MODERATING EFFECTS OF THE AGE
EFFECT ON ECSR PERFORMANCE
Yvette (E.T.) de Graaf
10740260 Prof. dr. V.R. O’Connell MSc Accountancy & Control, specialization Accountancy Faculty of Economics and Business, University of Amsterdam Word count: 12703
JUNE 24, 2018
Statement of Originality
This document is written by student Yvette (E.T.) de Graaf who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Abstract
This study examines the age effect on environmental corporate social responsibility (ECSR) found by Post, Rahman and Rubow (2011). Additionally, this study
investigates three variables that possibly moderate this effect. These moderating effects are director gender, director independence and firm financial performance, proxied by the firm’s return on assets. This study is relevant as the U.S. government has been indecisive towards environmental policies and firms can anticipate tighter environmental laws if the government rejoins the Paris Climate Accord. This study reconfirms the age effect on ECSR performance, proxied by the total number of environmental strengths. There is sufficient evidence to conclude director gender moderates the age effect on ECSR with a significance level of 5%. However, additional research is needed to draw conclusions about the hypothesized
moderating effects of director independence and firm financial performance. This study contributes to existing literature by remeasuring the age effect on ECSR performance found by Post et al. (2011) and finding director gender moderates this effect.
Keywords: Environmental corporate social responsibility (ECSR), board of directors,
Table of Contents
1 Introduction 6
2 Literature review 9
2.1 Board composition 9
2.2 Age effect on ECSR 10
2.3 Director gender 13
2.4 Independent directors 15
2.5 Firm performance 16
3 Research Design 20
3.1 Dependent and independent variables 20
3.2 Moderating variables 21 3.2.1 Director gender 21 3.2.2 Director independence 21 3.3 Control variables 22 3.4 Empirical model 23 3.5 Sample description 24 4 Results 26 4.1 Descriptive statistics 26 4.2 Correlation of variables 28 4.3 Regression Results 30 4.3.1 OLS regressions 30 4.3.2 Sensitivity analysis 33 4.3 Summary of findings 38
5 Discussion and conclusion 40
6 References 43
Appendix A: Pearson’s correlation matrix 47
List of figures and tables
Figure 1: Hypothesized relationships 19
Table 1: List of variables 23
Table 2: Sample collection 25
Table 3: Sample distribution per sector 26
Table 5: Average number of environmental strengths 28
Table 6: Spearman’s correlation matrix 29
Figure 2: Relationship between age and the number of environmental strengths 30
Table 7: OLS regression results 33
Table 8: Alternative OLS regression one 35
Table 9: Alternative OLS regression two 37
Table 10: Overview of the results of H2 38
Table 11: Overview of the results of H3 39
1 Introduction
After several environmental crises which have been linked by scientist to increasing levels of CO2 in the Earth’s atmosphere, like extreme drought in Australia and the
southern regions of the United States, record breaking heat waves and floods in Europe and the appearance of diseases transferred by mosquitos that previously were only known to be transferred to humans in the Southern Hemisphere, there is a growing demand for a more sustainable world (Senge, Smith, Kruschwitz, Laur, & Schley, 2008).
To turn our world into a more sustainable one, the UN introduced the 2030 agenda for sustainable development in 2015. This agenda includes seventeen sustainable development goals (United Nations, 2016). In addition to these goals, on November 4th, 2016, the Paris Agreement entered into force. The Paris Agreement is
ratified by 170 parties and its aim is to fight climate change by keeping the rise in global temperature for this century below two degrees Celsius, but aims to keep this rise under 1.5 degrees with additional efforts and also wants every party involved to be CO2 neutral by 2020, which means a country produces as much sustainable
energy as it uses (United Nations Framework Convention on Climate Change, 2014). The United States have formally joined the Paris Climate deal in 2015 under Barack Obama’s presidency, which was seen as a major event for the Paris Accord, as the United States are the second largest emitters of carbon dioxide in the world (BBC, 2016). However, on June 1st, 2017, the new U.S. president, Donald Trump, a
known climate sceptic, announced that the United States would withdraw from the Paris Accord, claiming the accord would economically disadvantage the U.S. (Cama & Henry, 2016). A study conducted by the Yale Program on Climate Change
Communication states that the majority of the U.S. citizens say that the United States should participate in the Paris Climate Accord. The study shows that only three percent of democrats are opposed to the Paris Agreement, whereas only the conservative republicans are split about the participation of the U.S., as 40% is in favor of the accord and 34% is against (Marlon, Fine, & Leiserowitz, 2017).
Therefore, it is likely a possible democratic successor will reinforce planned climate policies. U.S. based companies should take this into account and anticipate on possible changes in climate laws by planning how to do business environmental friendly and act on it, to prevent any surprises. One way to do so is by hiring an
ecological problems, academics and corporate managers need to work together. Organizations need to incorporate environmental policies in their normal course of business as organizations have the resources and power to make an impact on the ecological problems the earth faces.
A survey executed by the Pew Research Center (2011), states that the so called ‘Millennials’ also known as ‘Generation Y’, which consists of those born between the early 1980s and the mid 1990s, are the most environmentally aware generation and show greater support for environmental laws, tax regulation and are willing to pay more for sustainable products.
Post et al. (2011) researched the effect of age on a company’s environmental corporate social responsibility (ECSR). They hypothesized age had a curvilinear, quadratic effect on a company’s environmental practices, suggesting relatively young and old directors are more environmentally oriented. Instead of finding the
hypothesized curvilinear relationship, the authors found a curvilinear, quadratic relationship between director age and ECSR which suggests that relatively young and old directors are less environmentally oriented, with a peak in ECSR governance when directors are, on average, 56 years old.
This thesis continues to build on this insight and tries to find empirical
evidence for variables that could modify the age effect on ECSR performance. There are three variables that will be checked for a modifying effect. These variables are gender, director independence and firm performance. These variables have been chosen after reviewing relevant literature on general corporate social responsibility (CSR). CSR consists of four components, economic, legal, ethical and philanthropic responsibilities. A firm’s environmental responsibilities are a part of the philanthropic component, which is described as the responsibility to be a good corporate citizen (Carrol, 1991).
Boulouta (2013) and Zhang (2012), among others, empirically found that female directors are positively related to CSR performances for respectively S&P 500 and Fortune 500 firms. Director independence has also been empirically linked to CSR performance, for both positive (Krüger, 2009) and negative (Zhang, 2012) relationships between the two. Financial performance, in this thesis measured by a company’s return on assets (ROA), has been empirically found to be positively related to CSR performance.
The sample of this research consists of 56,095 director observations from 2007 to 2013. The results show a significant quadratic relationship between director age and the total number of environmental strengths, with a peak when directors are, on average, 60 years of age. Additionally, there is sufficient evidence to conclude director gender has a moderating effect on the age effect. Because the models used to measure the moderating effect of director independence and firm financial
performance gave mixed results, additional research is required to be able to draw any conclusions on the existence of the moderating effects of the two variables.
This study contributes to existing literature by reconfirming the age effect found by Post et al. (2011) and finding director gender moderates this effect. Firms can use these findings to appoint board members who are more likely to implement environmental friendly policies.
This study is structured as follows. The next section summarizes prior literature and explains the underlying theory. In section 3 the data and research methodology are discussed, followed by the results in section 4. Section 5 consists of the discussion and conclusion. This thesis ends with a list of the used references and an appendix.
