Impact of US Multinational Enterprises Scope of Internationalization on its Corporate Social and Financial Performance.

Master Thesis

Impact of US Multinational Enterprises Scope of

Internationalization on its Corporate Social and Financial

Performance.

by

Leonid Riazanov

University of Groningen — Newcastle University Business School

Dual Award in International Business Management &

Dual Award in Advanced International Business Management

Supervisors: Dr. Rudi de Vries and Dr. Elizabeth Alexander Student Number: S3906868 – 180514410

Abstract

The relationship between Corporate Financial Performance and Corporate Social

Responsibility Performance has been a topic of discussion in literature for the last several decades. This study explores the relationship between Corporate Social and Corporate Financial Performance, while introducing the level of a company’s internationalization of business activities as a moderator. This thesis will argue that high and low scopes of internationalization will negatively affect the results that CSR could have on a firm’s performance and will present a set of arguments that identifies the relationship between CSR and CFP as non-linear. Analysis of sample of 139 American multinational firms found support for existence of positive linear relationship between CSR performance and CFP. However, study found limited evidence in support of moderation effect of

1

Acknowledgements

I would like to express my deepest gratitude to my supervisors, Dr.

Alexander and Dr. de Vries with all the help, feedback and support that

guided me in writing this work. I would also like to thank prof. Muller

for his advice regarding topic of CSP – CFP relationship. I also want to

thank all members of my family and friends, whose constant support was

important for me when I was writing this thesis. Finally, I would like to

express gratitude to all members and lecturers of university of Newcastle

and University of Groningen from 2018/2019 dual award program, for

their work and support with both academic and admirative sides of the

program.

Thank you!

2

Table of Content

1. Introduction………4

2. Theoretical Framework……….7

2.1 Relationship Between Corporate Social Responsibility and Corporate Financial Performance………..7

2.2 The Impact of a firm’s internationalization on Corporate Social Responsibility and Corporate Financial Performance………12

2.3 Conceptual model……….19

3. Research Methodology……….20

3.1 Empirical Setting………..20

3.2 Data Collection………...21

3.3 Sample Profile………..……22

3.4 Variables and Measurement……….…………..…22

3.4.1 Independent Variable………...…22 3.4.2 Dependent Variable………..24 3.4.3 Moderator………...…..25 3.4.4 Control Variables………...……..26 3.4.5 Analysis Strategy………..27 3.5 Preliminary Analysis………29 3.5.1 Normality………..29

3.5.2 Skewness and Kurtosis……….30

3.5.3 Heteroskedasticity………30

3.5.4 Multicollinearity and Autocorrelation……….….31

4. Results………...….31 4.1 Descriptive Statistics………...….31 4.2 Correlation………..…………..34 4.3 Regression Results………..…..35 4.4 Robustness Check……….……38 5. Discussion………..………39 5.1 Theoretical Implications………...………41 6. Conclusion………..……….……..42

6.1 Limitations and Future Research………...……….42

Bibliography………..44

Appendices………...…….54

Appendix 1………54

Appendix 2………55

3

List of Figures

Figure 2.2 – Conceptual model Framework………...…19

Figure 3.1 – ASSET 4 pillars……….24

List of Tables

Table 3.2 – Description of variables………..……..…………27

Table 4.1 – Descriptive Statistics………...33

Table 4.2 – Pearson Correlation Matrix……….34

Table 4.3 – Regression results for H1a and H1b………36

Table 4.4 – Regression results for H2………...….38

Table 4.5 – Robustness check results for H2, year 2016………...……39

Appendix 1.1 – List of the firms in the sample………..…54

Appendix 2.1 – Normality tests, years 2012 – 2016………,,,…...55

Appendix 2.2 –Test on skewness and kurtosis, years 2012 – 2016………...57

Appendix 2.3 – Scatterplot of residuals, years 2012 – 2016………..………59

Appendix 2.4 – VIF tests on Multicollinearity, years 2012 – 2016………...…61

Appendix 3.1– Regression Results for H1a and H1b, years 2012 – 2016……….64

4

1. Introduction

While it’s is commonly accepted by some that business in our society has learned to ignore geographical boundaries, evidence often shows a much more complicated picture. In particular, our world currently is neither focused on its states’ borders nor is it

completely borderless – it is set in between, as “distances—geographic, cultural, administrative/political, and economic—increase, cross-border interactions tend to decrease” (Ghemawat, 2011). For this particular reason, a company can rely on its Corporate Social Responsibility policies as a way to try to overcome cultural, political or economic distances in order to gain trust and acceptance from local stakeholders and consumers.

With that said, the quality of the company’s relationships with local firms, manufacturers, government representatives and labor organizations determines the performance of a newly established subsidiary of a multinational enterprise in a host country (Oehmichen & Puck, 2016). This quality determines how fast subsidiaries can learn and adapt to a new environment and what image they might project. For that purpose, Corporate Social Responsibility policies could be used to reduce risks associated with cultural, political or economic distances (Husted & Allen, 2006), and to earn the trust and acceptance from local stakeholders and consumers. However, while firms might gain moral capital and enhance their images from strong Corporate Social Responsibility policies, the cost of maintaining consistent socially responsible policies require investments of a firm’s

administrative and financial resources (Bouquet & Deutsch, 2008), which could otherwise be used in projects with more certain financially successful outcomes.

The field around the Corporate Social Responsibility and Corporate Financial

Performance relationship is well studied (Orlitzky, et al., 2003; Margolis & Walsh, 2003), with a considerable number of authors presenting theories that argue for a positive

5

managers in pursuit of their personal goals despite a firm’s bad performance (Preston & O’Bannon, 1997). With that said, part of the recent debate argues that the link between CSR and CFP needs to be more complex than strictly linear (Kim, et al., 2018; Barnett & Salomon, 2006). In particular, from this point of view, several publications find evidence that suggest a curvilinear relationship between a firm’s socially responsible activities and its performance (Barnett & Salomon, 2012; Wang, et al., 2008).

Meanwhile, due to an increasing variety of stakeholders and pressure from headquarters, it becomes harder for a company to maintain consistent CSR policies throughout all of its subdivisions as it expands globally into more diverse markets (Bondy & Starkey, 2014; Surroca, et al., 2012), each of those with its own unique conditions and rules. This is particularly noteworthy in relation to the field of FDI management, which argues a firm’s gradual geographical expansion into new regions would exceed its internal capabilities at some point, thus leading to additional costs and a decline in profit. (Gomes &

Ramaswamy, 1999). Considering that management of socially responsible policies is part of the internal functions of the firm, it also affects a firm’s incapability to match its internal structure with its global presence. Furthermore, taking into account various negative consequences that excessive international expansion has on a firm’s financial performance (Lu & Beamish, 2004; Hoskisson & Turk, 1990) and misalignment of socially responsible policies with increasing internationalization of the firm (Laudal, 2011), the level of internationalization affects the impact that socially responsible policies could have on corporate performance. With that in mind, the discussion of both costs and benefits that the implementation of socially responsible policies could have on a firm’s financial performance, as well what to what effect a firm’s level of internationalization could have on the hypothesized link between CSR and CFP might contribute to the field of research.

