“CORPORATE PHILANTHROPY IN
THE NETHERLANDS”
An investigation on the effect of firm size, board composition
and sector on corporate philanthropy
Frederique Schansman 1323350
Supervisors: Prof. dr. H. van Ees Dr. D.H.M. Akkermans
RijksUniversiteit Gronigen Faculty of Economics and Business
Abstract
Table of Content
1 Introduction ... 4
1.1 Defining Corporate Social Responsibility... 5
1.2 Research Purpose ... 6
2 Literature Review... 8
2.1 Theoretical Background and Previous Research... 8
2.2 Concepts ... 10 2.2.1 Firm Size ... 11 2.2.2 Board Characteristics ... 12 2.2.3 Industry... 19 3 Methodology ... 22 3.1 Description Sample ... 22
3.2 Description Dependent Variable ... 23
3.3 Description Independent Variables ... 24
3.4 Description Research... 26 3.4.1 Model 1 ... 27 3.4.2 Model 2 ... 28 4 Data Analysis ... 30 4.1 Model 1 ... 30 4.1.1 Descriptive Analysis ... 30 4.1.2 Diagnostic Checks... 31
4.1.3 Regression Results and Hypothesis Testing... 34
4.2 Model 2 ... 37
4.2.1 Descriptive Analysis ... 37
4.2.2 Diagnostic Checks... 39
4.2.3 Regression Results and Hypothesis Testing... 41
5 Discussion ... 44
6 Conclusion... 49
References ... 51
Table Overview
Figure 2.1: Carroll’s Pyramid of CSR... 9
Table 3.1: Reasons Not to Answer My Request ... 23
Figure 3.1: General Assumptions of Multiple Regression Model ... 26
Table 4.1: Descriptive Statistics – Model 1 (unlogged variables) ... 30
Table 4.2: Correlation Coefficients – Model 1 ... 32
Table 4.3: VIF-values Philanthropic Expenditure Equation ... 33
Table 4.4: Results of Multiple Regression Analysis – Model 1... 35
Table 4.5: Mean Philanthropic Expenditures – Model 2 (unlogged variables) ... 37
Table 4.6: Descriptive Statistics – Model 2 (unlogged variables) ... 38
Table 4.7: Correlation Coefficients – Model 2 ... 40
Table 4.8: VIF-values Philanthropic Expenditure Equation – Model 2... 40
Table 4.9: Results of Multiple Regression Analysis – Model 2... 41
Table 5.1: Hypothesis Results... 44
Appendix Text Box 1: Email to Companies ... 56
Table 1: BIK Code ... 57
Table 2: White’s Test for Heteroskedasticity – Model 1 ... 58
Figure 1: Normality of Residuals – Model 1... 59
Figure 2: Normality of Residuals – Model 2... 59
Table 3: White’s Test for Heteroskedasticity – Model 2 ... 60
1 Introduction
In recent years ‘Corporate Social Responsibility’ (CSR) has become more and more part of normal daily business. How socially responsible a company is and how it incorporates this social behaviour into corporate culture can be found in a company’s annual report or even a separately published CSR report (i.e., in addition to the annual report). An increasing trend can be observed in the publication of separate CSR reports of companies. In 2001, only 4 of the 25 AEX companies1 in the Netherlands published a CSR report. Over the years the number of companies issuing CSR reports has increased, from 32 % of the 25 AEX companies in 2003 to 64 % in 2005 and over 70 % of 24 AEX companies in 2006. Also the incorporation of a section on Corporate Social Responsibility in the annual report has revealed an increasing trend: in 2001, 23.8 % of the remaining AEX companies (i.e., that did not have a separate report on CSR) had a chapter on CSR in their annual report, increasing to 47.0 % in 2003 and 100 % of the remaining AEX companies in 2006. Therefore, for the largest Dutch companies, CSR can be seen as a trend, as these companies are increasingly integrating CSR in their daily business.
These CSR reports try to answer the questions of how companies integrate corporate social behaviour into their daily business. Companies report, for instance, on how they are dealing with the environment (Royal DSM, 2006); how they are taking care of their employers (TNT, 2006); whether they are taking good care of the local community they operate in (Royal Dutch Shell, 2006); whether they have a company foundation (Royal Dutch Shell, 2006); whether and how much they donate to charity (Unilever, 2006); and so on. Through these CSR reports companies are not only being judged on their financial key performance indicators, but also on how socially responsible they behave.
Corporate Social Responsibility has, for many businesses, influenced their corporate practice and become a new way of doing business, it has been integrated into their traditional business model. Unilever, for example, practices social responsibility by trying to fight obesity as they are supporting a Dutch project concerning food and movement2. TNT is being socially responsible as they are an important partner of the ‘World Food Program’ and thereby they
are trying to fight hunger all over the world3. Because of these activities CSR is seen as sympathetic, it may create a good image and a positive corporate reputation. This is one of the reasons why the number of companies that are integrating CSR into their corporate strategies is constantly increasing.
1.1 Defining Corporate Social Responsibility
As companies are increasingly aware of CSR, it is important to know what it exactly is and how it can be measured. What, to begin with, is the exact definition of Corporate Social Responsibility? Several articles start with this question (Bais, 2005; Carroll, 1979, 1999; Ebner and Baumgarten, 2006; Snider et al., 2003). Most authors (Carroll, 1979; Snider et al., 2003) draw the conclusion that a precise definition of CSR depends on the perspective from which and the timeframe in which you are looking at CSR. Carroll (1979) gives an overview of how the viewpoints of several authors towards CSR have changed over time: from ‘profit making only’ by Friedman (1962) to ‘giving way to social responsiveness’ by Sethi (1975).
A general definition of CSR starts from the notion that organizations have a certain responsibility towards the society they are operating in, especially towards local communities and the environment. This is quite a wide-ranging definition, which can be specified to different social issues, such as human rights, quality of employment, health, durability of the climate, fight against poverty, etc. Every company chooses for itself which topics it finds most important and how it deals with these issues. It then again reports about these subjects in its annual or CSR report.
According to CSR Netherlands4 Corporate Social Responsibility can be typified according to
multiple forms (MVO Nederland, 2004): CSR as a source for profits, CSR as a motivator, CSR as good citizenship, CSR as a stimulus to volunteer work among employees and CSR as charity. In this thesis the focus will be on the last form: CSR as charity, as this thesis wants to contribute to the developing stream of research on this subject.
3 http://www.tntpost.nl/
4 CSR Netherlands, or MVO Nederland, is a Dutch organization which provides knowledge and information to
CSR as charity can be defined as donating money or goods to charity organizations or the community, through which the company tries to contribute in the solution of social problems. The way in which this is done may vary, for instance through directly donating money to a local community, giving (spare) products or other goods, setting up a social project, sponsoring a sports club, etc.