2 Literature review
The second section of this thesis summarizes previous research into the field of ECSR and the effect of the chosen moderating variables on the general concept of CSR. The literature review starts with a summary of how different board
characteristics can influence a firm’s policies and practices in section 2.1. Section 2.2 covers the age effect on ECSR practices, followed by prior literature on the effect of director gender in section 2.3. The fourth subsection covers the effect of director independence. This literature review will conclude with a subsection on the effect of firm performance in the fifth subsection.
2.1 Board composition
Hillman and Daziel (2003) state that the board of directors serve two main functions in an organization. The first one is to monitor management on behalf of the firm’s shareholders. Their second main function consists of providing resources for the organization. They concluded this after examining both the agency and the resource dependence theory. The agency theory states that because of the separation of ownership and control of an organization, conflicts of interest arise (Hillman & Dalziel, 2003). Because of the conflicts of interest, the board must focus on monitoring
management to protect the interest of the shareholders (Jensen & Meckling, 1976). Fama (1980) states that monitoring management directly links to a firm’s overall performance, as it reduces the costs that are incurred when managers pursue their own interests, rather than the firm’s.
The resource dependence theory states that when someone is chosen to be on an organization’s board, that person is expected to be concerned with the
company’s performance and will try to do his or her best to aid it. There are four ways in which a director can benefit the organization: advice and counsel, legitimacy, channels for communicating information between external organizations and the firm, and preferential access to commitments or support from important elements outside the firm (Hillman & Dalziel, 2003).
A board’s resources are based on the overall experience and expertise of all the board members, including knowledge of company strategy and operations, but also specialist knowledge about legal and regulatory affairs. A board can also benefit from having diverse members on its board of directors. Diversity in expertise
could also amplify network ties (Bear, Rahman, & Post, 2010). Diversity on the board not only increases a firm’s resource providing function, it also has a positive effect on its monitoring function, because it needs the right set of skills, experience, expertise and knowledge to provide advice and counsel and help assess the skill of top
management (Hillman & Dalziel, 2003). Additionally, Hemingway and Maclagan (2004) study motives used for implementing CSR policies and state that managers often follow their own personal morals instead of participating in CSR practices because it would benefit the reputation of the firm. This suggests that firms and shareholders can appoint directors with specific characteristics in order to influence the policies of the firm.
2.2 Age effect on ECSR
ECSR is a component of CSR, which has been defined in multiple ways in different studies. Dahlsrud (2008) analyzed 37 definitions of the CSR concept and came to the conclusion that the existing definitions are mostly congruent. The author established five dimensions and checked if the definitions covered the five dimensions. The stakeholder, social, economic and voluntariness dimensions got a dimension rate of 80 to 88 percent while the environmental dimension got a score of 59 percent, meaning the environmental dimension is less often included in the definition of CSR. The author has given two possible explanations for the lower dimension rate. The first one states that in early definitions, the environmental dimension was left out, which caused authors of other studies to exclude the environmental dimension from the CSR definition as well. The second explanation is that although not mentioned separately, the environmental dimension is considered to be included in the CSR concept. Carrol (1991) has included four components of CSR in his research into why organizations create CSR policies, which are economic, legal, ethical and philanthropic responsibilities. The author developed a pyramid to explain how firms first need to fulfil their economic, legal, and ethical responsibilities, before they can commit to their philanthropic responsibilities, which involve being a good corporate citizen and improve the quality of life, which involves acting environmentally aware. From these two studies into the structure of CSR practices, it can be concluded that ECSR is incorporated in the voluntariness (Dahlsrud, 2008) and philanthropic (Carrol, 1991) dimensions of CSR.
Apart from cost related issues that constrain organizations from participating in CSR, firms also have several tensions in sustainability practices, which are: differences in personal and organizational CSR interests, differences in short and long-term vision, similarity between organizations versus technological change, and efficiency versus the flexibility of socioeconomic systems, which can either be accepted or resolved by dividing or combining the two contrasting circumstances (Hahn, Pinkse, Preuss, & Figge, 2015).
Martin (2002) constructed a virtue matrix to show why organizations engage in CSR related activities. The author states that the reasons behind participating in CSR lie in norms and customs, for firms that actively choose themselves, and rules and regulations, for firms that are required by law to participate in CSR activities. The main difference between these two, are the beneficiaries of the organization’s practices. When firms choose to participate in CSR themselves, it benefits both shareholders and society, but when firms are required by laws or regulations it solely benefits society, which creates a barrier to innovate CSR practices.
Purdy, Alexander and Neill (2010) constructed a basis for the implied social contract between U.S. based firms and the U.S. society. In this basis the authors address 3 institutional pillars that function as the support system for the contract. The first institutional pillar is the regulatory one. The regulatory pillar concerns the general laws and social constructs in the U.S., which main focus points are the minimal labor market policies, high capitalism and individuality, big separation between firms and the government and a shareholder model that enhances short term profit seeking. The second pillar is the normative one, which concerns the duty of firms to act as a good corporate citizen. The normative pillar is of bigger importance for U.S. based firms than it is for European firms, as Europe based firms generally pay a higher tax rate than U.S. firms, and therefore are already considered good citizens by society. The third pillar of the contract is the cognitive one, which reflects the extent to which managers feel responsible for implementing CSR policies. This pillar is generally smaller for U.S. based firms compared to European firms, as managers from U.S. based firms tend to separate the social and economic role of an organization and implement specific CSR policies, depending on the industry they are in. On the other hand, European top managers are more focused on long-term sustainability and are often influenced by values of lower managers and employees. This suggests that CSR policies in the U.S. are mainly driven by pressure of society. Bulkeley and Mol
(2003) confirm this line of thought by stating that participation of stakeholders is highly important in facing environmental issues.
Garriga and Melé (2004) studied the reasons why corporations actively
choose to participate in CSR. They find that in instrumental theories, CSR is seen as means to achieve economic goals. This can be done in three ways. Firstly, to
maximize shareholder value and thus share price. Secondly, by using CSR to gain a competitive advantage. And thirdly, to use CSR as a marketing strategy. Apart from the instrumental theory, the authors explain sets of political, ethical, integrative and value theories as well, which state that firms engage in CSR practices because they have moral, political or societal incentives to do so. Additionally, Shrivastava (1995) listed several ways in which organizations benefit from implementing environmental friendly policies. Firstly, policies that reduce energy use and materials can also reduce cost. Furthermore, organizations can create a competitive advantage by drawing ‘green’ customers towards them and create a positive reputation by promoting their environmental friendly policies. Also, environmental sustainability could reduce long-term risks and could help firms get ahead of environmental rules and regulations that could be costly to implement on a short-term basis.
Previous research in the field of having particular young or particular old directors on a firm’s board has found that directors often pursue personal interests and moral beliefs in the case of CSR participation, instead of participating in the interest of the firm’s reputation (Hemingway & Maclagan, 2004). Hafsi and Turgut (2013) found that, in regard to CSR practices, older managers have a greater focus towards societal issues, while younger managers lean towards focusing on
environmental issues first. This is in line with sociologic research in the field of
differences in opinion on environmental issues in regard to different generations. The Pew Research Center (2011) conducted a large survey in the United States and found that the youngest surveyed generation, the Millennials (early 1980s to mid 1990s), are most positive about regulations that enhance alternative, natural, energy sources, and most negative about regulations that expand the exploration of fossil fuels. For Generation X (early 1960s to early 1980s), the Baby boomers (early 1940s to early 1960s) and the Silent Generation (mid 1920s to early 1940s) these findings decrease with each generation, with the Silent Generation as the least
Wiersema and Bantel (1992) examined the relation between the demography of top management teams and corporate strategic change and found that management teams with a lower average age are more willing to pursue a new corporate strategy, whereas teams with an older average age showed less flexibility to change. On the contrary, younger managers tend to make less ethical decisions and are easier influenced by external factors than older managers (Peterson, Rhoads, & Vaught, 2001).
