International Diffusion of Corporate Social Responsibility: The United Nations Global Compact

(1)

-

Master Thesis Double Degree Program

M.Sc. International Economics and Business – University of Groningen M.A. International Economics – University of Göttingen

International Diffusion of Corporate Social Responsibility:

The United Nations Global Compact

Name: Hanna Sophie Schmidt

E-Mail: hanna.s.schmidt@gmx.de Std.Number: S2417812 (Groningen)

21157755 (Göttingen)

Affiliation: Faculty of Economics and Business (Groningen) Faculty of Economic Sciences (Göttingen)

Submission: 8th July 2013

(2)

LIST OF ABBREVIATIONS

CSO Civil Society Organization CSR Corporate Social Responsibility

DESA Department of Economic and Social Affairs ECOSOC Economic and Social Council

FDI Foreign Direct Investment

FE Fixed Effects

FEVD Fixed Effects Vector Decomposition

GDP Gross Domestic Product

ILO International Labor Organization ISO International Standard Organization

NGO Non-Governmental Organization

OLS Ordinary Least Squares

RE Random Effects

UN United Nations

UNGC United Nations Global Compact

VIF Variance Inflation Factor

(4)

LIST OF TABLES

Table 1: The 10 Principles of the UNGC ... 5

Table 2: Descriptive Statistics ... 22

Table 3: Correlation Matrix ... 23

Table 4: UNGC Launch (Probit) ... 25

Table 5: Diffusion of UNGC (FE) ... 28

Table 6: Early vs. Late UNGC Launch (Probit) ... 30

Table 7: Early vs. Late UNGC Diffusion (FE) ... 32

Table 8: UNGC Launch incl. Interactions (Probit) ... 36

Table A1: Variable Description ... 53

Table A2: List of Countries Included in Analysis... 54

Table A3a and A3b: VIFs for UNGC Launch and Diffusion (Full Sample) ... 54

Table A4: Average Partial Effects (Full Sample) ... 55

Table A5a and A5b: VIFs for Early and Late UNGC Launch ... 55

Table A6a and A6b: VIFs for Early and Late UNGC Diffusion ... 56

Table A7: Average Partial Effects (Early and Late UNGC Launch) ... 56

(5)

ABSTRACT

This paper examines the drivers of the diffusion of corporate social responsibility (CSR) standards using the example of the United Nations Global Compact as one of the largest and most prominent examples of CSR. The analysis is embedded into the framework of institutional theory, modeling the diffusion of standards following efficiency and legitimacy reasons. As an expansion of this theoretical model, stakeholder pressures are considered as a further national legitimating factor impacting the diffusion of CSR on a firm level. The relevance of stakeholder pressures for the adoption of CSR standards is confirmed by highly significant empirical results.

Keywords

(6)

1. INTRODUCTION

The discussion about the need for non-state mandated regulatory institutions such as voluntary corporate social responsibility (CSR) standards has been refreshed by the recent occurrences in Dhaka, Bangladesh. The collapse of the textile factory building in Dhaka in April 2013 with more than 1000 casualties and more than 2000 injured persons is categorized as one of the worst industrial accidents worldwide. This tragic incident in the textile industry, representative for many other cases in the manufacturing industry, illustrates that workers are still treated systematically in inhumane ways. Corruption and avarice at leading positions add to the violation of workers’ human and labor rights. In case of the factory caving-in in Dhaka, workers were forced to report to work, although major deficiencies had been detected at the basic structures of the building which had been constructed partly illegally (Tagesschau, 2013).

After the tragedy the question of responsibility for this incident and the bad working conditions in general has arisen in the media. The accusation of the firm itself, the government and consumers is driven forward by labor unions, critical consumers and non-governmental organizations (NGO). Activists like the Clean Cloth Campaign blame Western firms for not advocating safe working conditions in their suppliers’ factories. Most of Bangladesh’s textile exports go to Europe where labor legislation has already reached a different level. Other voices like the US government claim that it is a government’s responsibility to advance the current situation, settle the grievances and improve the workers’ conditions (Spiegel Online, 2013). After the caving-in in Dhaka Bangladesh’s government started allowing independent labor unions in the textile industry. Although this can be evaluated as a small progress in the field of workers’ rights, many other basic rights still remain not granted. It can be furthermore criticized that after a catastrophe like in Bangladesh generally a lot of discussions are provoked by the media and improvements of the current situation are planned but in the end no major changes are imposed.

(7)

reluctant to regulate in an area of public concern (Scherer & Palazzo, 2011). Business firms are said to engage in this form of self-regulation in order to fill the regulatory vacuum on an international level (Scherer & Palazzo, 2011). By choosing voluntary standards instead of obligatory laws, firms do not face the increasing regulatory burdens that might weaken their competitiveness on global markets (Vogel, 2008). Nevertheless, laws are generally rated to generally provide more effective outcomes. Thus, voluntary standards represent a second best option and are helpful until the role of the government as well as the public attitude change (Nordmann, 2011).

Tragedies like in Bangladesh might be prevented provided that multinational companies themselves intervene on their workers’ behalf. Within the context of economic globalization, there are more demanding stakeholder expectations and a growing pressure of the civil society on firms to expand CSR activities. Negative publicity with respect to corporate abuses, environmental disasters and child labor has amplified the calls for social responsibility among business entities. Due to growing media exposure, the firms’ vulnerability has increased. CSR certification arrangements have become more proactive in time and NGOs have been trying to improve corporate behavior and to prevent accidents and misbehavior in the first place (Gereffi, Garcia-Johnson, & Sasser, 2001; Lim & Tsutsui, 2012; Scherer & Palazzo, 2011). For the effective working of CSR practices they need to be spread globally, otherwise the danger of a downward spiral in social and environmental conditions exists (Lim & Tsutsui, 2012; Scherer & Palazzo, 2011). One reason for this is that transnational corporations are able to choose among various legal systems that differentiate in jurisdiction and locations and compete for investors (Surroca, Tribó, & Zahra, 2013). Thus, firms can escape local jurisdictions if necessary. They are said to largely operate in a legal vacuum in global affairs. Although production and trade have spread worldwide, there is a lack of institutions and international laws that are able to regulate businesses on a global level (Detomasi, 2011). To overcome this gap, transnational CSR initiatives like the United Nations Global Compact (UNGC) can help to face externality problems that single firms and countries cannot solve unilaterally (Scherer & Palazzo, 2011).

(8)

factors make a firm start adopting the UNGC standards in a country where these were not standard beforehand. With the adoption of the UNGC for the first time in a country, the adopting firm fulfils the role of an innovator. Then, the country-wide diffusion of the UNGC is analyzed and the factors that impact the number of firms that are members of the UNGC per country are examined.