The biannual report of ‘Geven in Nederland 2007’ (Schuyt et al., 2007) states that in 2005 Dutch companies gave 1.5 billion Euros to charity by means of donations and sponsoring, which is almost 0.3 % of Dutch GDP. Most of this money is given to sports and recreation, but education and research have also received a substantial amount. In the same study, they found that 78 % of the companies researched either sponsored or gave money in the year 2005. Of the total amount that companies have given, 1.1 billion Euros was in the form of sponsoring and 0.4 billion Euros in the form of donations. Another interesting finding is that of the 1000 companies researched only 27 % declared to have a specific policy concerning CSR.
1.2 Research Purpose
Previous research on CSR as charity has mainly focussed on the UK. One of the reasons for this is that data for British companies are readily available. In the UK, it is required by law since 1967 to disclose information on the amounts spent on donations and sponsoring in a company’s annual report. This has been adopted in the Companies Act of that year and states that companies should disclose any charitable payment over ₤200 in their annual report; however, details of these payments do not have to be given (Cowton, 1987). Therefore, many researches on the link between CSR (as charity) and for instance firm size, organizational visibility, etc. have been limited to the UK (Adams and Hardwick, 1998; Brammer and Millington, 2005, 2006; Campbell et al., 2002; Campbell and Slack, 2006).
compare the results of studies in the UK with the results of this study, which focuses on the Netherlands. It wants to find out whether the fact that it is compulsory in the UK to disclose the amounts spent on charity in a company’s annual report has any effect on the relationship between CSR and various company and institutional characteristics, such as organizational size, board characteristics, etc. Does the fact that disclosing charity expenses is not mandatory in the Netherlands have any effect on what companies have spent on charitable donations and therefore also affects the relationship between CSR and several institutional characteristics? No difference in this relationship implies that the policy concerning disclosure does not have an effect on charity expenses, whereas a difference implies that disclosure does have an effect, signifying that a policy change for the Netherlands might be plausible. No research before has considered the implications of this difference in obligation, so this study will therefore try to fill this research gap. The possible implications of this disclosure difference will be based on a qualitative comparison between the results of this study and the results of precious studies in the UK. This comparison and its potential implications will be discussed in chapter 5.
The observable trends that CSR is getting an increasing amount of attention and that companies seem to be more and more aware of their social role, which can be concluded from the increasing amount of CSR reports, make this study interesting. Also, this study adds to previous literature by exploring the effect of several institutional characteristics that have not yet been explicitly explored on CSR as charity, such as the proportion of outsiders on the TMT and sector of the economy. Thus, the purpose of this study is to examine the relationship between several institutional characteristics and CSR as charity in the Netherlands. Another purpose of this research is to find out whether the obligation to disclose information about a company’s charitable expenses has any effect on what companies spend and thus on the relationship to be researched.
2 Literature Review
In this chapter the theory behind the relationships to be researched will be explained by means of an extensive literature review. First of all, the general background of CSR will be discussed. Thereafter, the different institutional characteristics that may influence corporate philanthropy will be considered. At the end of each characteristic a hypothesis will be posed with its proposed relationship with corporate philanthropy.
2.1 Theoretical Background and Previous Research
A classification of CSR referred to in many articles is the categorization made by Carroll (1979). His definition of CSR is in terms of four responsibilities, which are not mutually exclusive, i.e. they may occur at the same time. These four responsibilities are economic, legal, ethical and discretionary. With economic responsibility Carroll refers to the responsibility of firms to produce goods and services that society needs at an acceptable and profitable price. With legal responsibility he denotes that companies need to comply with “(…) the rules of the game” (Jamali, 2007), while with ethical responsibility Carroll means that firms are expected to act above and beyond legal requirements. The final category, discretionary responsibility, can be defined as philanthropic activities or contributions made to give something back to society being completely voluntary in nature.
Carroll has revised his definition of CSR in 1991 by making it a historical construction by means of a pyramid, with economic at the bottom and philanthropic5 at the top (see figure 2.1). This modification means that a company was historically first concerned about the economic responsibilities of the company towards society, implying that its principal role was to produce goods and services for the consumer and to make an acceptable profit accordingly. This responsibility was followed by the notion that profit making had to comply with the laws and regulations of the local government, known as legal responsibility. These legal responsibilities are expected to be incorporated in every business, as society is built upon the law. Economic and legal responsibilities are currently seen as coexisting responsibilities towards the society which should be exercised simultaneously. After the legal responsibility the ethical responsibility came, which ensues the economic and legal responsibilities as it also
5 In 1979 it was denoted as ‘discretionary responsibility’, but Carroll revised this definition to ‘philanthropic
comprises what a company’s stakeholders and the society consider fair and just. It is the requirement to do what is right for the stakeholders. Finally, the concept of philanthropic responsibilities arose. This responsibility entails the perception that the firm is expected to be a good corporate citizen.
Figure 2.1: Carroll’s Pyramid of CSR
Source: Carroll, 1991 PHILANTHROPIC Responsibility ETHICAL Responsibility LEGAL Responsibility ECONOMIC Responsibility
For companies it may be desirable to engage in CSR when its competitors do so, as otherwise their competitors may have a strategic advantage6. Also, today’s consumers are increasingly concerned and conscious about social and environmental issues, which results in consumers demanding from companies to behave socially responsible as well (Mohr and Webb, 2005). Moreover, companies might see CSR as a good marketing tool (Brønn and Vrioni, 2001; Porter and Kramer, 2002). Companies that undertake CSR activities as a means to create goodwill and to increase their brand awareness may increase their market share (in the long run).
Social behaviour may be exposed by means of the philanthropic responsibility of Carroll (1991). One way of expressing this philanthropic responsibility is by donating to charitable organisations. Porter and Kramer (2002) argue that philanthropy should be part of a company’s strategy. They state that “(…) the more a social improvement relates to a company’s business, the more it leads to economic benefits as well”. They reason that when a
company improves the social conditions of the locations where they operate (e.g. improving education, preserving employer safety and health, etc.), they may improve the company’s competitiveness in the end ass well. Therefore, a company should pursue a strategy where corporate philanthropy converges with shareholder interest, noted by Porter and Kramer as ‘A Convergence of Interests’. Giving to charity institutions in their local community may enhance the company’s goodwill within its local market. It may also enhance the company’s goodwill with (potential) consumers (Lantos, 2001). These may be reasons to incorporate philanthropy into the company’s strategy.
Henceforth, this thesis will shed some light on the apex of Carroll’s pyramid: it will research the organizational characteristics that may influence philanthropic behaviour. The corporate socially responsible (CSR) behaviour of Dutch companies will be investigated, though limited to the philanthropic aspect. The focus will be on what Dutch companies spend on donations, sponsoring, social projects, etc., in both the Netherlands and in other countries. The relationships to be analyzed are whether this amount can be related to the size of the firm, board characteristics (proportion of outsiders and proportion of women on the TMT), and the sector of the economy in which they are active. This research has been limited to these aspects, as they represent different characteristics of the company, which are likely to influence CSR. The following sub-chapters will provide the theoretical background and explanation of the relationships to be researched. The research will be concentrated on the year 2006, as this is the most recent year with available data.