The study of Post et al. (2011), which found the age effect on ECSR,
hypothesized to find a curvilinear relationship, with both particularly young and old managers engaging in more ECSR practices, because of respectively, more concern about the environment and more developed moral reasoning. However, instead of finding a curvilinear relationship with older and younger directors participating in more CSR practices, they found a curvilinear relationship where younger and older directors participate in less CSR practices, with a peak when directors are on average, 56 years of age. These findings contradict prior research into age differences in knowledge of environmental issues and behavior towards them. However the findings relate to past events, like the first celebration of Earth Day in 1970, when directors of the optimal age of 56 would have been in college (Post et al., 2011). The differences between the results of academic research and practice
suggest that generational differences are of greater impact than age. As this thesis continues to build on this relationship and tries to find variables that moderate this effect the first hypothesis is as follows:
Hypothesis 1: There is a curvilinear relationship between director age and ECSR, such that firms with younger and older boards participate less in ECSR practices
2.3 Director gender
The first hypothesized moderating variable this thesis will look into is director gender. The influence of female directors on CSR practices has been researched extensively. One of these studies is conducted by Bernardi and Threadgill (2011), who found that the number of female directors on a board is positively associated with CSR
The relationship between female board directors and CSR practices is underpinned by the signaling theory, which assumes that there is asymmetric information which can be reduced by giving signals to the other, less informed, party. In this case firms can signal being socially responsible by having a number of female directors on its board and showing it pays attention to women and ethnic minorities. This is reflected in the differences in presentation of the board in annual reports, where firms with a higher percentage of female directors are more likely to include pictures of their board members (Bernardi, Bean, & Weippert, 2002)
There are multiple ways in which female directors differ from Caucasian male directors. Hillman, Cannella Jr. and Harris (2001) described the main differences between female and Caucasian male board directors as female directors having higher educational degrees, more often coming from non-business backgrounds, and joining multiple boards faster than Caucasian males do. Bear et al. (2010) described the two biggest strengths of having female directors on the board as an increased sensitivity towards CSR and having participative decision-making styles. However, these strengths may not show when there is only one woman on the board, which makes it difficult to voice their opinion.
Galbreath (2011) confirms these findings, stating that women are better in engaging with multiple stakeholders because of their relational abilities which benefits a firm’s environmental sustainability, but they often have to deal with practical issues like sex-based biases and stereotyping. Boulouta (2013) found that more gender diverse boards participate in more CSR practices and explained this by the dimensions of CSR having the possibility to appeal to emphatic caring, which female directors tend to be more susceptible to.
The theory suggesting female directors have a greater sensitivity towards CSR and the empirical evidence that boards with multiple female directors participate in more CSR practices leads director gender being one of the hypothesized moderating variables of the age effect on ECSR. As there is little known about the age
differences between males and females, the direction of the moderating effect cannot be predicted. This leads to the second hypothesis being formulated as follows:
2.4 Independent directors
The second hypothesized moderating variable is the presence of independent directors on the company’s board. Independent directors are members of the company’s board that are not related to the company, as opposed to executive directors, who are employed by the company on which they have their board seat (Arens, Elder & Beasley, 2014). On most company boards there are both
independent and executive directors. The executive directors are more informed about the company’s activities, whereas independent directors can add specific expertise and a certain degree of objectivity to the board’s characteristics and therefore better exercise their fiduciary responsibilities. Together, the board can exercise its expertise, independence and legal power to monitor the decisions made by managers (Byrd & Hickman, 1992).
Zhang (2012) uses agency theory to argue that executive directors are too closely related to top management and therefore act in the interest of top
management instead of the interests of all stakeholders. This leads executive
directors to focus solely on profit maximization and have little interest in setting CSR policies for the organization they work for. Independent directors, on the other hand, are not as closely related to top management and are therefore more likely to act in the interest of non-shareholder stakeholders. However, the results showed only a negative relationship between CSR weaknesses and independent directors and no positive relationship with CSR strengths.
In line with the theory used by Zhang (2012), Jo and Harjoto (2011) found that socially responsible firms have a significantly higher percentage of independent directors on their boards. More positive relationships between CSR performance and independent directors have been found by Webb (2004) and Ibrahim and Angelidis (1995), among others.
However, Krüger (2009) argued that companies with a higher percentage of executive directors also have a positive relationship with CSR practices. Executive directors could use CSR practices as an entrenchment device to enhance
relationships with non-shareholder stakeholders. However, the results showed companies with a higher percentage of executive directors only have less negative CSR effects instead of more positive ones and therefore provide more evidence for the risk management hypothesis, which argues that executive directors are more
aware of the company’s operational risks and therefore reduce the occurrence of negative CSR effects.
Kassinis and Vafeas (2002) studied the relationship between independent directors and the likelihood of being involved in environmental litigation and found that the number of independent directors of a firm’s board decrease the chance of being taken to court over an environmental issue. Walls, Berrone and Phan (2012) contributed to these findings by finding a significant negative relationship between environmental concerns and the percentage of independent directors on the entity’s board. However, the percentage of independent directors did not have a significant relationship with the entity’s environmental strengths, suggesting a high percentage of independent directors could reduce a firm’s negative environmental reputation, but does not impact a positive one.
The combined results of the studies into the relationship between director independence and CSR practices give an unclear view of this relationship and do not give a clear explanation as to why these differences in the results exist. However, since the positive and negative relationships largely exceed the non-significant ones director independence is the second hypothesized moderating variable to the age effect. Like director gender, the direction of the moderating effect cannot reliably be predicted, which leads to the third hypothesis being formulated as follows:
Hypothesis 3: Independent directors moderate the age effect on ECSR 2.5 Firm performance
The third and final moderating variable of this study is firm performance. In the past decades, the goal of organizations, as seen by society, has shifted from the sole focus on profit maximization to a shared focus which also includes the organization’s responsibility in society to help improve non-economic welfare (Tsoutsoura, 2004). This shift in focus has been picked up by organizations by preparing sustainability reports, and more recently, integrated reports, which focuses on integrated thinking and connectivity of different types of information. Integrated reporting also gives organizations the opportunity to clearly communicate their CSR related activities in a concise manner, which improves availability and accessibility of information for stakeholders (International Integrated Reporting Council, 2013).
According to the CSR pyramid designed by Carrol (1991), a company must first fulfil their responsibilities towards the shareholders, which is to be a profitable
organization, before it can participate in any ECSR related activities. This is
emphasized by Tsoutsoura (2004), who points out that participating in CSR related activities involves several costs and organizations should not participate in said activities if they result in negative cash flows, as it is not in the best interest of shareholders, who, at least, expect the highest risk adjusted return. However the author also states that CSR practices in the corporate governance field reduce chances of bribery and corruption. The costs of implementing these CSR policies will most likely differ per industry and will be significant for the organization. However, if firms get ahead of environmental regulation and make these investments over several years, the costs are incurred over a longer time and will have a smaller impact on the firm.