The research is embedded into the theoretical approach of institutional theory that differentiates between the operationalization of efficiency needs and the operationalization of legitimacy pressures that can lead to the adoption of CSR standards (Aguilera & Cuervo-Cazurra, 2004; DiMaggio & Powell, 1983). In line with prior studies about the diffusion of voluntary standards, I measure the diffusion of the UNGC on a country level. This choice of level is justified because many pressures are exerted on a country level and a firm’s policy is strongly influenced by the national background (Berliner & Prakash, 2012). Additionally, various stakeholder groups are incorporated in the analysis as influencing factors for the implementation of CSR standards. By considering the individual company’s environment, the firm’s perspective is taken into account for membership decisions and the so far established level of analysis in the literature is amplified. As implied by the definition of CSR above, a firm’s actions also affect a firm’s external environment. The stakeholder approach assumes that those groups and the firm interact: the stakeholders both affect the company's actions and are affected by them. Therefore, stakeholders impose responsibilities on a firm (Papasolomou-Doukakis, Krambia-Kapardis, & Katsioloudes, 2005). Stakeholder theory is an important scientific approach when analyzing the diffusion of CSR. Consequently, in the analysis of the UNGC’s diffusion, various stakeholders and their needs and abilities to exert influence on a corporation are taken into account. Thus, the analysis considers three different influence factors that relate to efficiency reasons and two different forms of legitimacy pressures: external pressures stemming from an international context and national political pressures in the form of stakeholder legitimacy. By examining stakeholder pressures in a precise way, a more detailed analysis of the legitimacy pressures a firm faces is provided.

(9)

the UNGC into separate stakeholder categories it can be observed which stakeholders have the highest bargaining power and consequently the highest influence on a firm in terms of CSR adoption (Jamali, 2008). Based on the insights gained in this study, the current status of UNGC diffusion is assessed from a new, broadened perspective. Furthermore, some ideas on how to promote the diffusion of CSR and which stakeholder groups to be supported are presented that can help to enhance the imposition of worldwide accepted social norms.

The paper is structured as follows. The next section provides information about the UNGC, the CSR standard of interest. Section 3 presents the concept of CSR and the theoretical model of CSR diffusion. Section 4 deals with the specification of the data sources and the data collection process. Section 5 provides an overview of the concrete measures and the applied econometric techniques. Moreover, the empirical results are presented. In section 6, I amplify the baseline model and embed its analysis into a more holistic environment that takes into consideration interactions between institutions and stakeholders. In this way, a broader and more realistic perspective is depicted. Section 7 explains the outcomes, discusses the findings and presents limitations of the research. In the last section I draw a conclusion and show suggestions for future research.

2. THE UNITED NATIONS GLOBAL COMPACT

(10)

principles are based on the Universal Declaration of Human Rights (1948) and the International Labor Organization’s (ILO) Tripartite Declaration of Principles concerning Multinational Enterprises and Social Policy (1977). These principles are intended to lead the behavior of corporations, but they do not prescribe formal procedures as, for example, the ISO 9000 standards and its environmental counterpart, the ISO 14000 standard, sponsored by the ISO (Lim & Tsutsui, 2012; Perkins & Neymayer, 2010; Whitehouse, 2003).

Table 1: The 10 Principles of the UNGC

Area Principle

Human Rights 1. Businesses should support and respect the protection of internationally proclaimed human rights.

2. Businesses should make sure that they are not complicit in human rights abuses. Labor 3. Businesses should uphold the freedom of association and the effective recognition of the

right to collective bargaining.

4. Businesses should uphold the elimination of all forms of forced and compulsory labor. 5. Businesses should uphold the effective abolition of child labor.

6. Businesses should uphold the elimination of discrimination in respect of employment and occupation.

Environment 7. Businesses should support a precautionary approach to environmental challenges, 8. Businesses should undertake initiatives to promote greater environmental responsibility. 9. Businesses should encourage the development and diffusion of environmentally friendly

technologies.

Anti-Corruption

10. Businesses should work against corruption in all its forms, including extortion and bribery.

Source: UNGC, 2013.

To become an official and publicly-listed participant, a company needs to send a letter of commitment, signed by the firm’s chief executive officer, to the UN Secretary-General. The participating companies are asked to publicize their commitment to the UNGC by annually providing the UN with a report. As it is a voluntary initiative, the UNGC is not legally binding and its members’ commitment is not monitored or measured externally (Whitehouse, 2003). Thus, the UNGC relies on self-reporting by firms and voluntary monitoring by civil society actors. Supporters emphasize that the costs of participation for firms are non-trivial and that firms attract more media attention when being members. Therefore, they are more likely to be a target of criticism (Berliner & Praksh, 2012). However, the effectiveness of voluntary standards in general is highly debated. Nevertheless, the discussion about

bluewashing1 and other forms of figurative participation exceed the frame of the present paper (Corbett & Kirsch, 2001; Lim & Tsutsui, 2012).

In the literature on the diffusion of CSR standards, other authors like Lim and Tsutsui (2012) differentiate between different levels of commitment concerning the adoption of CSR

1

(11)

practices. In contrast, the focus of this study lies on the density of business entities adopting the UNGC per country and concentrates on the diffusion of the UNGC. As the UNGC does not count with external monitoring mechanisms, the commitment is probably always weaker than for standards as the ISO 14000 and therefore the distinction is not very meaningful.

3. THEORY AND HYPOTHESIS DEVELOPMENT

3.1 The Concept of Corporate Social Responsibility

Due to the course of globalization in the last decade, firms have been acting in more and more international settings. In order to regulate the diverse interests of a widespread environment, global voluntary frameworks of CSR have gained attractiveness for firms (Lim & Tsutsui, 2012). A question of growing importance is how firms can design their corporate behavior to take into consideration the political consequences of an action. According to Scherer and Palazzo (2011), “political CSR suggests an extended model of governance with business firms contributing to global regulation and providing public goods” (901).

(12)

is a multitude of global actors like governments, civil society and firms. Finally, the authors claim that a moral incentive for governments to cooperate on a global level for the provision of public goods is still lacking. Kaul et al. (1999b) affirm that due to these gaps, nation states face problems to implement policies concerning the provision of public goods. Despite the problems that are associated with intergovernmental agreements at a global level, the nation state has been so far considered as the main actor responsible to ensure the provision of global public good in contrast to firms.

However, as indicated above by Scherer and Palazzos’s (2011) definition of CSR, the relation between firms and the state has been changing in the context of firms acting increasingly in a multinational setting. This should be considered in the context of shifting perceptions of global governance, the “the process of defining and implementing global rules and providing global public goods” (Scherer and Palazzo, 2011: 900). Since the end of the 20th

(13)

anymore in times of globalization. Nevertheless, considered by itself, without any further binding constraints, the changed role allocation with a stronger focus on firms reinforces the likelihood of underprovision of public goods. However, there exist opposing forces like a firm’s reputation that positively impact the provision of public goods and that have to be considered in the context of a globalizing environment.