2.2 Concepts
2.2.1 Firm Size
The first line of research concerns the effect of firm size on corporate philanthropy. Why firm size is a variable that is very likely to influence CSR will be explained in the following paragraphs. Useem (1988) classifies firm size as the most important institutional factor that determines corporate giving decisions. He states that “larger firms contribute more money, regardless of profits”. The underlying theory here is professionalization, which indicates that larger companies operate more formal programs and that their strategies and policies are more exposed to values of the community they operate in. Related to this theory is that of Cowen et al. (1987), who state that larger firms are more probable to be noticed by the general public which leads to greater pressure in order to behave socially responsible.
In line with this reasoning is the stakeholder theory (Freeman, 1984), a theory which identifies the different groups that are stakeholders of a company and which illustrates and suggests several approaches that management of a company may use in order to preserve the diverse interests of these groups. Cowen et al. (1987) reason that larger firms have more stakeholders, who may be concerned about the social behaviour of the company, than small companies. Adams and Hardwick (1998) also consider stakeholder theory as the main reason why companies may make charitable contributions and may undertake social activities. According to Adams and Hardwick (1998), larger companies are more probable to be scrutinized by the government and the general public than smaller companies and they also are more likely to have stakeholders that are interested in the company’s corporate social behaviour. Not surprisingly, they found strong evidence for the fact that ‘larger companies engage in more social activities than smaller companies’. Other previous researches have also found a strong positive link between the size of the company and the amount spent on charitable donations. Brammer and Millington (2006) find that, on average, both larger and more profitable firms make higher levels of philanthropic contributions. This finding is consistent with general assumptions.
Based on stakeholder theory, it is assumed that there is a positive relationship between the size of a company and the amount this company spends on charitable donations. Therefore, the following is hypothesized:
Hypothesis 1: The level of corporate philanthropy (of Dutch companies) is positively related to company size.
2.2.2 Board Characteristics
Another line of research concerns the effect of board demographic characteristics on corporate social responsibility. Board characteristics may influence CSR as corporate strategy is often determined by the top management team (TMT) of the company (Hambrick and Mason, 1984; Wiersema and Bantel, 1992). Each TMT is alternatively composed and will therefore have a different view on strategy, different incentives with regard to CSR, make different choices etc. It has to be determined whether composition of TMT influences corporate philanthropic strategy as well. Therefore, it is interesting to find out whether there are certain characteristics within this TMT that may increase or decrease corporate expenditure on philanthropic activities. Examples of board characteristics may be gender, age, function, etc.
Inside versus Outside Directors
The first board characteristic to be researched is the proportion of outsiders on the TMT, more specifically the amount of outside versus inside directors on a company’s TMT. An inside director can be defined as a member of the board of directors who holds another position within the company, usually in management or as a major shareholder7. An outside director, on the other hand, is defined as an independent member of the board of directors that does not hold any operational responsibility towards the corporation8. The relationship to be researched is whether either inside or outside directors have more affiliation with CSR and may therefore increase the expenditure on corporate philanthropy. Does a higher proportion of outside directors on the TMT result in a higher corporate philanthropic expenditure? Or is it the other way around and does a higher proportion of inside directors on the TMT result in a higher expense on charitable donations? Is there a relationship at all?
The importance of this relationship can be explained by the ‘upper echelons perspective’. According to Hambrick and Mason (1984) this theory asserts that corporate strategy and performance can be partially determined by the background characteristics of the top management team. The authors thus state that the strategic decisions made by a company and the direction the company is going can be - to some extent - explained by an outline of the company’s ‘upper echelon’. Consistent with this upper echelon perspective is the view that strategic decisions regarding philanthropy may be influenced by the composition of the TMT. When looking at previous researches on the subject of CSR Orientation (CSRO), many were primarily based on the conceptualization by Carroll (1991), namely whether inside and outside directors tended to be more CSR oriented towards economic, legal, ethical or discretionary responsibilities. In many studies (Ibrahim and Angelidis, 1995; Ibrahim et al., 2003; O’Neill et al., 1989) directors were asked to fill out a questionnaire based on an instrument developed by Aupperle et al. (1985), which consists of a four-part construct of Carroll’s responsibilities (1979). By using this construct it was measured how CSR oriented each director9 was – economical, legal, ethical or discretional10. These studies found that outside directors tend to be more CSR oriented towards philanthropic activities, while inside directors are more economically driven (Ibrahim and Angelidis, 1995; Ibrahim et al., 2003; O’Neill et al., 1989). The main explanation the authors give for this finding is that outsiders
7 www.genesisexchange.com; https://www.woodmen.com 8 www.joegriffith.com
may ‘(…) have a broader range of experience and interests’ (Ibrahim et al., 2003). This implies that outsiders may have other – more discretionary related - concerns than insiders, who may primarily be involved in profit making. Another explanation that the authors (Ibrahim et al., 2003; Ibrahim and Angelidis, 1995) pose is that it is more probable to find outsiders in companies that are more profitable, which may therefore have the funds for ‘the luxury’ of being philanthropically involved. This explanation can be related to the relationship priory stated11: larger firms are more capable of making charitable contributions, implying that the level of corporate philanthropy may be positively related to company size. Another potential explanation for the CSR orientation of outsiders to be more philanthropic is that outside directors have different background characteristics and that they have less financial interests in the company than do inside directors (O’Neill et al., 1989).
The fact that outsiders are more sensitive to philanthropic activities than insiders is an interesting finding. It implies that outsiders may play an important role in the process of determining the philanthropic strategy. However, how significant is their influence? Can it be stated that the more outsiders in a company’s top management team, the higher the corporate contributions to philanthropic activities? This is the relationship that is going to be investigated in this thesis.
However, in order to research this relationship for the Dutch corporate structure, we need to adjust the one-tier (Anglo-Saxon) system12 to the two-tier Dutch system. In the Netherlands a company’s top management team is build up of the following bodies: the upper layer is the executive board, in Dutch called ‘Raad van Bestuur’. This executive board is being controlled by a supervisory board. This supervisory board monitors the strategy, financials and other decisions made by the executive board. Often, members of the supervisory board are on multiple supervisory boards of different several companies. This way, they are able to remain impartial. These members of the supervisory board do not have a financial stake in the company.
11 See chapter 2.2.1
12 This has to be done because the previous researches (Ibrahim and Angelidis, 1995; Ibrahim et al., 2003;
For this Dutch corporate governance structure, the members of the executive board can be said to be the members of the inside directors and the supervisory board can be said to be the outside directors. However, some members of the supervisory board have previously been on the executive board of the company in question, in which case they are called quasi-insiders. These members can be considered as less independent and less objective than the members of the supervisory board that have not worked for the company before.