The relationship between financial performance and CSR practices have been researched by multiple authors and have mixed results. Freedman and Jaggi (1982) found a negative relationship between firm performance and pollution performance for the top quartile of firms ranked on size, which could be caused by pollution performance being used to rationalize poor economic performance. In contrast, Stanwick and Stanwick (1998) found that large size organizations’ economic and environmental (pollution) performance are positively related to the organization’s CSR score, which means that firms with high economic performance and low pollution emissions score higher on their CSR related activities. In addition to this, they also found a positive relationship between sales and CSR score for large firms. These results support Carrol’s (1991) view that being profitable increases CSR opportunities for the organization.
Flammer (2015) found a positive relationship between CSR and financial performance for both accounting-based and market-based measures in which CSR performance is the independent variable. Also, after announcing a firm would
implement more CSR practices, productivity and sales increased significantly. In the study of Van der Velde, Vermeir and Corten (2005) CSR has been found to be positively correlated with market-based measures. Results show that investment portfolios with a high sustainability rating outperformed investment portfolios with a low sustainability rating. These results suggest that investors are willing to pay a premium for firms that have CSR policies in place and that firms can expect to
recover some of the costs made in implementing the CSR policies. The studies of, for example, Carrol (1991), Freedman and Jaggi (1982) and Stanwick and Stanwick (1998) examine the relationship between financial performance and CSR and chose financial performance as the independent variable, while Flammer (2015) used financial performance as the dependent variable, which suggests the relationship between financial performance and CSR performance is a two-way relationship.
Orlitzky, Schmidt and Rynes (2003) conducted a meta-analysis of 52 prior studies to construct a clear picture of the relationship between financial performance and CSR performance. Overall, they found a positive relationship between financial performance and CSR performance across different industries from prior studies. They found that CSR performance has a higher correlation with accounting-based performance than market-based measures. In addition to this, the meta-analysis rejects the neo-classical idea that firms should only focus on profit maximization for shareholders and that the effectiveness and efficiency of an organization is improved by adequate CSR practices. Given the results that firm performance has a significant influence on CSR policies, financial performance is hypothesized to have a
moderating effect on the age effect. Again, the direction of the moderating effect cannot be reliably predicted, which leads to the fourth hypothesis of this thesis being formulated as follows:
Hypothesis 4: Financial performance moderates the age effect on ECSR Figure 1 gives an overview of the four hypothesized relationships. The first hypothesized relationship is that of the quadratic relationship of age on a firm’s ECSR performance. The second, third and fourth relationship are of the three hypothesized moderating variables, director gender, director independence and financial performance. These moderating effects will be tested individually, to prevent the moderating variables from influencing each other.
Figure 1: Hypothesized relationships
H4 H3
H2
H1
Age performance ECSR
Director Gender
Director
3 Research Design
3.1 Dependent and independent variables
In the last two decades, the field of CSR has been researched extensively and numerous studies, including the study conducted in the paper by Post et al. (2011), used the KLD database for research into CSR performance (Rao & Tilt, 2016). The KLD database collects data on all components of CSR on a firm-level basis.
Nowadays, the KLD database has been merged with the GMI database into the MSCI database.
For the dependent variable, which concerns a firm’s environmental practices, the total number of environmental strengths from the MSCI database is chosen
(ENVPOS), as seen in table 1. MSCI has established a threshold for each
environmental strength and awards points to firms if they meet these thresholds. The environmental strengths include (MSCI, 2015):
• Clean Tech – An assessment of how companies take advantage of environmental friendly technologies.
• Waste Management – An indicator in which companies with small toxic
emissions and that proactively reduce the impact of packaging material score higher.
• Climate Change – A variable which looks at how companies invest in low-carbon technology and the low-carbon efficiency of facilities and products. It also assesses the energy efficiency of the firm and its awareness of climate change risk.
• Natural Resource Use – A variable which looks at water stress, raw material sourcing, financing environmental impact, green buildings, renewable energy and electronic waste.
• Other Strengths – An indicator that looks at a company’s additional environmental efforts, that are not covered by the above variables.
The independent variable in this research concerns director age (DAGE). The data for this variable has been collected through the Institutional Shareholder Services (ISS) database. The ISS director database includes various variables of individual directors of listed firms in the United States (University of Amsterdam, 2017).
3.2 Moderating variables
This thesis focuses on various moderating variables of the director age effect on ECSR performance, empirically found by Post et al. (2011). The three main moderating variables in this research are: director gender (DGEN), director independence (DIND), and firm performance (FIRMP).
3.2.1 Director gender
The influence of director gender on a firm’s CSR practices has been widely
researched. Rao and Tilt (2016) have summarized research on this topic and show that most quantitative research find either a positive or an insignificant relationship between the two variables.
The variable needed to research the moderating effect of gender on the age effect will be collected from the ISS database, which is made available by the University of Amsterdam through WRDS. For this moderating effect, a dummy variable will be created. A score of 0 will be given to male directors and a score of 1 will be given to female directors.
3.2.2 Director independence
Director independence is another board variable that has been the subject of many studies in the field of CSR. This moderating variable clarifies whether an employee is either employed by the organization, or holds over fifty percent of voting power, or works independently and has no connection to the organization other than a place on its board (WRDS, 2018). For this moderating effect, a dummy variable will be
created. A score of 0 will be given to directors that are employed by the company and a score of 1 will be given to independent directors.
3.2.3 Firm performance
The third and final moderating variable of this thesis focuses on firm financial performance. Like the other two moderating variables, firm performance has often been connected to a firm’s CSR performance. Orlitzky et al. (2003) have reviewed 52 studies on the effect of corporate financial performance on corporate social
performance and concluded that corporate financial reporting is positively correlated with corporate social performance. Across the 52 studies the authors have reviewed both accounting-based and market-based measures for firm performance have been
used. The most common accounting-based measures include return on equity, return on assets and earnings per share. In line with the works of McGuire, Sundgren, and Schneeweis (1988) and Turban and Greening (1997), this study will use the return on assets as a proxy for firm performance. The return on assets has been chosen over the return on equity to avoid the variable being influenced by a firm’s capital
structure. For this proxy, the firm’s income before extraordinary items and total
assets have been collected through the Compustat database in WRDS and these are used in Stata to calculate the firm’s return on assets, as seen in table 1 (Standard & Poor, 2011).
3.3 Control variables
To avoid results that are influenced by variables not included in this study several control variables have been added. The first control variable concerns a director’s ethnicity (DETH). Zhang (2012) researched the link between board demographic diversity and corporate social performance and found a positive relation between representation of racial minorities on the board and corporate social performance strengths. However, when Hasfi and Turgut (2013) researched the same topic, the relationship was found to be insignificant. This control variable will be a dummy variable. A score of 1 will be given to directors with any ethnicity other than
Caucasian and a score of 0 will be given to directors of Caucasian descent. This data has been collected through the ISS database in WRDS.
The second control variable controls for directors who have attended less than 75% of board meetings (DATT). This variable will also be a dummy variable, where directors who have attended less than 75% of board meetings will be given a score of 1 and directors who passed the 75% benchmark a score of 0. This data has been collected through the ISS database.