Therefore, this study seeks to answer the following research question:

“Does the degree of a company’s internationalization have an impact on the non-linear

relationship between a company’s CSR and its financial performance?”

6

relationship is formed due to the initially disadvantageous position a firm has owing to investment into socially responsible policies compared to firms not concerned with CSR. This leads to an initial downwards slope. However, either with an accumulation of support from the stakeholders (Barnett, 2007) or from a competitive advantage through the implementation of CSR policies (Kim, et al., 2018), a firm would start to gain

financial return from its CSR investment, reversing the direction of the downwards slope. Following that, this study will hypothesize that there is a moderating effect of a firm’s level of internationalization on the relationship between CSR and CFP. By engaging in internationalization and expanding into new markets, a firm would improve its financial performance (Ruigrok & Wagner, 2003) and give opportunities to form a network of connections with a larger pool of stakeholders and firms. These, in turn, would provide greater benefits via the exchange of knowledge and technology, thus stimulating

innovation (Almeida & Phene, 2004).The company’s success in aligning to and tackling the demands of the stakeholders within this new network (Jacqueminet & Trabelsi, 2018) would allow a company to receive additional benefits from its stakeholders’ satisfaction (Freeman, 1984). However, as a firm’s expansion goes beyond a certain point, it would start to incur additional costs (Lu & Beamish, 2004; Hoskisson & Turk, 1990; Rugman & Verbeke, 2005) that push down its performance, as the internal structure of the firms is unable to a adjust to its increasing scope of activities. Consequently, as a company grows, it becomes harder for the firm to control all of its functions and align them with

implemented CSR policies (Laudal, 2011), and to meet often conflicting demands of an increasing pool of stakeholders (Jacqueminet & Trabelsi, 2018).

The analysis of the relationship between a firms’ CSR and financial performance is done by comparing the given ratings of socially responsible behavior of these companies with their accounting-based financial performance. In the later stages of analysis, the level of internationalization is introduced as a moderator, which would be measured by the

number of subsidiaries and the company’s geographical presence. The potential impact of internationalization on the CSR-CFP relationship could be established by using available data on selected MNEs from the results of the analysis of the CSR-CFP relationship, structuring them into several groups based on available data on MNE’s market

7

2. Theoretical Framework

This section covers the relevant literature and explores existing arguments in order to formulate the hypothesis. In the first part, the concept of CSR is discussed and presented from the perspective of potential costs and benefits for international businesses.

Following that, the main theoretical arguments regarding the relationship between CSR and CFP are highlighted. Furthermore, the implication of internationalization for a

company’s CSR and financial performance is outlined. Finally, a conceptual model of the hypothesized relationships between variables is presented.

2.1 Relationship Between Corporate Social Responsibility and Corporate Financial Performance

The relationship between CSR and CFP has undergone considerable research over the years and includes a significant amount of academic publications, with Bragdon & Marlin (1972) and Moskowitz (1972) being among the first to look into this relationship. After analysis of 127 major publications in this field, Margolis and Walsh, (2003) found evidence in favor of a positive association “between a company's social performance and its financial performance” (p. 277). Orlitzky, et al (2003), who conducted a meta-analysis of available work and theoretical contributions shows more evidence in favor of a

positive relationship between CSP and CFP. Following several papers that found a link between CSR and a firm’s financial performance, Hirigoyen and Poulain-Rehm (2015) further investigate this link with an analysis of a relationship based on a more detailed description of existing theories regarding positive, negative, and neutral relationships between the variables, as well as calculating several indicators of both CSR performance and CFP. The analysis was done by implementing the Pearson correlation coefficient on data from 329 firms. It revealed a negative impact of CSR values such as “human and social rights in the workplace”, “community commitment” and “market behavior” on financial indicator of “creation of market value” (p. 38). In their research, Hirigoyen and Poulain-Rehm refer to previous contributions, which helped form potential explanations on the nature of the discussed relationship.

8

Considering that the results drawn on the link between CSR and potential financial performance were not always compatible, Grewatsch and Kleindienst conducted a more recent analysis of available research on the relationship between CSP, which is viewed as a consequence of firm’s CSR policy (Carroll, 2018), and the financial performance of companies in an attempt to find similar patterns between them. An analysis of a sample of papers, taking into account ones more recently added to the field, has shown that the relationship found by various authors was more divided, ranging from “positive, insignificant, negative, U-shaped, inverted U-shaped, or asymmetric” (Grewatsch & Kleindienst, 2017, p. 383). From the papers analyzed, 59% of the studies show a positive relationship, 9% suggest a negative relationship, and 32% report “other relationships including non-findings or mixed results” (Grewatsch & Kleindienst, 2017, p. 398).

Description

Social Impact Hypothesis / Instrumental stakeholder

theory

Freeman (1984); Clarkson (1995); Donaldson & Preston (1995); Mitchell, et al.,

(1997)

Positive

Theory explains how meeting expectations of various diverse stakeholder groups improves the performance of

the firm Name of the Theory Authors Nature of the relationship

Firms with low risk are more inclined to participate in CSR

due to stable income Positive

Positive Synergy Hypothesis

Waddock & Graves

(1997) Positive

Theory connects the Social Impact model with Slack Resources by arguing that CSR leads to an increased revenue that is later reinvested into

CSR, creating a CSR cycle Slack Resources Hypothesis Waddock & Graves

(1997)

Negative Synergy Hypothesis

Preston & O’Bannon

(1997) Negative

Large levels of investment in CSR reduces CFP, which further leads to constraints in

resources and a fall of investment in CSR Trade-off Hypothesis Friedman (1962); Vance

(1975) Negative

The implementation of CSR leads to additional costs and

causes negative impact on a firm’s profits, which in turn results in a descrease of a

firm’s share price

Managerial Opportunism Hypothesis

Preston & O’Bannon

(1997) Negative

Top management of the firm would try to extract wealth

for personal needs by reducing expenditure on CSR when financial performance is

high and increasing investment in CSR in days of bad performance for personal

9

Studies take into account firm characteristics such as its size, structure of ownership, strategic orientation or degree of innovation, as well as behavior of managers and their actions and firm’s commitment to sustainability as internal moderators. In terms of external factors, the papers commonly identify the relationship with stakeholders,

industry characteristics and the business and use them as moderators. Considering the fact that “scholars have called for more research on the contingencies—moderators and

mediators—affecting the CSR–CFP relationship” (Grewatsch & Kleindienst, 2017, p. 384), the focus of the question shifted from proving that CSR creates financial benefits to investigating what conditions CSR could be beneficial to an MNE.