In general, globalization has facilitated the international presence of firms building up globally scattered value chains and manufacturing facilities are often located in countries with flexible labor legislation. Usually, the legal regulations in developing and emerging economies differ from Western standards. On the one hand, these differences can lead to cost advantages in production due to lower social costs associated with labor contracts, the lack of pension systems, flexible labor contracts etc. On the other hand, they can be a burden to firms: As a consequence of the changed role allocation and the increased expectations imposed on firms, the taking advantage of those differences, i.e. the exploitation of lacks in the regulation of human rights, labor and environmental law, leads to criticism and bad publicity. Consequently, corporations start to take on responsibilities in the social and political area that go beyond the legal requirements in order to raise or maintain their reputation. They adopt concepts of CSR. This adoption of socially and environmentally responsible behavior in form of standards follows both ethical and economic reasons. Firms can simply strive for social justice on a national and international level. However, most authors present more calculating motives for the adoption of CSR. Cetindamar and Husoy (2007) characterize the obtainment of an improved corporate image as a main goal of entering CSR initiatives. Doh and Guay (2006) stress the wish of avoiding bad publicity and therefore losing customers as a main motive for firms to adopt the UNGC. Vogel (2008, 2010) also points out the fact that firms mainly try to protect their reputation, often as a reaction to critical campaigns by NGOs against certain firms.

(14)

enforcement and sanctions (Abbot & Snidal, 2000). Examples of soft law can be found in diverse fields that deal with voluntary standards, e.g. quality or environmental management standards of the International Standard Organization (ISO) such as the ISO 9000 and ISO 14000, but also in the area of human rights as the UN Declaration of Human Rights. The adoption of soft law on a country level signals to other states that for example human rights are respected. In case of corporate responsibility standards on a firm level, a company shows to stakeholders and other companies that it follows a certain set of standards in different steps of production (Perkins & Neumayer, 2010).

Often the idea behind promoting the introduction of soft laws as the UNGC is to encourage firms to embed a set of universal, shared values into their strategy and operations, and to make them part of a compulsory regime. Then, step by step, these values are made part of internationally accepted law that institutionalizes social and environmental norms (Perkins & Neumayer, 2010; Whitehouse, 2003). By contrast, other authors like Abbott and Snidal (2000), argue that soft law itself is the preferable form of law. It should not be considered an interim step toward hard law, but its advantages such as, for example, being easier to achieve and having lower costs should be stressed. Soft law can also help to balance competition and to come to a compromise among states and private actors. Thus, it can be a foundation for efficient international contracts. Vogel (2010) takes a mediating position and suggests integrating voluntary standards more closely into national laws in order to make them more effective in their impacts. In this way standards do not substitute the legal policy, that is considered to be, in general, more effective, but rather complete it. In conclusion, the effectiveness of soft law depends on the cooperation and reinforcement of voluntary standard initiatives and public authorities. Therefore, the future development concerning the individual roles of different actors in the dynamic process of public good provision such as the state, firms, and the civil society is of unknown outcome (Vogel, 2008). Scherer and Palazzo (2011) describe that neither national states nor international institutions are individually able to sufficiently provide global public goods. Thus, neither national governments nor firms have been able so far to effectively handle most of the negative social and environmental impacts they cause by their businesses (Vogel, 2010). The current situation is hence not considered to be necessarily stable.

(15)

voluntary commitment is expected to address social and environmental issues more efficiently. Further positive aspects associated with CSR initiatives are networking and stakeholder management. A benefit of networking is seen in the diffusion of knowledge between private and public actors and the rise of new communication channels. Once initiated, the adoption of soft law leads to learning processes and further changes over time (Whitehouse, 2003). Additionally, the consideration of all stakeholders’ needs is assessed to address the ethical side of CSR (Cetindamar & Husoy, 2007). A further advantage of voluntary standards in comparison to hard law is that they are described to be more flexible and allow for a timely action. Voluntary standards are, for example, not subject to rules of international organizations as the WTO that restrict the legal implementation of adherence to social standards as a condition for suppliers. In general, the introduction of legally binding regulations for businesses is very demanding and costly (Vogel, 2008, 2010).

Nevertheless, other scholars precisely criticize the aspect of voluntarism as being ineffective and being incapable of solving the problems of regulatory failures such as the underprovision of public goods. Since the regulations are not legally binding, firms are accused to parade their commitment in order to disguise their real businesses practices and without planning to fulfill the CSR standards. Voluntary standards do not have proper enforcement mechanism and are often considered to lack legitimacy (Corbett & Kirsch, 2001; Lim & Tsutsui, 2012; Vogel 2008, 2010). However, although voluntary standards are not a perfect solution for the problem of public good provision, they have to be compared to the existing political alternatives. Especially in developing countries, hard law is rated as being ineffective in addressing issues like human rights and environmental protection. Thus, voluntary initiatives can be considered as the second best solution which is able to partially overcome governance deficits and legal failures (Vogel, 2010).

3.2 Diffusion of Corporate Social Responsibility

(16)

form of conformism, is a source of survival (DiMaggio & Powell, 1983). Homogenization due to legitimacy motivation can be caused by three mechanisms: coercive, mimetic and normative isomorphism. Isomorphism is understood as a “constraining process that forces one unit in a population to resemble other units that face the same set of environmental conditions” (DiMaggio & Powell, 1983: 149). In this way, organizational characteristics are made more compatible with the firm’s environment in order to survive on the market (DiMaggio & Powell, 1983). An example for this threefold distinction in research is provided by Guler, Guillén and Macpherson (2002) who analyze the cross-national diffusion of organizational quality standards.

Coercive isomorphism deals with the problem of legitimacy and political influence. These forces cause pressures upon an organization to justify its behavior to other organizations and the society within which it operates. Often it is also the state which acts as a demanding actor (Boxenbaum & Jonsson, 2008; Guler et al., 2002). Mimetic isomorphism is the answer to uncertainty as firms tend to restructure, following similar firms in their field which they perceive to be more legitimate or successful. Normative isomorphism is associated with professionalization. Workers collectively try to gain control over the production and build a base for legitimacy for their occupational autonomy. Furthermore, a similar education leads to similarities in the perception of values and moral duties (DiMaggio & Powell, 1983).

(17)

practices is widely recognized in the literature (e.g. Detomasi, 2011). The lack of institutional design leads firms to adopt a form of soft law, because additional rules are needed in order to overcome inefficiencies. Codes of good governance as well as CSR can complement the legal system and therefore increase efficiency when firms are operating in an inefficient system of governance. As a consequence, CSR is applied in case of internal institutional failures. Hence, particularly in countries with institutional deficiencies, the seeking for efficiency is a pronounced driving force of adopting CSR standards. Also Scherer and Palazzo (2011) as well as Scherer, Baumann-Pauly and Schneider (2013) state that companies that are engaged in social security, health and educational projects often operate in countries with failed state agencies where legal regulation and moral orientation are lacking. They suggest that CSR initiatives are established to substitute for these gaps of defective national institutions. Firms are seen as state-like actors with a political role that support state systems when these systems fail and fill the gaps of governance on a global level (Scherer & Palazuo, 2011; Scherer et al., 2012).