The third control variable concerns a firm’s capital structure (FIRMC). The proxy for the capital structure will be the debt to equity ratio. The data on total
liabilities and total equity has been collected through Compustat in WRDS and will be used in Stata to calculate the firm’s debt to equity ratio, as seen in table 1 (Standard & Poor, 2011).
The last two control variables are firm size (LNSIZE), with the natural
calculated by dividing the market value of equity by the book value of equity, as seen in table 1. The market value of equity is available through Compustat and the book value of equity has been calculated by subtracting total liabilities from total assets (Standard & Poor, 2011).
Table 1: List of variables DAGE Director age measured in years
DATT Dummy variable, if director attended less than 75% of board meetings ‘one’
DETH Dummy variable, if director is of other than Caucasian descent ‘one’ DGEN Dummy variable, if director is female ‘one’
DIND Dummy variable, if director is classified as independent ‘one’ ENVPOS Total number of environmental strengths, positive score for a firm’s
environmental policies FIRMC Total liabilities / Total equity
FIRMP Income before extraordinary items / Total assets LNSIZE The natural logarithm of total assets
MTOB Market value of equity / Book value of equity
3.4 Empirical model
This thesis will use different regressions to examine the hypotheses. Hypothesis 1 measures the age effect on ECSR performance and hypotheses 2 to 4 measure the hypothesized moderating effects in isolation from each other, to prevent the
interaction terms influencing each other.
For hypothesis 1, a linear regression will be used to measure the director age effect on a firm’s CSR performance. In this model a quadratic term has been added to measure slope and steepness of the curvilinear effect. The empirical model used will be the following:
(1) 𝐸𝑁𝑉𝑃𝑂𝑆 = 𝛽-+ 𝛽/𝐷𝐴𝐺𝐸 + 𝛽3𝐷𝐴𝐺𝐸3+ 𝛽
4𝐷𝐸𝑇𝐻 + 𝛽7𝐷𝐴𝑇𝑇 + 𝛽8𝐹𝐼𝑅𝑀𝐶 + 𝛽>𝐿𝑁𝑆𝐼𝑍𝐸 + 𝛽A𝑀𝑇𝑂𝐵 + 𝜀/
To examine director gender as a possible moderating variable of the age effect on CSR practices, an interaction variable has been included in this model to measure
the hypothesized moderating effect. The following empirical model has been established:
(2) 𝐸𝑁𝑉𝑃𝑂𝑆 = 𝛽-+ 𝛽/𝐷𝐴𝐺𝐸 + 𝛽3𝐷𝐴𝐺𝐸3 + 𝛽
4𝐷𝐺𝐸𝑁 + 𝛽7𝐷𝐴𝐺𝐸 ∗ 𝐷𝐺𝐸𝑁 + 𝛽8𝐷𝐸𝑇𝐻 + 𝛽>𝐷𝐴𝑇𝑇 + 𝛽A𝐹𝐼𝑅𝑀𝐶 + 𝛽F𝐿𝑁𝑆𝐼𝑍𝐸 + 𝛽G𝑀𝑇𝑂𝐵 + 𝜀/
To examine director independence as a possible moderating variable of the age effect on CSR practices, an interaction variable has been included in this model to measure the hypothesized moderating effect. The following empirical model has been established:
(3) 𝐸𝑁𝑉𝑃𝑂𝑆 = 𝛽-+ 𝛽/𝐷𝐴𝐺𝐸 + 𝛽3𝐷𝐴𝐺𝐸3+ 𝛽
4𝐷𝐼𝑁𝐷 + 𝛽7𝐷𝐴𝐺𝐸 ∗ 𝐷𝐼𝑁𝐷 + 𝛽8𝐷𝐸𝑇𝐻 + 𝛽>𝐷𝐴𝑇𝑇 + 𝛽A𝐹𝐼𝑅𝑀𝐶 + 𝛽F𝐿𝑁𝑆𝐼𝑍𝐸 + 𝛽G𝑀𝑇𝑂𝐵 + 𝜀/
To examine firm performance as a possible moderating variable of the age effect on CSR practices, an interaction variable has been included in this model to measure the hypothesized moderating effect. The following empirical model has been established:
(4) 𝐸𝑁𝑉𝑃𝑂𝑆 = 𝛽-+ 𝛽/𝐷𝐴𝐺𝐸 + 𝛽3𝐷𝐴𝐺𝐸3+ 𝛽
4𝐹𝐼𝑅𝑀𝑃 + 𝛽7𝐷𝐴𝐺𝐸 ∗ 𝐹𝐼𝑅𝑀𝑃 + 𝛽8𝐷𝐸𝑇𝐻 + 𝛽>𝐷𝐴𝑇𝑇 + 𝛽A𝐹𝐼𝑅𝑀𝐶 + 𝛽F𝐿𝑁𝑆𝐼𝑍𝐸 + 𝛽G𝑀𝑇𝑂𝐵 + 𝜀/
3.5 Sample description
The sample used for this research consists of listed firms based in the United States between 2007 to 2013. The range was chosen based on the availability of the data. The available data from the ISS database, which provides data on director
characteristics, starts from 2007 and the number of environmental strengths from the MSCI database, which provides data on CSR practices, has been updated until 2013. To collect all the needed data, one additional database has been used, the COMPUSTAT database, which provides data on a firm’s financials.
The initial sample consisted of 96,400 observations from the ISS database. After merging the sample with the other two databases using ticker symbols as company identifier 82,318 observations remained. After merging the databases missing and duplicate data has been removed, which totalled 26,217 observations that were excluded from the sample. Finally, inaccurate data has been excluded from the sample. This was done by removing all directors with an age below 10. This
resulted in one observation that had to be excluded with an age of 1. Also, five observations with a ROA value of -121.0 have been deleted, as this number differed from the other data points by over 100, which raises doubts about the trustworthiness of the observation in its entirety. After all exclusions 56,095 observations remained in the final sample. To prevent the results from being skewed due to outliers in the data, the 1st and 99th percentile of FIRMP, FIRMC and MTOB have been altered with
winsorization. Due to this there 1,112, 1,105 and 1,115 changes have been made to respectively, FIRMP, FIRMC AND MTOB.
Table 2: Sample collection
Selection Criteria Observations
ISS: definitive proxies filed between 2007 and 2013
96,440
Less: No match with ISS and MSCI (1,996) Less: No match with ISS and
COMPUSTAT
(12,126)
82,318 Less: Duplicates and unavailable data (26,217)
Less: Inaccurate data (6)
4 Results
This section will cover the results of the empirical analysis. Section 4.1 covers the descriptive statistics, followed by the correlation of variables in subsection 2. The regressions and sensitivity tests are covered in section 4.3, followed by a summary of the results in the fifth and last subsection.
4.1 Descriptive statistics
Table 3 provides an overview of the distribution of the sample across different sectors, as defined by the Global Industry Classification Standard (2016). The sample consists of firms across all eleven sectors, with the most represented sector being the industrials sector with 16.50%, closely followed by the information
technology sector with 15.24%. The two least represented sectors are the
telecommunications services and real estate sectors with respectively, 1.31% and 2.82% of the sample.