In order to be certain that the members of the supervisory board remain impartial, the Dutch Corporate Governance Code (Tabaksblat, 2003) has secured several best practice recommendations that apply to all Dutch listed companies13. These rules include principles and best practice provisions for the executive board and the supervisory board. One of those best practice provisions is that a company may have a maximum of one supervisory board member that is dependent (i.e., a quasi-insider), the rest should be independent (i.e., outside directors). If the company does not comply with this rule it has to explain why this is so in its annual report14. An important notion is that this code does not hold for companies that are not
listed on the Dutch stock market.
Dependent directors are defined by the Dutch Corporate Governance Code (Tabaksblat, 2003) according to several principles15. A supervisory board member is considered to be independent if the following criteria do not apply to him or to any relative up to the second degree16:
1. Employment or executive board membership of company in five years prior to appointment on supervisory board;
2. Receipt of personal financial payment from the company other than compensation received for normal course of business;
3. Has had notable business relation with the company in the year prior to appointment; 4. Membership of executive board of a company in which an executive board member of
the company which he supervises is a supervisory board member; 5. Holding more than ten percent of the shares of the company;
13 These are companies that are statutory established in the Netherlands and are listed on the Dutch stock market. 14 Also known as the ‘comply or explain’ principle of the Corporate Governance Code
15 For a complete and thorough overview of the criteria, see the Dutch Corporate Governance Code by
Tabaksblat, 2003.
16 First degree family is parents or children; second degree family is brothers and sisters, grandparents or
6. Membership of the executive or supervisory board of (or representing in any way) a corporation which holds over ten percent of the shares of the company, unless the corporation is a member of the same group as the company;
7. Temporarily managed the company throughout the last year where executive board members were not present or not capable to discharge their responsibilities.
This thesis will refer to the top management team (TMT) as the supervisory board and the executive board together. In this TMT there are insiders, outsider, and quasi-insiders. The insiders are the dependent directors on the executive board; the quasi-insiders are the dependent directors on the supervisory board; and the outsiders are the rest of the supervisory board - the independent directors as defined by the Dutch Corporate Governance Code. What is of interest here is the proportion of outside directors to the total number of directors on the TMT. Theory suggests that outsiders are more philanthropically oriented than insiders, which implies that a higher proportion of outsiders on the TMT can be associated with a higher level of philanthropic expenditures.
Therefore, it will be researched whether the proportion of outsiders to the total number of directors on the TMT can be said to influence the amount spent on philanthropic causes. Based on the upper echelon perspective that board characteristics may determine corporate strategy and based on other theory that suggests that more independent directors on the TMT lead to more philanthropic expenditures, the following is hypothesized:
Hypothesis 2: The level of corporate philanthropy (of Dutch companies) is positively related to the proportion of outside directors on the top management team.
Women
on their empirical findings, Ibrahim and Angelidis (1994) state that women are less associated with being economically (that is: profit) driven than men, whereas they are more associated with being interested in philanthropic causes. The importance of this implication is that women may bring a different objective to the company, which may improve the decision-making process of a company17.
Some researches have even found that women in corporate boards have a positive influence on both economical and social corporate performance. A recent research performed by Catalyst (2007) points out that companies with at least three women in their top management team perform notably better than the average company. A same conclusion is drawn in a research performed by McKinsey and Company, called ‘Women Matter’ (Desvaux et al., 2007). This study concludes that corporations with women on the top management team perform significantly better than corporations that do not have women in top positions. These ‘women on top’ corporations tend to have an EBIT18 of 48 % higher than other companies. This raises the interest to which other effects women may have on the company. One of those interests is the effect of women on CSR. Many researches on the link between women on corporate boards and the amounts spent on charitable donations have been conducted (Wang and Coffey, 1992; Williams, 2003). As a result of such findings, an increasing awareness in the possibilities to increase the number women in top management positions is the topic of many news articles (Rengers and van Uffelen, 2006; Elsevier, 2008; Jonkhout, 2008; NRC, 2008).
An example of the increased attention to women on corporate boards is a law introduced in Norway, concerning the presence of women on corporate boards. The Norwegian government has introduced a minimum quota of 40 % of women on every (public and private) company board in 2005. The last couple of years, this has lead to the additional voting of 400 women on corporate boards and has made Norway the country with, by far, the highest representation of women on boards19. Many countries have approved of this new law and are trying to introduce it in their home markets as well20 (NRC, 2008).
17 http://www.mt.nl/experts/1009639/Vrouwen_in_topfuncties_economische_noodzaak.html 18 Earnings Before Interest and Tax
Looking at previous studies on the relationship between women on corporate boards and corporate philanthropy mostly positive correlations are found. Williams (2003) finds strong evidence for the fact that firms with higher proportions of women on their boards engage in charitable giving to a greater extent than firms that have lower proportions of women on their boards. He makes a distinction between several types of charitable programs, namely educational, community service, and arts and cultural programs. He finds support for the fact that a higher number of women on corporate boards has a positive effect on charitable giving to community service programs and arts as well as cultural programs. No support is found for the assumption that it has a positive effect on educational programs, which is an unexpected finding, as women are generally associated with being involved in educational issues. An explanation for this unpredicted finding may be that educational institutions, as opposed to community and cultural organizations, might more often receive substantial governmental support.
Wang and Coffey (1992) also find that the proportion of female board members is positively related to the level of a company’s charitable contributions. They hypothesize that since female directors are more receptive to corporate social responsibility their presence on executive boards should have a positive effect on corporate philanthropy. They also argue that women are not as profit-driven as men and are more sensitive to a larger variety of stakeholders than men. They find support for these claims, and conclude that women have a positive impact on the level of charitable donations.
Based on the theory that women seem to be more philanthropically-oriented than men, the following is hypothesized:
2.2.3 Industry
The reason why industry may be of influence on corporate giving may be explained by means of stakeholder theory21. First of all, the consumer as of a company’s stakeholders determines the public exposure of the company. Stakeholder (i.e., consumer) pressure to behave in a socially responsible way might push companies in similar industries to contribute to philanthropic institutions with the intention of enhancing corporate reputation and the external perception of the firm (Brammer and Millington, 2006). This suggests that corporate giving strategies might be determined by differences between industries which may be generated by these stakeholders (Amato and Amato, 2006). Secondly, the government may be an important stakeholder in the determining of a company’s CSR strategy. Local and state governments are currently employing strategies to promote CSR (European Commission, 2002). An example is industries with socially damaging externalities that are encouraged by the government to behave in a socially responsible way (Royal Dutch Shell, 2006). These encouragements may even be strengthened through legislation to support philanthropic activities (Brammer and Millington, 2006). In conclusion, companies are exposed to stakeholder pressures that require firms to behave socially responsible, but these pressures may vary considerably among industries.