One particularly noteworthy study comes from Grewatsch and Kleindienst (2017), who reported that 32% of established relationships between CSR and CFP are either show mixed findings or no definitive results. Comparison of their paper with the publications by Margolis and Walsh (2003) and Orlitzky, et al (2003) shows how the perception of the CSR-CFP relationship has changed. While authors of the aforementioned papers and theories in this field, such as the Clarkson and Instrumental stakeholder theory (1995), Friedman and Trade-off hypothesis (1962) and Preston and O’Bannon with Managerial Opportunism hypothesis (1997), have mostly argues that the relationship is linear and either positive or negative, an increased number of publications with mixed results are indicating a shift. Papers by Barnett & Salomon (2006), (2012); Kim, et al. (2018); (Muller, 2018) recognize there is a less linear and complex nature to the relationship. The curvilinear relationship between CSR performance and CFP, which is discussed in the publication by Barnett & Salomon in 2006, incorporates evidence of positive effects of social sreening for SMI funds, such as filtering fuction of social sreening for high-quality firms which are able to afford CSR investments, and are therefore, more likely to be stable and have above-average returns, and accounts for evidence of negative impact of too stringent social screening, which cuts off significant amount of profitable firms and reduces pool of investment opportunities. After finding support for a curvelinear

realtionship with partial support for the hypothisized realtionships behind social screening and financial market returns, the authors expanded upon the topic of a curvelinear

10

ability of a firm to identify, act on, and profit from opportunities to improve stakeholder relationships through CSR.” The accumulation of this might explain why stakeholders consider certain firms more credible. Because of this, firms with active CSR policies, that were initially at a disadvantageous position compared to firms not

concerned with CSR can start to turn their CSR activities into material returns, which would change the initial downward slope of a curvelinear realtionship, pushed down by costs of inversment in CSR, to an upward slope (Barnett & Salomon, 2012).

The concept of SIC, described in Barnett & Salomon paper (2012), bares a resemblance to the concept of copetitive action (CA) , defined by Kim, et al. (2018, p. 1098) as “externally directed, specific, and observable competitive moves to enhance a firm’s competitive position”. Furthermore, Kim, et al. consider both concepts of CA and CSR as “firm actions that do not only reinforce a firm’s competitive position but also fulfill its responsibilities as a business entity to society” (p. 1098) and through reasoning, based on potential overlap of economic and ethical responsibilities that correspond to CA and CSR accordingly, the authors integrate the concepts of CA and CSR into their discussion of the CSR-CFP relationship. This idea of acquiring competitive advantage through the

fulfillment of ethical responsibilities and later converting them into tangible returns partially corresponds with Barnett & Salomon’s description of the CSR-CFP relationship. While Kim, et al and Barnett & Salomon consider similar ideas on the relationship of CSR and CFP, Kim et al chose to separate the concept of CSR into positive and negative CSR (PCSR and NCSR, where the strength of each’s relationship with CFP is moderated (measures by the level of a firm’s competitive action., where the strengths of both the positive PCSR – CFP realtionship and the negative NCSR - CFP relationships are moderated by the level of the firm’s Competitive Action.

Similar to previously mentioned studies (Barnett & Salomon, 2012; Kim, et al.,2018), Wang, et al. consider that stakeholder perception of the firm in a competitive or dynamic environment will have an effect on the relationship between Corporate Financial

11

separate departments that are devoted to corporate philatropy when the level of CSR activity at the firm increases (Brammer & Millington, 2003).

From the literature analyzed, it is clear that the majority of the authors are inclined to believe that certain relationships exists between a firm’s level of Corporate Social Responsibility and its Corporate Financial Performance (Margolis & Walsh, 2003; Orlitzky, et al., 2003; Grewatsch & Kleindienst, 2017). Furthermore, as discussed in the literature, CSR could be presented as beneficial for performance and could be used to improve performance by satisfying various groups of stakeholders (Freeman, 1984; Clarkson, 1995), reinvesting profit gain back into CSR programs in expectation of greater profits (Waddock & Graves, 1997). Futhermore, as evident from meta-analysis of significant amount of work by Margolis and Walsh (2003) and Orlitzky et al (2003), results of previosly done research strongly suggest existence of positive linear relationship between CSR and CFP. With that, hypothesis 1a is formulated as:

Hypothesis 1a: There is a positive linear relationship between a company’s

Corporate Social Performance and their Corporate Financial Performance.

12

scholars argue there is a negative impact on performance, due to additional costs to the firm from CSR investment (Friedman, 1962; Vance, 1975), opportunistic usage of CSR by managers for personal gains (Preston & O’Bannon, 1997) and a depletion of resourses from CSR investment which would result in a lower profit in the future and therefore, lesser CSR investment (Preston & O’Bannon, 1997). Unlike the aforementioned arguements, Barnett & Salomon (2006) illustrate how CSR could be benefitial and damaging for performance at the same time with their example of SMI funds and the mechanisms of social screening. Taking into account both sides of the argument, several publications approached the relationship between CSR and CFP from a non-linear perspective (i.e. Wang, et al., 2008; Barnett & Salomon, 2012; Kim, et al., 2018). Therefore, it would be sensible to argue that certain levels of CSR under specific

conditions could lead to a positive or a negative effect on a firm’s financial performance. With that, hypothesis 1b is formulated as:

Hypothesis 1b: The relationship between a company’s Corporate Social

Performance and their Corporate Financial Performance is curvilinear.

13

CSR as a concept is identified by Wood (2010) as “a set of descriptive categorizations of business activity, focusing on the impacts and outcomes for society, stakeholders and the firm itself” (p. 51). The issue of CSR is considered important for international business, which has led to the research such as Bondy and Starkey (2014) on topic of how CSR activities of MNEs tackle local issues that firm would have in foreign environment.

Bondy & Starkey acknowledge that MNEs “face unique opportunities and challenges

resulting from operating across national borders, where they are embedded in different contexts with different structures and relational networks” (2014, p. 5). Based on Husted and Allen’s work (2006), the authors created a conceptual model that explores the

relationship between CSR strategies and international management that defines pressures on the choice of CSR, CSR internationalization strategies, the variation of those strategies and the solutions of global-local CSR issues. According to Husted and Allen (2006), effective CSR management may lead to acquiring a competitive advantage and reduce risks, consequently leading to financial performance benefits for the company. However, managing CSR is often accompanied with “strong pressures of institutional isomorphism that attenuate the strategic logic” (Husted & Allen, 2006, p. 838), which would add further complexity to managing CSR for companies that operate in multiple countries. With either global or local strategies, pressures from consumers and competitors within multinational markets and from the requirements for investment demand a company’s integration in the product market. Results of the interviews, used by authors for to

construct a framework that highlighted differences between what effect CSR strategy has in creating legitimacy at home or in other countries, could be considered as showcase of issues that MNEs face in their international business activities and implementation of their CSR policies (Bondy & Starkey, 2014). The result of the research by Bondy & Starkey determines how potential differences between countries affects what local stakeholders and institutions perceive as desirable in terms of a firm’s activities and its socially responsible policies. In particular, these differences represent a unique

combination of each country’s institutional capabilities. For example, a country’s level of technological capabilities impacts technological development for firms within that

14

subsidiaries and make a firm work as a whole on global level, an MNE’s headquarters need to take into account the organizational practices that have evolved in these diverse institutional environments (Kostova, 1999).

