Corporate Social Performance,
Corporate Financial Performance and differences among industries:
evidence from the S&P 500
By: Sven Mol
Corporate Social Performance, Corporate Financial Performance and differences among
industries: evidence from the S&P 500
Author : Sven Mol
Date : 14 April 2011
Student number : 1434551
Study : MSc International Business & Management/
International Financial Management University : University of Groningen
Faculty : Economics and Business First Supervisor : Dr. J.H. von Eije
Second Supervisor : Dr. W. Westerman
Acknowledgements
I would like to express my gratitude to the following people who helped me during the
writing of this thesis. First, I would like to thank Dr. J.H. von Eije, who “let me be”, what I
really appreciated, but also provided me with insightful thoughts, constrictive feedback and
support whenever I needed, especially during the final stages of writing my thesis. Second, I
would like to thank Dr. W. Westerman for evaluating my thesis. Third, I would like to thank
Prof. Dr. H. van Hees, of the University of Groningen, for sharing the KLD data and Joost
Castelein, Investor Relations Analyst at KPN, for granting me access to the Bloomberg
Database. Last but not least, I thank my family and friends who supported me during the
writing of my thesis and my studies in general.
Abstract
In the last three decades, a lot of research has been conducted on the relationship between corporate social performance (CSP) and corporate financial performance (CFP). This has however resulted in inconclusive outcomes. On an industry level we examine the relationship between CSP and CFP. We will use KLD data to measure the independent variable CSP and use accounting based measures, ROA and ROE, to measure the dependent variable CFP. We use size, as measured by total sales and total number of employees; and risk, measured by leverage, as control variables. First, we find that the level of CSR is different for different industries. Moreover, we find a significant relationship between CSP and ROA and find an almost significant interaction-effect between sector and CSP on ROA.
We do not find a significant relationship when regarding ROE as the dependent variable. In addition, if CSP is categorized in low, medium and high, we find that companies with a low CSP have a significantly lower ROA than companies with a medium CSP and companies with a high CSP. We conclude that companies, and especially the companies in the IT industry that show low responsibility, could invest in CSR to attain a higher ROA.
Keywords: Corporate Social Responsibility, CSR, Corporate Social Performance, CSP, Financial
Performance, CFP, Industries.
Contents
Acknowledgements... 3
Abstract... 4
Contents... 5
1. Introduction ... 7
2. Literature Review ... 10
2.1 CSR and CSP... 10
2.2 The CSP-CFP Relationship ... 12
2.2.1 The shareholder approach and the stakeholder approach ... 12
2.2.2 Mediators... 13
2.2.3 Empirical Evidence ... 15
2.3 Differences in measurement... 16
2.3.1 Measuring CSP ... 16
2.3.2 Measuring CFP ... 19
3. Literature analysis ... 21
3.1 Analysis CSP-CFP literature ... 21
3.1 Results relationship... 22
3.2 Results CSP and CFP ... 23
4. Hypotheses... 25
4.1 CSP-CFP ... 25
4.2 Industry effects ... 25
5. Results ... 27
5.1 Inconclusive results... 27
5.2 Variables: CSP... 27
5.3 Variables: CFP... 28
5.4 Control variables ... 29
5.5 Sample and industries... 30
5.6 Methods ... 31
6. Results ... 32
6.1 Sample and variables ... 32
6.2 Descriptives... 32
6.3 Relationship between CSP and CFP ... 33
6.3.1 Results ... 33
6.3.2 Relationship between aspects of CSP and CFP ... 33
6.4.1 Results ... 35
6.5 Relationship between CSP, CFP and Industry ... 35
6.5.1 CSP and ROA/ROE ... 36
6.5.2 Results... 36
6.5.3 Industry effects ... 39
7. Conclusion and Discussion ... 40
8. Limitations and future research... 44
Appendix I: Relationship between CSP and CFP in 93 studies focused on US and published after 1990... 46
Appendix II: Number of CSP and CFP combinations in 92 studies focused on US and published after 1990 sorted by sort of relationship... 49
Appendix III: ANOVA ... 50
Appendix III: ANOVA ... 50
Appendix IV: Regression ... 52
Literature... 53
Internet and databases ... 66
1. Introduction
Corporate Social Responsibility (CSR) has gained momentum in the last decade. CSR is increasingly becoming a leading issue in worldwide business. A growing number of companies embrace the concept and feel the need to make clear what it actually means.
They take a variety of initiatives all aimed at making sense of CSR (Cramer, 2004). According to a research of The Economist among executives and investors on CSR 85 percent said that CSR is an “important” or “central” consideration in their investment decision. Compared to 5 years earlier this was only 44 percent (Heal, 2008).
From a theory perception it is not clear that engaging in CSR activities will lead to financial and economic benefits (Beurden and Gossling, 2008). And according to the news headlines this seems true; there seems to be a conflict between principled behaviour and profitable behaviour. In previous years we have seen mortgage companies enriching themselves by providing loans to credit unworthy families, the Enron scandal, the American automobile industry lobbying aggressively to allow continued emissions of greenhouse gases, agencies and investment bankers rating bad bonds as high quality etc. These examples suggest that companies think that acting in an unprincipled manner pays. However, now we have seen that the market has punished these companies. Due to the credit crisis that started in 2008 many banks got bankrupted. General Motors, till 2008 the biggest automobile manufacturer, was also close to bankrupt. On the other hand the now biggest and most profitable automobile manufactory is Toyota, which has been positioning itself as a green car company. And sure news headlines do not tell us the whole story. For example we do not see headlines that tell us that Starbucks reduces the environmental impact of the coffee production and that Nike is improving their labour conditions in third world countries (Heal, 2008).