When determining the industry in which a company is present, the company is classified according to its principal business activity, which is the product or service that represents the largest percentage of total sales. Brown et al. (2006) state that companies in the same industry group tend to adopt similar practices when it comes to corporate giving. They divide firms in three types of industries: ‘regulated industry’ (i.e. non-financial industry that is confronted with entry and rate regulations), ‘environmental impact industry’ (i.e. industry that may cause significant damage to the environment) and ‘financial regulated industry’ (i.e. insurance and banking sector which is subject to regulatory inspection). They find that firms in the regulated industry donate significantly more to charity. Another interesting result of this study was that the mining and construction, retail, and transportation industries give notably less to charitable institutions than the manufacturing industry. On the other hand, petroleum, pharmaceutical and utility industries give significantly more than the manufacturing companies.
The effect of industry on corporate philanthropy has been researched in diverse ways. Griffin (2003) states that corporate philanthropic contributions increase in the first year after a merger or acquisition made in the same industry. The effect was even so strong, that the contributions continued to increase in the same industry until three years after the acquisition. Amato and Amato (2006) find that differences in giving habits between industries can be explained by strong industry effects, such as differences in sectoral matters and public exposure. They find support for their hypothesis that the proportion of charitable donations to revenues fluctuates across industries and learned that twenty to twenty-two percent of this variation can be explained by means of these industry effects. Their explanation for this finding is that within industries the ‘(…) giving culture may create an environment that requires firms to meet or exceed competitor philanthropy in order to maintain customer and community goodwill’. This implies that companies within the same industry adopt similar giving practices, which may again differ across industries.
Another way of differentiating between industries is by means of the sector division of the economy. This sector division segregates the economy in 4 different sectors: the primary, secondary, tertiary, and quaternary sector. The primary sector contains companies that change natural resources into primary products, such as agriculture, quarrying industries, etc. The secondary sector consists of companies that create finished, useable products, such as manufacturing companies or construction companies. The tertiary sector is also known as the service sector, as it comprises companies that provide services to businesses and to consumers, such as transport, retail, insurance, etc. The quaternary sector is an extension of the tertiary sector, as it includes intellectual services, such as IT, R&D and information services. There may be a difference in the giving patterns between these four sectors. The theory behind this difference - that the sector in which a company finds itself may determine CSR - is that the stakeholder groups differ per sector. According to Jones (1999) the secondary sector has the most stakeholder groups and he therefore hypothesizes that this group is most likely to engage in CSR practices. Another explanation is that this sector produces most consumer goods, an industry group which is found to be most sensitive to CSR as this is demanded by their stakeholders22.
Based on previous theory, the relationship between sector and corporate charity will be researched. Can it be said that companies in a certain sector spend significantly more on charitable donations than companies in the other sectors? Do companies within the same sector have the same spending pattern when it comes to philanthropic expenditures? Do some sectors give absolutely nothing to charity? These are questions of interest, which will be dealt with in this research.
It is expected that there will be sectors that give more to charity than others. As stakeholder pressure is found to influence corporate behaviour regarding philanthropy, it is expected that sectors that are more directly involved with the consumer (business-to-consumer), i.e. the secondary sector, will donate more to community charities than the primary, the tertiary, and the quaternary sector23, as these sectors are found to be more business-to-business oriented. Therefore, it is assumed that there will be significant differences in the amounts spent on charitable donations between the different sectors of the economy:
Hypothesis 4a: The level of corporate philanthropy (of Dutch companies) is higher for companies in the secondary sector of the economy.
Hypothesis 4b: The level of corporate philanthropy (of Dutch companies) is lower for companies in the primary and tertiary sectors of the economy.
23 However, the quaternary sector is left out of consideration in this study. For an explanation see chapter 3.3
3 Methodology
In this chapter the methodology used for this research will be discussed. First a description of the sample will be given, followed by specifications of the dependent and the independent variables. Finally, an explanation of the research will be given, which consists of descriptions of the first and the second model.
3.1 Description Sample
This research has investigated a sample of 44 large Dutch companies, based on the list of FEM Business 50024 for the years 2005 and 2006. The 100 largest companies of 2005 and the 100 largest of 2006 were taken. These two lists together comprised of 113 companies. As the FEM Business list does not contain all AEX and AMX listed companies25, the ones that were not yet on either FEM Business top 100 list were added to the sample. This added up to a total of 124 companies. These companies are a representation of the largest companies26 in the Netherlands.
The companies have been contacted27 with the request to provide the amount spent on charitable donations and sponsoring for the year 200628. Of these 124 companies, 64 answered that they could not provide the data requested. This had several reasons: some companies said that the data were not centrally registered; others said that they did not make these data publicly available; again other companies did not want to cooperate with my research, etc. Table 3.1 gives an overview of the (negative) responses given.
24 http://www.fembusiness.nl/
25 These AEX and AMX listed companies are also considered to be large sized companies. For unknown reasons
these companies are not recorded in the FEM Business list. However, these companies are of interest for this study and are therefore added to the sample.
26 According to the definition of the European Commission, a large company is any company that has over 250
headcounts and a turnover of more than 50 million Euros or total assets of over 43 million Euros (European Commission, 2005)
27 The companies have been contacted by email: see the appendix for the content of this email (text box 1). 28 This was done when this amount was not specifically mentioned in the annual report or CSR report (which
Table 3.1: Reasons Not to Answer My Request
Reason Number of
Companies
Percentage Data not centrally registered 14 21.88 %
Data not publicly available 32 50 %
Did not want to cooperate 3 4.69 %
Other 15 23.44 %
Total 64 100 %
The companies that did or could not answer the request were immediately removed from the research. 16 companies never responded the request, so it was unable to include those companies in the sample as well.In the end, the sample consists of 44 companies.
The fact that not every company includes the expenses on philanthropic causes in its annual report or CSR report, introduces a selection bias to my research. No random sample selection of all companies in the Netherland can be made, so a distortion of the data may arise and the research is limited in this way. As this selection bias is known beforehand, it is taken into account while performing the research.
3.2 Description Dependent Variable
In many articles (Adams and Hardwick, 1998; Brammer and Millington, 2006) the natural logarithm of all variables is taken. This enhances the interpretation possibilities of the research, as the variables can be interpreted as elasticities and heteroskedasticity is tried to be eliminated. For this research, however, the dependent variable sometimes takes the value 0, which makes it impossible to take the natural logarithm. However, this problem may be evaded by taking the natural logarithm of corporate philanthropy plus 1, i.e. ln(1 + x) 29. Thus, for this research the dependent variable is expressed by ln(PHIL + 1). From now on (PHIL + 1) will be denoted by PHIL′.
3.3 Description Independent Variables
Firm size
Firm size is measured by the total assets of a company. This measurement was chosen, as it is the most common measure in previous research (Brammer and Millington, 2006; Adams and Hardwick, 1998). The annual reports of 2006 for the 44 companies in this sample were used to obtain the total assets of each company. The variable firm size remains in nominal value.
Board composition
Board diversity is measured by both the proportion of outside directors on the top management team as well as by the proportion of women on the top management team (TMT). TMT here is defined by the number of members on the executive board and the number of members on the supervisory board together.
