 209 instruments were mentioned in 21 or less articles. The three most frequently mentioned instruments were the United Nationals Global Compact (n = 215, 54%), the Global Reporting Initiative (n = 155, 39%) and the OECD Guidelines for Multinational Enterprises (n = 104, 26%).  19821985198819911994199720002003200620092012 Corporate social responsibility practices 4.3 Categorizing international corporate social responsibility instruments 4.3.1 Producing organization. Most of the instruments from our sample were produced by INGOs (n = 173, 76%), followed by IGOs (n = 47, 21%) and, finally, both IGOs and INGOs (n = 9, 4%). 4.3.2 Participation. The majority of the instruments from our sample were open (n = 80, 35%). The following most common instruments were IFAs (n = 56, 24%), membership (n = 47, 21%) and assurance (n = 46, 20%). All the IFAs were produced by INGOs. Of the remaining assurance and membership instruments, the following should be noted. For INGOs, assurance instruments were the most common, followed by membership instruments. For IGOs and both IGOs and INGOs, membership instruments were the most common, followed by assurance.

Stakeholder involvement.
Only 63 (28%) out of the 229 instruments from our sample specified the need for stakeholder involvement. The number of instruments with stakeholder involvement produced by IGOs and INGOs are 13 out of 47 (28%) and 47 out of 173 (27%), respectively.  4.3.5 External orientation. Only a small number of the instruments from our sample featured an external orientation (n = 41, 18%), and these were proportional between IGOs and INGOs. The results of the producing organization, participation, activity focus, stakeholder involvement and external orientation classifications are displayed in Table 3. 4.3.6 Corporate social responsibility focus. Labor practices and worker protection was the most common CSR focus subcategory in our sample (n = 142, 62%). This was followed by governance and legal compliance (n = 125, 55%), human rights (n = 112, 49%), environmental performance and protection (n = 108, 47%), social performance and protection (n = 85, 37%), economic performance and contribution (n = 58, 25%) and product quality and safety (n = 57, 25%). These results are illustrated in Figures 5 and 6.
For instruments produced by both IGOs and INGOs, the most common CSR subcategories in our sample were governance and legal compliance and human rights, both being addressed by six (67%) of the nine instruments. These were followed by labor practices and worker protection (n = 5, 56%), environmental performance and protection (n = 4, 44%), social performance and protection (n = 2, 22%) and both economic performance and contribution (n = 1, 11%) and product quality and safety (n = 1, 11%). 4.3.7 Industry application. We identified 21 industry application categories in our sample, displayed in Table 4. The category with the most instruments was "any," meaning the instruments could be applied in any industry. This category accounted for 69 (30%) of the 229 instruments. Similarly, there were 14 (6%) instruments that could be applied in multiple industries. Of the instruments applicable to specific industries, the four most common were agriculture (n = 19, 8%), manufacturing and engineering (n = 17, 7%),  SAMPJ 13,4 minerals and fossil fuels (n = 16, 7%) and banking finance and investments (n = 16, 7%). The three least common were chemistry (n = 2, <1%), sports (n = 2, <1%) and pharmaceutical (n = 3, 1%). A noteworthy finding is the different proportions in industry application between IGOs and INGOs. For INGOs, instruments applicable to any industry accounted for 40 (23%) out of 173 instruments. There were also instruments applicable to every industry category. For  Corporate social responsibility practices IGOs, instruments applicable to any industry accounted for 26 (or 55%) out of 47 instruments. In contrast, however, instruments applicable to specific industries covered only eight industry categories.

Discussion
The findings revealed that most of the ICSRIs reflected in our sample were produced by INGOs. Given that this category is broad, consisting of instruments produced by industry associations, NGOs and private companies, this is not an unexpected finding (Perez et al., 2019;Fransen et al. (2019). These organizations and associations are more numerous than IGOs and, therefore, it stands to reason that they produce more instruments. Moreover, the audiences and objectives of INGOs can be appreciated as more specific (Perez et al., 2019), requiring instruments that suit their specific needs and objectives. In contrast, given their mandate to represent a wide range of constituents with mutual concerns, IGOs need to produce only a few instruments that apply broadly (Berliner and Prakash, 2015). With respect to the lower number of instruments coproduced by IGOs and INGOs, these may only be needed on the rare occasions when the interests of both types of organization converge and the competencies of each are required. Regarding instrument participation, most ICSRIs were open. This could be anticipated from Berger-Walliser and Scott (2018), who explained that the low liability is highly appealing to businesses. IFAs were the second most common type of instrument in our sample. This could be an indicator of its growing popularity as a mechanism to control MNC behavior (Bourguignon et al., 2020). The research also revealed that open and assurance instruments were the most popular types produced by IGOs and INGOs, respectively. This reinforces the work of previous researchers (Cutler and Lark, 2020;Eller, 2017;Omoteso and Yusuf, 2017;Tamvada, 2020)  The majority of instruments from our sample were management instruments, followed by reporting instruments. These findings reinforce the work of Dankova et al. (2015) and Siew (2015) who identified a comparatively higher proportion of these instruments in their studies. In general, the activity focus subcategory numbers highlight the activities that are typically at the discretion of the corporation, but which outside parties would like to influence, that is to say, the areas that cannot be covered by law or civil society monitoring. Hence, the relatively low number of other types of instruments in our sample (i.e. assessment, communication and marketing) is understandable.
