The effect of product diversification on Corporate Social Performance in the non-renewable energy industry:

(1)

Master Thesis International Business and Management

The effect of product diversification on Corporate

Social Performance in the non-renewable energy industry:

Exploring the moderating effects of host country development

and the Sustainable Development Goals

Author: Renske Jongsma Student number: S3492435 Email: r.jongsma@student.rug.nl

Supervisor: Dr. B.J.W. Pennink Co-assessor: Dr. C. Schlägel Date of submission: January 20th, 2020

Word count: 14,989 (excluding abstract, tables and references)

Faculty of Economics and Business University of Groningen

Duisenberg Building, Nettelbosje 2, 9747 AE Groningen, The Netherlands P.O. Box 800, 9700 AV Groningen, The Netherlands

(2)

(3)

ABSTRACT

Research on the relationship between product diversification on corporate financial performance is well-established, but the way in which product diversification influences a firm’s behavior towards stakeholder demands and social concerns remains largely unexplored. Therefore, building upon stakeholder and institutional theory, this study investigates the relationship between product diversification and corporate social performance (CSP), thereby attempting to make essential contributions to the current literature. Based on an extensive literature review, it was expected that related, unrelated and total product diversification are positively related to CSP. Moreover, it was hypothesized that the exposure to weak institutional host country environments negatively affects the relationship between diversification and CSP, and that the Sustainable Development Goals (SDGs) have a positive effect on the relationship. The sample selected for this research is the non-renewable energy industry, since the industry shows great divergence in terms of corporate social responsibility (CSR) performance. In addition, the industry is highly susceptible to regulatory changes, while the Sustainable Development Goals have an enormous focus on the reliability and sustainability of energy, making it a highly relevant industry to study. By analyzing 40 non-renewable energy firms over a time frame of seven years, the OLS regression results reveal that unrelated diversification is positively related to CSP, while the other forms of diversification show insignificant results. Contrary to expectations, the Sustainable Development Goals negatively affect the relationship between product diversification and CSP, while the moderating effect of exposure to weak institutional environments is insignificant. Accordingly, the results of this study challenge existing theories while adding more context to the existing relationship, and in turn provide promising avenues for future research.

Key words: stakeholder theory, institutional theory, product diversification, corporate social

(4)

2.2.4. Conceptual model ... 14 3. METHODOLOGY ... 15 3.1. Data ... 15 3.2. Sample ... 15 3.3. Measurements ... 17 3.3.1. Dependent variable ... 17 3.3.2. Independent variables ... 17 3.3.3. Moderating variables ... 19 3.3.4. Control variables ... 20 3.4. Conceptual model ... 23 3.5. Data analysis ... 23 4. RESULTS ... 26 4.1. Descriptive statistics ... 26

4.2. Assumptions of regression analysis and robustness check ... 28

4.2.1. Normal distribution of the variables ... 28

4.2.2. Heteroskedasticity ... 28

(5)

4.3. Regression results ... 30

4.3.1. The relationship between product diversification and CSP ... 30

4.3.2. The moderating effect of exposure to weak institutional environments ... 33

4.3.3. The moderating effect of the Sustainable Development Goals and lagged analyses 35 5. DISCUSSION ... 40

5.1. Overview of the results ... 40

5.2. Discussion of the results ... 40

5.3. Theoretical contributions and managerial implications ... 45

5.4. Limitations and directions for future research ... 47

6. CONCLUSION ... 49

REFERENCES ... 50

APPENDICES ... 59

Appendix A: List of developing and least developed countries ... 59

(6)

i

LIST OF TABLES, FIGURES AND FORMULAE

Tables: Page:

TABLE 3.1: Overview of selected firms and industries within sample 16

TABLE 4.1: Descriptive statistics hypothesis 1 and 3 27

TABLE 4.2: Descriptive statistics hypothesis 2 27

TABLE 4.3: Breusch-Pagan test for heteroskedasticity 28

TABLE 4.4: Correlation matrix 29

TABLE 4.5: VIF results for independent variables 30

TABLE 4.6: OLS regression results for hypothesis 1 32

TABLE 4.9: OLS regression results for lagged analysis of hypothesis 3 39

TABLE 5.1: Hypothesis results 40

Figures: Page: FIGURE 2.1: Influence of institutional and stakeholder perspectives in the introduction of socially responsible behaviors (Barrena-Martínez et al., 2015). 9

FIGURE 2.2: Conceptual model 14

FIGURE 3.1: Conceptual model including control variables 23

FIGURE 4.1. Moderating effect of unrelated product diversification x SDGs 36

FIGURE 4.2. Moderating effect of related product diversification x SDGs 36

Formulae: Page: FORMULA 3.1: Related diversification within main industry group 18

FORMULA 3.2: Total related diversification 18

FORMULA 3.3: Unrelated diversification 19

FORMULA 3.4: Total diversification 19

FORMULA 3.5: Exposure to weak institutional environments 20

FORMULA 3.6: Market-to-book ratio 21

FORMULA 3.7: Return on Assets 22

FORMULA 3.8: Regression formula – hypothesis 1 and 3 24

(7)

ii

LIST OF ABBREVIATIONS

MNEs Multinational Enterprises

CSP Corporate Social Performance

CSR Corporate Social Responsibility SDGs Sustainable Development Goals

ESG Environmental, Social, and Governance Score WESP World Economic Situation and Prospects UN United Nations

ROA Return on Assets

(8)

1

1. INTRODUCTION

Firms operating in the non-renewable energy sector do not necessarily appear in a good light within society. Worldwide media widely reports industry debacles such as oil tanker accidents by Exxon-Valdez, the involvement of BP in human rights abuses in Colombia, and indigenous unrest such as the anti-Shell protests in Nigeria. These publicized events put excessive pressure on the non-renewable energy industry, which causes difficulties to manage the relationship with society. In the past few years, these firms seem to have paid more attention to being socially responsible and some multinational enterprises even started to invest in renewable energy sources (Frynas, 2009).

As energy firms are currently subjected to disruptive change (Steen & Weaver, 2017), they may spread their risk by becoming more diversified. This new development is important because diversified firms have a considerable impact on the society as a whole (Kang, 2013). Moreover, energy is one of the most important resources in the world, while societal challenges and stakeholder demands are growing. Consequently, much research has been conducted on corporate social responsibility, as it is nowadays a crucial element in firms’ strategy (Chan, 2014). It is argued that diversified MNEs are currently the most influential force in the world (Kang, 2013), which means their CSR strategy can be considered as the most important in terms of setting an example. To be more specific, research from the Nederlandse Emissie Autoriteit shows that 50% of the emissions in the Netherlands originate from only ten firms and energy firms have a huge share in this list (NOS, 2018).

(9)

2

Dastgir is limited to the manufacturing industry in Indonesia. Both Kang (2013) and Patrisia and Dastgir (2017) did not control for country development differences.

The relationship between product diversification and CSP is clearly understudied but there is also an international aspect to it. Research points out that industry characteristics often determine the degree to which firms adopt CSR practices, while interpretations and other concerns may differ per country (Brammer & Millington, 2008; Frynas, 2009; Hawn & Kang, 2013). The energy industry shows a similar pattern, it seems one of the leading industries in terms of CSR but practices are adopted very unevenly within the industry (Frynas, 2009). Therefore, this study will take both a stakeholder as well as an institutional perspective in order to answer the first part of the research question:

(1) “What is the effect of product diversification on the corporate social performance

within energy firms and how is this relationship moderated by the strength of the institutional environment of the host country?”