The first variable is measured by the proportion of outside managers on the TMT. To obtain this proportion, first the number of outsiders is determined. This number is calculated by subtracting the number of quasi-insiders on the supervisory board from the total number of members on the supervisory board, as all members that are not qualified as quasi-insiders are thought to be outsiders. Then this number is divided by the total members on the TMT, which results in the proportion of outside directors on the TMT. The variable for the proportion of outsiders remains in nominal value.
The women variable is measured by the proportion of women on the top management team (TMT). The proportion of women on the TMT is calculated by dividing the total number of women on the executive board and the supervisory by the total number of members on the TMT. The variable for the proportion of women remains in nominal value.
Again, all information with regard to the composition of the board composition (both the number of insiders, outsiders and quasi-insiders and the number of women) were obtained from the annual reports of 2006 for the 44 companies in this sample.
Industry
The industry variable is determined using the BIK codes provided by the Dutch Chamber of Commerce (Kamer van Koophandel, 2005). This list comprises of all possible activities a company can undertake. The companies in this sample are coded on the basis of their main/prime activity (defined as the activity that represents the largest percentage of sales). The primary sector consists of all companies with a code between 0100 and 1450; the secondary sector consists of the companies that are coded between 1500 and 3720, and between 4500 and 4550; and the tertiary sector contains companies with a code between 4000 and 4100, and between 5000 and 990030.
Every company is classified into one of the sectors listed above, based on their BIK code. The number of companies per sector differ significantly, which implies that, for this sample, the Netherlands is better represented (in relatively large companies) in some sectors of the economy than in others. For this sample, there are three companies in the primary sector, seventeen companies in the secondary sector, and twenty-four companies in the tertiary sector.
As this research has three sectors, the research will contain two dummy variables31. This suggests that when a company is present in a certain sector, this sector dummy is marked by a 1, while the other sector dummies are denoted by a 0. For this research, the secondary sector is used as the reference group (the other sectors will be analyzed in comparison to the
30 The quaternary sector, with BIK codes ranging from 7200 to 7260, is left out of consideration in this research
for practical reasons, as there was only one company in this sample that qualified for this sector. The quaternary sector is therefore added to the tertiary sector, as both sectors are service sectors.
31 It is not possible to include a dummy variable for all three sectors, as then exact collinearity would exist.
secondary sector) and this sector is therefore omitted from the research32. The industry effects are only incorporated in model 2.
3.4 Description Research
In order to find an answer to the hypotheses posed in this research, two different models will be used – one excluding industry effects and one including industry effects. Two models enhance the possibility to check for robustness, as when both models show more or less the same results, the model is said to be robust as it remains effective under different conditions. The hypotheses can be researched simultaneously using a multiple regression analysis33. The multiple regression model has some general assumptions with which this research model has to comply. These general assumptions can be found in figure 3.1. When MR1 till MR5 of these general assumptions hold the least square estimators of βK (i.e., bK) are the Best Linear
Unbiased Estimators (BLUE)34, also known as the Gauss-Markov theorem. When MR6 also holds, we can state that the dependent variable is also a normally distributed random variable (Hill et al., 2001). However, MR6 is not needed for the least square estimators to be BLUE.
Figure 3.1: General Assumptions of Multiple Regression Model
Source: Hill et al., 2001
MR6. yt ~ N [ (β0 + β1xt1 + … + βKxtK), σ2 ] ↔ εt ~ N (0, σ2 )
MR4. cov(yt, ys) = cov(εt, εs ) = 0
MR5. The values of xtK are not random and are not exact linear functions of the other explanatory variables.
MR2. E(yt) = β0 + β1xt1 + … + βKxtK ↔ E(εt) = 0
MR3. var(yt) = var(εt, εs ) = σ2
MR1. yt = β0 + β1xt1 + … + βKxtK + εt, t = 1, ……., T
32 This means that there is no variable for the secondary sector in the second model, as the effect of the
secondary sector is measured by the constant term.
33 This multiple regression analysis will be conducted in the statistical program Eviews 3.0. 34 This means that they have the smallest variance of all linear and unbiased estimators of β
K. The estimators are
3.4.1 Model 1
For the first model the regression analysis is estimated using the following equation:
ln(PHIL′ ) = β0 + β1SIZE + β2OUTS + β3WOMEN + ε (1.1)
where PHIL is the level of corporate philanthropic expenditures, SIZE is the firm size measured by the level of total assets, OUTS is the proportion of outside managers on the top management team, WOMEN is the proportion of women on the top management team, and ε is the error term. In this model, it is assumed that βK will be given by a parameter bK, based on
the least squares principle.
The model is expressed as a semi-log model, implying that the dependent variable is expressed in natural logarithm, and the independent variables remain linear. The interpretation of this model is as follows: the coefficient βK of any independent variable represents the
percentual effect of a unit change of that independent variable on the dependent variable. More specifically, β1 measures the percentage change in PHIL that follows from a unit change
in SIZE35. This response of PHIL is identical for companies of all sizes, identical for all proportions of outsiders and for all proportions of women on the TMT.
Using a multiple regression model means that the independent variables are regressed on the dependent variable. All statistical assumptions and rules will be checked and considered in order to safeguard the statistical value of the research. The signs of the parameters and the corresponding significance levels will then be analysed and discussed in order to detect their relationship with the dependent variable.
35 The fact that the dependent variable PHIL was changed to (PHIL+1) does not change the interpretation of the
3.4.2 Model 2
The second model concerns the industry effect and is an extension of the first model. The equation for philanthropic expenditures therefore changes to the following:
ln(PHIL′ )= β0 + β1SIZE + β2OUTS + β3WOMEN + δ1S1 + δ2 S3 + ε (1.2)
where again PHIL is the level of corporate philanthropic expenditures, SIZE is the firm size measured by the level of total assets, OUTS is the proportion of outside managers on the top management team, WOMEN is the proportion of women on the top management team, and Si
is a dummy variable for sector (where i stands for either primary, secondary or tertiary) and ε is the error term. Again, it is assumed that, for the independent variables, βK will be given by
a parameter bK. For the dummy variable Si, δi will be given by a parameter di, based on the
least squares principle. The general assumptions regarding the multiple regression model are the same as mentioned above (figure 3.1).
The parameter di measures the expected corporate philanthropy expenditure differential
between companies that are in sector i and companies in the reference group (sector). One of the sectors is omitted from the research, which is the reference group, and all results will be related to this group. The coefficients of the dummy variables represent the expected corporate philanthropy expenditure differentials relative to the reference group. For this research, the secondary sector (S2) will be used as the reference group.