This aligns with a publication by Surroca, Tribo, & Zahra (2012), which discusses the existence of differences between each country’s environment, as well as how a lack of knowledge of these differences or even deliberate ignorance and transfer of Corporate Social Irresponsibility (CSiR) could damage a subsidiary’s reputation in that geographical region. One of their relevant findings is how pressure from a company’s headquarters can put constraints on a subsidiary’s attempt to participate in CSR when it is necessary and demanded by local stakeholders, which can put the subdivision in a considerably weaker position. Furthermore, considering how differences institutions between an MNE’s home country and its subsidiary’s country can form institutional distances and further challenge a firm’s ability to legitimize and successfully conduct business there (Kostova & Zaheer, 1999; Peng, 2012), misuse of CSR or transfer of CSiR practices to subsidiaries could sabotage attempts of that subsidiary to gain required legitimacy (Surroca, et al., 2012), and lead to decline in profit or market share (Graafland & Smid, 2004). Additionally, the benefit that a company receives from CSR would be dependent on formal and informal institutional distances of countries within the geographical location portfolio, where a firm would have higher CSR performance if its subsidiaries were located in countries with lower formal and informal institutional distance (Keig, et al., 2019).

Bouquet and Deutsch (2008) also acknowledge that “enhancing the CSR of a firm also involves a set of attendant costs that can be high for MNEs, suggesting the need for careful strategic thinking” (p. 755). These authors find evidence in support of their hypothesis of a non-linear relationship between corporate social responsibility and the multinationality of a firm from data on social responsibility from the KLD data set and data on accounting and finance from Standard & Poor’s Research Insight database. Among the required components for maintaining CSR around the world, the authors list adoption of global ethical codes of conduct, the monitoring of business partners for non-compliance and employing specialized senior executives tasked with developing CSR. However, the costs of having CSR could be compensated by gaining competitive

15

The field of research on International Business and MNE’s management of Foreign Direct Investments (FDI) elaborates on potential reasons as to why firms aim to internationally expand. In particular, Hymer (1976) explains the drive for

internationalization with the concept of firm-specific advantages (FSA), which explains how firms would make foreign direct investments if their FSAs allow them to cover additional costs from business conducted abroad. “Internalization theory” further explores MNE behavior under external influences, analyzing the acquired costs from the management of the network of links among subsidiaries and between parent and

subsidiary companies (Rugman & Verbeke, 2008; Rugman, 1981). Among the issue that could be within the MNE’s network of links is how the degree of internationalization or the substantial size of an MNE’s subsidiary could cause difficulties in maintaining control over subsidiaries (Peng & Beamish, 2014); (Lu & Beamish, 2004), in the ways a

subsidiary could manage connections and relationships with local stakeholders and its headquarters (Oehmichen & Puck, 2016); (Almeida & Phene, 2004) and what impact it all has overall of an MNE’s performance. Particularly, in publications on the process of internationalization and the further development of a firm as its size increases, scholars often discuss financial growth from internationalization until a certain climax, after which the benefits are outweighed by the acquired costs from expanding into unfamiliar territories and the firm’s performance decreases. (Ruigrok & Wagner, 2003; Gomes & Ramaswamy, 1999). An example of such acquired costs could be those associated with the liability of foreignness and coordination costs, as new subsidiaries in a foreign environment would have higher costs and a harder time competing with established local firms. Meanwhile, the headquarters would struggle to coordinate their growing operations in an increasing number of countries, each with its own unique business environment (Lu & Beamish, 2004). These costs include costs associated with limitations of internal governance and control structures, which would cause an emphasis on maintaining

16

internal structures and systems, which no longer reflect the external context of the firm’s global operations, unless the firm is successful in reconfiguring its internal setting to match its global role, in which case it might again experience positive performance (Tushman & Romanelli, 1985; Benito & Welch, 1997).

Interestingly, regarding the relationship between Corporate Financial Performance and the degree of internationalization, Ruigrok & Wagner (2003) established a non-linear link between the two. Their article considers two major groups of theories, foreign direct investment and theories of the multinational firm, which explain benefits of

internationalization. Foreign direct investment aims to explain benefits of portfolio diversification of a firm with risk reduction (Lessard, 1976), while theories of the multinational firm explains benefits through an increased completeness of the firms through utilization of resources that are not available domestically (Fayerweather, 1978).While taking into account the cost of internationalization, the article names Siddharthan and Lall (1982) among the first to indicate the potential large-scale costs of internationalization for the firm, arguing that “increasing degrees of internationalization and concomitant organizational and environmental complexity may eventually exhaust managerial capacity” (Ruigrok & Wagner, 2003, p. 67). With an analysis of German firms, the authors show a convex relationship between Return on Assets and a firm’s degree of internationalization, and a concave relationship between operating costs of sales and the degree of internationalization. Adding to the building body of work, a publication by Riahi-Belkaoui (1998) uses the non-linear relationship between internationalization and a firm’s performance to analyse the effect of a company’s international expansion, dividing degrees of internationalization into a high, middle and lower range. With this, they found a negative relationship between variables at both the low and high ranges and positive in the middle range.

17

the factors that have put additional constraints on a firm for doing business abroad are market driven differences in foreign exchange, higher insurance costs, trade barriers and a potential lack of knowledge on regulation. Among non-market constraints, one factor could be the existing skepticism with local stakeholders. Together, all those constraints could create frustrations that might lead to lower productivity within the firm, while increased international presence could reduce the firm’s total profit through lower margin foreign sales. Furthermore, the publication by Muller considers the relationship between CSP and financial performance as U-shaped, with both strong investments or non-existent investments in CSR leading to better financial performance for the company.

With the establishment of subsidiaries on a larger geographical scale, a firm would choose to either tailor its CSR policies according to the global strategy of the whole enterprise or it would allow subsidiaries to implement strategies adjusted to the local environment in which they operate (Muller, 2006). Considering that, subsidiaries of MNEs, in order to legitimize in a host environment (Graafland & Smid, 2004), would try to satisfy demands of stakeholders through a localized CSR approach. This is

significantly important for subsidiaries located geographically and institutionally far from the company’s country of origin as there is a high probability, they experience lack of attention and supply of resources (Surroca, et al., 2012). Furthermore, with a company’s expansion into new regions, investment into a positive relationship with stakeholders would provide not simply legitimization for a newly established subsidiary, but would allow the formation of a network of connections, from which the subsidiary could gain knowledge or exchange technology (Almeida & Phene, 2004). With the successful

alignment of demands with the majority of stakeholders within this network, a firm would be able to benefit from its CSR policies (Jacqueminet & Trabelsi, 2018). However,

18

operations (Laudal, 2011) and the necessity to meet requirements of stakeholder groups from various countries across the globe, each of which could have their own

understanding of what it means to be socially responsible (Gjølberg, 2009). According to Campbell, et al. (2012), despite the potential ability of CSR policies to aid the subsidiary in establishing itself in the eyes of local institutions from countries with high cultural and institutional distance, an MNE would avoid engaging in CSR due to higher costs.