The study demonstrated that most of the ICSRIs from our sample do not include stakeholder involvement. While this finding supports Liubachyna et al. (2017), it contrasts with Dankova et al. (2015). This difference can be attributed to their smaller, purposive selection of instruments. It is surprising that stakeholder involvement, or the lack thereof, is proportional between instruments produced by IGOs and INGOs. This could imply that there are issues with the practicality of stakeholder involvement, or that international organizations are more interested in establishing universal practices rather than getting MNCs to develop tailor-made solutions for their stakeholders, as previous researchers have argued (Berliner and Prakash, 2015;Brown et al., 2018).
The results also revealed that few ICSRIs feature an external orientation. As this implies a relatively weak contribution to local area development, this finding contrasts with Kocher's (2021) presumption that voluntary regulation is designed to replace the role of governments or IGOs.

SAMPJ 13,4
In our study, we identified 21 industry application categories. While the instruments applicable to any or multiple industries are not the most common in Raynolds et al.'s (2014) study, it is a finding that is reflected by the instruments included in Knudsen's (2017) study. In this study, a large difference in the coverage of the instruments between IGOs and INGOs was identified. There were INGO instruments for every category, whereas IGO instruments were only in eight industry categories. Moreover, there was a large proportion of instruments applicable to any industry between IGOs and INGOs. These findings support the argument that INGOs need more instruments to suit a wider audience (Perez et al., 2019), while IGOs need fewer instruments that can be applied more broadly (Berliner and Prakash, 2015).
Concerning CSR focus, this study demonstrated that a high number of instruments include governance criteria. This finding supports Kaymak and Bektas (2017). The high representation of governance in the instruments could indicate that IGOs and INGOs identify this subcategory as especially critical for social impact or harm, as Ucar and Staer (2020) had suggested. The relationships between the remaining six CSR focus subcategories in this study closely resembled the findings of Dankova et al. (2015), except for environmental and social criteria, which were equal in their study.
Labor was the most common area of CSR focus. This is unsurprising as there has been public concern for labor rights at least since the labor struggles around the world of the 1930s (Marens, 2008), predating the notion of CSR by about 30 years (Agudelo et al., 2019;Carroll, 1999). By comparison, concern for the impact of business on human rights and the environment is relatively recent, only gaining popularity in the 1970s (Agudelo et al., 2019;Mantilla, 2009). Along these lines, it is understandable that long-standing issues of public interest feature more strongly in ICSRIs. It is unexpected, however, that the inclusion of human rights and environmental issues have not caught up to labor issues, given that human rights (Stohl and Stohl, 2010) and environmental (Doh and Guay, 2006) organizations were among the first to promote the concept of CSR.
The varying inclusion of social performance and protection criteria between the instruments is also worth investigating. It is one of the highest focus areas in the instruments produced by IGOs, and one of the lowest focus areas in the instruments produced by INGOs and coproduced by IGOs and INGOs. While public concern for the impact of business on society is not as old as concern for labor rights, it can be placed within the same time period as public concern for the impact of business on human rights and the environment (Agudelo et al., 2019). The mismatch may be due to differing priorities. INGOs might be more concerned with eliminating harmful MNC behaviors. A firm's lack of support for nearby local communities may be seen as negative, but not necessarily harmful. Instead, the abuse of workers and indigenous peoples may be seen as causing direct harm and therefore be of greater concern. IGOs might be more concerned with fostering beneficial behaviors to generate useful economic and social outcomes at the macro level.
Similar reasoning could be used to explain the low inclusion of economic criteria in the identified instruments. For IGOs, getting businesses to contribute to national and regional issues may be a crucial milestone in achieving their own goals (Hohnen, 2008). INGOs could be more concerned with reducing everyday human rights and labor violations and holding companies accountable for their harmful behaviors (Denedo et al., 2017). It may be that regulating product safety and quality is considered a legal rather than a CSR issue, explaining the all-around low attention to product issues in the instruments.