The following equation (1.3) represents the expected values of corporate philanthropic expenditure for each sector:
Again, the model is expressed as a semi-log model, so the dependent variable is expressed in natural logarithm, and the independent variables remain linear. The interpretation of the independent variables SIZE, OUTS, and WOMEN is the same as for the first model. The interpretation of the dummy variables is as follows: the dummy variable δi represents the
differential in the level of corporate philanthropy for sector i relative to the secondary sector. When this variable is positive, sector i spends relatively more on corporate philanthropy than the secondary sector, while when this variable is negative, sector i spends relatively less than the secondary sector. Again, this response of PHIL is identical for companies of all sizes, identical for all proportions of outsiders and for all proportions of women on the TMT, and identical for all companies in either the primary or the tertiary sector.
4 Data Analysis
In this chapter the data from the sample will be analyzed using a multiple regression analysis and the results will be discussed. Also, the hypotheses will be tested in order to be able to either confirm or reject the proposed relationships. First model 1 will be discussed, including hypothesis 1 till 3. Thereafter, the second model will be considered, which contains the industry effects so as to test for hypotheses 4a and 4b.
4.1 Model 1
4.1.1 Descriptive Analysis
The philanthropic expenditures equation is estimated by least square estimators and some extra diagnostic tests were performed. Table 4.1 shows the descriptive statistics of the sample for the first model, equation (1.1):
Table 4.1: Descriptive Statistics – Model 1 (unlogged variables) Mean Standard Deviation PHIL (× 1,000 Eur) 10,560.99 27,590.75 SIZE (× 1,000,000 Eur) 101,502.0 268,202.6 OUTS 0.582533 0.153292 WOMEN 0.052570 0.064603
The average percentage of outside directors on the top management team is 58.25 %, with a minimum of 0 % outsiders and a maximum of 100 % outsiders on the TMT. It can be assumed that the top management team that contains no outsiders at all (i.e., only quasi-insiders) is not listed on the Dutch stock exchange, since it a best practice provision of the Dutch Corporate Governance Code36 that there is a maximum of one quasi-insider on the supervisory board.
The average percentage of women on the TMT is 5.26 %, with a range from 0 % (lowest) to 21.43 % (highest). To be exact, 22 companies in the sample did not have any women on their TMT at all. This is quite remarkable and in far contrast with the minimum quota of 40 % recently introduced in Norway, which many countries have approved of and are trying to introduce in their home countries as well. It shows that the Netherlands has a long way to go when it comes to this subject.
4.1.2 Diagnostic Checks
To check for the assumptions of the multiple regression model, some diagnostic checks have to be performed. These diagnostic tests are used to ensure the statistical significance of the model and to find out whether the least square estimators are BLUE.
One of the assumptions is about the homoskedasticity of the errors (MR3), which implies that the variances for all observations should be the same. When this assumption is violated, the errors are said to be heteroskedastic37. Heteroskedasticity may be detected using White’s test38. From table 2 in the appendix we can see that the hypothesis of homoskedasticity can not be rejected39 and it is assumed that the errors are homoskedastic.
36 See chapter 2.2.2 for clear description of the best practice provision of the Dutch Corporate Governance Code. 37 Heteroskedasticity poses a problem to the interpretation of the model, as the standard errors of the least square
estimators of the parameters may be incorrect.
38 White’s test regresses the squared residuals from the regression model to the independent variables, the
cross-product of the independent variables and the squared value of the independent variables, so for this model it is: ε2 = β
0 + β1SIZE + β2 (SIZE)2 + β3 (SIZE×OUTS) + β4(SIZE×WOMEN)+ β5OUTS + β6(OUTS)2 + β7(OUTS
×WOMEN) + β8WOMEN + β9(WOMEN)2 + υ. From this regression model, R2×N (where N is the sample size)
is observed. When this value (also known as the LM statistic) is larger than the critical value, the null hypothesis has to be rejected.The null hypothesis to be tested states that the errors are homoskedastic.
39
In table 2 in the appendix we find that the result for the test is 6.249895, which is the sample size (N) times the R-squared. This LM statistic for White’s test follows a chi-squared distribution, so the critical value here is χ2
c = 9.488 (5% significance level with 4 degrees of freedom). The test statistic is smaller than the critical value
Another diagnostic test to be checked for is autocorrelation (MR4). Autocorrelation measures whether there is a linear relationship between two separate instances from the same random variable. When autocorrelation exists, the least square estimators are no longer BLUE, as one of the general assumptions of the multiple regression model fails to hold.
Autocorrelation can be detected using the Durbin-Watson statistic. The null hypothesis to be tested is that there is no autocorrelation, against the alternative hypothesis that there is. In the absence of a program that calculates the p-value corresponding to the Durbin-Watson statistic, the bounds test is performed40. For this model, the critical value for the lower bound dLc is 1.391 and the critical value for the upper bound dUc is 1.600 (at a significance level of 5 %). From table 4.4 it can be seen that the Durbin-Watson statistic d for this model is 1.649, which exceeds the critical value for the upper bound dUc, implying that the null hypothesis cannot be rejected and that there is thus no autocorrelation.
Another important check is to test for multicollinearity (MR5). The existence of multicollinearity causes problems in interpreting the regression results, as its presence indicates a linear relationship between two or more independent variables and will make it more difficult to consider the individual effects of the independent variables on the dependent variable. When multicollinearity exists, the least square estimators are no longer BLUE, as one of the general assumptions of the multiple regression model fails to hold.
Table 4.2: Correlation Coefficients – Model 1
ln(PHIL′) SIZE OUTS WOMEN
ln(PHIL′) -
SIZE 0.485175 -
OUTS 0.074480 -0.050439 -
WOMEN 0.362352 0.101001 0.232275 -
40 The bounds test considers two other statistics d
U (the upper bound) and dL, (the lower bound) which depend on
N and K; N being the sample size and K being the number of parameters. When the Durbin-Watson statistic d is
smaller than dLc, the null hypothesis is rejected; when d is larger than dUc, the null hypothesis cannot be rejected;
The correlations in table 4.2 represent the correlations between the dependent and the independent variables, but also between the independent variables reciprocally. These correlations (between the independent variables SIZE, OUTS and WOMEN) show no indication of multicollinearity, as they are all smaller than .23241. An extra examination of multicollinearity can be done using the Variance Inflation Factor (VIF) Analysis42. For this measure, a VIF-value of 5 or higher is associated with multicollinearity. Table 4.3 shows the VIF-values for equation (1.1). It may be concluded from both multicollinearity tests that multicollinearity will not pose a problem on the interpretation of the regression results in this research.
Table 4.3: VIF-values Philanthropic Expenditure Equation Independent
Variable
Parameter R-squared VIF
SIZE b1 0.015974 1.0162
OUTS b2 0.059469 1.0632
WOMEN b3 0.066689 1.0715
A final diagnostic check is for the normality of errors, i.e., that the errors have a normal probability distribution (MR6). When this test cannot be confirmed, this has no implication for the least square estimators to no longer be BLUE, but it does imply that the dependent variable is not a normally distributed random variable.