Table 3: Sample distribution per sector
Table 4 shows the descriptive statistics for all used variables. ENVPOS has a
minimum value of 0 and a maximum value of 6. The mean of 0.67400 shows that the number of environmental strengths is relatively low across the sample. Only 0.72% of observations have a value equal to or higher than 5, while 64.79% of the sample received a score of 0, which suggests firms have room for improvements in their environmental practices.
GICS Sector
Key Sector Name Observations Percentage
10 Energy 3538 6.31% 15 Materials 4331 7.72% 20 Industrials 9253 16.50% 25 Consumer Discretionary 8385 14.95% 30 Consumer Staples 3539 6.31% 35 Health Care 5874 10.47% 40 Financials 6596 11.76% 45 Information Technology 8549 15.24% 50 Telecommunication Services 736 1.31% 55 Utilities 3714 6.62% 60 Real Estate 1580 2.82%
The age of directors in the sample ranges from 27 to 100, with the average age being 62.63 and the range between the 25th and 75th percentile lying between 57 and 68,
which suggests most directors are from the baby boomer generation (Pew Research Center, 2011). The summary of the moderating variables shows that 13.8% of
directors are female and 79.5% are independent and have no other affiliation to the company other than their seat on the board. The average ROA of the sample is 0.048 with the 25th percentile starting at 0.015, which means at least 75% of the
directors in the population works at a profit generating firm. The descriptive statistics for the control variables show that 10.2% of directors of the listed U.S. firms included in this sample are of non-Caucasian descent and are considered an ethnic minority. Only 0.68% of directors attended less than 75% of board meeting and the average debt to equity ratio is 2.51, with 50% of firms having a debt to equity ratio reaching between 0.75 and 2.61. The average market to book ratio of the sample is 2.73.
Table 4: Descriptive statistics
Variable N Mean sd p25 Median p75 Minimum Maximum
ENVPOS 56,095 0.67400 1.11768 0 0 1 0 6 DAGE 56,095 62.62895 8.02798 57 63 68 27 100 DGEN 56,095 0.13800 0.34490 0 0 0 0 1 DIND 56,095 0.79458 0.40401 1 1 1 0 1 FIRMP 56,095 0.04792 0.06565 0.01546 0.04595 0.08263 -0.24180 0.21507 DETH 56,095 0.10218 0.30289 0 0 0 0 1 DATT 56,095 0.00676 0.08192 0 0 0 0 1 FIRMC 56,095 2.51271 3.36151 0.74783 1.35291 2.60727 -3.96804 19.37618 LNSIZE 56,095 8.49402 1.71159 7.28224 8.36943 9.59330 3.50745 14.69750 MTOB 56,095 2.73144 2.62413 1.29914 1.98404 3.15053 -0.39710 17.45331
Table 5 provides an overview of the average number of environmental strengths per year. After 2009, there has been a significant increase in the year average, which only lasted for two years, as the average dropped to levels below the 2007 score in
2012. One explanation for the increase of the 2010 score relative to the score of the year before is the 7.0 magnitude Haiti earthquake that occurred on January 12, 2010, which could have caused more environmental awareness in directors (USGS, 2010). However, the average number of strengths dropped significantly in 2012, even though the United States were hit with the largest, costliest tornado outbreak in history in April 2011. Over the course of three days, the Eastern part of the United States were hit with over 360 tornados, which caused 321 fatalities (National Oceanic and Atmospheric Administration, 2012). There is no reason to believe economic shocks are related to the big differences in the yearly averages of environmental strengths. The economic crisis of 2008 is said to have ended in fall 2011, which does not explain the drop of environmental strengths in 2012 (Weisenthal, 2013).
Table 5: Average number of environmental strengths
4.2 Correlation of variables
To get a clear view of how the variables in this thesis correlate with each other a Spearman correlation matrix has been included in table 6. The Spearman correlation coefficient has been chosen over the Pearson coefficient, as the Pearson coefficient is only sensitive to a linear relationship between two variables and the Spearman coefficient has a higher sensitivity to nonlinear relationships (Hauke & Kossowski, 2011). Because this thesis hypothesizes a nonlinear, quadratic relationship between director age and the number of environmental strengths, the Spearman correlation coefficient is more appropriate to use.
In table 6, the correlation coefficients that are significant at a significance level of 1% are presented in italics. However, Hauke and Kossowski (2011) researched the differences between Pearson’s and Spearman’s correlation coefficient and concluded that Spearman’s coefficient should not be overinterpreted, which is why
Year Observations Average
ENVPOS 2007 3,774 0.43190 2008 5,974 0.50084 2009 6,410 0.45101 2010 6,775 1.23469 2011 11,322 0.92951 2012 11,598 0.39533 2013 10,242 0.66930
bold to clarify the relationship between the variables. The complete Pearson’s correlation matrix can be found in appendix A.
Table 6 shows that when Pearson’s coefficient is significant, this is also the case for the Spearman coefficient, but two of the presented coefficients are only significant for the Spearman coefficient. Nevertheless, the biggest difference between the two lies in the correlation between FIRMC and MTOB, which are
negatively correlated in Spearman (r = -0.0636, p < 0.0001), but positively correlated in Pearson (r = 0.1887, p < 0.0001).
It is worth noting that the independent variable of this research is not significantly correlated with the dependent variable (r = -0.0042, p = 0.3237), unlike the three moderating variables, which are all positively correlated with ENVPOS. The control variables are all significantly correlated with ENVPOS for Spearman, with the exemption of DATT, which is not correlated with any of the other variables.
Table 6: Spearman’s correlation matrix
* - Pearson correlation coefficient of 0.1887
1Significant Spearman correlation coefficients are presented in italics (significance level of
1%)
2Correlation coefficients that are significant in Pearson’s coefficient are presented in bold
(significance level of 1%)
Variable ENVPOS DAGE DGEN DIND FIRMP DETH DATT FIRMC LNSIZE MTOB
ENVPOS 1.0000 DAGE -0.0042 1.0000 DGEN 0.0668 -0.1720 1.0000 DIND 0.0627 0.1796 0.1304 1.0000 FIRMP 0.1359 -0.0178 0.0036 0.0016 1.0000 DETH 0.0691 -0.0890 0.0677 0.0659 0.0146 1.0000 DATT -0.0111 -0.0031 -0.0109 0.0075 -0.0097 0.0074 1.0000 FIRMC 0.0943 0.0125 0.0626 0.0475 -0.4103 0.0473 -0.0022 1.0000 LNSIZE 0.4787 0.0368 0.0782 0.0642 -0.1324 0.0838 -0.0079 0.4964 1.0000 MTOB 0.1278 -0.0455 0.0187 0.0103 0.5795 0.0140 0.0027 -0.0636* -0.1198 1.0000
4.3 Regression Results
This subsection covers the analysis of the regressions performed to find evidence for testing the hypotheses of this thesis. The subsection is divided into two parts. Part 4.3.1 summarizes the results of the OLS regression and part 4.3.2 summarizes the sensitivity tests performed to make sure the right conclusions are drawn from the results. A two-tailed significance level of 5% has been used as a basis for all regressions.