Within the reviewed literature on CSR, it is argued that a firm’s international activity should be carefully tailored due to existing differences of demands from each stakeholder (Bondy & Starkey, 2014) (Surroca, et al., 2012). With low levels of internationalization, a firm would get fewer benefits from CSR (Bouquet & Deutsch, 2008) and have lower performance than firms that exploited opportunities of new markets (Fayerweather, 1978; Lessard, 1976). Consequently, the geographic expansion of a firm would allow access to a greater pool of resources (Fayerweather, 1978) and reduce risk through portfolio diversification (Lessard, 1976). In turn, a growing number of stakeholders from firm’s expansion into new regions would allow firm to accumulate more Stakeholder Influence Capacity (Barnett, 2007) through succesful implementation of socaily responsible policies, with that gaining recognition and goodwill from stakeholders, on which firm could later later rely for greater financial gain, thus futher stimulating the relationship between CSR and CFP. However, once a firm’s expansion goes beyond a certain point (Ruigrok & Wagner, 2003; Gomes & Ramaswamy, 1999), the firm’s internal systems would no longer be able to properly coordinate its external activities, leading to

accumulation of costs from mismanagement (Lu & Beamish, 2004; Hoskisson & Turk, 1990; Rugman & Verbeke, 2005) and misalignment of a firm’s actions with its CSR policies (Laudal, 2011), thus reducing any potential benefits from corporate social activities on a firm’s financial performance. That leads to Hypothesis 2 being:

Hypothesis two: The level of a company’s internationalization and international

19

2.3 Conceptual model

The presented conceptual model highlights the expected relationships between the discussed variables. The model shows that within the assumed relationship between CSR policy and financial performance, CSR is taken as an independent variable and CFP as a dependent variable. Furthermore, the level of the company’s internationalization is predicted to moderate the relationship defined in hypothesis 1.

Figure 2.2 – Conceptual model

20

3. Research Methodology

3.1 Empirical Setting

In order to find the hypothesized relationship between Corporate Social Responsibility and Corporate Financial Performance, MNEs located in US were used for data collection and analysis. Then, appropriate firms were selected based on size, as well as the expected availability of data on their Corporate Social activities from specialized CSR databases, such as ASSET4 or KLD. One of the reasons this study chose to select companies located in the US is the evident FDI flow from the US to the rest of the world, which is estimated to be around 240 000 Million USD. This places USA on the second place in outwards FDI flow compared to other countries before 2016, with subsequent first place after 2016 (OECD, 2017). The substantial FDI flow from the US into other countries’ industries indicates that US businesses have a positive perception on investing abroad. This could affect the views of American MNEs on internationalization. The substantial FDI flow from the US into other countries’ industries indicates that US businesses have a positive perception on investing abroad. This could affect the views of American MNEs on internationalization.

Another factor that highlights suitability of MNE from USA is the general perception towards CSR inside United States, which in turn would determine how consequential investment in CSR is for firm’s business activity. With that said, firms with U.S. tend to “freely acknowledge their ethical and social obligations” (Forte, 2013, p. 816), which manifests in fulfillment of discretionary responsibilities – voluntary service before the society. This is particularly reflected on U.S firm’s motivation toward CSR principles as being part of their core values, as well as on their readiness to “ address CSR principles, processes, and issues” (Forte, 2013, p. 818). Furthermore, as a country with a large concentration of MNEs and a high level of environmental regulation, investment in CSR for US firms would allow them to maintain their reputation in the eyes of domestic government institutions and amongst consumers, which in turn allows for a better comparison of MNEs at different stages of internationalization. In addition to that, an analysis of similarly sized companies from one country would allow for a better

21

3.2 Data Collection

The research will center around quantitative methods and available secondary data. The chosen time range for the dataset is 2012 to 2016, with the goal of capturing firms’ performance dynamics from the 2008 financial crisis until the present day. Data

collection was done through several resources, such as Compustat, Orbis and Thomson Reuters ASSET4 database.

Following that, stock tickers of firms with ASSET4 scores in the 2012-2016-time range were used to recover data from the Compustat database on financial indicators of a firm’s performance relevant to this study. These indicators are Total Assets, Total Sales,

Research and Development Expense, Net Income or Loss, and a firm’s Standard Industry Classification code. Firms with data missing in any of these categories between the years 2012 to 2016 were removed from the dataset. This resulted in an initial sample size of 255 companies with an equal number of observations for each company within the given time range.

In order to account for similarities in the companies’ sizes, the firms from the resulted sample were cross-referenced with the 2016 Fortune 1000 list of US companies. 255 companies from dataset were present in this list.

Further data on the Foreign Sales of these 255 companies’ geographical segments was taken from Compustat. The initial data reported by companies in the sample included each segment’s Export Sales, Net Sales by geographical region, the classification given to the segment and the name of the geographical region in which the foreign affiliates were located. In order to avoid mistakes in combining data from different years into one, data from each year was downloaded and analyzed separately. Following that, in order to improve precision of the data (Ruigrok & Wagner, 2003; Geringer, et al., 1989), Export Sales were accounted for where reported, with segments identified as “domestic” and disclosed as located in the USA or North America were removed. Finally, data on Net Sales from the remaining segments was summarized for each company to be used later in calculating the company’s ratio of Foreign Sales to Total Sales.

22

classification and industry classification was collected. After combining all available data, it was revealed that certain firms had missing data in several categories, which led to their removal, reducing the sample to 139 companies for each year, with total of 695

observations.

3.3 Sample Profile

The resulting sample includes collected data on each firm’s performance indicators, equally weighted ASSET4 ratings for each company’s CSR, the total number of

subsidiaries, the number subsidiaries located abroad and controlled subsidiaries as well as the number of countries where the firm has a presence. As was previously mentioned, the initial sample of 255 companies was reduced to a pool of 139 companies after combining data from ASSET4 and COMPUSTAT, as well as data on each company’s geographical segments and information from Orbis, and from removing companies with missing datasets.

Considering how various characteristics of the industrial field in which each company operates could influence its financial performance and R&D expenditure (Waddock & Graves, 1997), as well as its engagement in CSR (Cai, et al., 2012) and its

internationalization strategy (Andersson, 2004), in order to account for the factor of industry for the purposes of this study companies were selected from the same industrial field. The industry was determined by taking the first two digits of the companies’ North American Industry Classification Codes (NAICS) and cross-referencing them with United States Census Bureau (2017). Results for the sample showed that predominant majority of companies operate within the manufacturing sector, which allowed to exclude 29

companies from different industries from the sample.

3.4 Variables and Measurement

3.4.1 Independent Variable

23

and the adoption of a responsiveness strategy or posture’. One of the methods of measuring CSP is highlighted by Wood (2010) and uses KLD, ARESE or Vigeo for multidimensional measurements on CSP, the process of conducting CSR though current management practices and the impact of CSR through potential reputation effects. In particular, an article by Orlitzky, et al. (2003) separates CSP into “CSP disclosures,” “CSP reputation ratings,” “social audits, CSP processes, and observable outcomes” and “managerial CSP principles and values” (p. 408). In return, CSR is defined by Carrol as encompassing “economic, legal, ethical, and discretionary expectations that society had of organizations at a given point in time” (2018, p. 750).