An example of institutional deficiencies is the presence of corruption. The General Secretary of the Trade Union Advisory Committee to the OECD describes corruption as an increasing problem of worldwide reach. As it hinders development, disturbs the allocation of resources, destabilizes a government’s trustworthiness and discriminates against the weakest subjects in a society, it causes inefficiency. Especially in businesses, dealing with high amounts of money and competing with other firms for orders or profits, the danger of corruption exists. In a corrupt environment, corruption within a firm is more likely (Evans, 2006). Concluding, corruption can be considered an indicator for inefficiency in a country (Scherer & Palazzo, 2011). Firms can be attributed a role of agents of institutional change against corruption. They can actively advocate for anti-corruption and in this way prevent material damages and an undermining of a society’s foundation caused by corruption (Scherer & Palazzo, 2011). Due to the above explained reasons, I investigate whether the presence of corruption leads to UNGC participation in order to overcome the problems associated with weak institutions.

Hypothesis 1: The greater the lack of institutional efficiency in a firm’s national environment, the more likely is a firm in this country to adopt the UNGC membership (bonding hypothesis).

(18)

coercive exogenous pressures lead to the adoption of CSR practices. Aguilera and Cuervo-Cazurra (2004) name external forces as economic openness and the presence of foreign institutional investors that can force firms to internationally legitimate the national corporate governance system. Transferred to the topic of CSR, it means that the adoption of CSR can provide legitimacy by other nations. This is especially relevant if the country context in which a firm operates is highly interconnected with other countries. Thus, exogenous pressures can lead to the introduction of widely recognized and accepted practices. This effect is also known as spillover via transnational linkages, or rather trading-up and investing-up dynamics (Aguilera & Cuervo-Cazurra, 2004; Perkins & Neumayer, 2004).

Guler et al. (2002) also stress in their paper on the impact of a country’s exposure to international trade for the international diffusion of corporate governance systems, here in form of the adoption of the ISO 9000 code. According to their argumentation, intense trade relationships cause cohesion which stimulates the likelihood of having similar numbers of certificates, due to the influence of similar customer requirements. Guler et al. (2002) further use inward foreign direct investment (FDI) over gross domestic product (GDP) to account for the extent to which each country’s firms are internationally oriented. Like this, the coercive effect of foreign multinationals investing in the country on domestic firms adoption behavior is considered. The authors emphasize that a developed country that is tightly internationally connected, displaying economic openness, faces greater pressures to harmonize its corporate governance and is therefore more likely to adopt the ISO 9000 code. Although the paper analyzes the diffusion of a quality certification and not a CSR initiative, it contains similarities in some aspects of its institutional design and the findings can be consequently transferred. Both standards are voluntary certifications that promote management systems as opposed to product standards (Berliner & Prakash, 2012). The link between international exposure to trade and CSR standards is also found in Corbett and Kirsch (2001) who investigate the diffusion of the ISO 14000 standard concerning the environmental aspects of CSR. Thus, the impact of international pressures on the introduction of CSR can be measured by high trade shares and inward FDI and is expected to enhance the diffusion of CSR.

Hypothesis 2: The higher the international economic pressures a country faces, the more likely is a firm in this country to adopt the UNGC membership (legitimacy hypothesis).

(19)

Stakeholder theory has become popular in the last years. According to this stream of literature, different influence groups pursue varying targets associated with a business firm. All of them try to advance their interests and receive the public goods they are striving for. As a consequence, they compete in power and assertiveness. The incorporation of stakeholders into the decision making of a firm is regarded as a broader responsibility to the social system that surrounds a corporation, as opposed to the narrow definition which only concerns stockholders’ economic welfare. Stakeholder and shareholder approach are opposed concerning the allocation of resources within a firm.2 Stakeholders are not only naturally influenced by firms in terms of employment relationships or product choices but they also influence the firms’ behavior. Hence, they are seen as imposing them different responsibilities. Stakeholders can therefore be categorized as an exogenous impact, although non-owning stakeholders are often not formally represented in a firm’s corporate behavior, depending on the governance structure. The broader forms of value redistribution that are part of the stakeholder view can lead to a firm’s contribution to an increased social welfare. The search for a greater stakeholder value is therefore believed to benefit economy and society as a whole (Aguilera & Jackson, 2003; Papasolomou-Doukakis et al., 2005; Perkins & Neumayer, 2010).

A further point of conflict is that CSR aims to integrate economics and ethics. Consequently, firms often have to face conflicting aims of shareholders and other stakeholder groups (Cetindamar & Husoy, 2007). Jamali (2008) suggests an interrelated link between CSR and stakeholder theory: while CSR describes a firm’s social duties, stakeholder theory defines the groups for which the firm should take these responsibilities. This perspective is helpful as an indicative basis for further issues of stakeholder theory. A further paper stressing the need for an actor-centered approach is the one by Aguilera and Jackson (2003). The authors theoretically discuss which stakeholder groups impact a firm’s corporate governance decisions.

A connection between legitimacy perspective and stakeholder approach is provided by Scherer and Palazzo (2011). In their approach, legitimacy is considered a socially constructed concept that is based on subjective perceptions and defined as conformity with social rules, traditions and norms. Consequently, each group of stakeholders stands for its own perceptions, diverging from the other groups’ perceptions and aims. According to Freeman (1984), the traditional pool of stakeholders consists of shareholders, customers, employees and suppliers. These traditional stakeholders are completed by so called silent stakeholders

2

(20)

such as local communities and the environment (Simmons, 2004). Corporations have limited resources to fulfill the wishes of all stakeholders equally. Therefore, they run so called stakeholder prioritization or classification, based on the evaluation of the stakeholders’ bargaining power (Mitchell, Agle, & Wood, 1997). Khan, Munir and Willmott (2007) suggest that the unevenly distributed power among stakeholders can suppress other participants’ demands and concerns. Although silent stakeholders are influenced by firm decisions, they have a low, if existing, impact on those decisions. Often they are more scattered and less organized and thus their incoherently formulated aims are less intensely taken into consideration by part of the firm (Jamali, 2008). Since it is difficult to measure the silent stakeholders’ impact on a firm’s decisions due to their low level of organization and their diversified and inconsistent intentions, this paper focuses on the traditional set of stakeholders. Within the group of traditional stakeholders, there are economic stakeholders, such as customers and employees. These stakeholders receive more attention because they directly influence a firm’s performance. Considering the suppliers’ impact on the adoption of CSR would mean to incorporate a global value chain analysis as companies are highly interconnected on a national and international level. It is expected that relationships of dependence can positively influence the adoption of the UNGC and that suppliers can pass on pressures to distributors. A related aspect is the role that multinational enterprises play in diffusion. According to Guler et al. (2002) they can be seen as key agents in the international diffusion of organizational practices because they translate standards of their home country production to the foreign host countries where they operate. Since data on global value chains and supplier relationships are hardly to obtain, suppliers have to be omitted as a stakeholder group impacting the diffusion of the UNGC. Further groups that are not explicitly modeled, despite sometimes being characterized as stakeholders, are managers and the state. It is assumed that the managers’ behavior is solely influenced by the other stakeholders. Thus, their decision taking which is reflected in the firm’s acting displays the outcome of stakeholder prioritization. The state, despite being present as a direct influencing actor in some firms and even whole industries, is only modeled at the institutional stage (Aguilera & Jackson, 2003).