Within this relationship, there will be distinguished between related, unrelated and total product diversification because it defines the amount and diversity of stakeholders to take into account (Brammer & Millington, 2008; Kang, 2013).

Moreover, Patrisia and Dastgir (2017) suggest a longitudinal study on the relationship between product diversification and CSP in order to measure the consistency and validity of the relationship.This makes it possible to investigate changes in the relationship in anticipation of the Sustainable Development Goals and following their adoption in 2015. Since the SDGs are exogenous drivers of sustainable performance and achieving the SDGs would fulfill the long-term goals of energy firms, this leads to the second part of the research question:

(2) “To what extent is the relationship between product diversification and CSP of energy

firms moderated by the adoption of the Sustainable Development Goals?”

(10)

3

2. LITERATURE REVIEW

The objective of this chapter is to provide an overview of the existing literature by bringing together several research streams in order to create a solid base for the hypothesis development. The theoretical background starts off with a general understanding of the foundation of this study by introducing the main concepts and theories on the topic. The theories are then linked and extended by other explanatory mechanisms in order to provide a deeper understanding of how they explain the relationship between product diversification and CSP.

2.1. Theoretical background

2.1.1. Product diversification in the energy industry

(11)

4

of incumbents is influenced by the institutional environments they are operating in as well as the perceived opportunities and barriers for growth.

Li, Wang, Lou, Cheng and Yang (2016) studied the effect of diversification on corporate (financial) performance of China’s energy companies. They found that Chinese energy companies are not suited to product (industrial) diversification and that it in fact will hinder their corporate performance. This is especially the case for small private companies without government funds. Since this research was conducted in China only, it is not generalizable for energy firms worldwide, but the authors argue that famous energy firms in developed countries usually switch from specialization to diversification first, but in the end refocus on their core businesses.

2.1.2. Stakeholder theory

Stakeholder theory was first described by Freeman (1984) stating that “A stakeholder approach

is about creating as much value as possible for stakeholders, without resorting to trade offs”

(Freeman, Harrison, Wicks, Parmar, & Colle, 2010: 28). Freeman (1984) dictated that firms create value by taking all different groups related to the company into account rather than just shareholders. Thus, stakeholders can be any party, individual or group, that affects or is affected by activities that firms perform (Freeman et al., 2010). As firms can have a diverse set of stakeholders for different reasons, it is important to balance the diverse range of interests of stakeholders when taking a stakeholder perspective (Clarkson, 1995).

Cornell and Shapiro (1987) state that successful stakeholder management is the key to success for firms in general. The argument of the authors is that the total value of a firm depends on their ability to accomplish the social obligations it has towards its stakeholders. Consequently, not confirming to these social obligations results in reputational damage and losing legitimacy (Cornell & Shapiro, 1987) which underlines the importance for firms to engage with their stakeholders.

2.1.3. Institutional theory

(12)

5

direct effect on the successfulness of international operations (Chan, Isobe and Makino, 2008; North, 1990).

Firms try to adapt to the institutions in host countries in order to gain legitimacy from their stakeholders, which in turn often leads to isomorphic processes (Dimaggio & Powell, 1983). Isomorphism can be described as the process of homogenization within the external environment of the firm. It is a constraining process where firms facing the same set of environmental conditions imitate each other (Dimaggio & Powell, 1983). Institutional isomorphism breaks down into three processes: coercive isomorphism, mimetic isomorphism and normative isomorphism. First, coercive isomorphism relates to the pressure from external factors such as powerful or critical stakeholders to change the institutional practices of a firm. This leads to homogeneity of firms because the powerful stakeholders on which the organization is dependent might have the same demands and expectations for other firms operating in the same environment. Second, mimetic isomorphism relates to firms trying to obtain competitive advantage in terms of legitimacy. Organizations will try to copy other organizations’ practices encouraged by uncertainty in a different environment. Finally, normative isomorphism relates to common values shared by a range of formal and informal groups to which decisionmakers belong. These values lead to the adoption of particular similar institutional practices due to for example culture or shared working practices (Dimaggio & Powell, 1983).

Carpenter and Feroz (2001) argue that institutional theory is based on the premise that firms within the same environment are responding to institutional pressures by adopting structures and procedures that are socially accepted as the suitable organizational choice.

In conclusion, all of isomorphic processes together lead organizations operating in the same institutional environment to adopt similar structures and management practices, regardless of their actual usefulness or organizational efficiency (Carpenter & Feroz, 2001; DiMaggio & Powell, 1983).

2.1.4. Defining corporate social responsibility and corporate social performance

(13)

6

definition (Aguinis & Glavas, 2012). The difficulties in finding a common definition for CSR arise from the dynamic and complex character of the concept (Carroll, 1999; Moon, Crane & Matten, 2005).

The most commonly used framework of CSR is the pyramid of Carroll (1991). Carroll argues that in order for CSR to be accepted by businesses, it is important that the concept is framed in such a way that the entire responsibility of businesses is embraced. The pyramid consists of economic, legal, ethical and philanthropic responsibilities and Carroll (1991) argues that the total corporate social responsibility of business entails the simultaneous fulfilment of all responsibilities.

Matten and Moon (2008) argue that for the sake of a multinational study, a broad rather than a specific definition seems to be most appropriate. CSR in a broad sense can be described as

"actions that appear to further some social good, beyond the interests of the firm and that which is required by law” (McWilliams & Siegel, 2001: 117). This definition covers all relevant

components as well as it captures an MNEs’ CSR activities in host countries (Campbell, Eden & Miller, 2012).

A related concept to CSR is corporate social performance. CSP is primarily an extension of CSR, however, CSP is focused on actual achieved results rather than the general notion of the responsibility businesses bear (Sage Publication, 2012). Thus, CSR and CSP are interrelated concepts and CSP can be seen as a natural consequence of CSR. This study will follow the definition of Wood (1991: 693): “a business organization's configuration of principles of social

responsibility, processes of social responsiveness, policies, programs, and observable outcomes as they relate to firm's societal relationships”. Thus, this definition places CSR into

a broader context and constitutes the social performance as the outcome of CSR activities undertaken by a firm (Ioannou & Serafeim, 2012).

(14)

7

2.1.5. Stakeholder perspective on CSR

As discussed before, a wide variety of beliefs and attitudes led to various definitions of CSR (Hill, Stephens and Smith, 2003). However, in more recent literature, there tends to be a focus on the responsibility of a firm towards its stakeholders (Jones, 2005; Spence, Coles & Harris, 2001; Vos, 2003, as cited in Sweeney & Coughlan, 2008). This is important because there is a natural fit between the idea of stakeholder theory and CSR, since it explains towards whom the organization should be “socially responsible” (Carroll, 1999). Organizations have a responsibility towards a multitude of parties and in order to balance and prioritize their demands and expectations within a logical order, stakeholder can be categorized into two types. Primary stakeholders are essential for the proper functioning of the firm and generally have some sort of formal contract with the firm, while secondary stakeholders are not directly involved in the economic activities of the firms but can have a significant influence on its activity (Clarkson, 1995; Freeman, 1984; Goodpaster, 1991).