Corporate social responsibility practices 6. Implications 6.1 Theoretical When examined through stakeholder theory, it is apparent that, by and large, the ICSRIs are not aligned with Dmytriyev et al.'s (2021) tenets of: creating value for all stakeholders; basing decisions on ethical motivations; and developing mutually beneficial stakeholder relationships.
Regarding the first tenet, the stark difference in CSR focus, especially for instruments developed by INGOs, suggests that certain stakeholders are prioritized. For the second tenet, the high attention to governance in most instruments indicates a strong attention to accountability. Yet, the limited inclusion of external orientation criteria implies that these instruments are not intended to encourage ethical decision-making. Concerning the last tenet, the lack of stakeholder inclusion criteria suggests that improving stakeholder interactions is not a core purpose of these ICSRIs.
To measure SLO, Gehman et al. (2017) promote the 14 questions developed by Black (2017). Our results suggest that the ICSRIs from our sample reflect only seven of these measures. Accordingly, there is only weak support to suggest that the instruments are aligned with the SLO concept. From a more general perspective, however, there appears to be some agreement with the SLO concept. Gehman et al. (2017) explain that the SLO is a unidirectional concept. In other words, an evaluation of a firm from external individuals or groups. In our results, this is implied by the activity focus of the instruments. If they were more bidirectional, the presence of another activity focus category for stakeholder relationship management would be expected (similar to the AccountAbility AA1000 Stakeholder Engagement Standard).
Prioritization of the issues and stakeholders that are most likely to damage reputation (Hurst et al., 2020) or profitability (Voyer and van-Leeuwen, 2019) are also features that have been associated with the SLO concept. Our findings suggest that this is case for the instruments produced by INGOs, given the uneven attention to CSR focus criteria. Labor and human rights issues would quickly be picked up by civil society actors (Ullah et al., 2021), who could catalyze public disapproval (Henry et al., 2019). The same goes for reports of bribery, corruption or tax avoidance (Ucar and Staer, 2020), which would also attract interventions from local governments. Dissatisfied employees could strike, disrupting operations and revenue. Moreover, given that the environment is a "hot topic" (de Freitas Netto et al., 2020), this issue may be prioritized. On the other hand, this does not appear to be the case for the instruments produced by IGOs, which have more balanced CSR criteria.
Institutional theory offers valuable insights into the isomorphic mechanisms at work in ICSRIs. Coercive mechanisms, which apply external pressure on organizations, are best represented by the assurance and IFA instruments. Mimetic mechanisms, which involve the dissemination of best practices, are best represented by the open instruments. Normative mechanisms, which concern an internally generated desire to professionalize, are best represented by the membership instruments. From our sample, most of the instruments produced by INGOs use coercive mechanisms (n = 99, 57%). This is followed by instruments using mimetic (n = 45, 26%) and normative (n = 29, 17%) mechanisms. In contrast, it appears that most of the instruments produced by IGOs use mimetic mechanisms (n = 31, 66%), followed by normative (n = 15, 32%) and coercive (n = 1, <1%) mechanisms. The higher number of coercive instruments suggests that INGOs believe that CSR can only be guaranteed through external verification, perhaps because of the threat of SAMPJ 13,4 "blue-washing" (Berliner and Prakash, 2015). This is reinforced by the greater range of industry-specific instruments produced by INGOs. The higher number of mimetic instruments suggests that IGOs believe that to improve international CSR, there is a need to establish universal basic practices (Berliner and Prakash, 2015;Brown et al., 2018). This statement is supported by the higher number of cross-industry instruments produced by IGOs.

Practical
In support of voluntary regulation, this study has demonstrated that most of the analyzed instruments focus on worker protection, human rights, governance and environmental protection. As such, they are useful contributors in establishing and solidifying worldwide norms that international law has, so far, struggled to achieved independently (Cutler and Lark, 2020). However, the instruments in our sample rarely oblige corporations to address wider developmental issues beyond their own impacts or establish a process to engage with stakeholders. This suggests that voluntary regulation has failed to normalize the idea that corporations are inclusive institutions that should contribute to local and regional development (Agudelo et al., 2019). Furthermore, instruments from our sample seldom focus on issues related to product quality and safety, economic contribution and social protection. As such, it appears as though many of these instruments have not been developed to fulfill a corporation's obligations in line with stakeholder theory (Freeman, 2010) or the CSR pyramid (Carroll, 2016). Instead, many instruments concentrate on improving reputation and appealing only to those stakeholders who may affect profitability, in line with the SLO concept (Hurst et al., 2020;Voyer and van-Leeuwen, 2019). Based on the above, four recommendations are made below.