With that said, in their 2012 publication, Ioannou & Serafeim, name KLD, Bloomberg and Thomson Reuters ASSET4 as sources of CSR, with the authors themselves relying on an ESG dataset in their research. The authors mention that data for an ESG score is

collected by analysts through “stock exchange filings, CSR and annual reports, non-governmental organizations’ websites, and various news sources” (Ioannou & Serafeim, 2012, p. 21). It should be mentioned that only data from DataStream ASSET4 database was available at the moment of the study (without direct access to DataStream as well). Consequently, the score system from the ASSET4 database was an equally weighted score of economic, environmental, social and corporate governance pillars. This equally weighted score was taken as measurement of a firm’s CSP.

24

Figure 3.1 – ASSET 4 pillars ( QSG Equity Research Team, 2009)

3.4.2 Dependent Variable

According to Orlitzky, et al. (2003, p. 407), a firm’s financial performance could be categorized into three subdivisions: “market-based (investor returns), accounting-based (accounting returns), and perceptual (survey) measures.” Elaborating on these categories, the authors discuss the use of price per share or share price appreciation as indicators of shareholder satisfaction. This, in turn, would determine the market value of the company, or the reliance on accounting-based measures such as Return on Assets (ROA), Return on Equity (ROE) or Earnings per Share, as those reflect “internal decision-making

capabilities and managerial performance” (2003, p. 408) and thus, reveal a firm’s internal efficiency to a degree. In addition to that, in their description and analysis of major publications in the field of CSR and CFP relations (2017), Grewatsch and Kleindienst observe that publications relied on indexes such as ROA, ROE, or ROS for analysis of financial accounting performance and on measures of Tobin’s q for financial market-based performance.

25

increasing internationalization and consequential limitations of organizational internal capacities, accounting-based measures are best suited for the analysis. Therefore, Return on Assets was chosen as the measurement of CFP.

ROA is used as an “indicator of how profitable a company is relative to its total assets” (Investopedia, 2019) and could give an understanding of how efficiently the company’s assets are used for a firm’s activities in creating profit. According to Masa'deh, et al (2015, p. 135), the ROA ratio “measures the return by using assets to produce income” and is calculated by:

ROA = (Net Profit before Interest and Tax / Total Assets) * 100

In particular, ROA is used for financial performance measurement in several works that looked into the relationship between CSR and CFP from a non-linear perspective (Barnett & Salomon, 2012; Wang, et al., 2008). Furthermore, ROA is commonly taken as an accounting based financial performance indicator within academic literature that explores relationship between a firm’s CFP and its degree of internationalization. For example, in Lu & Beamish ( 2004), the authors use ROA to measure a firm’s accounting-based perfomance in their general linear squares random effect model. Riahi-Belkaoui (1998) also use ROA as a firm’s performance indicator in their piecewise linear regression model.

3.4.3 Moderator

According to the literature, a moderator is a qualitative or quantitative variable that “affects the direction and/or strength of the relation between an independent or predictor variable and a dependent or criterion variable” (Baron & Kenny, 1986, p. 1174). This study uses the level of a firm’s internationalization as a moderator. Within the academic literature that looks into the degree of internationalization (DOI), the methods of

measuring the level of internationalization varies from measuring structural, financial and psychological attributes (Ruigrok & Wagner, 2003), with reliance on a company’s

26

In turn, the ratio of foreign sales to total sales (FSTS) has been commonly used within literature on internationalization of MNEs to indicate a firm’s engagement with foreign markets (Muller, 2018; Ruigrok & Wagner, 2003; Sullivan, 1994). The characteristic of the FSTS ratio suits purposes of this study, as it would allow for the determination as how much each firm has globalized their business activity. This is further categorized within a high, medium or low level of internationalization. Therefore, the FSTS ratio was taken as the measurement of the firms’ DOIs.

3.4.4 Control Variables

Among the variables that could influence the CSP-CFP relationship are additional

financial performance ratios, the firm’s size, measure of company’s risk and research and development intensity. These are all considered as potential control variables. In addition to that, Margolis and Walsh name a firm’s risk, industry and size as “most common control variables” (2003, p. 14) within the academic literature on the CSR-CFP

relationship. The authors explain usage of industry as a control variable with existence of differences between industries in ecological impact on environment from business

activity, or in enforced regulatory rules and scrutiny. Furthermore, taking into account a firm’s size would be necessary as “larger firms may have greater resources for social investments, attract greater pressure to engage in CSP or—just the opposite—succumb to a diffusion of responsibility” (Margolis & Walsh, 2003, p. 14). Finally, control for risk is relevant due to evidence of more active CSR participation from firms within the lower risk category (Cochran & Wood, 1984). This shows a potential relationship between CSR and risk category of the firms (Orlitzky & Benjamin, 2001). Additionally, adding

Research and Development Intensity to the control variables as an indicator of the company’s innovative activities would help this study establish a more precise relationship between CSR and CFP, as creation of socially responsible products by company was often preceded by process and product innovation (McWilliams & Siegel, 2000).

27

debt to equity ratio (Lu & Beamish, 2004). Control for industry will be done through selecting companies for the analysis that operate within one industry. A firm’s innovative activities are controlled through data on Research and Development Intensity.

Table 3.2 – Description of variables

Type of Variable Name Measurement Measurement Description Data Source

Independent Variables Corporate Social Performance (CSP) ASSET4 Rating

Calculated from 250 key performance indicators Thomson Reuters ASSET4 database Dependent Variables Corporate Financial Performance (CFP)

ROA = (Net Income / Total Assets) *

100% Compustat

Moderator

Degree of Internationalisation

(DOI)

FSTS Ratio = (Foreign Sales / Total Sales) *

100% Compustat

Control Variables

Firm's Size Number of

Employees Number of employees in the firm Compustat

Risk Debt – to – Equity Ratio

= (Debt in Current Liabilities + Long-Term Debt / Shareholder’s

Equity) * 100% Compustat Research and Development R&D Intensity

= (R&D Expenditure / Total Sales)

* 100% Compustat Robustness Check Variables Geographic Diversification of Internationalisation Ratio of Geographic Dispersion

Average of ratio of number of firm’s subsidiaries to maximum number of subsidiaries within the

sample in sum with ratio of number of countries with FDI to maximum of FDI countries within

the sample.

Orbis

3.4.5 Analysis Strategy

28

subsequent application of a polynomial regression in from of quadratic model. Based on the comparison of the results from the two regressions, the nature of the relationship will be established.