Another way to evaluate different stakeholder claims, are three vital criteria: legitimacy, power and urgency (Mitchell, Agle & Wood, 1997). The most pertinent of these three is legitimacy, which relates to the extent to which the stakeholder group has a proper, desirable or appropriate claim (Carroll, 1999; Mitchel et al., 1997).

According to Mitchell et al. (1997) stakeholders have a direct influence on the behavior of the company and thus their CSR implementation, hereby referring to not only powerful (primary) stakeholders such as investors, shareholders and employees, but also outside (secondary) groups such as the media, press and activists. This influence manifests itself in the way stakeholders interpret corporate actions and challenge the legitimacy of organizations (Lamin & Zaheer, 2012). Therefore, the stakeholder approach can be seen as the first of two solid pillars to explain the incorporation of CSR activities of firms (Barrena-Martínez, Lopez-Fernández & Romero-Fernández, 2015) as illustrated in figure 2.1. Hereby, there will be distinguished between the macro-context, consisting of external stakeholders which determine the survival of the firm in the market, and the micro-context, consisting of interest groups with direct links to the company, able to exert power and influence decisions (Clarkson, 1995).

2.1.6. Institutional perspective on CSR

(15)

8

the diverse motives of stakeholders within different national environments, which is important because motives influence the corporate governance, shaping the CSR adoption of MNEs (Matten & Moon, 2008).

Organizational strategies and actions have been adapted to environmental pressures and requirements for decades (Fernández-Allés & Valle-Cabrera, 2006). Consequently, the response of firms to demands from the internal and external context is considered as pivotal for organizational survival (Dacin, 1997; Fuenfschilling & Truffer, 2014) and allows firms to achieve legitimacy (Cruz-Suárez, Prado-Román, & Díez-Martín, 2014; Suchman, 1995, as cited in Barrena-Martínez et al., 2015). Legitimacy is defined by Scott (2007: 45) as “a condition

that reflects the cultural alignment, normative support, or consonance with the rules or laws of the environment”, hereby underlining that it is not something that is simply possessed or

exchanged by firms. The legitimacy challenge leads to the uncovering of a natural and complementary theoretical link between institutional theory and stakeholder theory because of the significant influence (local) stakeholders can have on the CSR implementation of firms (Barrena Martínez et al., 2015). Higher legitimacy pressure from society will lead firms to try and enhance their legitimacy by acting more responsibly (Amaeshi, Adegbite & Rajwani, 2016). Therefore, institutional theory is the second solid pillar in explaining how external pressures may condition the incorporation of CSR activities of firms, as shown in figure 2.1 (Barrena Martínez et al., 2015).

However, gaining legitimacy in the global context leads to a dilemma since MNEs need to adapt to home country pressures as well as to host country institutions in order to gain both internal and external legitimacy, which is called institutional duality (Kostova & Roth, 2002; Hillman & Wan, 2005). Moreover, Matten & Moon (2008) argue that CSR is embedded in the institutional system of a country. Since CSR practices are interpreted differently across countries (Baughn, Bodie & McIntosh: 2006; Bondy, Matten & Moon, 2004; Welford, 2005; Wokutch, 1990, as cited in Yang & Rivers, 2009) and due to the fact that CSR practices are embedded in institutional systems (Matten & Moon, 2008), institutional differences might lead to challenges in the adoption of a CSR strategy for MNEs.

(16)

9

FIGURE 2.1: Institutional and stakeholder pillars in the decision of CSR activities (Barrena-Martínez et al., 2015).

Now the main theories of this study are discussed, chapter 2.1.7. will provide some background information on the Sustainable Development Goals.

2.1.7. Sustainable Development Goals

The Sustainable Development Goals are an effect of the aim to achieve a better future for all countries, adopted by all 193 members states of the United Nations. The plan is set for the upcoming 15 years and is aimed at ending extreme poverty, fighting inequality and injustice in order to protect the planet. This plan resulted in the evolvement of the 17 Sustainable Development Goals on which all countries need to actively contribute to make it work. The stand-alone goal that is dedicatedly focused on energy is goal 7: “ensure access to affordable,

reliable, sustainable and modern energy for all" (United Nations, 2019). According to the

Stakeholder Forum (2015), this goal is indicated as the most important transformational challenge, together with climate change and sustainable consumption and production.

(17)

10

to feel responsible for the SDGs/Global Goals, further research tells a different story. It demonstrates that firms really need the global goals since they offer compelling growth strategies for firms and the world economy. Aside from the benefits for our planet, it opens up major market opportunities for firms when the goals are achieved. Delivering the goals provides economic gains of a minimum of US$12 trillion for the private sector and can potentially be two to three times more. Consequently, the 17 largest business opportunities that arise from energy challenges have a potential value of over 4.3 trillion US dollar in 2030. This value arises partially from the expansion of renewables (Business and Sustainable Development Commission, 2017).

2.2. Hypothesis Development

2.2.1. Product diversification and corporate social performance

Several studies point out that industry characteristics often determine the degree to which firms adopt CSR practices (Brammer & Millington, 2008; Frynas, 2009; Hawn & Kang, 2013). CSR standards seem to be highly diverse between industries but are usually shared between countries on an industry level. Even though the key environmental and social concerns within industries are shared between countries, the interpretation and other concerns may differ per country. The energy industry shows a similar pattern, according to Frynas (2009) it is one of the leading industries in terms of CSR but practices are adopted very unevenly within the industry. This raises the question as to whether the CSP of firms within the energy industry is influenced by the product diversity of the firm due to the wider range of stakeholders.

(18)

11

Both Kang (2013) and Patrisia and Dastgir (2017) find a positive relationship between unrelated product diversification and CSP and a negative relationship between related product diversification and CSP. Kang (2013) argues that there are three reasons for diversification to affect the CSP of firms. First, diversification provokes risk averse behavior, therefore inducing managers to respond. Moreover, diversification lowers the employment risk of managers which allows them to allocate more attention and firm resources. Finally, diversification gives a stronger incentive for firms to invest in CSP because it creates an economy of scope for CSP related investments. According to the stakeholder theory, diversified firms in general have to deal with a larger amount of salient stakeholders with regards to legitimacy, power and urgency compared to focused firms (Mitchell et al., 1997).

Within the relationship, the level of unrelated diversification is expected to be more positively associated with the CSP of the firms than is the level of related diversification. The argument from both Kang (2013) and Patrisia and Dastgir (2017) is that unrelated diversification increases the amount of stakeholders and social demands more drastically compared to related diversification. Moreover, unrelated diversification is considered to have a stronger effect on managerial risk aversion compared to related diversification (Hitt, Hoskisson & Kim, 1997) which implies that firms will take decisions more cautiously. However, for the non-renewable energy sector specifically, one would expect that with increasing policies firms would invest in renewable (sustainable) energy (Lund, 2009; Steen & Weaver, 2017). This form of related diversification would in turn lead to a better corporate social performance and therefore the relationship is expected to be positive. However, the effect is expected to be less strong compared to unrelated diversification. These arguments translate into the following hypothesis:

Hypothesis 1a: Related product diversification is positively related to the corporate social performance of firms operating in the non-renewable energy industry.

Hypothesis 1b: Unrelated product diversification is positively related to the corporate social performance of firms operating in the non-renewable energy industry.

