Corporate social responsibility and corporate
financial performance in controversial industries:
The moderating role of culture
Thijsje Oenema t.j.oenema@student.rug.nl
s173406 12-6-2017
Supervisor: Prof dr. C.L.M. Hermes Msc Finance
Msc International Financial Management Faculty of Economics and business
University of Groningen
Abstract: This study examines the relationship of corporate social responsibility on corporate financial performance for controversial industries and how cultural dimensions affect this relationship. By examining a unique dataset of 550 firms over 41 countries over a period of 2007–2014, this study found a positive effect of CSR on CFP in controversial industries. Furthermore, the Hofstede cultural dimension uncertainty avoidance weakens this positive relationship. In addition, for the mining industry, the Hofstede cultural dimension power distance weakens as well the positive effect of CSR on CFP. Therefore, it can be concluded that cultural dimensions influences the CSR-CFP relationship in controversial industries. Jel classification: G30, L25, M14
2 1. Introduction
Climate change, depletion of national resources or increasing health costs have recently started gaining more awareness and attention from governments, NGOs and the public in general. This leads to the media and the government negatively pressuring industries that are controversial in nature, as they produce unhealthy or addictive products like tobacco or alcohol. Such industries manufacture products while inflict disproportionate damages on the environment (Cai et al., 2012; Rodrigo et al., 2016). Anti-tobacco campaigns, the “war on sugar” or NGOs boycotting oil companies are a small section of the negative attention surrounding these controversial industries. The negative attention as well as the unethical and irresponsible behaviour towards society affects the firms with related stakeholders more than the ones in the non-controversial industries (Moura-Leite et al., 2014). Hence, to improve the relationship with the stakeholders and enhance corporate financial performance (CFP), corporate social responsibility (CSR) activities are required to strengthen corporate reputation and goodwill (Aguinis and Glavis, 2012; Lindgreen et al., 2012). Cai et al. (2012) argued that controversial industries especially need to engage in CSR activities compared to other industries because of their controversial activities. Using these CSR activities, controversial industries try to repair the damages caused by their wrongdoings and unwanted behaviour (Cai et al., 2012; Orlitzky et al., 2011; Wilson & West, 1981). However, in scientific research, there is not yet a clear consensus on the values or meaning of CSR and how such activities affect the needs of a firm’s stakeholders (Aguinis and Glavas, 2012; Margolis and Walsh, 2003; Wood and Jones, 1995). According to the classic view endorsed by Friedman (1970), the only focus of managers should be profit maximization for the firm’s shareholders. Under this view, CSR engagement destroys firms’ value when profits dwindle. In contrast, the stakeholder view argues that the primary objective of a firm should be satisfying the interests of the stakeholders (Freeman, 1984). In this context, CSR engagement may improve the relationship with stakeholders, resulting in reduced transaction costs (Joyner and Payne, 2002) and improved CFP (Barnet, 2007).
3 2016). In this context, CSR can be explained differently for different informal institutional factors, like culture (Bénabou and Tirole, 2010; Matten & Moon, 2008). How cultural values affect the CSR has gained a growing attention in academics in the last few years. Despite the differences in specific outcomes, the overall conclusion is that culture strongly influences CSR in different firms (Ho et al., 2012; Peng et al., 2012; Ringov and Zollo 2007; Thanetsunthorn, 2015; Waldman et al., 2006). Cultural background shapes a manager’s personal values, and subsequently, their influence on CSR (Hemingway and Maclagan, 2004). Additionally, Rodrigo et al. (2016) argued that culture could be one the reasons why CSR has different effects on the CFP between different countries. Still, evidence is needed to ground this argument (Rodrigo et al., 2016). The importance of understanding how culture affects the CSR–CFP relationship in controversial industries increases with the recent enhanced pressure on controversial industries to exhibit good corporate behaviour (Campbell, 2007; Ioannou and Serafeim, 2012; Rodrigo et al., 2016). Therefore, this study attempts to provide a better understanding of how cultural dimensions influence the CSR–CFP relationship in controversial industries. Hence, the following research questions are drawn:
Do cultural factors influence the relationship between CSR and CFP in controversial industries?
To provide an answer to this research question, there will be an investigation on how CSR affects CFP in controversial firms. Thus, the five cultural dimensions of Hofstede (1980, 2001), which are power distance, individualism, masculinity, uncertainty avoidance and long-term orientation, are used to examine the effect of culture on the CSR–CFP relationship in controversial industries. Based on a sample of 550 firms spread over 41 countries within the alcohol, biotech, cement, gambling, mining, oil, weapon, tobacco and sugar industries, this study finds a positive effect of CSR on CFP. Furthermore, the cultural dimension uncertainty avoidance weakens this positive relationship between CSR and CFP in controversial industries. After an adjustment in sample size with excluding small countries and for the mining industry, only the cultural dimension power distance weakens the positive effect of CSR on CFP. Therefore, I can conclude that cultural dimensions have influence on the CSR-CFP relationship in controversial industries.
4 in this field was based on a smaller selection of controversial industries and countries. To my knowledge, this is the first study of its kind to use a large panel, i.e., 41 countries, to examine the effect of CSR on CFP, as well as to create, a subsample with one industry to investigate if the effect of CSR on CFP differ for the mining industry. My motivation is based on prior research suggesting the investigation of differences between industries in the CSR–CFP relationship (Waddock and Graves, 1997) and because of the different selection of controversial industries between Cai et al. (2012) and Rodrigo et al. (2016). The second contribution is the inclusion of the “new” controversial industries, as characteristics and behaviour towards these industries have changed over time (Fam et al., 2004; Waller et al., 2005). This study includes the sugar industry in the controversial industry sample. The third and most important contribution is the empirical examination of the moderate effects that national cultural dimensions have on the CSR–CFP relationship. After examining the differences in the CSR–CFP outcomes between studies based on culture, the need for more empirical research has been requested by Mar Miras-Rodriguez et al. (2015). This study is the first one to my knowledge, which performs an empirical analyses of the moderate effects culture has on the CSR–CFP relationship in controversial industries.
This study provides valuable insights for managers to get a better understanding of whether CSR activities increase CFP and how certain cultural dimensions influence this relationship. Culture does not change over time and cultural characteristics provide a better understanding of how the public opinion reacts to negative media attention in controversial firms and CSR activities. This research supports the decision-making process of managers regarding CSR initiatives. If cultural values influence this relation, certain strategic decisions regarding CSR initiatives could be implemented to improve CFP, as well as protect the firm from negative consequences regarding the increased negative media attention.
5 2. Literature review
This section starts with a description of the controversial industries, followed by a description of the CSR, based on the legitimacy theory, which questions why controversial industries engage in CSR. This section is followed by several theories and views around the CSR–CFP relationship. Thereafter, institutional theory is described, followed by an explanation of the cultural dimensions and the subsequent hypotheses development of this theory affecting the CSR–CFP relationship.
2.1 Defining controversial industries
6 2.2 Defining corporate social responsibility
7 On the contrary, due to their immoral characteristics, CSR activities of controversial industries are seen by some as untrustworthy and distrustful, given the nature of activities in the past, thereby decreasing organisational legitimacy (Cai et al., 2016; Jo and Na, 2012; Reast et al., 2013).