4.3.1 OLS regressions
Hypothesis one expects to find a quadratic relationship between ENVPOS and DAGE, as has previously been established in Post et al. (2011). Table 7 provides an overview of the found coefficients of all models and shows the p-levels for each coefficient. The coefficients of AGE (t = 10.05, p < 0.001) and (AGE)2 (t = -10.40, p <
0.001) are both significant at the 5% significance level. These results suggest there is a quadratic relationship between ENVPOS and AGE, with a peak in environmental strengths when directors are, on average, 60 years of age, when all control variables are 0. The relationship between director age and the total number of environmental strengths are presented in figure 2. These results support the first hypothesis and the relationship found by Post et al. (2011) and rejects the idea that younger generations are more environmentally oriented. The peak in age shifted from 56 in the Post et al. (2011) study to the 60 that was found in this thesis.
The remaining three hypotheses expect to find moderating effects of the age effect on the number of environmental strengths. The second hypothesis focuses on the possible moderating effect of director gender. The results of the regression are found in table 7. The moderating effect of gender is portrayed in DAGE*DGEN and has a significant coefficient of -0.00044962 (t = -2.69, p = 0.007). This coefficient suggests director gender moderates the effect of age on the total number of environmental strengths. This coefficient is also significant for a one-tailed t-test, with a significance level of 5%, which means that the more female directors on a board, the less
prominent the age effect is. This could suggest that women are more environmentally oriented than men and therefore less susceptible to the age effect. These results support hypothesis 2, which states that there is a moderating effect of director gender on the age effect.
Hypothesis 3 concerns the possible moderating effect of director
independence, which is expected to influence the age effect on environmental strengths after prior literature found a relationship between the independence of directors and the firm’s CSR policies. The outcomes of the regression used to assess the significance of the interaction are presented in table 7. The coefficient of
DAGE*DIND, which portrays the moderating effect of director independence on the age effect on environmental strengths, has a coefficient of 0.0027118 that is
significant for a significance level of 5% (t = 2.29, p = 0.022), which suggest that more independent directors on a board enhances the effect of director age on
environmental strengths. This result confirms hypothesis 3, which states that director independence moderates the age effect.
The last moderating variable this thesis measures, is firm financial
performance, proxied by a firm’s ROA. Hypothesis 4 expects to find a moderating effect of financial performance on the age effect. The moderating effect of financial performance is portrayed in the coefficient of DAGE*FIRMP, which has a value of -0.0148426 (t = -7.58, p < 0.001), shown in table 7. This coefficient shows that firm financial performance moderates the age effect by reducing the effect age has on the total number of environmental strengths when the ROA increases. This can mean that when a firm performs well, directors tend to make more environmental friendly decisions, regardless of age, or that when firms perform poorly, only directors who are not, on average, particularly old or young make more environmental friendly decisions even though there are costs associated with them. This result confirms
hypothesis 4 by proving the moderating effect firm performance has on the age effect.
The control variables, director ethnicity, attendance of board meetings, capital structure, firm size and the market-to-book ratio have been included in all empirical models. Director ethnicity, firm size and the market-to-book ratio are in all four models positively associated with the number of environmental strengths, while the capital structure, proxied by the debt-to-equity ratio, is negatively associated with the number of environmental strengths. The attendance of board meetings is not
Table 7: OLS regression results Variable H1a H2a H3a H4a DAGE 0.0541267* 0.0566002* 0.0493646* 0.0544308 <0.001 <0.001 <0.001 <0.001 (DAGE)2 -0.0004493* -0.0004612* -0.0004327* -0.0004462 <0.001 <0.001 <0.001 <0.001 DGEN 0.3444106* 0.001 DAGE*DGEN -0.0044962** 0.007 DIND -0.0904306 0.213 DAGE*DIND 0.0027118** 0.022 FIRMP 1.6242 <0.001 DAGE*FIRMP -0.0148426 <0.001 DETH 0.0884157* 0.0845469* 0.0817531* 0.0861067 <0.001 <0.001 <0.001 <0.001 DATT -0.0591462 -0.0552964 -0.0612558 -0.0531832 0.229 0.260 0.212 0.279 FIRMC -0.0812815* -0.0812992* -0.0811864* -0.0767489 <0.001 <0.001 <0.001 <0.001 LNSIZE 0.3625781* 0.3614295* 0.3617236* 0.3605145 <0.001 <0.001 <0.001 <0.001 MTOB 0.0808722* 0.0806696* 0.0805499* 0.074155 <0.001 <0.001 <0.001 <0.001 Intercept -4.029653 -4.137176 -3.853681 -4.07101 Observations 56,095 56,095 56,095 56,095 Adjusted R2 0.2723 0.2729 0.2730 0.2737 F-value 2999.53 2340.20 2341.86 2350.27
Bold coefficients are significant at a p < 0.05 significance level (two-tail)
*,**,*** significant at, respectively a significance level of 0.01, 0.05, 0.1
4.3.2 Sensitivity analysis
To make sure the chosen empirical models did not result in any wrongly drawn conclusions, additional regressions have been run with alternative empirical models.
Firstly, a regression has been run with all available variables and no interaction variables, to make sure the age effect is still significant when more variables are included in the model. Secondly, the empirical model has been changed to replace the interacting variable with (DAGE)2, instead of DAGE.
Table 8 provides an overview of the first alternative regression that includes all variables of the thesis. Both DAGE (t = 9.27, p < 0.001) and (DAGE)2 (t = -9.66, p <
0.001) are significantly related to ENVPOS in empirical model 1b. This result suggests that there is an age effect on the total number of environmental strengths and therefore confirms hypothesis one. Apart from that, all hypothesized moderating variables are significantly related to the total number of environmental strengths. Consistent with empirical model 1a, all control variables with the exception of director attendance of board meetings are also significant at a significance level of 5%.
Table 8: Alternative OLS regression one (1𝑏) 𝐸𝑁𝑉𝑃𝑂𝑆 = 𝛽-+ 𝛽/𝐷𝐴𝐺𝐸 + 𝛽3𝐷𝐴𝐺𝐸3+ 𝛽 4𝐷𝐺𝐸𝑁 + 𝛽7𝐷𝐼𝑁𝐷 + 𝛽8𝐹𝐼𝑅𝑀𝑃 + 𝛽>𝐷𝐸𝑇𝐻 + 𝛽A𝐷𝐴𝑇𝑇 + 𝛽F𝐹𝐼𝑅𝑀𝐶 + 𝛽G𝐿𝑁𝑆𝐼𝑍𝐸 + 𝛽/-𝑀𝑇𝑂𝐵 + 𝜀/ Variable H1b DAGE 0.050102* <0.001 (DAGE)2 -0.0004184* <0.001 DGEN 0.0639282* <0.001 DIND 0.0668299* <0.001 FIRMP 0.528759* <0.001 DETH 0.0766946* <0.001 DATT -0.0544719 0.267 FIRMC -0.0774944* <0.001 LNSIZE 0.3584138* <0.001 MTOB 0.0746507* <0.001 Intercept -3.944081 Observations 56,095 Adjusted R2 0.2741 F-value 2118.69
Bold coefficients are significant at a p < 0.05 significance level (two-tail)
*,**,*** significant at, respectively a significance level of 0.01, 0.05, 0.1
Table 9 shows the outcomes of the second additional regression. In this regression the interaction variable between director age and the moderating variable has been replaced by a quadratic interaction variable with (DAGE)2. In all three models the age
effect on the total number of environmental strengths is significant at a significance level of 5%. The included control variables are also significant in all three empirical models, with the exception of DATT, consistent with the models used in the original regression of this thesis.