In order to investigate the combined effect of unrelated and related product diversification, which was found to be insignificant in Patrisia and Dastgir’s (2017) study, the following hypothesis was formulated:

(19)

12

2.2.2. Exposure to weak institutional host country environments

Similarly to expanding a business to product diverse markets, operating a business in different countries increases the amount of stakeholders. Kang (2013) argues that a firm’s geographic diversification has a positive effect on its CSP based on the fact that firms expanding their international markets will face a more diverse set of stakeholders.

However, Yang and Rivers (2009) argue that engaging in different institutional environments poses challenges for MNEs because the CSR attitudes in different (geographical) markets might differ from the firms’ CSR standards. This relates to the challenge of institutional duality for firms operating internationally (Hillman & Wan, 2005; Kostova & Roth, 2002). Indeed, several studies have suggested that the management and orientation of CSR differs significantly across different countries (Baughn, Bodie & McIntosh, 2006; Bondy, Matten & Moon, 2004; Welford, 2005; Wokutch, 1990, as cited in Yang & Rivers, 2009). Welford (2005) found that these differences are related to economic development, with developed countries having a predominantly higher occurrence of CSR-related activities. Similarly, Baughn et al. (2016) argue there is a relationship between a company’s behavior towards CSR and the economic and social conditions of a country.

Strong institutional contexts can be seen as an imperative for CSR practices (Matten & Moon, 2008), where developing countries are predominantly characterized by institutional voids, which increases the opportunities for corporate social irresponsibility (Mair & Marti, 2008; Matten & Moon, 2008). Even though it can be argued that firms diversify into countries with weak institutional environments to fill institutional voids, it is more likely that firms imitate the lower levels of CSR commitment of the host country competitors to decrease uncertainty and costs (Reimann, Rauer, & Kaufmann, 2015). This phenomenon can be referred to as isomorphic processes according to institutional theory (Dimaggio & Powell, 1983). Following this line of reasoning, it is expected that diversification into weak institutional environments weakens the relationship between product diversification and CSP.

Hypothesis 2: Weak institutional host country environments weaken the positive relationship between product diversification and corporate social performance in the non-renewable energy industry.

2.2.3. Sustainable Development Goals as a moderating variable

(20)

13

new revenue, ensure investor interest in addition to recruiting and retaining talent (Busco, Granà, & Izzo, 2018). A survey of the PWC (2015) shows that the SDG awareness among the business community is very high (92%), and that 71% of the organizations are planning on responding to and engaging with the SDGs (Busco et al., 2018). In addition to the financial benefits, focusing on the SDGs will further strengthen the relationship between organizations and stakeholders. This is because developing and delivering solutions for the achievements of SDGs improves relationships with regulators and stakeholders and lowers the costs of compliance. Hence, when organizations successfully ingrain the SDGs in their strategy, this will enhance legitimacy in the form of improved credibility with the society and reduced future liability for any kind of environmental damage (Busco et al., 2018). In line with this argument, Schrettle, Hinz, Scherrer-Rathje and Friedli (2013) argue that exogenous and endogenous drivers lead firms to more sustainable efforts. Exogenous drivers can be divided into three stakeholder clusters: environmental regulation, societal values and norms, and market drivers. According to Busco et al. (2018), firms are externally driven by the SDGs to set goals regarding their impact in the future.

(21)

14

Therefore, the expectation is that when a firm focuses on the SDGs, which are long-term goals and an exogenous driver of sustainable effort (Schrettle et al., 2013), this focus positively moderates the relationship between product diversification and CSP:

Hypothesis 3: The positive relationship between product diversification and corporate social performance in non-renewable energy industry was strengthened by the adoption of the Sustainable Development Goals in 2015.

2.2.4. Conceptual model

In order to visualize the hypotheses explained in the previous sections, a conceptual model was constructed as illustrated in figure 2.2:

(22)

15

3. METHODOLOGY

3.1. Data

Quantitative analysis will be conducted using secondary data on the firm-level over a time span ranging from 2011 - 2018, in order to investigate changes in anticipation of the SDGs and following their adoption. The data was accessed through Thomson Reuters’ Eikon, Bureau van Dijk’s Orbis and Compustat IQ.

3.2. Sample

The selected sample for this study consists of stock-listed, non-renewable energy firms. The energy industry in general has been selected because there seems to be great divergence in terms of CSR adoption in the industry. According to Frynas (2009), negative publication on non-renewable energy firms has put excessive pressure on the industry which in turn makes them pay more attention to CSR. However, even though the energy industry is one of the leading industries in terms of CSR, the practices are adopted very unevenly within the industry. Additionally, Steen and Weaver (2017) argue that the ‘greening’ process of energy systems implies that many non-renewable energy firms are subjected to (potentially) disruptive change. This implication is confirmed by a report of the Stakeholder Forum (2015) which states that the Sustainable Development Goal related to sustainable energy is indicated as the most important transformational challenge, together with climate change and sustainable consumption and production. These arguments make non-renewable energy firms a highly relevant sample for the sake of this study.

(23)

16

TABLE 3.1: Overview of selected firms and industries within sample SIC

Code

NAICS Code

Description of the industry Number of firms in selected sample 1221 2121 Bituminous coal and lignite surface

mining

0

1222 2121 Bituminous coal underground mining 0

1241 2131, 2389 Coal mining services 0

1311 2111 Crude petroleum and natural gas 14

1321 2111 Natural gas liquids 0

1381 2131 Drilling oil and gas wells 3

1382 2131, 5413 Oil and gas fields exploration services 7 1389 2131, 2371,

2389

Oil and gas fields exploration services, not elsewhere classified

9

2911 3241 Petroleum refining 7

4612 4861 Crude petroleum pipelines 0

4613 4869 Refined petroleum pipelines 0

The original sample in Compustat, used to access the independent and control variables, presented 410 firms within the above industries. Similarly, Eikon was used to access the dependent variable, in addition to further control variables, and consisted of 323 firms and after the removal of missing data left with 157 firms. Finally, the Orbis database, used to access the moderating variable, exposure to weak institutional environments, consisted of 794 firms, however after the removal of purely domestic firms left with 712 firms.

(24)

17

3.3. Measurements

3.3.1. Dependent variable

The dependent variable of this study is the corporate social performance of non-renewable energy firms. This variable will be measured by means of the combined score of economic, environmental, social and governance (ESG) measures, using the Thomson Reuters’ Eikon database. This measure is used by a number of recent prior studies (e.g. Eding & Scholtens, 2017; El Ghoul, Guedham & Kim, 2017; Ioannou & Serafeim, 2012; Maon, Swaen, & Lindgreen 2017; Tarmuji, Maelah & Tarmuji, 2016) and is considered to be comprehensive and standardized as it is collected through a consistent strategy across national boundaries (Tarmuji et al., 2016). The Eikon ESG score is chosen as measure for CSP for several reasons. First of all, it is a global dataset which covers more than 7000 companies which makes it much more internationally diversified compared to the KLD index, another widely used measure of CSP, which only captures data on US firms (Eding & Scholtens, 2017). The second reason is that the data is highly objective, easily accessible and very usable for quantitative analysis (Ioannou & Serafeim, 2012).