2.3 Influence of corporate social responsibility on corporate financial performance Different theories are developed to describe the effects of CSR on CFP. This chapter clarifies these theories with the accompanying views. It starts with the stakeholder theory, supporting the positive relationship between CSR on CFP. Stakeholders are different parties with interest in a firm, without whom a firm would not survive (Clarkson, 1995). Thereafter, the negative relation between CSR on CFP is described based on the shareholder theory, followed by the neutral relationship based on the supply and demand framework. Accordingly, of the empirical evidence regarding this relationship is described in chapter 2.3.3, followed by the hypothesis development.
2.3.1 Positive influence of corporate social responsibility on corporate financial performance
To increase CFP, managers need to allocate the firms’ scarce resources to create value. Cornell and Shapiro (1987) stated that to make profit, firms’ allocation of resources needs to satisfy explicit contracts and implicit claims purchased by the stakeholders. Implicit claims are promises towards stakeholders to guarantee certain services, like a promise to a life time job or software updates. Without adequately allocating these scarce resources, stakeholders would become dissatisfied, which would negatively influence the CFP. Based on the stakeholder theory (Freeman, 1984) and the transaction theory (Williamson, 1989), higher levels of CSR satisfy the implicit claims of stakeholders and decrease transaction costs by signaling that they are “doing the right thing” (Joyner and Payne, 2002). If the firm does not meet the needs of non-financial stakeholders, with sufficient CSR activities, they might turn their back towards the firm (Mitchell et al., 1997; Preston and O’Bannon, 1997).
8 an improved relationship with the community can lead to fewer rules and regulations or a lower tax burden set by the local government (Waddock and Graves, 1997). Especially for controversial industries this is important as they produce products that cause negative externalities, like high pollution levels or depletion of natural resources, for the local community (Rodrigo et al., 2016). Thus, CSR in controversial industries, can signal “good management” and increase the relationship with key stakeholders and increase CFP (Cai et al., 2012).
Cai et al. (2012) further built on this and created the value-enhancement theory that clarifies the positive influence CSR has on CFP, which is based on the strategic view of creating shared values by Kramer and Porter (2011). This theory assumes that the market is shaped by social instead of economic needs and with the exact knowledge of how social needs influence firms’ actions, shared values can be created. Shared values are a firm’s actions that are simultaneously beneficial for the firm as well as for the environment affected by these actions. These shared values will reduce frictions between society and controversial industries, resulting in less transaction costs and more social and economic benefits.
Based on these theories, CSR will increase reputation, improve the relation with stakeholders and create shared values between society and controversial firms, resulting in lesser transaction costs. These CSR activities will support the organisational legitimacy in controversial industries, which leads to easier access to resources, and therefore, have a positive influence on CFP (Reast et al., 2013)
2.3.2 Negative influence of corporate social responsibility on corporate financial performance
9 will involve higher financial costs compared with the financial benefits and this disadvantage the firm more compared with firms not engaging in CSR. In controversial industries, this negative influence can be strengthened by immoral managers that make decisions for personal, social and self-esteem reasons, which are not based on what is right or ethical (Cai et al., 2012). These immoral managers place their personal benefits above the risk of destroying shareholders’ wealth with their initiated CSR activities. This so-called insider-initiated corporate philanthropy reasons to incorporate CSR activities are beneficial for the managers and detrimental to other stakeholder’s claims and the firms’ society (Bénabou and Tirole, 2010).
Rodrigo et al. (2016) suggested that controversial industries are prone to a higher punishment from stakeholders as their activities have higher negative social and environmental impacts. Moreover, stakeholders might get the idea that the CSR activities in a controversial industry are trying to cover up the negative social and environmental impacts and trying to build a corporate reputation (Cai et al., 2012; Moura-Leite et al., 2014). Cai et al. (2012) argued that, eventually, stakeholders will notice the real intentions of these immoral managers in controversial industries and will punish these managers, which affect CFP. This view is empirically supported by El Ghoul et al. (2011). They concluded, after examining the effect of CSR initiatives at the cost of equity across US-based firms, that CSR initiatives in controversial industries, like tobacco, are associated with a higher cost of equity capital, while in other industries, CSR activities reduce the cost of equity.
2.3.3 Empirical evidence for the relationship between CSR and CFP in controversial industries
10 More empirical studies have been conducted without a specific focus on controversial industries. Margolis et al. (2009) examined 251 published studies between 1972 and 2007 in the field of CSR and CFP. They concluded, after performing a meta-analyses, that corporate wrongdoings are costly for the firm. Although, no negative significant effect was found between CSR and CFP and therefore they concluded that CSR is not destroying shareholders’ wealth. On the contrary, they found the opposite, a small positive partial correlation (mean r=.13) between CSR and CFP. Orlitzky et al. (2003) examined 52 studies with a total sample size of 33,878 firm-year observations and found a small but positive correlation (0.18) between CSR and CFP. Therefore, in accordance with the empirical studies and the existing literature on controversial industries (Cai et al., 2012; Margolis et al., 2009; Orlitzky et al., 2003), I expect that the positive influence of CSR outpaces the negative influence of CSR. Therefore, I constructed the following hypothesis:
H1: CSR has a positive influence on CFP in controversial industries. 2.4 Institutional theory
The influence CSR has on CFP in controversial industries differs between countries and is influenced by country-level institutions (Ioannou and Serafeim, 2012; Rodrigo et al., 2016). Institutions are “stable, valued, recurring patterns of behavior” defined by there “adaptability, complexity, autonomy and coherence” (Huntington, 1969, p 12). Institutional theory is based on formal rules that are imbedded in socially shared beliefs, values, constraints and informal rules, which shape the behaviour of people and firms altogether (North, 1990). The core assumption of this theory is that the firm will not grow or survive if it is not aligned with the environmental and cultural values of its surroundings (Frynas and Yamahaki, 2016).
11 These differences are in line with the findings of Matten and Moon (2008) who argues that in Europe CSR is driven by the social demands of the community or the government while in the United States, CSR is used for creating shareholder value.
For controversial industries, institutional differences shape stakeholder attitudes towards their products and services. Waller et al. (2005) argued that geographical factors don’t play a major role in shaping people’s attitudes towards controversial industries, but historical, religious and especially cultural factors could play an important role. This idea is supported by Ioannou and Serafeim (2012), who argued that culture plays an important role in shaping the relation between CSR and CFP.
2.5 Culture and its influence on the firms’ behaviour
As Waller et al. (2005) argued, culture plays a certain role in shaping attitudes towards controversial industries. Culture is defined as values, ideas, patterns and beliefs shared by members of a certain group based on symbols, which are embedded in social norms and practices (Adler and Gundersen, 2008; Kroeber and Kluckohn, 1952). These social norms and practices define what is essential and desired in society, thereby influencing preferences. These preferences and certain expectations shape people’s decisions and actions, and in turn, influence their behaviour (Hofstede, 1980). Therefore, culture may be one of the main determinants in corporate decision-making.