First, both IGOs and INGOs balance the CSR focus subcategories that are addressed in their instruments to reflect the categorizations used in this study. Although CSR instruments need to be applicable to the distinct social, political and business interests of the localities or industries they are intended for, CSR is about meeting social expectations (Carroll, 2016). Accordingly, while ICSRIs are expected to be diverse, that diversity should be in the form of different approaches to each CSR focus subcategory rather than the specific targeting of one subcategory over another. The reconciliation of CSR focus subcategories could be achieved by updating existing instruments, or by developing new ones to compensate for neglected focus areas.
Second, in this same vein, instruments need to be revised or developed to include stakeholder involvement and external orientation criteria.
Third, the creation of a meta-standards organization, like ISEAL, but for ICSRIs is required. The study revealed that certain criteria were considerably underrepresented in the instruments, such as stakeholder involvement and external orientation, as well as certain CSR focus subcategories. This implies two critical threats to voluntary regulation. First, that certain instruments may have been designed to allow adopting firms to give the impression of social responsibility. Second, that the competition between voluntary regulation developers is high (Bartley, 2022). As a result, there is a risk that instrument developers will make their standards less rigorous to encourage more companies into adopting them. A meta-standards organization would go far in promoting the most effective instruments. This organization could include the participation of IGOs, such as the United Nations Economic and Social Council. It could give its seal of approval to the instruments that best address universally important and industryspecific criteria.
Fourth, we recommend that a website be established, managed and maintained through an international collaboration involving universities and civil society organizations under the supervision of an IGO. This website should list and classify all the ICSRIs according to Corporate social responsibility practices the framework developed in our study. This would be useful for a number of reasons: it would support efforts to solidify global CSR norms; it would facilitate the navigation of ICSRIs; and it would increase public and commercial awareness of the properties and comprehensiveness of different ICSRIs.

Conclusion
The aim of this study was to understand the properties of the instruments that regulate and govern CSR at the international level. This was achieved by using a systematic literature review method to collect journal articles containing references to ICSRIs; the use of a content analysis method to extract and place those instruments into a list; and the adaption of a classification framework to categorize the listed instruments according to their properties. This approach has been successful, and from 396 highly relevant articles, an extensive list of 229 ICSRIs was compiled, a classification framework was applied and the research question was answered.
In conclusion, most ICSRIs in our sample were produced by INGOs. Instruments produced by IGOs and coproduced by IGOs and INGOs are, respectively, the second and third most common. Most instruments were open, placing no obligations on the producer or targeted user. IFAs were the second most common, while assurance and membership instruments were the least common types. The majority of the instruments concentrated on management activities. Reporting, assessment and communication and marketing instruments were the least common activity types. Most instruments did not specify stakeholder involvement and only concentrated on the internal functions and impacts of the targeted user.
Overall, labor practices and worker protection were the most common focus of the instruments from our sample. This was followed by human rights, governance and legal compliance and environmental performance and protection. The least common issues covered by the instruments were product quality and safety, economic performance and contribution and social performance and protection. There were considerable differences in focus between IGO and INGO instruments. The majority of the instruments were targeted at specific industries. The remaining instruments were targeted at multiple or any industries. There are some large differences in industry application between the instruments produced by IGOs and INGOs.
This study has made several contributions to the knowledge of international CSR. New broad-based understandings about voluntary regulation have been generated, and valuable information has been brought into the CSR literature. In addition, our research draws attention to previously overlooked dimensions of CSR instruments by using external orientation as a classification category, communication and marketing as an activity focus subcategory and governance and legal compliance as a CSR focus subcategory. Furthermore, the quantitative approach used in the study has set a useful precedent, laying out a road map that future researchers can easily follow.
The timeliness of the study is emphasized by the increasing CSR-related regulation that the recent years have witnessed. After more than 20 years since the international community resigned itself to the idea of voluntary regulation, questions of its efficacy, and the need for legal mechanisms, are again being revisited (Tamvada, 2020). The Modern Slavery Act of 2015 in the UK, the Loi de Vigilance of 2017 in France, the Child Labour Due Diligence Law of 2019 in The Netherlands and the Lieferkettengesetz of 2021 in Germany are all evidence of increasing public concern for MNC behavior. Along these lines, the findings of the study empower policymakers, industry practitioners and civil society actors to make informed decisions regarding the future of international CSR.

Future research
This study examined the properties of ICSRIs using broad categories. We compared the nature of the instruments according to their producing organizations. Future research should investigate the differences between instruments according to their participation types and industry applications. For example, examining how CSR focus differs between membership and certification instruments, or how external orientation differs between instruments designed for the textile and banking industries. Moreover, the categories used in this study could be further fragmented to reveal further instrument differences. For example, the INGO category could be subdivided into industry associations, civil society organizations and private companies. Furthermore, although the classification framework was developed following an extensive literature review, there are still categorization types that could enhance its comprehensiveness. Future researchers can enhance the classification framework by identifying and including additional categorizations.