A linear regression would be formulated as:

Y=b1X + b0 +e; where Y – dependent variable; X- independent variable; b0 – constant variable;

b1 – slope;

According to Wood (2012 ), quadratic models are one of the methods which are used to assess U-shaped relationships in the field of management and economics. Moreover, quadratic model with the application of quadratic term in the regression is consistent with other studies in this area (Muller, 2018). Thus, model will be initially formulated as:

Y= b1*X1 +b2*X1^2 + b0 + e; where, Y – dependent variable; X- independent variable; b0 – constant variable; b1 + 2*b2*X1 –slope; e- error;

With incorporation of all previously mentioned variables into operating model for the analysis of hypothesized relationship and the assumption of linear nature of the control variables (Aiken & West, 1991), linear model, with timed fixed effects for regression of the combined data from 2012 to 2016, is presented as follows:

ROA = b1*CSP + b2*Risk + b3*R&D + b4*Size + δ*t + b0 + e;

Where quadratic model would add quadratic term of the variable of CSR

29

Regarding the approach of measuring the effect of internationalization on the CSR-CFP relationship, according to the rules of moderated regression (Aiken & West, 1991; UNC Kenan-Flagler Business School, 2019), DOI and the product of all DOI variables from the CSR-CFP quadratic model will be added in the moderated regression analysis

(Prameswari, et al., 2016), with independent variable of CSR interacting with DOI moderator. For analysis of the moderating relationship on low, medium and high values of DOI, data sample will be analyzed through regression, in order to determine whether moderator has any significant effect on the data. In case if significant moderation effect is found, regression would be done on centered terms with using PROCESS developed by Andrew F. Hayes (2019), in order to assess and visualize effect of DOI at low, medium and high levels. Following model describes moderated regression with the assumption of the quadratic model accepted as best fit:

ROA = b0+ b1*CSP + b2*Risk + b3*R&D + b4*Size + b5* CSP^2 + b6* DOI + b7*CSP^2*DOI + δ*t + e;

3.5 Preliminary Analysis

To ensure that the selected strategy for the analysis provides the most accurate results, the data sample should be analyzed on several criteria before conducting further regression analysis for both linear and quadratic models.

3.5.1 Normality

In order to estimate if data from the sample is normally distributed, both Shaprio – Wilk and Kolmogorov – Smirnov tests have been done for more precise results. Each of the discussed tests are commonly used to assess data on the criteria of normality of the data (Mardia, 1980; Rosenthal, 1968), however, as both tests has their own disadvantages (Yazici & Yolacan, 2007), such as initial restrictions for Shaprio – Wilk for sample size less than 50 or relatively low power of Kolmogorov – Smirnov test compared to other available tests (Razali & Yap, 2011 ), this study will make conclusions from the results of both test.

As can be seen from the results, most variables are not normally distributed. Such

30

several authors have identified to be not normally distributed (Barnes, 1982; Ezzamel, et al., 1987). However, normal distribution of the data from the dependent and independent variables is not always a necessary condition for linear regression and for its variation, polynomial regression (Habeck & Brickman). Nevertheless, such data would be difficult to assess on confidence intervals, as well as conduct goodness of fit test (Tellinghuisen, 2008). For the purposes of increasing confidence level from the regression, a

bootstrapping method, which is a random resampling of data through significant amount of times by using statistical software, will be implemented in the further analysis

(Henderson, 2005; Marais, 1984; Noh, et al., 2011). In addition, bootstrapping method has been used within the literature on firm’s financial performance (Lin & Chang, 2011) and CSR (Zhou, et al., 2018).

3.5.2 Skewness and Kurtosis

While normality tests show that data have non-Gaussian distribution, a test on skewness and kurtosis will indicate to what extend data is asymmetrically distributed. Tests can be found in the Appendix 2.2 section.

Results from the test indicate that data on Return on Assets, Employees, Leverage and Research and development is highly skewed and show signs of kurtosis, with Equal-Weighted Rating being moderately skewed and FTST being symmetrical. Noteworthy observation is that skewness and kurtosis results of Return on Assets change from 2012 to 2016 toward less asymmetrical distribution.

3.5.3 Heteroskedasticity

In order to test assumption of heteroskedasticity of the variables, a scatterplot of residuals was plotted. Construction of the residual scatterplot with Standardized Predicted Value and Standardized Residual at x and y axis would allow to visually assess heterogeneity of variances across the data (Davidian & Carroll, 1987; Kutner, et al., 2005). Created

scatterplots shows that residuals form a clearly identifiable pattern in “Industry and CFP”, “Firm size and CFP” and “R&D Intensity and CFP” from 2012 to 2016. Moreover, a pattern that could indicate heteroskedastic distribution is identifiable in CSP-CFP

31

& Bocheng, 2008). In order to further assess heteroskedasticity of the data, a Breusch‐ Pagan test will be implemented, which is used to assess whether “model errors are associated with any of the model predictors” (Astivia & Zumbo, 2019, p. 5).Results from the test for 2012 show p=0.221, with p=0.100 in 2013, p=0.001 in 2014, p=0.011 in 2015 and p=0.156 in 2016, which could be interpreted as presence of heteroskedasticity in the data.

3.5.4 Multicollinearity and Autocorrelation

Test of the data on multicollinearity could be done through VIF test. Results for

continuous variables within the regression model for years 2012 -2016 could be found in the Appendix 2.4 section. Results of the test indicate that there is no multicollinearity within the regression model.

Variables were tested on autocorrelation through Durbin-Watson test. Results for the test throughout the years were 1.893 for year 2012, 1.858 for year 2013, 1.715 for year 2014, 2.040 for year 2015 and 1.957 for year 2016, indicating that there is no significant autocorrelation detected.

4. Results

4.1 Descriptive Statistics

This section presents description of variables that are used in regression analysis. Figure 4 summarizes data used on 696 observations for 139 companies over 5 years, starting from 2012 and ending at 2016.

The firms’ average financial performances within the sample shifted from the low of 6.25% in 2012 to the high of 7.32% in 2014. After which, the average performance fell to 6.50% in 2016, showing an average performance of 6.81% over the total time period. Descriptive statistics for independent variable of CSP shows that firms within the sample predominantly have scored highly in terms of CSP. Due to steady increase in firm’s CSP from average of 71.69% in 2012 to 86.61% in 2016, overall mean value of CSP for 695 observations is on high level of 78.00%. Furthermore, higher median values in

32

97.03% to lowest of 9.98%, it could be said that majority of the firms have improved their CSP scores within 5 years, with 33.26% being lowest CSP score present in the sample in 2016.

33

Table 4.1 – Descriptive Statistics

Variable Measurement 2012 2013 2014 2015 2016 All

34

4.2 Correlation

As predicted from the literature, the Pearson correlation matrix identifies the existence of a weak but significant correlation (=0.107) between CSR and ROA at a level of 0.01. There is an additional weak but significant correlation between ROA (=0,075) and CSR (=0,119) and DOI. Based on theoretical literature, this would be predicted to be higher. The relatively moderate correlation between CSR and the size of the firm, the control variable, derived from the number of employees, would imply a potential connection between growth of the firm and its CSR performance. The existence of a positive correlation between R&D Intensity and DOI implies the existence of a relationship mentioned by several authors. The results have not identified any strong correlations that would interfere with the regression analysis.