As noted by Ioannou and Serafeim (2012), culture plays an important role in shaping CSR actions in firms. Thus, differences between cultures change the fundamentals of CSR engagement. In some cultures, CSR is based on norms and values, and in other cultures, voluntary actions are more important for shaping the CSR (Matten & Moon, 2008). This has been confirmed in extended empirical research examining the influence of culture on CSR engagement (Peng et al., 2012; Ringov and Zollo, 2007; Thanetsunthorn, 2015; Waldman et al., 2006). It can be concluded that culture affects CSR activities and this impact may vary between cultural dimensions, as elaborated in the next chapter.
12 2.6 Hypothesis development: Moderate effect culture
As mentioned in the previous paragraph, different cultural dimensions have different impacts on CSR. In this section, I elaborate on the five (power distance, masculinity, uncertainty avoidance, individualism and long-term orientation) cultural dimensions of Hofstede (1980, 2001) used in most empirical studies to clarify the impact culture can have on CSR (Shi and Wang, 2011).
In the next paragraphs, each cultural dimension is described and what the influence of the specific characteristic of each dimension is on the positive relation that CSR has on CFP, as stated in hypothesis 1. This forms the basis for the hypothesis development described at the end of each paragraph.
2.6.1 Power distance and its effect on the positive CSR–CFP relationship
The first cultural dimension is power distance. Power distance is the degree of acceptance of unequal authority by individuals (Hofstede, 1980). Societies with “high power distance scores” accept orders and listen only to highly placed people with power (House et al., 2004; Waldman et al., 2006). In low power distance cultures, stakeholders are more involved in the decision-making process (Jensen and Meckling, 1976; Waller et al., 2005). Hence, firms face higher pressure in doing good for stakeholders, and therefore, social and environmental issues appear early on and are on a higher ranking in the agenda of low power distance societies (Ringov and Zollo, 2007).
Products and activities of controversial industries negatively influence social and environmental behaviour. As controversial industries incorporate CSR activities, this actvities will be obtained easily in lower power distance cultures, where stakeholder will acknowledge the efforts towards CSR. This acknowledgement can lead to increased reputation and trust, which can lead to higher CFP. As noted earlier, for controversial industries reputation and trust from stakeholders are important factors for the survival of the firm.
13 H2a: The cultural variable power distance negatively influences the positive relationship between corporate social responsibility and corporate financial performance in controversial industries.
2.6.2 Masculinity and its effect on the positive CSR–CFP relationship
Cultures that score high on masculinity like to have power are assertive and materialistic, and are competitive (Hofstede, 2001). This ensures that cultures that are focused on scoring high on masculinity are, in general, not eager to help each other (Steensma et al., 2000). These characteristics have various effects on the CSR–CFP relationship in controversial industries.
One of the reasons to hypothesis that CSR has a positive effect on CFP is to create a positive image and increase reputation and trust to cover up their unethical and immoral activities and products. However, masculine cultures assign less value to ethics and may not value “good” behaviour compared to cultures that score low on masculinity (Steensma et al., 2000; Vitell and Festervand, 1987). If CSR activities increasing the CFP are based on awarding controversial firms by exploiting good behaviour, this will not uphold in cultures scoring high on masculinity. The more selfish behaviour of masculine cultures will strengthen this negative effect. Cultures that exploit selfish behaviour do not care much for environmental or social engagement with the accompanying CSR activities. This is empirically supported by Peng et al. (2012) and Ringov and Zollo (2007) that found a negative relation between masculinity and CSR. They argued that shareholders in masculine countries only think about their own welfare. They mark every CSR activity as destroying the CFP or every action to increase managers’ self-esteem. Therefore, correcting wrongdoings with CSR in controversial industries may not be valued more by shareholders. This view is supported by Mar Miras-Rodriguez (2015) that found studies showing a negative relationship between CSR and CFP when examined on countries scoring high on masculinity. Therefore, I have composed the following hypothesis:
H2b: The cultural variable masculinity negatively influences the positive influence of CSR on CFP in controversial industries.
2.6.3 Uncertainty avoidance and its effect on the positive CSR-CFP relationship
14 Cultures with high uncertainty avoidance prefer rules and regulations to minimize the level of uncertainty. Low uncertainty avoidance amplifies flexible, easy-going and risk-taking behaviour (Javidan et al., 2006). High risk-taking behaviour is associated with actions that are less social or moral (Vitell et al., 1989). Some individuals think that unethical or illegal actions will lead to higher benefits for themselves (Rallapalli et al., 1994). Controversial industries produce products associated with asocial or immoral characteristics, and regarding CSR activities, they reduce the risk-taking behaviour more than in conventional industries (Jo and Na, 2012). Cultures scoring high on uncertainty avoidance may value the risk reduction via CSR in controversial industries. When firms are aware of the restrictions and penalties made by the law that influence CFP negatively, through CSR activities, the risk of penalties may be lowered, and people scoring high on uncertainty avoidance will value this even more. This view is supported by the results of Mar Miras-Rodriguez (2015) that concluded that studies with a positive relation between CSR and CFP is conducted under firms’ residents in countries scoring high on uncertainty avoidance. Therefore, I draw the following hypothesis:
H2c: The cultural variable uncertainty avoidance positively influences the positive relationship between corporate social responsibility and corporate financial performance in controversial industries.
2.6.4 Individualism and its effect on the positive CSR-CFP relationship
Individualism is the degree of how close individuals are with each other (Hofstede, 2001). In high individualistic societies, people do not value social interaction much, are more self-interested and are less ethically responsible (Akaah, 1990; Ho et al., 2012). On the contrary, low individualistic (collectivist) cultures put much value on working together and place the group’s welfare above their individual needs (Peng et al., 2012).
15 not much need for controversial industries to engage in CSR activities. High individualistic cultures do not stress on social and environmental damage caused by controversial industries, and therefore, high individualistic cultures will not value corporations more if they enhance CSR activities (Ioannou and Serafeim, 2012).
Therefore, CSR activities in high individualistic cultures are based more on voluntary actions by social motives of some stakeholders. These voluntary actions, along with less value on ethical behaviour, are not in line with CSR. Thus, CSR activities in controversial industries in high individualistic countries may not be related to increased CFP. On the contrary, in low individualistic cultures, CSR activities are based on requirements by rules, values and norms (Matten & Moon, 2008). These requirements are based upon societal expectations of legitimacy (Matten and Moon, 2008). Therefore, I draft the following hypothesis:
H2d: The cultural variable individualism negatively influences the positive influence of CSR on CFP in controversial industries.
2.6.5 Long-term orientation and its effect on the positive CSR-CFP relationship
Long-term orientation refers to the time horizon of individuals in a society (Hofstede 2001). Cultures with a long-term oriented view put more value on what is good in the long run. In short-term oriented cultures, the past and the present are more important.
16 Corporate Social Responsibility Corporate Financial Performance National Culture - Power Distance - Masculinity - Uncertainty Avoidance - Individualims - Long- Term Orientation
CSR. This view is supported by Mar Miras Rodriguez (2015) who found that firms with a small but positive relation between CSR and CFP are situated in cultures that are high on future orientation. Therefore, I construct the following hypothesis:
H2e: The cultural variable long-term orientation positively influences the positive relation between corporate social responsibility and corporate financial performance in controversial industries.