(21)

in the form of labor rights as Principle 6 of the UNGC that attempts to eliminate discrimination in terms of occupation (Jamali, 2008).

The role of labor in a country is characterized by its union organization, its representation rights and its skill formation. The better the employees are organized, the more powerful they are and the more intensely a company has to consider their demands. Thus, the more firms per country adopt the UNGC principles. Employees can also be present in corporate boards of firms and like this directly exert pressure. These pressures work like a normative isomorphism. Thus, the employees’ ability to exert influence on firms’ decisions is conceptualized by labor unions (Aguilera & Jackson, 2003).

Hypothesis 3a: The higher the level of organization of the labor force in a country, the more likely is a firm in this country to adopt the UNGC membership (workers’ participation hypothesis).

Another aspect to be considered is that corporate governance structures are often believed to concentrate especially on shareholders’ striving for profit maximization. This focus does not appropriately allow for the consideration of other stakeholders’ needs since it is motivated mainly by economic incentives (Scherer et al. 2012, Chung & Talaulicar, 2010). Hence, shareholders and stakeholders compete to achieve their aims. The shareholder view has been widely discussed as being unethical. Although recently the research about socially-driven shareholder activism has grown, the focus in this analysis remains on the financially-driven activism that is mostly discussed in literature (Chung & Talaulicar, 2010).

(22)

Hypothesis 3b: The better shareholder rights are protected on a national level, the less likely is a firm in this country to adopt the UNGC membership (shareholder value hypothesis).

Scherer and Palazzo (2011) characterize the growing pressure of the civil society on corporations, acting as a form of coercive isomorphism, to be one of the main drivers behind the expansion of CSR activities. In the present model, the civil society and a firm’s consumers are combined. The civil society bases its legitimacy for companies on moral judgments in an environment of culturally caused taken-for-granted assumptions in order to legitimate a corporation’s behavior. If a society presents a high consumer sophistication, i.e. is highly educated, has a high consciousness about human rights and the environment and is organized in NGOs that explicitly deal with these aspects, the firms in a country are more likely to adopt the UNGC in order to be legitimated. Scherer and Palazzo (2011) characterize the cooperation between firms and NGOs as central for the increasing importance of moral legitimacy. Lim and Tsutsui (2012) draw a link between the national and the global civil society through membership in international NGOs. These memberships empower the national civil society and endorse progressive ideas that positively influence UNGC participation. The authors depict that the world society is an important source of pressure on firms to adopt CSR policies. Moreover, a higher education promotes a better understanding and interpretation of environmental and social abuses and activates a stronger demanding behavior for standards (Potoski & Prakash, 2005). As the society is composed by consumers and potential consumers, firms face economic pressures to put the claims into practice because consumers can injure companies through boycotts (Shaw, Newholm, & Dickinson, 2006). These ethical consumption decisions can also be seen as voting (Shaw et al., 2006). Spitzeck and Hansen (2010) even suggest that a high stakeholder power can only be achieved by consumers. A further aspect concerning pressures of the society on firms is related to the role that the media and especially the internet play. Controversies and criticism about business practices are increasingly spread via the internet. The civil society faces a greater accessibility of information about companies and this raises the level of consumer empowerment and activism. The internet is classified as a powerful tool of information and communication that is labeled by its range and simplicity to obtain information (Taubman, 2007).

(23)

Hypothesis 3c: The higher the level of organization and education of a country’s society, the more likely is a firm in this country to adopt the UNGC membership (societal codetermination hypothesis).

Several studies stress the differences in temporal occurrence of the driving forces efficiency and legitimacy (e.g. Aguilera & Cuervo-Cazurra, 2004). They imply that in the beginning of the adoption process firms try to improve their performance while later on the net benefits become less important and adoption occurs due to legitimacy reasons. Thus, early adopters pursue efficiency aims and try to win the competition against other firms. On the other hand, later adopters have to face strong pressures from several (stakeholder) groups and have to strongly legitimate their behavior in order not to lose their market position (DiMaggio & Powell, 1983). This perspective brings dynamics into the analysis and differentiates between firms that shortly after the creation of the UNGC in the year 2000 become members and put into practice the UNGC principles and those that only later on have decided to participate. As the UNGC was already created more than ten years ago, the time period allows the split of the analysis period. Due to the explanations above it is likely that there are different impacting factors that lead to the adoption of the UNGC and its diffusion in an early and a later time period.

Hypothesis 4: Firms that early become members of the UNGC follow efficiency reasons while firms that later become members pursue legitimacy aims (dynamic hypothesis).

4. DATA

4.1 Data Collection

(24)

Department of Economic and Social Affairs (DESA) (2013) provides a database, the CSONet, where CSOs are listed. The website is meant to facilitate interactions between civil society organizations and DESA. The CSONet serves the currently 3,735 NGOs in consultative status with the UN ECOSOC. An important source for data is furthermore the homepage of the UNGC (2013) itself. A further source is the Polity IV Project by Marshall (2013) that covers Political Regime Characteristics from 1800-2011 and codes democratic and autocratic patterns of authority following expert judgments. In comparison to the use of mono source data, my collection of multiple source data strengthens its robustness. It is unlikely to suffer from biased data due to a single source.

4.2 Variables

Although the decision for UNGC membership is taken at a firm level, the unit of analysis of the paper is the country level because most institutions and economic forces operate at the country level. The firms of a country are embedded into the national environment which is considered to contain the driving forces for membership decisions that are analyzed in the following (Lim & Tsutsui, 2012). As precise firm level data is lacking, motives on the firm level are proxied by stakeholder variables. This approach allows modeling the impact of firm level factors on a country level.