The combined ESG score is a result of the overlaying the ESG score, which measures a company’s ESG performance based on publicly reported data, with ESG controversies in order to provide a comprehensive evaluation on the impact and conduct of the company’s sustainability (Refinitiv Reuters, 2019). Data on over 400 metrics is derived by research analysts from publicly available sources such as CSR and annual reports, NGO websites and stock exchange filings and consequently transformed into consistent units from qualitative to quantitative data (Ioannou & Serafeim, 2012; Refinitiv Reuters, 2019). The ten categories remaining serve as the basis for the three pillars of the ESG score: (1) Environment; (2) Social; (3) Governance. The value of the combined ESG score ranges from 0-100 with 100 being the highest possible score (Refinitiv Reuters, 2019).

3.3.2. Independent variables

The independent variables of this study represent related, unrelated, total product

diversification. Based on extensive literature on diversification (Baysinger & Hoskisson, 1989;

(25)

18

which a firm operates as well as the relative importance of each segment in terms of sales (Palepu, 1985). The advantages of the entropy measure are the high levels of objectivity, reliability and the ability to consider the type and level of diversification concurrently (Martin & Sayrak, 2003; Patrisia & Dastgir, 2017; Sambharya, 2000). Moreover, similar studies (Kang, 2013; Patrisia & Dastgir, 2017) used the same entropy measurement. The data required for the independent variable was derived from the Compustat database.

Related diversification (DR) is measured as the diversification that arises from operations in

four-digit segments within the same two-digit industry group. Based on the main SIC code the company familiarizes itself with and the primary SIC codes within the same two-digit industry group, formula 3.1 and 3.2 were utilized to measure related diversification:

FORMULA 3.1: Related diversification within main industry group

𝐷𝑅𝑗 = ∑ 𝑃𝑗_𝑖 𝑛 𝑖𝜀𝑗 ln 1 𝑃𝑗 𝑖

𝐷𝑅𝑗 is the related diversification in several segments within the main industry group, whereas

𝑃𝑗

𝑖 can be defined as the share of segment i of group j in the total sales of the industry group.

n is defined as the number of industry segments. In order to calculate the weighted average of

the total related diversification (𝐷𝑅) in case the firm operates in several industry groups, the following formula was utilized:

FORMULA 3.2: Total related diversification

𝐷𝑅 = ∑ 𝐷𝑅𝑗 𝑚

𝑗=1

𝑥 p𝑗

In formula 3.2, which measures total related diversification (DR), 𝑃𝑗 is defined as the share of

the jth group sales in the total sales of the firm. m is the aggregation of n industry segments into

m industry groups.

Unrelated diversification (DU) is measured as the diversification that arises from a firm

(26)

19

FORMULA 3.3: Unrelated diversification

𝐷𝑈 = ∑ 𝑃_𝑖

𝑚

𝑖=1

𝑥 ln 1 𝑃_𝑖

In formula 3.3, measuring unrelated diversification, 𝐷𝑈 is defined as the unrelated diversification in all entire group shares, whereas 𝑃𝑖 is the share of the segment i of group j in

the total sales of the firm.

Finally, total diversification is calculated as the sum of related and unrelated product diversification, as illustrated in formula 3.4:

FORMULA 3.4: Total diversification 𝐷𝑇 = 𝐷𝑈 + 𝐷𝑅

3.3.3. Moderating variables

Sustainable Development Goals

In order to examine how the Sustainable Development Goals, which serves as a moderating variable, affect the relationship between product diversification and CSP, the main relationship will be studied over a time period. The SDGs are adopted in September 2015, and officially came into force on January 1st_{, 2016 (United Nations, 2019), which means that panel data is}

able to capture the differences over the years. The selected time period is from 2011-2018 in order to be able to investigate the possible differences between the relationship four years before and the three years following from the adoption of the Sustainable Development Goals. This is a similar approach to Jimenez-Parra, Alonso-Martinez and Godos-Diez (2018) who used a time frame of eight years (2006-2013) to investigate the effect of regulation.

In order for the enforcement of the SDGs to be recognized by STATA, a dummy variable was created where 1 equaled the years 2016-2018 and 0 equaled the years 2011-2015. This way, it is possible to investigate whether there was a change detected in the relationship between product diversification and CSP before and after the SDGs became effective.

Exposure to weak institutional environments

The second moderating variable of this study represents the exposure to weak institutional

environments. Since the literature does not present a specific measure for this variable, the

(27)

20

context of CSR research. Oftentimes, internationalization is measured as the number of unique countries a firm operates in (Bansal, 2005; Keig, 2013; Strike, Gao & Bansal, 2006), but this measure does not take into account the intensity and depth of exposure to foreign host country environments. Therefore, an additional measure is to count the number of foreign subsidiaries that a firm has formed (Chetty, Erikson & Lindbergh, 2006; Strike et al., 2006). Since the aim of the moderating variable is to measure the effect of exposure to weak institutional host country environments, the measure was adapted by specifying the presence of subsidiaries in developed and least developed countries. However, only counting the subsidiaries in weaker institutional environments would provide as a biased view since the firms in the sample size differ in terms of size as well as in the number of foreign subsidiaries. Therefore, the measurement of exposure

to weak institutional environments resulted in the following ratio (formula 3.5):

FORMULA 3.5: Exposure to weak institutional environments 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑠𝑢𝑏𝑠𝑖𝑑𝑖𝑎𝑟𝑖𝑒𝑠 𝑖𝑛 𝑤𝑒𝑎𝑘 𝑖𝑛𝑠𝑡𝑖𝑡𝑢𝑡𝑖𝑜𝑛𝑎𝑙 𝑒𝑛𝑣𝑖𝑟𝑜𝑛𝑚𝑒𝑛𝑡𝑠

𝑇𝑜𝑡𝑎𝑙 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑓𝑜𝑟𝑒𝑖𝑔𝑛 𝑠𝑢𝑏𝑠𝑖𝑑𝑖𝑎𝑟𝑖𝑒𝑠

The subsidiary data was derived from Orbis and after the removal of subsidiaries in every firm’s home country, the subsidiaries were labeled according to the weakness of the host country institutional environment. The development of countries was defined in line with the United Nations World Economic Situation and Prospects (WESP) country classification (WESP, 2019). The WESP classification defines all nations in the world as three broad categories: developed economies, economies in transition and developing economies. These categories are intended to reflect the basic economic conditions of each country. Since these categories are quite broad and only based on basic economic conditions, other distinctions are made such as full importing/exporting countries, economies by per capita GNI, and least developed countries. For the purpose of measuring the moderating variable in this study, the countries classified within the categories developing and least developed countries were considered as weak

institutional environments. An overview of the countries classified as developing or least

developed can be found in Appendix A.

Unlike the other variables, the moderating variable will only be tested in the year 2018 since the required subsidiary panel data was not available in the Orbis database.

3.3.4. Control variables

(28)

21

performance, firm age, and market-to-book ratio have been included as control variables on the firm level, where CSP means have been included on the industry- and country level.

Firm size

The first control variable included is firm size as larger firms generally have a higher CSP due to the fact that they have the resources available to invest in socially responsible behavior (Liang & Renneboog, 2017; Perrini, Russon, & Tencati, 2007, Useem, 1988). In addition to larger firms having more opportunities to invest in CSR, larger firms are more visible to the public which means they face higher levels of stakeholder pressure which in turn might lead them to behave more responsibly (Brammer et al., 2009). In line with previous studies, the number of full-time employees was used as a proxy for firm size (Baumann-Pauly, Wickert, Spence, & Scherer, 2013; Kang, 2013; Perrini, Russo, & Tencati, 2007), which was logarithmically transformed in order to ensure normality. The data was derived from the Eikon Thomson Reuters database.