2.7 Conceptual Model
This conceptual framework illustrates the hypotheses of the previous sections:
17
3. Data and Methodology
To test the hypotheses, I need information about CSR, CFP and culture, along with the accompanying control variables. This chapter provides information about these variables, as well as the sample construction, selection and the methodology used for testing the hypotheses.
3.1 Sample and variable construction
Data for the dependent, independent and control variables were obtained from Worldscope, Asset4 database, Hofstede database and the World Bank database. An overview of the variables and their sources are in the table 1 of the Appendix. This study examines the effect CSR has had on CFP between 2007–2014 in an unbalanced panel dataset with 2729 firm-year observations over 550 firms within 41 countries. The next section describes the variables used in the models, followed by a description of the methodology including the proposed regression models.
3.1.1 Controversial industries
For the sample selection of controversial industries, I selected industries that create addictive and or unhealthy products or are related to moral, environmental or social issues. In line with Cai et al. (2012), I included the tobacco, alcohol, gambling, cement and biotech industries. Furthermore, I included the mining industry in the sample, as it associated with the depletion of natural resources (Rodrigo et al., 2016). Thereby, it should be noted that the selection process of controversial industries in the empirical research between CSR and CFP between Anglo-Saxon countries in the study of Cai et al. (2012) differ from the selection process of Rodrigo et al. (2016) in Latin American countries. Within this study, firms from different cultures and countries are included, and therefore, we base our selection of controversial industries on both studies. Additionally, due to recent discussions about sugar companies being unhealthy and addictive in nature, I included the sugar industry in the sample. These industries together are called controversial industries.
18 group 7, includes sic codes 3240–3241, biotech firms, group 46, with sic codes 2833–2836. For the gambling industry, sic codes 7933 and 79991 are included. Mining industry with sic
codes 1011–1099 and 1221–1241 were further included in the sample. For the sugar industry, I included companies with sic codes 2061–2068.
3.1.2 Dependent variable: Corporate financial performance
Following the literature concerning the relation between CSR and CFP, this study measures CFP with Tobin’s Q (Cai et al., 2012, El Ghoul et al., 2017). Tobin’s Q represents a firm’s total market value of assets divided by the replacement costs. The exact replacement costs are difficult to obtain, and therefore, Tobin’s Q is measured by the market value of equity and debt scaled by total assets (Lioui and Sharma, 2012).
Tobin’s Q = (Market Value of Equity + Total Debt) / (Total Assets) (1) The advantage of Tobin’s Q over accounting measures is that it reflects historical CFP as well as expectations of future CFP. Another advantage of using Tobin’s Q as a proxy for CFP is that it has been used in the existing literature for controversial industries in relation with CSR, which makes the results of this study more comparable.
3.1.3 Independent variable CSR performance
This study used the Thomson Reuters ASSET4 database to measure CSR. This database is used in CSR studies with non-U.S. and U.S firms (Cheng et al., 2014; El Ghoul et al., 2017; Ioannou and Serafeim, 2012). This database contains environmental, social and corporate governance information for more than 6000 companies. Information is collected from publicly available sources like annual reports, NGO websites or CSR reports. Companies are rated based on the collected information using more than 750 individual data points to calculate more than 250 key performances. These ratings are equally weighted valuations of the companies’ performances. Ratings are normalised and standardised to scores between 0 and 100 for each company. These scores are organised into four pillars: economic, corporate governance, environmental and social (Reuters, 2012). Every year the company receives a z-score, which compares its performance with a benchmark based on the rest of the companies in the Asset4 database. In this study, we used three different CSR scores to measure the CFP. The first CSR score is the CSR index. In line with Ioannou and Serafeim, (2012), the CSR
19 index is an equally weighted score of the environmental and social pillar of the Asset4 database.
CSRi=(CSRenv+CSRsoc)/2 (2)
where CSRi is the CSR index, CSRenv the environmental pillar and CSRsoc the CSR social
pillar.
The environmental pillar examines factors such as emission reduction, process innovation and the use of resources. The social pillar contains information about health and safety issues, human rights and product responsibility. These two pillars are also used to construct the environmental score and the social score to measure the CSR. The deviations in the measurement methods are because cultures value environmental and social performance of a company differently (Thanetsunthorn, 2015). In this study, the economic and governance pillars are excluded from the sample because the two pillars have no direct relationship with the CFP.
3.1.4 Moderate variable: Culture
To measure culture, Hofstede’s (1980, 2001) national cultural dimensions, from the Hofstede database2, were used. Hofstede’s cultural framework has widely been adopted in research of
CSR (Ringov and Zollo, 2007; Peng et al., 2012) and in relation with the CFP (Newman and Nollen, 1993)3. Hofstede (1980) analysed employee attitudes in different countries and
identified the following dimensions: power distance, masculinity, uncertainty avoidance and individualism. Analyses differences between the East and West world countries, Hofstede (1991) introduced the fifth dimension, the long-term orientation4. These dimensions score
from 0 till 100, whereby a high score means a high value to that dimension. The database includes information for over 100 countries. Because of the time-invariant nature of Hofstede’s cultural dimensions, and given that cultural values become stable over time, the
2 For Cyprus, Sri Lanka, Kazakhstan, Oman and Zimbabwe, no Hofstede dimensions are available and the firms based in these countries
were excluded from the sample. For Channel Islands and Guernsey, no Hofstede dimensions are available, and thereafter, they were assigned to Great Britain. This classification is grounded on the reason that before the year 2006, firms from Channel Island or Guernsey were allocated to Great Britain. For Qatar, Dubai and Bahrain, the Hofstede dimensions of Arab Countries are given.
3 The work of Hofstede has also been criticised, as it is outdated and based only on empirical research. As a reaction, House et al. (2004)
developed another cultural framework based on values and practices. However, due to data restriction, this has not been used in this research.
4The sixth-dimension indulgence versus restraint, which is not evaluated in this study, is associated with happiness (Hofstede et al., 2010).
20 dimension scores do not change over the years and only one score per dimension per country was included.
3.1.5 Control variables
This study includes firm- and country-control variables to control for unobservable firms and country-specific characteristics that influence the CFP. The included control variables are consistent with previous empirical studies to investigate the relation CSR has on CFP (Cai et al., 2012; El Ghoul et al., 2017; Rodrigo et al., 2016; Waddock and Graves, 1997).
3.1.5.1 Firm-control variables
To control for specific firm variables, firm size (size), the level of risk (risk), return on assets (ROA), research and development (R&D), sales growth (SAG), capital expenditure ratio (CAPEX), and multinationalism (FORSA) are included in the models.
Size: Firm size is used as a control variable because, in general, large firms tend to face lower growth opportunities that influences CFP. Firm size negatively influences the CFP (Fama and French, 1992). In line with Cai et al. (2012), firm size is calculated as the natural logarithm of total assets.
R&D: Research and Development expenditures are associated with higher CFP. With R&D, firms can differentiate their products compared to other similar products, which will lead to higher CFP (McWilliams and Siegel, 2001). According to McWilliams and Siegel (2000), not including R&D in the regression model would lead to upward bias results. Due to a lack of R&D data, this study uses intangible assets divided by total assets as a proxy for R&D. Chan et al. (2001) stated that R&D intensive firms have few tangible assets.