In order to test the theoretical relationships discussed above, I run two separate regressions that differ in their dependent variables. The dependent variables are concerned with the adoption of the UNGC by firms per country. The focus lies on the corporate sector. Businesses participants from all sectors with any types of ownership adopting the UNGC principles are counted, disregarding other non-business participants as NGOs or academic members from the UNGC homepage.3 The first dependent variable, , is a binary choice variable.4 It indicates whether at least one firm in a country has launched the UNGC or not. Thus, the dummy variable takes the value 1 for the years where at least one firm in the country is member in the UNGC (starting in the year of the first launch) and zero when no firm in a country is a UNGC member. The second dependent variable is . The total number of member firms per country and year is counted from the list of participants on the UNGC homepage. Thus, the first regression with as dependent variable denotes the

3

It is surprising that Lim and Tsutsui’s (2012) count for the top ten UNGC members, that are counted as the corporate participants from the UNGC’s office diverge from my counts of business participants based on the participant search function of the UNGC homepage. Lim’s and Tsutsui’s (2012) counts are consistently higher.

4

(25)

arrival of the UNGC in a country, whereas the second model analyzes which factors lead to a high number of firms adopting the UNGC.

The independent variables are classified into four groups, following the description of the theoretical model, whereby the fourth group includes the control variable. In addition, a fifth category contains information on the variable included in the model extension in Section 6. The first category contains the efficiency variable . Transparency International’s Corruption Perception Index for 2012 is an aggregate indicator. The scale ranges from 0 for highly corrupt countries to 100 for countries with very low corruption.

The second group of explanatory variables consists of the exogenous international pressures. These are measured by high trade shares, displaying trade openness, and inward FDI. is calculated as the sum of exports and imports of goods and services as a percentage of GDP, in order to control for a country’s size (Aguilera & Cuervo-Cazurra, 2004). Pressures from international investors are measured in form of inward FDI over GDP to account for the extent to which each country’s firms are internationally oriented (Guler et al., 2002). thus refers to the net inflows of FDI into a country in current US-$. It is normalized by GDP in order to control for the size of a country’s economy. Both variables are calculated on base of the WDI.

(26)

outstanding of listed domestic companies, as the percentage of GDP, taken from the WDI (Aguilera & Cuervo-Cazurra, 2004). The impact of the civil society on UNGC membership is depicted in the variables , and . is measured as secondary school enrollment to official secondary school age population, following the WDI (Lim & Tsutsui, 2012). Since neither data about the number of NGO members per country nor the total number of NGOs is (freely) available, has to be proxied with the number of CSOs that currently maintain official relations with the UN ECOSOC which focuses on the improvement of development challenges in the areas of employment, education and health. From the CSONet database, the actual number of CSOs per country is counted. is operationalized as the percentage of people with access to the worldwide network per country from the WDI.

The last group of variables refers to the control variable. is the gross domestic product in current US-$. It is employed in order to control for a country’s size (Lim & Tsutsui, 2012). The variable , employed in the model extension, “examines concomitant qualities of democratic and autocratic authority in governing institutions” (Marshall, 2013). The scale adopted from the Polity2 variable of the Polity IV Project ranges from -10 for highly autocratic countries to 10 for strongly democratic nations.

The final panel data sample covers the period of analysis from 2000, the year the UNGC was established, until 2011. Most of the databases provide data for around 170 countries. The limiting factors are trade union members and the ASD Index that only cover 54, respectively 72 countries. However, the regressions are mostly based on a sample of 29 countries, due to the inconsistent availability of data.5

5. METHOD OF ANALYSIS AND EMPIRICAL RESULTS

In the literature typically two distinct models are employed for analyzing diffusion patterns of CSR, mainly in form of codes of good governance, ISO 9000 or 14000 certifications or UNGC members. Many authors first analyze the determinants that lead to the first adoption in a country and then, in a second model, they examine the factors that lead to the further diffusion in a country (Aguilera & Cuervo-Cazurra, 2004; Haxhi & van Ees, 2010). I follow the approach of employing two different modelling techniques as the dependent variables have different distributional characteristics.

5

(27)

5.1 Descriptive Statistics

Before starting with the regression analysis, I look at the summary statistics to examine whether there is reasonable variation in the data. Moreover, I examine the distributions of the variables. If independent variables are found to have highly skewed distributions, they are entered into the analysis in logarithmic form (Lim & Tsutsui, 2012).

Table 2 - Descriptive Statistics

Variable Obs. Mean Median Std. Dev. Min Max

launch 2352 0.34 0.00 0.47 0.00 1.00 ln_gccount 808 2.10 1.95 1.46 0.00 6.57 ln_corruption 2076 3.65 3.61 0.48 2.08 4.50 ln_trade 2034 4.39 4.40 0.55 -1.18 6.13 ln_fdi 1998 -3.57 -3.38 1.39 -18.10 0.55 ln_union 648 -2.06 -1.75 1.25 -5.90 -0.22 ln_investor 852 -0.99 -0.89 0.60 -2.53 0.00 ln_stock 1279 3.38 3.54 1.33 -3.33 6.41 ln_educ 1078 4.10 4.34 0.59 1.17 4.60 ln_ngo 2136 3.72 3.80 1.56 0.00 8.16 ln_internet 2230 2.07 2.44 1.90 -8.15 4.57 ln_gdp 2216 23.72 23.49 2.30 17.97 30.34 ln_democracy 1887 2.50 2.83 0.71 0.00 3.04

The analysis of skewness and kurtosis reveals asymmetric distributions for all independent variables and the dependent variable . Thus, the logarithmic transformation is applied to these variables. In case of the index scale has to be redefined (from 1 to 21) in order to eliminate negative values. The summary statistics of the newly created variables is presented in table 2. The standard deviations confirm that there is sufficient variation in the data. The comparison between mean and median reveals that the variables’ distributions remain only slightly skewed after taking the logarithmic values. Skewness is thus considered to be only a minor problem and can be disregarded. The remaining information about minimum and maximum values loses part of its explanatory power due to the logarithmic transformation.

As indicated by Corbett and Kirsch (2001) and Haxhi and van Ees (2010), multicollinearity is likely to be an issue and can reduce the quality of the estimations’ results. Thus, I conduct a correlation analysis to get an overview of the correlation patterns between the variables included in the analysis.6 Table 3 presents pairwise correlations and reveals that the independent variable is highly correlated with Furthermore,

(28)

is highly correlated with and and is correlated with .7

When correlations among the explanatory variables are very high, i.e. exceeding the threshold value of 0.6, it is difficult to estimate independent effects of these variables on the dependent variable. The precision of the estimated parameters in the model can be negatively affected. To verify my findings, I calculate the variance inflation factors (VIF).8 For both the launch and the diffusion model, all values are below 10, which can be considered as rule the of thumb value indicating serious problems due to collinearity. Thus, despite substantial bivariate correlations among some of the independent variables, multivariate correlation causing severe collinearity problems appears to be a minor issue. Still, the coefficients of the following analysis need to be analyzed with caution given the fact that variables reflecting the general level of development of a country are likely to be related to each, such as corruption, education and internet.