For empirical model 2b, the interacting variable (DAGE)2*DGEN is negatively
associated with environmental strengths (t = -2.70, p = 0.007), which, like in empirical model 2a, results in a negative moderating effect of director gender on the age effect on environmental strengths and reconfirms hypothesis 2.
In empirical model 3b, there are some differences in the outcomes of the regression compared to the empirical model of 3a. The interaction between director independence and director age was significantly related to environmental strengths in model 3a, which suggested a positive moderating effect of director independence. However, in empirical model 3b this moderating effect is no longer significant and therefore rejects hypothesis 3 (t = 1.82, p = 0.068). This moderating effect was present when there was an interaction between DIND and DAGE, which could suggest that director independence has an influence of the linear trend, whether the effect is positive or negative, of the relationship, but not on the direction and
skewness of the curvature.
For the last moderating variable, a significant negative moderating effect has been established in empirical model 4a. For model 4b, the moderating effect of firm performance is insignificant (t = 0.48, p = 0.633). These findings reject hypothesis 4, however, like with director independence, firm financial performance could influence the linear trend of the age effect, but have no influence on the direction and
Table 9: Alternative OLS regression two (2𝑏) 𝐸𝑁𝑉𝑃𝑂𝑆 = 𝛽-+ 𝛽/𝐷𝐴𝐺𝐸 + 𝛽3𝐷𝐴𝐺𝐸3+ 𝛽 4𝐷𝐺𝐸𝑁 + 𝛽7𝐷𝐴𝐺𝐸3∗ 𝐷𝐺𝐸𝑁 + 𝛽8𝐷𝐸𝑇𝐻 + 𝛽>𝐷𝐴𝑇𝑇 + 𝛽A𝐹𝐼𝑅𝑀𝐶 + 𝛽F𝐿𝑁𝑆𝐼𝑍𝐸 + 𝛽G𝑀𝑇𝑂𝐵 + 𝜀/ (3𝑏) 𝐸𝑁𝑉𝑃𝑂𝑆 = 𝛽-+ 𝛽/𝐷𝐴𝐺𝐸 + 𝛽3𝐷𝐴𝐺𝐸3+ 𝛽 4𝐷𝐼𝑁𝐷 + 𝛽7𝐷𝐴𝐺𝐸3∗ 𝐷𝐼𝑁𝐷 + 𝛽8𝐷𝐸𝑇𝐻 + 𝛽>𝐷𝐴𝑇𝑇 + 𝛽A𝐹𝐼𝑅𝑀𝐶 + 𝛽F𝐿𝑁𝑆𝐼𝑍𝐸 + 𝛽G𝑀𝑇𝑂𝐵 + 𝜀/ (4𝑏) 𝐸𝑁𝑉𝑃𝑂𝑆 = 𝛽-+ 𝛽/𝐷𝐴𝐺𝐸 + 𝛽3𝐷𝐴𝐺𝐸3+ 𝛽 4𝐹𝐼𝑅𝑀𝑃 + 𝛽7𝐷𝐴𝐺𝐸3∗ 𝐹𝐼𝑅𝑀𝑃 + 𝛽8𝐷𝐸𝑇𝐻 + 𝛽>𝐷𝐴𝑇𝑇 + 𝛽A𝐹𝐼𝑅𝑀𝐶 + 𝛽F𝐿𝑁𝑆𝐼𝑍𝐸 + 𝛽G𝑀𝑇𝑂𝐵 + 𝜀/
Variable H2b H3b H4b DAGE 0.0562042* 0.0510282* 0.0539534* <0.001 <0.001 <0.001 (DAGE)2 -0.000458* -0.0004428* -0.0004495* <0.001 <0.001 <0.001 DGEN 0.2125063* <0.001 (DAGE)2*DGEN -0.0000378** 0.007 DIND 0.0089284 0.812 (DAGE)2*DIND 0.0000174 0.068 FIRMP 0.4142977 0.087 (DAGE)2*FIRMP 0.0000278 0.633 DETH 0.0845926* 0.0815602* 0.0868352* <0.001 <0.001 <0.001 DATT -0.552862 -0.0615229 -0.0543518 0.260 0.210 0.269 FIRMC -0.0813042* -0.081188* -0.0776103* <0.001 <0.001 <0.001 LNSIZE 0.361444* 0.3617289* 0.3602157* <0.001 <0.001 <0.001 MTOB 0.0806738* 0.0805536* 0.0751634* <0.001 <0.001 <0.001 Intercept -4.125126 -3.916236 -4.01667 Observations 56,095 56,095 56,095 Adjusted R2 0.2729 0.2730 0.2730
F-value 2340.21 2341.57 2341.52
Bold coefficients are significant at a p < 0.05 significance level (two-tail)
*,**,*** significant at, respectively a significance level of 0.01, 0.05, 0.1
4.3 Summary of findings
In section 4, two empirical models have been used to test whether to reject or confirm the hypotheses. The first model included an interaction variable with DAGE, while the second model an interaction variable with (DAGE)2. As there have been
differences in the outcomes of the significance of the expected moderating variable, these have been summarized per model in tables 10, 11 and 12, for respectively director gender, director independence and firm financial performance.
Before drawing any conclusions about moderating variables, the results must conclude that the age is significantly related to the total number of environmental strengths. Tables 7 and 8 show that the age effect is significant in empirical model 1a and 1b. This means hypothesis one is confirmed by the outcomes of the empirical models. This implies that there is a significant age effect on the total number of environmental strengths.
Table 10: Overview of the results of H2 Empirical model
2a
Empirical model 2b
Age effect Significant Significant
Gender – moderating effect Significant, negative Significant, negative Adjusted R2 0.2729 0.2729
The second hypothesis expects to find director gender to moderate the age effect. The outcomes the two empirical models have been summarized in table 10. In models 2a and 2b, the moderating effect was found be significant. These results together lead to the conclusion that there is a negative moderating effect of the age effect by director gender and confirmation of the second hypothesis, suggesting females are less susceptible to the age effect as they are naturally more
Table 11: Overview of the results of H3 Empirical model
3a
Empirical model 3b
Age effect Significant Significant
Independence – moderating effect Significant, positive Insignificant Adjusted R2 0.2730 0.2730
The third moderating variable this thesis expects to find, is director independence. Table 11 provides an overview of the two empirical models used to collect evidence to determine whether to reject or confirm hypothesis 3. The evidence has mixed results, and the adjusted R2 cannot help determining the model that best fits the
purpose. However, the result might suggest that director independence moderates the linear term of the age effect, but does not moderate the quadratic term. Both models are generally accepted in research, which is why a single best model cannot be determined without further research. To avoid type 1 and type 2 errors,
conclusions cannot be drawn for hypothesis 3.
Table 12: Overview of the results of H4 Empirical model
4a
Empirical model 4b
Age effect Significant Significant
Performance – moderating effect Significant, negative Insignificant Adjusted R2 0.2737 0.2730
The last hypothesized moderating variable of this thesis is firm financial performance. The results of the two empirical models have been summarized in table 12. Like the moderation effect of director independence, the empirical results show differences in the significance of the outcomes for the moderation effect of firm financial
performance. The moderating effect is supported by the data in empirical model 4a but is found insignificant in model 4b. This time, the adjusted R2 is slightly lower for
the model that does not support the hypothesis. However, the difference in R2 is not
big enough to conclude it is a weaker model. To avoid type 1 and type 2 errors conclusions cannot be drawn for hypothesis 4.