Sales Growth: Firms with higher sales growth (SAG) have lower fixed costs and increase market power, which have a positive influence on CFP (Brush et al. 2000). Therefore, this study includes sales growth as the ratio of total sales of the previous year divided by the total sales (Cai et al., 2012).
Capital expenditure: An increase in capital expenditures will lead to higher stock returns (McConnell and Muscarella, 1985). To capture this effect, this study includes capital expenditure as a percentage of total assets as a control variable.
ROA: Return on Assets (ROA) is used as a proxy for profitability. Profitability is associated with higher CFP (Varaiya et al., 1987). ROA is calculated as EBIT/total assets.
21 (Myers, 1977). On the contrary, based on the “free cash flow theory”, debt can have a positive influence on CFP (Jensen, 1986). Hence, total debt divided by total assets will be used to control for risk.
Multinationalism: The degree of multinationalism of a firm influence CFP. To a certain degree, multinationalism increases CFP. Thereafter, the cost will outweigh the benefits and multinationalism would decrease CFP (Kotabe et al., 2002). For this reason, in line Tallman and Li (1996), foreign sales as a percentage of total sales is used as a proxy for multinationalism.
3.1.5.2 Country control variables
To control for country differences, this study follows Rodrigo et al. (2016) and includes the level of income (GDP) and institutional differences (GOV).
GDP growth: GDP growth is used as a proxy for the level of income, as this influences the CFP (Krugman, 1991; Rodrigo et al., 2016).
Institutional differences: Institutions’ level of strength plays a certain role in CFP (Hansen and Wernerfelt, 1989). Strong market-supporting institutions are positively associated with higher CFP (El Ghoul et al., 2017). In line with Rodrigo et al. (2016), the level of strength of institutions is measured by the mean of the six Worldwide Governance Indicators (GOV). The GOV are cross-country indicators to measure governance (Kaufmann et al., 2011). These indicators are the rule of law, which capture the trust and confidence of agents by the rules of society, voice and accountability. These indicate the freedom of media and the freedom of selecting the government, political stability and no violence, evaluating the degree of destabilisation of the government, government effectiveness, assessing the quality and credibility of the government policies and services, regulatory quality, evaluating the soundness of the policies and regulation of the government and corruption control, evaluating the extent to which public power is used for private interests.
3.2 Methodology
22
increase CSR activities (Campbell, 2007; Waddock and Graves, 1997). If this is the case, OLS can produce biased coefficient estimates. In line with various empirical resources (El Ghoul et al., 2017; Nelling and Webb, 2009, Rodrigo et al., 2016; Waddock and Graves, 1997), this study controls for this possible endogeneity problem with taking the first lag of CSR and the control variables.
Another problem with OLS estimation in the panel data can be the unobserved heterogeneity between cross-sections (Nelling and Webb, 2009). For example, unobserved heterogeneity can arise between firms if managers have different personalities, skills of informal code of behaviour influencing work ethics of employees (Hansen and Wernerfelt, 1989). Regression models with fixed effects control for this unobserved heterogeneity. The importance of including firm-fixed effects over industry control variables is noted by Khaled and Paton, (2005). They argue that empirical research is less reliable without including firm-fixed effects because the heterogeneity between firms’ play an important role in testing the relationship between CSR and CFP. However, a random effect model would be preferred, considering the time-invariant variable culture used to test hypothesis 2. Nevertheless, the Hausman test5 shows a significant p-value, indicating that including random effects or the
pooled OLS model would lead to biased results. Therefore, and in line with Nelling and Webb (2009), El Ghoul et al. (2017) and Khaled and Paton, (2005) this study controls for firm-fixed effects to test hypothesis 1 and time-fixed effects as well. White standard errors are used to overcome potential problems with heteroscedasticity (Brooks, 2014).
To test hypothesis 1, the following regression models are estimated:
1. ToQi,t = α0 + α1CSRi,t-1 +α2RISKi,t-1+ α3R&Di,t-1 +α4ROAi,t-1 +α5SAGi,t-1+α6SIZEi,t-1
+α7GDPi,t-1 +α8GOVi,t-1 +α9CAPEXi,t-1 + α10FORSAi,t-1 + νi + ct + εi,t
2. ToQi,t = α0 + α1SOSCi,t-1 +α2RISKi,t-1+ α3R&Di,t-1 +α4ROAi,t-1 +α5SAGi,t-1+α6SIZEi,t-1
+α7GDPi,t-1 +α8GOVi,t-1 +α9CAPEXi,t-1 + α10FORSAi,t-1 + νi + ct + εi,t
3. ToQi,t = α0 + α1ENVSCi,t-1 +α2RISKi,t-1+ α3R&Di,t-1 +α4ROAi,t-1 +α5SAGi,t-1+α6SIZEi,t-1
+α7GDPi,t-1 +α8GOVi,t-1 +α9CAPEXi,t-1 + α10FORSAi,t-1 + νi + ct + εi,t
23 where FV is a measure of CFP, CSR is a measure of CSRindex, Sosc and Envsc are used as proxies for CSR, Risk is the risk ratio, R&D is the Research and Development ratio, ROA is the return on asset, SAG is the sales growth, Size is the firm size, GDP is GDP growth, GOV is an indicator for institutions, CAPEX is the capital expenditure ratio, FORSA is foreign sales as a percentage of total sales, εis the standard error, i is index for firm, t is the time index, νi is the firm-fixed effect and ct is the time-fixed effect.
3.2.1 Moderating effect culture
To test hypotheses 2a, 2b, 2c, 2d and 2e, the moderating variable culture is included as an interaction term in model 1. As mentioned in the above paragraph, a problem arises when including a time-invariant variable within a fixed-effects model. Hofstede’s cultural dimensions do not move over time and in a fixed effect this dimension will be absorbed by the intercept and would lead to an omitted culture coefficient. However, Wooldridge (2009) and Boyce and Wood (2011) argued that the interaction coefficient can still be interpreted when one of the underlying variables is omitted. Thus, to test hypothesis 2, the main parameter to interpret is the interaction coefficient between the cultural dimensions and the CSR index, and the cultural variable is dropped out of the regression models.
Another problem concerning our regression models is the multicollinearity between the interaction variable and the CSR. Therefore, I centred the independent variables to enhance interpretability and efficient estimation of the interaction variables (Aiken and West, 1991; Jaccard and Turrisi, 2003). This technique, to overcome multicollinearity between the
interaction term and the CSR, is heavenly debated. Katrichis (1992) stated that mean-centring the interaction and independent variables leads to biased results of the main effects. Kromrey and Foster-Johnson’s (1998) argued that results of centred and non-centred regression models are identical. However, to overcome the high correlation between CSR and the interaction variable, the independent variables are mean-centred.