Firm age

Firm age is included as a control variable as it has been proven to have a positive effect on CSP (Withisuphakorn & Jiraporn, 2017). Firm age has been measured as the logarithm of the number of years since the company was founded from 2011-2018. The data was derived from Orbis as well as annual reports of the firms within the sample.

Market-to-book ratio

The market-to-book ratio was included to control for the existence of intangible assets such as R&D capability and brand strength since it may affect the CSP of firms (McWilliams & Siegel, 2000; Kang, 2013). The market-to-book ratio is able to determine the growth opportunities or the potential to grow (Choi & Moon, 2016). The market-to-book ratio was included as a control variable rather than other predicting measures such as R&D data since this data was unavailable for the firms within the sample. One possible explanation for this missing data is that traditional industries usually do invest in these areas but they often have no specific R&D budget available (Moncada-Paternò-Castello, Ciupagea, Smith, Tübke & Tubbs, 2010). Formula 3.6 was used to calculate the market-to-book ratio:

FORMULA 3.6: Market-to-book ratio 𝑀𝑎𝑟𝑘𝑒𝑡 𝑐𝑎𝑝𝑖𝑡𝑎𝑙𝑖𝑧𝑎𝑡𝑖𝑜𝑛

(29)

22

The data on Market Capitalization was derived from the Eikon database and represents the total sum of the market value for all relevant issue level share types. The Net Book Value is equal to total assets minus total liabilities and this data was accessed through Compustat.

Firm performance

The fourth control variable included represents firm performance as it is expected to have a positive effect on CSR commitment. As CSR might be considered as a costly choice for firms, it is rather sensitive to the existence of slack resources (Jackson & Apostolakou, 2010). Thus, firms with a higher amount of slack resources are more likely to invest in CSR (Waddock & Graves, 1997) and consequently have a higher CSP score (Jackson & Apostolakou, 2010). Following similar studies (Kang, 2013; Patrisia & Dastgir, 2017), firm performance is measured by calculating the Return on Assets (ROA), which reflects the operating performance of the firm by presenting the asset utilization (Griffin and Mahon, 1997; Vitezić, Vuko, & Mörec, 2012). The data was derived from the Compustat database and was calculated by using formula 3.7:

FORMULA 3.7: Return on Assets 𝑁𝑒𝑡 𝐼𝑛𝑐𝑜𝑚𝑒

𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠

Industry level effects

In order to control for industry differences in CSP, the mean CSP scores by industry (two-digit SIC code) are included, following prior related studies (Kang, 2013; Patrisia & Dastgir, 2017). Controlling for industry level effects is especially important for the energy industry in general since CSR practices seem to be adopted uniquely within the industry (Frynas, 2009). The data for this variable was calculated as the mean ESG score per industry, derived from the Thomson Reuters’ Eikon database.

Country level effects

(30)

23

3.4. Conceptual model

The following conceptual model (figure 3.1) illustrates the expected relationships while including the control variables discussed in this chapter.

FIGURE 3.1: Conceptual model including control variables

3.5. Data analysis

In order to gain insights in the relationship between product diversification and CSP, several regression analyses will be conducted. After collecting data on all variables from different databases, the data was matched in Excel and organized in a panel, which was then analyzed using STATA.

(31)

24

effect of the control variables. Based on the hypotheses, the following full formulae were utilized to run the regressions:

FORMULA 3.8: Regression formula – hypothesis 1 and 3

CSP𝑖𝑡 = 𝛽1RELATED_PD𝑖𝑡 + 𝛽2UNRELATED_PD𝑖𝑡 + 𝛽3 TOTAL_PD𝑖𝑡 + 𝛽4SDGS𝑖𝑡 + 𝛽5RELATED_PD𝑖𝑡 x 𝑆𝐷𝐺𝑆𝑖𝑡 + 𝛽6UNRELATED_PD𝑖𝑡 x 𝑆𝐷𝐺𝑆𝑖𝑡 + 𝛽7TOTAL_PD𝑖𝑡 x 𝑆𝐷𝐺𝑆𝑖𝑡 + 𝛽8CSPINDUSTRY𝑖𝑡 + 𝛽9CSPCOUNTRY𝑖𝑡 + 𝛽10FIRMSIZE𝑖𝑡 + 𝛽11FIRMAGE𝑖𝑡 +

𝛽12MARKETTOBOOKRATIO𝑖𝑡 + 𝛽13FIRMPERFORMANCE𝑖𝑡 + 𝛽14YEAR𝑖𝑡 + 𝛽15COUNTRY𝑖𝑡 + ε𝑖𝑡

Formula 3.8 represents the main relationship between product diversification and CSP including the moderating variable of the SDGs in order to test hypothesis 1 and 3. In this formula, i stands for firm in the year t, meaning that all firms have been analyzed over a timeframe of 2011-2018. CSP stands for corporate social performance as measured by the ESG score, UNRELATED_PD, RELATED_PD and TOTAL_PD stand for the product diversification as measured by the entropy formula, and the SDGs were included as a dummy variable. The moderating effect of the SDGs is denoted in the formula as

RELATED/UNRELATED/TOTAL_PD x SDGS. Additionally, the CSP industry and country means, firm size, firm performance, firm age, market-to-book ratio, year and country level effects have been included as control variables. Finally, 𝜀 denotes the error term.

FORMULA 3.9: Regression formula – hypothesis 2

CSP = 𝛽1RELATED_PD + 𝛽2UNRELATED_PD + 𝛽3TOTAL_PD + 𝛽4EXPOSURE_WEAK_ENVIRONMENT +

𝛽5RELATED_PD x EXPOSURE_WEAK_ENVIRONMENT + 𝛽6UNRELATED_PD x EXPOSURE_WEAK_ENVIRONMENT +

𝛽7TOTAL_PD x EXPOSURE_WEAK_ENVIRONMENT + 𝛽8CSPINDUSTRY + 𝛽9𝐶𝑆𝑃𝐶𝑂𝑈𝑁𝑇𝑅𝑌 + 𝛽10FIRMSIZE + 𝛽11COUNTRY + ε

(32)

25

RELATED/UNRELATED/TOTAL_PD x EXPOSURE_WEAK_ENVIRONMENT. Besides this difference, some control variables have been deleted from the regression formula, this will be elaborated on in chapter 4.3.2.

(33)

26

4. RESULTS

4.1. Descriptive statistics

Table 4.1 portrays an overview of the descriptive statistics of the variables used in this research. The sample consists of 40 individual firms originating from 15 countries, these firms have been analyzed over a time frame of 7 years which totals to 320 observations for every variable. An exception here is the exposure to weak institutional environments of which the sample consisted of 55 firms which were only analyzed over a 1-year time frame. This will be further explained in chapter 4.3.2. Table 4.2 presents the descriptive statistics of the variables used for hypothesis 2.