41 A correlation coefficient of higher than .80 is determined as high and may be related to multicollinearity. 42 For the VIF analysis each independent variable is regressed onto the other independent variables, after which
To check for the normality of errors the Jarque-Bera test43 can be used. From figure 1 in the appendix, we can see that the Jarque-Bera test statistic for this model is 5.327. The 5 % critical value from a chi-squared distribution with 2 degrees of freedom is 5.99, so the JB test statistic is smaller than the critical value, implying that the null hypothesis can not be rejected and that the errors are normally distributed. As a result, the dependent variable is normally distributed. This implies that the dependent variable has the correct functional form.
In conclusion, it can be said, that the statistical value of the model is guaranteed. There is no heteroskedasticity, no multicollinearity, no autocorrelation, and the errors are said to be normally distributed, so the least square estimators are said to be BLUE and therefore the Gauss-Markov theorem holds.
4.1.3 Regression Results and Hypothesis Testing
The parameter estimates and the test statistics for the philanthropic expenditure equation are summarized in table 4.4. The R-squared indicates the predictive ability of the model44, which here is 0.25930. This means that the regression model explains around 25.93 % of the variation in philanthropic expenditures around its mean. The F-test is used to test the overall significance of the model45. The F-statistic has a value of 4.66766, and is significant with a p-value smaller than .01, which means that the null hypothesis can be rejected and that the relationship between the dependent and the independent variables is significant.
43 The Jarque-Bera test for normality is based on two measures: skewness and kurtosis. Skewness refers to the
symmetry of the errors around zero (perfect symmetry has a skewness of zero); kurtosis refers to the ‘peakedness’ of the distribution of the errors (normal distribution has a kurtosis of three). The Jarque-Bera test statistic is calculates as follows: JB = (N/6) × ( S2 + [k – 3]/4 ). When the JB-statistic is smaller than the critical
value (based on a chi-squared distribution), the null hypothesis of normally distributed errors cannot be rejected (Hill et al., 2001).
44 A R-squared of 1 means that all sample data fall exactly on the fitted least square line, so the closer R-squared
is to 1, the better the sample data fit the regression model.
45 The F-test includes all the explanatory variables, instead of testing these variables separately. The null
hypotheses being tested is that all the parameter estimates are equal to zero (H0: β1 = β2 = β3 = 0), against the
alternative that at least one is nonzero. The 5% critical value is Fc(3,40) = 2.84, so F > Fc [4.66766 > 2.84] so the
Table 4.4 also shows the parameter estimates of the independent variables. The estimate of the coefficient for SIZE has the expected positive sign and it is significant at the 5 % significance level in a one-tailed t-test46. The parameter estimate of WOMEN also has the expected positive sign and it is also significant at the 5 % level for a one-tailed t-test. However, the parameter estimate of OUTS is not significant, which implicates that no conclusions on the relationship between the proportion of outsiders on the TMT and corporate philanthropy can be drawn from this sample.
Table 4.4: Results of Multiple Regression Analysis – Model 1 Independent Variables Parameter Parameter
Estimate Standard Deviation t-statistic Constant b0 4.848890 *** 1.770837 2.73819 SIZE (Total Assets) b1 0.0000052 *** 0.00000017 3.08334 OUTS (Proportion outsiders on TMT) b2 0.250301 3.000184 0.08343 WOMEN (Proportion women on TMT) b3 12.29602 ** 7.146525 1.72056 Test statistics N 44 R2 0.25930 F-statistic 4.66766 *** Durbin-Watson statistic d 1.649059
Note: The table shows the LSE parameter estimates for the philanthropic expenditures equation using data from a sample of 44 Dutch companies for the year 2006; * p < .10; ** p < .05; *** p < .01 for
one-tailed t-test.
46 For a one-tailed t-test the null hypothesis of a parameter equal to zero is tested against the alternative of a
The implications of the regression results for the hypotheses posed above will now be considered. Hypothesis 1 stated that there is a positive relationship between firm size and philanthropic expenditures. This hypothesis is supported in the regression analysis at a 5 % significance level, as the coefficient for SIZE is positive. It supports the view that larger firms engage in higher philanthropic expenditures than smaller firms. Therefore, it can be said that larger firms are more capable of spending money and goods on philanthropic causes than smaller companies. The parameter estimate should be interpreted as a semi-log relationship, so for every increase of one million Euros in assets47, a company is expected to spend 0.000517% more on charitable contributions (while disregarding the other two independent variables, i.e. to keep them at zero). And for every increase in assets of one billion Euros, a company is expected spend 0.517% more on philanthropic expenditures. These results suggest that firm size plays an important role in the levels of corporate philanthropic expenditures.
Hypothesis 2 stated that the relationship between corporate philanthropy and the number of outside directors on the top management team is positive. The parameter estimate of OUTS has a positive sign, but was not found to be significant at the 10 % level. Therefore, it cannot be said that that the parameter estimate is significantly different from 0. This insignificance makes it impossible to interpret the results as they are inconclusive. The fact that the sample is rather small may have caused the problem of insignificance.
Hypothesis 3 stated that the relationship between the proportion of women on the TMT and corporate philanthropy is presumed to be positive. This assumption is supported by the findings of the regression analysis at the 5 % level, as the parameter estimate for WOMEN has a positive sign. The WOMEN coefficient should be interpreted as follows: every percent increase of women on the TMT48 is associated with an increase of 12.296 % on corporate donations,so with every increase of 1 % of women on the TMT the expected amount spent on charitable contributions increases with almost 12.30%. Therefore, it can be said that the proposed positive relationship between the proportion of women on the TMT and corporate philanthropy is confirmed.
47 It has to be kept in mind that the SIZE variable is expressed in millions of Euros, so a unit change in size is
equal to a one million Euros change.
48 Note that the WOMEN variable is expressed in percentages itself, making it possible to observe the percentual
In conclusion, the coefficient for SIZE and WOMEN are found to be significantly different from zero and have the expected positive effect on corporate philanthropic expenditures. The coefficient for OUTS has a positive sign, but can not be said to be significantly different from zero. The implications of these results will be discussed in chapter 5, as first the results for the second model will be discussed, including the effects of industry sector on corporate philanthropy.
4.2 Model 2
4.2.1 Descriptive Analysis
The equation for the second model, incorporating the effect of sector, is estimated by least square estimators and some extra diagnostic tests were performed. From table 4.5 the mean philanthropic expenditures of each sector can be read. These values indicate the average amount spend on corporate philanthropy per sector. It can be seen that these averages differ rather a lot. The lowest mean value of philanthropy is spent in the secondary sector: almost 8.5 million Euros, which is the average of seventeen companies, while the highest mean value is spent in the primary sector: over 37.5 million Euros, which is the average of three companies.In each sector the lowest amount spent by a company was zero Euros. The highest amount was spent in the tertiary sector, an amount of over 125 million Euros by just one company.
Table 4.5: Mean Philanthropic Expenditures – Model 2 (unlogged variables)
Sector Mean PHIL
(x1,000eur)
Standard Deviation N
1 37,866.67 64,206.33 3
2 8,442.65 20,215.88 17