The following models are estimated:
4. ToQi,t = α0 + α1CSRi,t-1 + α2(PD*CSR)i,t-1 + α3RISKi,t-1+ α4R&Di,t-1 +α5ROAi,t-1
+α6SAGi,t-1+α7SIZEi,t-1 +α8GDPi,t-1 +α9GOVi,t-1 + α10GDPi,t-1 + α11CAPEXi,t-1
24
5. ToQi,t = α0 + α1CSRi,t-1 + α2(MAS*CSR)i,t-1 + α3RISKi,t-1+ α4R&Di,t-1 +α5ROAi,t-1
+α6SAGi,t-1+α7SIZEi,t-1 +α8GDPi,t-1 +α9GOVi,t-1 + α10GDPi,t-1 + α11CAPEXi,t-1
+α12FORSAi,t-1 + νi + ct + εi,t
6. ToQi,t = α0 + α1CSRi,t-1 + α2(UAI*CSR)i,t-1 + α3RISKi,t-1+ α4R&Di,t-1 +α5ROAi,t-1
+α6SAGi,t-1+α7SIZEi,t-1 +α8GDPi,t-1 +α9GOVi,t-1 + α10GDPi,t-1 + α11CAPEXi,t-1
+α12FORSAi,t-1 + νi + ct + εi,t
7. ToQi,t = α0 + α1CSRi,t-1 + α2(IDV*CSR)i,t-1 + α3RISKi,t-1+ α4R&Di,t-1 +α5ROAi,t-1
+α6SAGi,t-1+α7SIZEi,t-1 +α8GDPi,t-1 +α9GOVi,t-1 + α10GDPi,t-1 + α11CAPEXi,t-1
+α12FORSAi,t-1 + νi + ct + εi,t
8. ToQi,t = α0 + α1CSRi,t-1 + α2(LOST*CSR)i,t-1 + α3RISKi,t-1+ α4R&Di,t-1 +α5ROAi,t-1
+α6SAGi,t-1+α7SIZEi,t-1 +α8GDPi,t-1 +α9GOVi,t-1 + α10GDPi,t-1 + α11CAPEXi,t-1
+α12FORSAi,t-1 + νi + ct + εi,t
25 4. Results
In this section, we describe the total sample, which includes the descriptive statistics and the correlation model, followed by the results of the regressions of the proposed models.
4.1 Descriptive Statistics
The total sample with the controversial industries comprehends an unbalanced panel data set and covers the period of 2007–2014. After removing all the missing observations and matching all the variables used in this study, the total sample is 2729 firm-years over 41 countries. To reduce the effect of outliers, all variables are winsorized at the 2.5th and 97.5th
percentiles. The data is normally distributed (tested with Jarque-Bera test), and non-stationary (tested with the panel-data nit-root and tested where in the majority of the tests the a null hypothesis could be rejected, given the data was stationary).6 An overview of the total sample composition by industry and country is shown in Tables 3 and 4 of the Appendix. The total firm years are distributed over nine different controversial industries. The largest industry is the oil industry with 40.01 percent of the observations, followed by mining with 25.58 percent. The weapons and sugar industries are the smallest with 0.48 and 2.09 percent, respectively. The total sample represents 41 countries with the United States as the largest country with 18.60 percent and Egypt as the smallest country with 0.18 percent.
Table A presents the summary statistics of all the variables used in this study. The mean value of Tobin’s Q is 1.47, indicating that market value is higher than the book value of the firm. This means, on average, firms are not undervalued. This Tobin’s Q score is one point lower compared to the mean value of Tobin’s Q in the study of Cai et al. (2012). This indicates that the firms included in the sample of Cai et al. (2012) have, on average, higher market value scaled with the book value compared to this study. The ROA is, on average, positive, which indicates, on average, a positive accounting-based financial performance. In line with Cai et al. (2012), the standard deviation of capital expenditure and sales growth is higher than their mean, implying high variation within the variables. On average, firms have a CSR score of 50. They score higher on social engagement compared to environmental engagement. These findings are comparable with the sample of Ioannou and Serafeim, (2012). The cultural dimensions are described as their own rating and not as the interaction variable with CSR. The cultural dimension individualism has a mean of 70, indicating that more countries scoring high on individualism are included in the sample.
26 Table A: Descriptive statistics
This table reports the number of observations, mean, median, maximum, minimum and standard deviation for each variable. The sample consists of an unbalanced panel of 2729 firm-years from 550 companies for the period of 2007–2014. Toq is the firm’s Tobin’s Q for the annual period of 2007–2014. CSR represents the corporate responsibility score as an average of the social and environmental score. Sosc represents the social score. Envsc represents the environmental score. IDV is the degree of individualism of the origin country of the firm. Lost is the degree of long-term orientation of the origin country. Mas represent the degree of masculinity of the origin country. PD is the degree of power distance. UAI is the degree of uncertainty avoidance. Risk is measured as total debt over total assets. R&D is a proxy for research and development measured as intangible assets over total assets. ROA is the return on assets measured as EBIT divided by total assets. Sag represents the sales increases relative to the previous year. Size is the company’s size measured as the natural logarithm of total assets. GDP stands for GDP growth between the years measured in percentages. FORSA stands for foreign sales as a percentage of total sales. CAPEX is the degree of capital expenditures divided over total assets. All variables are winsorized at a 2.5 percent level.
Observations Mean Median Maximum Minimum Std. Dev.
Dependent variables 1.Toq 2729 1.46 1.14 5.86 0.42 1.03 Independent variables 2.CSR 2729 50.48 50.33 93.65 8.45 30.09 3.SOSC 2729 52.07 54.04 95.72 5.63 31.64 4.Envsc 2729 48.91 46.89 93.77 10.07 30.72 5.IDV 2729 71.87 80.00 91.00 18.00 23.18 6.LOST 2729 43.89 36.02 87.91 21.16 21.46 7.MAS 2729 58.77 61.00 95.00 26.00 13.68 8.PD 2729 46.90 40.00 93.00 31.00 15.57 9.UAI 2729 53.77 48.00 95.00 29.00 18.54 Control variables 10.RISK 2729 0.22 0.20 0.62 0.00 0.16 11.R&D 2729 0.13 0.05 0.65 0.00 0.17 12.ROA 2729 0.09 0.09 0.40 -0.60 0.13 13.SaG (%) 2729 22.07 10.52 429.39 -59.78 61.28 14.Size (LN) 2729 15.44 15.41 18.58 9.81 1.72 15.GDP (%) 2729 2.10 2.06 9.40 -3.16 2.52 16.Gov 2729 1.16 1.31 1.71 -0.54 0.61 17.FORSA (%) 2729 42.43 42.01 100.00 0.00 37.22 18.CAPEX (%) 2729 10.43 5.88 62.29 0.00 12.43 4.1.2 Correlation matrix
29
4.2 Regression outcomes
The regression results part consists of three sections. First, I describe the results of the regression model with respect to hypothesis 1. Second, I describe the results of the models regarding hypothesis 2. Third, I describe the various robustness checks to validate the models. 4.2.1 The effect of CSR on CFP in controversial industries (Hypothesis 1) Table C presents the results of the regression models 1–3 to test hypothesis 1 that CSR has a positive effect on CFP in controversial industries. The results show a positive, significant coefficient (P<0.01) for CSRindex, environmental score and social score after controlling for other factors influencing Tobin’s Q. These results corroborate the findings of Cai et al. (2012). It further shows the firm-specific control variable Size as a statistically significant (P<0.01) negative coefficient. This is in line with previous argued literature that large size lowers Tobin’s Q (Fama and French, 1992). The ROA coefficient is statistically (P<0.01) positive, supporting the theory that higher financial performance enhances Tobin’s Q (Varaiya et al., 1987). In model 2, the coefficient of Risk is statistically (P<0.1) negative, indicating that high leverage has a negative impact on Tobin’s Q. This supports the theory that high leverage creates an underinvestment problem with the high risk of default (Myers, 1997). Although models 1–3 show high R-squared values (0.749 on average), most included control variables show no significant relationship with Tobin’s Q. One of the reasons of high R-squared is the use of firm-fixed effects. Overall, the positive, significant coefficient of all three CSR proxies support hypothesis 1 that CSR has a positive effect on CFP in controversial industries.