(34)

27

TABLE 4.1: Descriptive statistics hypothesis 1 and 3

TABLE 4.2: Descriptive statistics hypothesis 2

(1) (2) (3) (4) (5)

VARIABLES N mean sd min max

CSP 55 48.095 13.10 20.32 75.23

Related Product Diversification 55 0.0390 0.956 0 0.366

Unrelated Product Diversification 55 0.153 0.237 0 1.055

Total Product Diversification 55 0.192 0.242 0 1.055

Exposure to Weak Environments 55 0.211 0.257 0 1

CSP Industry Mean CSP Country Mean 55 55 44.33 43.87 1.581 1.383 43.09 40.92 47.67 47.67 Firm Size 55 21,640 67,035 60 476,223

- Log Firm Size 55 8.256 1.836 4.174 13.07

Company 55 20.50 11.56 1 55

Country 55 10.13 5.601 1 15

Year 55 2,018 0 2,018 2,018

(1) (2) (3) (4) (5)

VARIABLES N mean sd min max

CSP 320 48.76 15.97 15.46 89.02

Related Product Diversification 320 0.0561 0.119 0 0.587 Unrelated Product Diversification 320 0.212 0.295 0 1.540

Total Product Diversification 320 0.268 0.300 0 1.578

SDGs effective CSP Industry Mean 320 320 0.693 45.06 0.462 4.133 0 40.68 1 56.73 CSP Country Mean 320 44.00 2.061 39.97 51.96 Firm Age 320 53.17 30.77 2 131

- Log Firm Age 320 3.767 0.712 0.693 4.875

Firm Size 320 31,274 84,181 115 552,80

- Log Firm Size 320 8.779 1.754 4.745 13.22

(35)

28

4.2. Assumptions of regression analysis and robustness check

4.2.1. Normal distribution of the variables

One of the assumptions of OLS regression is a normal distribution of the variables as non-normality might weaken the validity of statistical outcomes (Hair, Black, Babin, Anderson & Tatham, 1998). In cases of non-normality, a common solution for this problem is to take the natural logarithm of each variable in question (Baltagi, 2008; Poole & Farragi, 1971). Thus, firm age and firm size were logarithmically transformed while the other variables remain unchanged.

4.2.2. Heteroskedasticity

The assumption of homoskedasticity is that the variance of regression errors is constant and thus unrelated to any predictor. Contrastingly, heteroskedasticity exists when this assumption is violated and implies that the coefficients are skewed to one side of the regression, rather than the true regression line. This violation of constant variance can invalidate statistical inferences and should thus be tested for and being resolved if necessary (Hayes & Cai, 2007).

A simple way to identify heteroskedasticity is to graph a scatterplot of the fitted values of the dependent variable and heteroskedasticity was not identified in this test. A more formal way of testing for heteroskedasticity is the Breusch-Pagan test of which the outcomes can be found in table 4.3. The null hypothesis assumes a constant variance and the p-value tests whether the null hypothesis can be rejected. Since all p-values are >0.05, there is not enough evidence to reject the null hypothesis and it is thus safe to assume that heteroskedasticity is not found to be a problem in the dataset.

TABLE 4.3: Breusch Pagan test for heteroskedasticity

Independent variable Related PD Unrelated PD Total PD

Chi2(1) 0.05 0.36 0.18

Prob > chi2 0.8227 0.5496 0.6723

4.2.3. Correlations and multicollinearity

(36)

29

Table 4.4 represents the correlation matrix of all variables used in this study. Based on extensive research, Dormann et al. (2013) concluded that multicollinearity starts to severely distort the model at an absolute threshold of 0.7. Using this threshold, the correlation matrix shows that total product diversification and related product diversification are strongly correlated with a significant value of 0.921. However, this does not come as a surprise since total product diversification is calculated as the sum of related and unrelated product diversification. Since the independent variables will never be used in the same model, this breach is not expected to distort the results of this study. Furthermore, year and SDGs effective are highly correlated with a significant value of 0.845. This correlation was expected since the adoption of the SDGs was measured as a dummy variable of year. Since the SDGs are a time-dependent variable, both will still be utilized in this study.

TABLE 4.4: Correlation matrix

(37)

30

TABLE 4.5: VIF results for independent variables

VIF mean: 1.22

4.3. Regression results

Tables 4.6, 4.7 and 4.8 present the results of the Ordinary Least Squares regression analysis using fixed effects and robust standard errors. An overview of the variables and scales used in the regression for a better understanding of the unstandardized coefficients can be found in appendix B. As there are three hypotheses tested through regression analysis, each will be discussed individually. In addition, for hypothesis 3, 1-, and 2-year lags were conducted and analyzed which can be found in table 4.9.

4.3.1. The relationship between product diversification and CSP

Table 4.5 presents the results of the first regression analysis, where model 1 only includes the control variables for comparative purposes and in model 2, 3, and 4 the independent variables are included individually. Due to the high correlation between related product diversification and total product diversification, it was not possible to include a model with all the independent variables.

Model 1 shows that without the inclusion of the independent variables, three of the control variables are highly significant. First of all, the CSP industry mean has a highly significant positive effect on CSP (β=0.716, p<0.01), CSP country mean also has a highly significant and positive effect which seems to be a stronger predictor with β=2.212 and p<0.01. This indicates that a firm’s CSP is highly dependent on both industry as well as country standards. Finally, firm size has a highly significant and positive effect on CSP (β=2.899, p<0.01), meaning that

Variable Variance Inflation Factor Tolerance

Related product diversification 1.05 0.951987

Unrelated product diversification 1.46 0.682838

Total product diversification 1.59 0.626959

CSP industry mean 1.12 0.893731

CSP country mean 1.28 0.782596

Firm size (log) 1.17 0.857592

Firm age (log) 1.20 0.830348

Market-to-book ratio 1.12 0.891475

Firm performance 1.08 0.927091

Country 1.13 0.886767

(38)

31

the bigger the firm gets, the higher the CSR performance. On the contrary, the control variables firm age, market-to-book ratio and firm performance all have a negative but insignificant effect on CSP.

After the inclusion of the independent variable related product diversification in model 2, the control variables CSP country mean and firm size remain highly significant, and the significance level of CSP industry mean decreases to p<0.05. The inclusion of related product diversification evokes the R-squared to increase from 29.4% to 32%, thus slightly increasing the explanatory power of the model. However, related product diversification itself has a negative, though insignificant effect on CSP (β=-5.258, p>0.1), therefore not supporting hypothesis 1a.

(39)

32

TABLE 4.6: OLS regression results for hypothesis 1

CSP (H1) (1) (2) (3) (4)

VARIABLES Model 1 Model 2 Model 3 Model 4

Control variables

CSP industry mean 0.716*** 0.594** 0.630** 0.644** (0.242) (0.257) (0.254) (0.254) CSP country mean 2.212*** 2.242*** 2.194*** 2.169***

(0.720) (0.767) (0.785) (0.787) Firm size (log) 2.899*** 3.247*** 2.967*** 2.857***

(0.736) (0.744) (0.727) (0.732)

Firm age (log) -1.889 -1.199 -2.400 -2.070

(1.783) (1.783) (1.801) (1.783) Market-to-book ratio -0.0965 -0.097 -0.1906 -0.168 (0.440) (0.442) (0.447) (0.442) Firm performance -1.150 -0.616 -0.958 -0.819 Country fixed effects

(2.697) Yes (2.608) Yes (2.552) Yes (2.522) Yes Independent variables Related PD Unrelated PD -5.258 (7.700) 7.066* (3.899) Total PD 5.482 (3.883) Constant -86.31*** 35.39 37.84 36.29 (31.75) (34.55) (35.27) (35.28) Observations 320 320 320 320 R-squared 0.294 0.320 0.326 0.323

(40)

33

4.3.2. The moderating effect of exposure to weak institutional environments

Table 4.7 presents the results of the second regression analysis which is, unlike the other regressions, only conducted over the year 2018. However, due to missing data, 40 firms are not sufficient to run a regression with all the control variables included, since a proper regression analysis requires a minimum of 10 observations per variable (Tabachnic & Fidell, 2007). Therefore, the sample size was increased by 15 firms which provided the necessary data in the year 2018. This brings the total number of observations to 55, which means that 5 variables can be included in the analysis. Since the first regression revealed that only the CSP industry mean, CSP country mean and firm size are significant predictors of CSP, they have been included in the second regression while the insignificant control variables were left out.