Table 3: Correlation Matrix

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) 1. ln_gccount 1.000 2. ln_corruption 0.192* 1.000 3. ln_trade -0.187* 0.219* 1.000 4. ln_fdi -0.189* 0.171* 0.470* 1.000 5. ln_union 0.060 0.572* 0.259* 0.128* 1.000 6. ln_investor 0.018 0.331* 0.040 0.100* -0.101* 1.000 7. ln_stock 0.220* 0.460* 0.142* 0.072* 0.226* 0.405* 1.000 8. ln_educ 0.279* 0.510* 0.188* 0.149* 0.634* 0.099* 0.391* 1.000 9. ln_ngo 0.362* 0.021 -0.399* -0.180* 0.101* 0.125* 0.138* -0.015 1.000 10. ln_internet 0.315* 0.665* 0.302* 0.214* 0.508* 0.136* 0.466* 0.720* -0.042 1.000 11. ln_gdp 0.556* 0.387* -0.189* -0.065* 0.193* 0.291* 0.433* 0.438 0.602* 0.488* 1.000 Note: *p<0.05

5.2 Modeling the UNGC Launch

The models in the first part of the regression analysis, dealing with the launch of the UNGC, have a dichotomous variable as dependent variable. It captures whether a country has at least one firm participating in the UNGC in a certain year or not. Binary choice models estimate the likelihood that a country adopts CSR practices for the first time (Aguilera & Cuervo-Cazurra, 2004; Haxhi & van Ees, 2010).

To overcome problems of nonlinearity, heteroskedasticity and to restrict probabilities to the interval between [0, 1], I use a probit model. is the probability that at least one firm in a

7_{Since the explanatory variables and thus their bivariate correlations remain the same, the table for is}

not displayed.

8

(29)

country adapts the UNGC: . Probit models are estimated based on the Maximum Likelihood estimation technique that automatically accounts for heteroskedasticity (Hill, Griffiths & Lim, 2012). The model including all independent variables derived in the theory part and the control, i.e. the full model, is specified as follows:

A difficulty with the interpretation of the parameter estimates of probit regressions is that the magnitudes of the coefficients cannot be directly interpreted. The coefficients that are reported in the table predict the signs of the partial effects of each on the membership probability. To comment on the effect of on the probability of UNGC membership the average partial effects are computed. To determine the practical significance of the coefficients, the estimated partial effect of each independent variable is multiplied with its standard deviation (Haxhi & van Ees, 2010).

I conduct a stepwise regression procedure, following the general idea of testing the impact of certain predictors from the described set of independent variables on UNGC launch. I start with the control model including the control variable only. I then include the explanatory variables category-wise into the model and in a further step combine the categories to a full model. Table 4 presents the regression results for the probit models, the average partial effects are reported in table A4 in the appendix. Specification 1 (column 1) shows the expected positive effect of the control variable that is maintained in all specifications. The bigger a country in terms of size, the higher is the probability of UNGC launch. The pseudo R² is with 0.334 already relatively high.9

In column 2 to 4 the individual models are displayed for each impacting group identified in the theoretical model separately. The efficiency variable in specification 2 has a highly significant impact on the UNGC launch. Nevertheless, the positive result contradicts Hypothesis 1 that predicted a weak institutional environment to enhance the probability of adoption. Column 3 includes the variables measuring international pressures originating from trade and FDI. Both variables are highly significant, and the pseudo R² is with 0.331 slightly higher than in case of efficiency pressures. Nevertheless, only

9

(30)

presents the expected positive impact and confirms Hypothesis 2, while has a negative sign. Consequently, Hypothesis 2 cannot be accepted. Specification 4 includes the proxies for the three stakeholder groups and shows a (high) statistical significance for all variables except . The pseudo R² indicates with 0.477 a high goodness of fit. Hypothesis 3a about workers’ participation and 3b about shareholder values are accepted. Concerning Hypothesis 3c, the results allow a partial acceptance only. While the effect of is positive as expected, the coefficient for is negative, partially contradicting Hypothesis 3c about social co-determination.

Table 4: UNGC Launch (Probit)

(1) Control (2) Institutions (3) Int. Pressures (4) Stakeholders (5) Full Model ln_gdp 0.481*** 0.455*** 0.485*** 0.460*** 0.419** (0.0191) (0.0209) (0.0210) (0.157) (0.185) ln_corruption 0.227*** -2.146** (0.0820) (0.860) ln_trade -0.298*** -1.077*** (0.0822) (0.320) ln_fdi 0.159*** 0.339*** (0.0317) (0.112) ln_union 0.189* 0.488*** (0.111) (0.167) ln_investor -1.038*** -0.822*** (0.265) (0.302) ln_stock -0.000378 -0.0521 (0.161) (0.174) ln_educ -2.981*** -3.130*** (0.796) (0.955) ln_ngo 0.395** 0.431** (0.167) (0.179) ln_internet 1.245*** 1.793*** (0.217) (0.325) Constant -11.93*** -12.11*** -10.09*** -4.618 10.62* (0.463) (0.501) (0.661) (3.255) (6.023) Observations 2216 2025 1897 262 253 Log-Likelihood -964.323 -929.708 -852.489 -78.498 -66.934 Pseudo R² 0.334 0.313 0.331 0.477 0.546

Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

(31)

launch. Therefore, for the full model in column 5, Hypothesis 1 cannot be rejected. There is highly significant evidence that the lack of efficiency in a firm’s national environment leads to the adoption of the UNGC membership. The results concerning international legitimacy pressures remain for the full model similar to the individual model. Trade openness surprisingly hinders the UNGC launch whereas a high share of FDI to GDP increases the probability of UNGC launch. Thus, international economic pressures on a country’s membership probability are found to have mixed results and the legitimacy hypothesis is only partly confirmed. The measurement of stakeholder impacts in the full model again leads to ambiguous results. Hypothesis 3a about the level of organization of the labor force is accepted. The higher the level of organization of the work force in a country, the higher is the impact of employees on a firm’s decision and, as a result, the higher the probability that a firm in this country launches the UNGC. Concerning the shareholders’ impact, remains insignificant in the full model. In contrast, the results for are significant. Thus, in the full model there is again evidence found for the restricting impact of shareholders on the UNGC launch. The results for the level of organization and education of a country’s society remain inconsistent. While the estimated coefficients for and support Hypothesis 3c and show that a higher number of CSOs in a country and a higher share of internet users lead to a higher probability of UNGC launch, the results for are surprising. It is revealed that a country is less likely to adopt the UNGC for the first time if the share of secondary school enrollment is increasing, as already indicated in the specification in column 4. This result is highly significant.

5.3 Modeling the Diffusion of UNGC Membership

(32)

Based on the full model an appropriate modeling technique is determined. When a suitable estimation method is found, further variations of the model are run to analyze the diffusion of the UNGC. I start with standard tests to find out whether OLS leads to spurious results. The Jarque-Bera statistic leads to the conclusion that the errors are normally distributed. A White test reveals the existence of heteroskedasticity. In case of heteroskedasticity and serial correlation, the errors have to be corrected by applying panel-robust standard errors. Otherwise, the reliability of the estimator is overstated and interval estimates and test statistics are invalid.