Table C: Regression results of models 1–3
This table shows the results of an OLS regression with firm-fixed and year-fixed effects. Tobin’s Q is used as the dependent variable in all the models. Model 1 tests the effect of the proxy CSRindex on Tobin’s Q. Model 2 tests the effect of the proxy social score on Tobin’s Q and model 3 tests the proxy environmental score on Tobin’s Q. The rest of the variables expressed in abbreviations are outlined in Chapter 3 and in Appendix table 1. The robust standard errors of the variables are presented in the parentheses. Statistical significance is indicated with *p<0.01; **p<0.05 and ***p<0.01, respectively.
Variable Model 1 Model 2 Model 3
30 RISK(-1) -0.390 -0.391* -0.389 (0.238) (0.237) (0.239) R&D(-1) -0.342 -0.333 -0.354 (0.289) (0.288) (0.290) ROA(-1) 0.730*** 0.736*** 0.737*** (0.200) (0.199) (0.202) Sag(-1) 0.000 0.000 0.000 (0.000) (0.000) (0.000) Size(-1) -0.222*** -0.230*** -0.211*** (0.062) (0.062) (0.062) GDP(-1) 0.022** 0.023** 0.022** (0.011) (0.011) (0.011) Gov(-1) 0.238 0.233 0.248 (0.152) (0.151) (0.153) Capex(-1) 0.001 0.001 0.001 (0.002) (0.002) (0.002) Forsa(-1) 0.000 0.000 0.000 (0.001) (0.001) (0.001) Constant 4.343*** 4.487*** 4.278*** (0.946) (0.943) (0.948) Adj. R-squared 0.749 0.750 0.748 Observations 2729 2729 2729 4.2.2 The moderate effect of culture on the CSR–CFP relationship in controversial industries (Hypothesis 2)
31 rejected hypothesis 2b; masculinity has a negative influence on the positive effect CSR has on CFP in controversial industries. Also, this model supports hypothesis 1 with a positive, significant (P<0.01) CSR coefficient. Model 6 shows the results for testing hypothesis 2c. Hypothesis 2c argues that the cultural dimension uncertainty avoidance strengthens the positive effect of CSR on CFP. The results show a negative, significant (p<0.05) interaction term coefficient, indicating a positive, significant (p<0.01) CSR coefficient. This indicates that uncertainty avoidance negatively influences the positive effect of CSR on CFP in controversial industries. These findings are not supporting hypothesis 2c and are in contrast with the conclusion of Mar Miras-Rodriguez et al. (2015), that the CSR–CFP relationship becomes stronger in countries scoring high on uncertainty avoidance. Model 7 shows the results of testing hypothesis 2d regarding the influence individualism has on the CSR–CFP relation. The interaction coefficient is insignificant. Therefore, I rejected that individualism negatively influences the positive effect of CSR on CFP (hypothesis 2d). A similar conclusion can be drawn from the CSR coefficient when compared to models 4, 5 and 6. This conclusion can also be made for model 8, which presents the results of hypothesis 2e. This hypothesis states that cultural variable long-term orientation strengthens the positive effect CSR has on CFP. The insignificant interaction term leads to a rejection of hypothesis 2e. Overall, hypotheses 2a, 2b, 2c, 2d and 2e are rejected. Nevertheless, the results show a negative significant coefficient of the interaction term including uncertainty avoidance. Therefore, it can be concluded that high uncertainty avoidance cultures weakens the positive effect of CSR on CFP in controversial industries. The control variables Risk, ROA and Size show the same significant results as in models 1–3. In addition, models 6 and 7 show a positive, significant (p<0.1) coefficient for GOV. This supports El Ghoul’s (2017) theory, which argues that stronger institutions are associated with higher CFP.
Table D: Regression results of models 4–8: moderate effect
This table shows the results of an OLS regression with firm-year and year-fixed effects. Tobin’s Q is used as the dependent variable in all the models. Model 4 includes the PD interaction effect with CSR. Model 5 includes the masculinity interaction effect with CSR. Model 6 includes the uncertainty avoidance interaction effect with CSR. Model 7 includes individualism interaction effect with CSR and model 8 includes the variable long-term orientation with CSR as an interaction effect. The rest of the variables expressed in abbreviations are outlined in Chapter 3 and in Appendix table 1. The robust standard errors of the variables are presented in the parentheses. Statistical significance is indicated with *p<0.1; **p<0.05 and ***p<0.01, respectively.
Variable Model 4 Model 5 Model 6 Model 7 Model 8
33 4.3. Robustness checks
To verify the strength of the results and to detect any misspecification regarding the models used to test hypotheses 1 and 2, I performed robustness checks. These robustness checks include several model and sample adjustments.
4.3.1 Include industry control variables instead of firm-fixed effects within models 1-3 In the existing literature, there is no clear consensus regarding which methodology is best suited to test the CSR–CFP relationship. One of the differences between the studies is the decision of the fixed-effect model. In this study, firm-fixed effects are included to control for unobserved heterogeneity between cross-sections within models 1–3. However, Rodrigo et al. (2016) used industry-control variables to control for industry-specific unobserved shocks. This view is supported in other empirical studies conducted on the CSR–CFP relationship (Gregory et al., 2014; Waddock and Graves, 1997). To test if the industry-control variables have a different effect on the results and to make the results more comparable, I re-ran models 1–3, and included the industry-control variables instead of including the firm-fixed effects. The results are shown in table 6. of the Appendix. The results show no difference of the significance nor a change in the sign of the coefficient of CSR. Therefore, this will not change the conclusions made regarding hypothesis 1.
4.3.2 Sample adjustment regarding development status within models 1-3
34 Appendix, South Korea and Qatar are based on Worldscope database, where they are classified as emerging countries, while, for instance, in the IMF database, they are classified as developed countries.
4.3.3 Sample adjustment regarding included controversial industries
Another sample adjustment was made regarding industries, as time and culture can determine which industry is called controversial (Fam et al., 2004; Rodrigo et al., 2016; Waller et al., 2005. This was one of the reasons why this study included mining and sugar industries in addition of the original sample of controversial industries by Cai et al. (2012). To compare our results with Cai et al. (2012), we excluded the mining and sugar industries and re-ran the regression models 1–3. The results in table 9 of the Appendix shows that the CSR coefficient remains significant (P<0.01) and positive across all proxies for CSR. The outcomes of these results are comparable with Cai et al. (2012) and the results of our original model 1-3. CSR has a positive influence on CFP with or without the mining and sugar industries included in the sample.