Similar to the first result table, model 1 only includes the control variables and in model 2, 3, and 4 the independent variables are added individually. Additionally, model 5, 6, and 7 present the interaction effect of the exposure to weak institutional environments with the respective independent variables.

Since a relatively low number of individual coefficients in the regression show a significant effect on CSP, this section will not discuss all models separately but focus on the striking details. First of all, the CSP country mean has a steady significant effect on CSP across all seven models. Interestingly, the sign has changed to a negative effect, while it was positive in the first regression. Firm size still has a positive effect on CSP in the second regression, however, it is not significant in all models. The CSP industry mean takes on a negative sign as well, however, it is not significant anymore.

Though a lower amount of the individual coefficients are significant in the second regression, the R-squares are higher compared to the first regression models and increase steadily with the inclusion of the independent variables and interactions. However, none of the independent variables, nor the interactions are significant. The most logical explanation for the increased

R-squares is that the explanatory power is higher due to the small number of observations, being

(41)

34

TABLE 4.7: OLS regression results for hypothesis 2

CSP (H2) (1) (2) (3) (4) (5) (6) (7)

VARIABLES Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7

Control Variables

CSP industry mean -0.848 -0.847 -0.726 -0.721 -0.687 -0.501 -0.416

(1.150) (1.165) (1.224) (1.224) (1.332) (1.300) (1.376) CSP country mean -3.164*** -3.149*** -3.809*** -3.680*** -4.322** -5.102*** -5.136***

(0.906) (0.931) (1.031) (0.992) (1.600) (1.670) (1.831)

Firm size (log) 2.957** 2.932* 2.582* 2.441 3.418** 3.127 3.029

Country fixed effects Independent variables Related PD (1.457) Yes (1.496) Yes 1.833 (14.79) (1.510) Yes (1.609) Yes (1.663) Yes 4.221 (23.04) (1.894) Yes (2.008) Yes Unrelated PD 12.65 11.41 (10.19) (25.34) Total PD 11.96 9.953 Interactions (10.27) (21.66) Weak environments -9.278 -10.14 -11.30 (10.88) (10.17) (11.88)

Related PD x Weak environments -6.497

(98.25)

Unrelated PD x Weak environments 4.784

(54.68)

Total PD x Weak environments 9.204

(48.48) Constant 220.1*** 219.5*** 245.1*** 240.3*** 266.3*** 294.5*** 293.6***

(69.92) (71.34) (69.97) (70.08) (85.10) (84.77) (88.45)

Observations 55 55 55 55 55 55 55

R-squared 0.374 0.374 0.396 0.395 0.392 0.414 0.416

(42)

35

4.3.3. The moderating effect of the Sustainable Development Goals and lagged analyses

Table 4.7 presents the results of the third regression model, including the moderating effect of the enforcement of the Sustainable Development Goals in 2016. Since the control variables and the main relationship between product diversification and CSP were already tested in the first regression analysis, this table only includes the interaction effects as illustrated in model 1, 2, and 3.

Model 1 shows that only two of the control variables remain significant after the inclusion of the interaction effects. The CSP industry mean remains to have a positive and significant effect (β=0.583, p<0.05) and firm size still has a positive and highly significant effect on CSP (β=3.282, p<0.01). Interestingly, the CSP country mean is not significant anymore, just like firm age, market-to-book ratio, and firm performance are insignificant. After the inclusion of the interaction effect between related product diversification and the SDGs, there is a slight increase in the R-Squared of 32% to 32.3%. However, related product diversification remains to have a negative, though insignificant effect (β=-10.54, p>0.1), whereas the interaction effect has a positive, insignificant effect on CSP (β=13.91, p>0.1).

(43)

36

of unrelated product diversification were favorable before 2016. The two lines cross each other at a level of 0.5, being a moderate level of diversification.

Model 3 shows similar results as model 2 but the coefficients and significance levels are slightly lower. The R-Squared increases from 32.3% to 33.6% with the inclusion of the interaction effect between total product diversification and CSP. The effect of total product diversification is now positive and significant (β=9.387, p<0.5), and the adoption of the SDGs still has a positive and significant effect on CSP (β=8.513, p<0.5). However, similar to model 2, the interaction effect between total product diversification and the SDGs is negative (β=-12.94,

p<0.5). This effect was again graphically depicted in a margins plot in figure 4.2. As illustrated

in the margins plot and the coefficients in model 3, the interaction effect of total product diversification is a little less strong compared to unrelated product diversification but it follows a similar pattern.

FIGURE 4.1: Moderating effect of FIGURE 4.2: Moderating effect of total unrelated product diversification x SDGs product diversification x SDGs

(44)

37

The lagged analyses reveal some surprising and interesting results compared to the third regression in terms of the independent variables and interaction effects. First of all, model 1 and 4 reveal that related product diversification has a negative an significant effect on CSP (β=-24.62, p<0.05, and β=36.18, p<0.05). Thus, related product diversification is negatively significant in both models, but has a stronger effect in the 2-year lag analysis. Another striking result is that the individual effect of the SDGs is negative in model 1-4, and positive in model 5 and 6. However, the direct effect of the SDGs is not a significant predictor of CSP anymore in the lagged analyses. The general pattern in the two lagged analyses is that the effects strengthen in the 2-year lag, both in terms of the coefficient betas as well as in terms of significance.

Even though the SDGs do not seem to be an individual predictor of CSP anymore in the lagged analyses, it does result in significant interaction effects with all independent variables throughout all models. Model 1 and 4 reveal that the interaction effect between related product diversification and the SDGs is positively and significantly related to CSP, where the effect is stronger in the 2-year lag (β=22.00, p<0.1, and β=33.70, p<0.01). Contrastingly, model 2 and 5 show a positive and highly significant relationship between unrelated product diversification and CSP (β=13.45, p<0.01, and β=18.76, p<0.01). However, similar to the third regression analysis, the interaction effect between unrelated product diversification and the SDGs results in a negative, significant relationship (β=-12.79, p<0.01, and β=-18.98, p<0.01). Finally, total product diversification has a positive though insignificant effect in model 3 (β=7.901, p>0.1) but a positive and significant effect in model 6 (β=11.44, p<0.05). In terms of interaction, both model 3 and 6 show a negative and significant interaction effect between total product diversification and the SDGs (β=-9.209, p<0. 1, and β=-13.51, p<0.05).

The effect of product diversification on Corporate Social Performance in the non-renewable energy industry:

Master Thesis International Business and Management