Table 4.2 – Pearson Correlation Matrix

Financial Performance CSP Degree of Internationalisation Risk Firm's Size Research and Development Intensity Financial Performance 1 CSP 0,107** ₁ Degree of Internationalisation 0,075 * _0,119** ₁ Risk -0.047 -0.014 -0.050 1 Firm's Size -0.028 0,290** _{0.032 -0.021 1} Research and Development Intensity 0.067 0.070 0,240** _{-0.048 -0,135}** ₁

35

4.3 Regression Results

In order to establish the nature of the hypothesized relationship, a regression analysis was conducted in several steps. Through each of those steps, variables were gradually added to regression, expanding on the previous model. Initial regression is done for panel data, which encompasses in itself the whole period of 2012-2016, with consequent reiteration of each step for each individual year, to further analyze relationship between variables. As discussed in preliminary analysis, bootstrapping has been implemented for separate regression of each year samples to address existing problems with normality assumptions, with aim to make results of regression for individual years more precise. Because of neglectable improvements n significance levels and due to relatively large sample size, regression for the whole sample of 2012-2016 period do not have bootstrap implemented in it. Bootstrap results for yearly regressions are based on 1000 bootstrap samples, with confidence levels based on Bias corrected accelerated (BCa) and simple sampling. With that said, significance levels of most of the variables in presented regression results have been positively affected by the implementation of bootstrapping.

For the first model, regression is done with presence of control variables only. The results do not indicate any significant relationship between variables, with only constant in the regression model showing significant result (b=6.519, p<0.01). Variable indicating year 2015 was excluded by statistical software from the analysis, due to regression model reaching limit on collinearity statistic tolerance. Following that, independent variable of CSP is introduced into the model, in order to test Hypothesis 1a, which assumes positive relationship between variables. Results of that regression for the whole sample does show great significance for variable of CSP (b= 0.040, p<0.01), indicating support for

Hypothesis 1a. Following that, model 2b introduces quadratic term of independent

variable of CSR, thus testing the Hypothesis 1b (H1b). However, results of the regression for that model do not indicate any significance for both CSP (b= 0.090, n.s.) or CSP squared (b= 0.000, n.s.), which indicates rejection of H1b. Comparison of values of 𝑅2 between model 2a (𝑅2 =0.026, F(1, 686) = 2.293, p<0.05) and model 2b (𝑅2 =0.027, F(1, 685) = 2.090, p<0.05 ) with 𝛥𝑅2_{=0.001, n.s for model 2b shows no evidence of}

improvement of the model with the addition of quadratic term to it.

36

slightly significant linear relationship between ROA and CSP, however, results are consistent in their rejection of H1b. In particular, results of the regression for 2014 (b=0.036, p<0.1), and 2015 (b=0.053, p<0.1) shows slight significance of CSP variable, providing support for existence of positive linear relationship. Regression results for 2016 showed strong significance of CSP in Model 2a (b=0.165, p<0.01). With that said,

Hypothesis 1b in favor of quadratic model is rejected and linear model is taken for further regression analysis of moderating relationship between DOI and previously discussed variables.

Table 4.3 – Regression results for H1a and H1b

Financial Performance

Model 1 Model 2a Model 2b

Firm's Size -0.003 -0.007 -0.007 R&D Intensity 0.061 0.046 0.048 Risk -0.016 -0.016 -0.017 Year 2012 -0.443 -0.026 0.006 Year 2013 0.447 0.769 0.807 Year 2014 0.589 0.829 0.835 Year 2016 -0.274 -0.477 -0.442 CSP CSP- Squared 0.040*** 0.090 0.000 Constant 6.519*** 3.471** 2.094 0.104 0.161 0.163 0.011 0.026 0.027 0.011 0.015*** 0.001 _{1.078 10.695 0.481} Note: * p<0.10; ** p<0.05; *** p<0.01;

Due to rejection of Hypothesis 1b, moderated regression will be used with linear model, defined in hypothesis 1a:

37

The results of the regression for the whole sample without the interaction term indicates significance of independent variable of CSP (b=0.039, p<0.01), with proceeding addition of interaction term showing evidence of significance for CSP (b= 0.086, p<0.01), DOI (b= 0.094, p<0.05), as well as slight significance for the interaction term itself (b= -0.001, p<0.1). F-values and 𝑅2 for the model 2c (𝑅2 =0.028, F (2, 685) = 2.212, p<0.05) and model 3 (𝑅2 _{=0.033, F (1, 684) = 2.329, p<0.05) indicate significance of both models.}

However, while for the model 2c 𝛥𝑅2_{shows high level of significance (𝛥𝑅}2_=0.017,

p<0.01), 𝛥𝑅2 _{between model 2c and 3 indicates much lower level of significance}

(𝛥𝑅2_{=0.005, p<0.1). Due to uncertain results with levels of significance of interaction}

effect and 𝛥𝑅2 from comparison of the models without and with interaction effect, additional regression was done for each year, to examine if evidence of moderating effect could be confirmed. In return, results of those regressions showed no significance of interaction term or significant 𝛥𝑅2 between model 2c and 3, indicating rejection of Hypothesis 2.

38

Table 4.4 – Regression results for H2

Financial Performance

Model 1 Model 2c Model 3

Firm's Size -0.003 -0.008 -0.008* R&D Intensity 0.061 0.035 0.043 Risk -0.016 -0.016 -0.016 Year 2012 -0.443 -0.094 -0.126 Year 2013 0.447 0.718 0.715 Year 2014 0.589 0.786 0.816 Year 2016 -0.274 -0.479 -0.469 CSP 0.039*** 0.086*** DOI 0.015 0.094** CSP * DOI -0.001* Constant 6.519*** 2.998** -0.631 0.104 0.168 0.181 0.011 0.028 0.033 0.011 0.017*** 0.005* 1.078 6.127 3.309 Note: * p<0.10; ** p<0.05; *** p<0.01; 4.4 Robustness Check

In order to explore impact of internationalization on firm’s performance and CSR more deeply, a robustness check is performed by replacing FSTS measurement with calculated ratio of number of subsidiaries to geographical presence of each firm. Due to absence of verifiable data in time range of 2013-2015, on either number of firm’s subsidiaries or geographical presence, robustness check is done for only year of 2016. Results support rejection of Hypothesis 2, with interaction term showing no significance when

incorporated into regression (b= -0.002, n.s.), and with 𝛥𝑅2 _{showing no significant F}

39

Table 4.5 – Robustness check results for H2, year 2016

Financial Performance

Model 1 Model 2c Model 3

Firm's Size -0.001 -0.013 -0.013 R&D Intensity 0.119 0.086 0.088 Risk -0.042 -0.068* -0.073* CSP 0.166*** 0.177*** DOI (Robustness Check) -0.002 0.184 CSP * DOI (Robustness Check) -0.002 Constant 5.843*** -7.812* -8.876* 0.149 0.339 0.341 0.022 0.115 0.117 0.022 0.093 0.001 1.022 6.987 0.215 Note: * p<0.10; ** p<0.05; *** p<0.01;

5. Discussion