4.3.4 Model adjustment regarding fixed-effect model within models 4–8
As mentioned in the methodology, there are several ways to test the effects of a time-invariant interaction term on the CSR–CFP relation. The models 4–8 interpret the coefficient of the interaction term in a fixed-effect model, as the underlying variable culture is omitted (Boyce & Wood, 2011; Wooldridge, 2009). However, Brambor et al. (2006) argued that all variables that constitute the interaction term should be included in the model as omitted variables will lead to significant constructive errors. In line with these arguments, I re-ran models 4–8 and included all variables that constitute the interaction term and consequently not control for firm-fixed effects. The outcomes of this approach are presented in table 10 of the Appendix.
35 CFP would be strengthened by the cultural variables uncertainty avoidance, individualism and long-term orientation and weakened by the cultural variable power distance. Therefore I should reject hypothesis 2B and 2D and accept hypothesis 2A, 2C and 2E. However, the R-square score is 0.274, which is substantially lower compared to the average R-R-square score of 0.749 with the models including firm-fixed effects. These differences in results compared with models 4–8 may suggest that excluding firm-fixed effects would lead to omitted-variables-biased results and is less reliable as argued by Khaled and Paton (2005).
4.3.5 Sample adjustment regarding size within models 4-8
This robustness was performed regarding the used sample related to test how culture influences the CSR–CFP relationship. This is because previous mentioned countries with a few firm-year observations in the sample may bias the results concerning cultural interaction terms. Therefore, countries with less than 10 firms’ observations were excluded from the sample. The results are presented in table 11 of the Appendix. The results show that the coefficient of interaction term, CSR*PD, is negatively significant (p<0.05). This result does support hypothesis 2a, which argues for a negative influence on the positive effect of CSR on CFP. Models 5–8 show all insignificant coefficients for the interaction terms comprehending CSR and the accompanied cultural variable. This contradicts the main results drawn in Chapter 4, where uncertainty avoidance weakens the positive influence of CSR on CFP. This suggests that countries with low firm observations bias the results of models 4–8.
4.3.6 The effects of CSR on CFP in the mining industry
36 natural resources and other negative externalities. On the contrary, Cai et al. (2012) investigated an Anglo-Saxon country (United States) and based their selection of controversial industries on previous literature (Hong and Kacperczyk, 2009; Jo and Na, 2012). To test if CSR in the mining industry has the same positive effect on CFP in the complete sample as for Latin-American countries or for Anglo-Saxon countries, I re-ran model 1 with, first, the effect of CSR in the mining industry on CFP for the complete sample (model 1a), and second, with only Latin-American countries (model 1b) included and third, only with a selection (model 1c) of Anglo-Saxon countries (United States, United Kingdom and Canada). For model 1b the subsample has too few observations to perform a random- or fixed-effect model. The results are presented in table 12 of the Appendix and shows for model 1a, and model 1b positive yet insignificant CSR coefficients for testing the relationship on CFP. Model 1c, including only a sample of Anglo-Saxon countries, show a significant (p<0.1) CSR coefficient. For mining firms operating in Anglo-Saxon countries, CSR is positively influencing CFP. This does not differ with the previous results regarding the complete sample of controversial industries. Moreover, no conclusion can be made if these results differ for Latin-American countries. The most important reasons is the small sample size, which reduces the power of the regression models.
To further investigate why controversial industries are seen differently between countries, I re-ran the models 4–8, including only the mining industry for the total sample as well as for the subsample with Anglo-Saxon countries and a subsample with Latin-American countries. As mentioned by Rodrigo et al. (2016) and Waller et al. (2005), cultural dimensions shape the attitudes towards controversial industries and may differ between countries. These regression outcomes gain more insight in whether culture has different influences on the mining industry for the total sample and for specific regions’.
The results are presented in table 13 in the Appendix. The results for subsamples including Anglo-Saxon countries and Latin-American countries show insufficient results7.
Therefore, no conclusion can be drawn regarding the influence of different regions on the outcomes of how cultural dimensions influence the relation of CSR in the mining industry on CFP. This is caused by the small sample size and less variation in the cultural variables. This leads to high multicollinearity between the centred interaction term and centred CSR. Although, for the total sample, model 4 shows a negative significant (P<0.1) coefficient for the interaction term, including power distance and a positive significant (P<0.1) CSR
37 coefficient. This supports hypothesis 2a, which predicted a negative influence on the positive effect of CSR on CFP in controversial industries. Furthermore, Model 7 shows a positive, significant (P<0.1) coefficient for the interaction term, including individualism. However, no conclusion can be drawn on whether it strengthens the positive relation between CSR on CFP, as the CSR coefficient in this model is insignificant. However, interaction terms, including individualism and power distance, have a significant effect on CFP. This supports the view that different cultural variables influence CFP on different controversial industries.
4.4 Summary of the robustness checks
38 5. Conclusion
As controversial industries are scrutinised by the media and stakeholders put more value on CSR, it might become more important for controversial firms to engage in CSR activities. The goal of this study was to provide an understanding how CSR influences CFP in controversial industries and to show how cultural factors influence this relationship.
By employing panel data from 2729 firm-years in 41 countries over a period of 2007– 2014, this study finds a significant positive effect of CSR activities on CFP in controversial industries. This result holds across all proxies for CSR and is in line with the study conducted by Cai et al. (2012). This finding supports the proposed hypothesis that CSR has a positive effect on CFP in controversial industries. This result indicates that investors place a higher CFP on CSR in controversial firms compared to controversial firms that do not undertake CSR activities. This supports the stakeholder theory and transaction theory (Freeman, 1984; Williamson, 1989) that higher levels of CSR satisfy the implicit claims of stakeholders and decrease transaction costs, which results in an increase in CFP.
Furthermore, this study finds that the cultural variable uncertainty avoidance has a significant negative impact on the positive effect of CSR on CFP. The CSR activities might signal a higher-risk taking behaviour in high uncertainty cultures, and therefore, the firm is valued lower compared to firms in the same industry not engaging in CSR activities. This does not confirm hypothesis 2c that uncertainty avoidance positively strengthens the positive effect of CSR on CFP. However, when adjusting the sample size and excluding countries with less than 10 firms or including only firms in the mining industry, the interaction term, including uncertainty avoidance, becomes insignificant and the interaction term, including power distance, becomes significant and negative. Moreover, more significant interaction variables are found when excluding time-fixed effects in models 4–8.
39 and investigate, with this broader set of firms, the differences between countries. Another distinction between countries can be made to look at differences between market economies or liberal markets (Ioannou and Serafeim, 2012). There is a difference in CSR levels between those markets, which could provide further understanding on how CSR influences CFP in controversial industries (Jackson and Apostolokou, 2010).
Next to the small sample sizes to draw effective conclusions for subsamples, another limitation of this empirical research that could flaw results was the difficulty to measure the exact value of CFP. To calculate the exact value of Tobin’s Q, the replacement cost needs to be forehanded. Due to data restrictions in this study, I followed Liou and Sharma’s (2012) method to obtain Tobin’s Q. This method differs from the measurement followed by Cai et al. (2012).