To use a model that appropriately captures country heterogeneity, the pooled, the fixed effects (FE) and the random effects (RE) model are evaluated (Perkins & Neumayer, 2010). An important characteristic of the FE estimator is that it accounts only for variation within individuals. Thus, variables that do not vary over time, in my case , , and , are omitted. For these variables it is not possible to estimate an effect on the number of UNGC members. In general, RE models reveal more precise results as they take into account variation within and between individuals. A problem with RE models is that the assumption of the uncorrelated error term is often violated which leads to inconsistent estimates. I verify the assumption that there is substantial cross-country heterogeneity by applying a standard F-test and checking whether the fixed effects are jointly significant. A pooled model is therefore not adequate, as there are indeed significant differences between the individual intercepts. With the aim of testing whether both FE and RE estimators are consistent and the difference between FE and RE is not significant, I conduct a robust Hausman test according to Wooldridge (2002). Unlike the Hausman version, a test of overidentifying restrictions is applicable to heteroskedastic- and cluster-robust versions. The test reveals that the RE estimates are inconsistent. Consequently, I choose a FE model with cluster robust standard errors as a reference model, despite the drawbacks because of omitted time-invariant variables. The specification of the complete FE model is the following, with denoting the country, the time period and the country-specific FE:

(33)

, the impact of efficiency pressures via institutions cannot be captured with a FE model. Consequently, no statement about the bonding hypothesis can be made.

The specification in the first column includes only the control variable . Size seems to be an important factor involved when modeling the diffusion of the UNGC. The R² is already 0.606. Column 2 and 3 present the results for the sub-models looking at the impact of international pressures and stakeholders on UNGC diffusion. Concerning international pressures, shows the expected outcome on a high level of statistical significance, whereas presents the opposite sign but is also statistically significant at the 1% level. The legitimacy hypothesis is consequently only partly supported. The R2 for the model including the group of international legitimacy variables (column 2) is 0.640.

In the stakeholder sub-model (column 3), the employees’ position is not reflected (exclusion of ) and the shareholders’ impact can only partly be depicted (exclusion of ). Thus, the model loses part of its explanatory power, although the R² is with 0.680 very high. The only statistically significant variable is . The share of internet users has a statistically positive impact on the diffusion of UNGC membership. A population’s education is not significant. Thus, Hypothesis 3c about social co-determination can be partly accepted.

Table 5: Diffusion of UNGC (FE)

(1) Control (2) Int. Pressures (3) Stakeholders (4) Full Model ln_gdp 2.573*** 2.529*** 1.794*** 1.730*** (0.160) (0.143) (0.414) (0.306) ln_trade 1.160*** 2.872*** (0.428) (0.882) ln_fdi -0.233*** -0.219*** (0.0382) (0.0713) ln_stock -0.159 -0.00157 (0.163) (0.103) ln_educ 0.870 2.590 (3.063) (2.530) ln_internet 1.582*** 1.266*** (0.558) (0.449) Constant -63.66*** -68.37*** -54.66*** -72.75*** (4.088) (4.100) (18.48) (16.95) Observations 797 754 194 185 Number of country 118 115 29 29 R² 0.606 0.640 0.680 0.747

Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

(34)

Thus, the models are found to be robust. The coefficients for the exogenous legitimating variables remain highly significant as in column 2. has the highest statistical impact on the UNGC diffusion in the model. Although the inflows of FDI to a country seem to hinder diffusion slightly, a high global connectivity in form of trade supports the diffusion of the UNGC. Consequently, Hypothesis 2 is partly accepted. Coming to the impact of stakeholders on the diffusion of the UNGC, the reliability of the results is relatively low. The workers’ impact is not captured and although the stakeholders’ influence is partly included, is insignificant. While the impact of education is also insignificant, the share of internet users has a statistically significant and positive impact on the spread of UNGC membership reflecting the civil society’s power of codetermination. Hypothesis 3c is therefore partly accepted. The control variable is highly statistically significant and has a relatively high impact. Thus, a country’s size needs to be controlled for when assessing the UNGC membership in all models.

5.4 Early and Late Adopters

The examination of early and late adopters splits the time horizon of analysis into two parts: one from 2000-2005 and the other from 2006-2011. Perkins and Neumayer (2010) define the period until 2005 in their paper as an early stage of diffusion. Thus, the models about UNGC launch and its diffusion described above are each divided into two sub-models. The calculation of the variance inflation factors for the early sample reveals that for the launch model, all VIFs are below 10. In case of the diffusion model, the VIF of is above the threshold value of 10 and therefore likely to cause problems of multicollinearity. For late adopters, in both models, i.e. for UNGC launch and diffusion, multicollinearity is not an issue since the VIFs are all below the threshold value.10

As explained in section 5.2., the probit model is used to estimate UNGC launch. When trying to employ the full model for the sample of late adopters, a problem arises. The Stata output suggests that the outcome of the binary choice variable is perfectly separated by some of the explanatory variables, and thus the standard errors are undefined because of a flat likelihood. Therefore, I drop an explanatory variable that causes the problem of perfect prediction. The problem can be solved when excluding or . As the model fit is four times as high when including rather than , I drop the variable measuring shareholder rights. For a better comparison with the results from the early launch sample,

10

(35)

is also dropped in this regression. The results for the probit models of early and late launch are presented in table 6.

The early launch, from 2000-2005 seems to follow similar drivers as presented in table 4 when estimating the full sample.11 The international exogenous pressures provide the same unexpected results: while inflows of FDI enhance the launch, trade seems to hinder it. The stakeholders’ impact is mostly in line with the expectations. Whereas the presence of labor unions, NGOs and a widespread internet access in a country exert pressures on a firm in this country to legitimate its behavior in front of the civil society and workers, the effect of education is again negative and shareholders are not found to have a significant impact.

Table 6: Early vs. Late UNGC Launch (Probit)

(1) Early excl. ln_investor (2) Late excl. ln_investor ln_gdp 0.334 1.961** (0.206) (0.914) ln_corruption -1.865 4.670 (1.189) (4.192) ln_trade -1.177*** -4.814** (0.419) (1.914) ln_fdi 0.335* 0.102 (0.175) (0.256) ln_union 0.635*** 1.440 (0.231) (1.043) ln_stock 0.0831 -1.381** (0.245) (0.672) ln_educ -2.509** 2.627 (1.047) (5.212) ln_ngo 0.633*** 1.125 (0.216) (0.955) ln_internet 1.301*** 0.201 (0.413) (0.876) Constant 10.07 -49.78** (7.941) (21.27) Observations 159 136 Log-Likelihood -42.488 -10.908 Pseudo-R² 0.613 0.824