There is indeed a temptation by certain companies to cut in their activities of CSR to attain
higher profits in the short run. However, there is evidence that there are forces that punish a
company that misbehaves and reward a company that behaves. Is engaging in CSR and
profitable behaviour in conflict or can a company both do well (socially and environmentally)
Quite a lot of research has been conducted on this relationship between corporate social performance (CSP) and corporate financial performance (CFP) (Griffin and Mahon (1997);
Margolis and Walsh (2001); Orlitzky et al. (2003); Wu (2006) and Beurden and Gössling (2008)). This has however resulted in inconclusive results and disagreement concerning terminological and methodological issues.
This study will investigate this possible causal relation between CSP and CFP. Despite the fact that various researchers found that the industry influences the level of CSR (Waddock and Graves, 1997) researchers have neglected to examine CSR and the CSP-CFP relationship on an industry level. This paper will respond to this need; this research will focus on the differences among industries.
The research contribution that arises will be three-fold. First, next to a regular literature review, we provide a comprehensive and up-to-date analysis of the literature, since the literature on the CSR-CFP is extensive and inconclusive. We found that recent (meta-) studies (e.g. Orlitzlky et al., 2003) often refer to studies before 1990. Because of the modern character of CSR, these studies could be outdated and therefore we will primarily focus on recent literature published after 1990. Second, since the literature does not provide consistent outcomes, we will contribute to the validation of a positive, a negative or no relationship between CSP and CFP. Third, because literature on the CSP-CFP for separate industries is lacking, we will identify differences in the level of CSR and differences in the CSR-CFP relationship for different industries.
The research question for our research is:
Does a relation exist between Corporate Social Performance (CSP) and Corporate Financial Performance (CFP) and does this relation differ among industries?
This thesis is organized as follows. First, in the literature review we will provide the
background and the literature of the topic. Second, in the literature analyses we will analyze
all existing literature on the CSP-CFP relationship. This section will be followed up by our
hypotheses. In addition, we describe our methodology, variables and sample. Last, we come
up with the results and we will finalize with the conclusion and limitations.
2. Literature Review
2.1 CSR and CSP
The concept of CSR has been developing over time. It first emerged in the thirties and forties of the last century. This was the time that “Fortune” magazine first asked businessmen about their social responsibilities (Carroll, 1999). However, it was only in 1953 that Bowen published its book about the social responsibilities of businessmen. Carroll states that Bowen is the “Father of Corporate Social Responsibility”. The introduction of this book launched the modern period of literature on CSR. Bowen states that “the social responsibility of businessmen refers to the obligation of businessmen to pursue those policies, to make those decisions, or to follow those lines of action which are desirable in terms of the objectives and values of our society.”
Nowadays CSR is broadly adopted by companies. However, there is still no unambiguous definition of the concept of CSR (Frederick, 1994; Griffin 2000). In investigating the CSP-CFP relationship it is important to have a clear understanding of the concept of CSR. Wood’s (1991) definition of CSR is frequently used in the field: “CSR is a business organization’s configuration of principles of social responsibility, processes of social responsiveness, and policies, programs, and observable outcomes as they relate to the firm’s societal relationships”. Another often cited definition is the one of Mc Williams and Siegel (2001), they define CSR as “actions that appear to further some social good, beyond the interests of the firm and that which is required by law”. Although there are many different definitions in the CSR literature most agree that “Corporate Social Responsibility is a responsive obligation to serve society” (Brammer and Millington, 2003). By “obligation” they mean that companies should pursue socially responsible activities. If they do not, negative consequences are likely to emerge.
One of the most important scholars in the field of CSR is Archie B. Carroll. He states that
there are four kinds of social responsibilities that “constitute total CSR: economic, legal,
ethical and philanthropic” (1991). These four layers of CSR are shown in Caroll’s pyramid.
Figure 1: The Pyramid of Corporate Social Responsibility (adopted from Carroll, 1991)
The economic responsibilities are shown in the lower layer of the pyramid. These
responsibilities are the most important for a company. Historically, companies are designed
to produce services and goods for the public and have only one goal: being profitable. This
layer is the foundation upon which other layers rest. The second layer represents the legal
responsibilities. Next to making profit, companies must comply with the law and regulations
of the government. Together with the economic responsibilities this layer is regarded as the
fundamental responsibilities for a company and is mandatory. The third layer characterizes
the ethical responsibilities. Ethical responsibilities “embody those standards, norms, or
expectations that reflect a concern for what consumers, employees, shareholders, and the
community regard as fair, just, or in keeping with the respect or protection of stakeholders'
moral rights”. Ethical responsibilities are considered harder to define than legal
responsibilities and economic responsibilities, but it definitely means that organizations
should go beyond law with their activities. These responsibilities are expected by society but
responsibilities entail “those corporate actions that are in response to society’s expectation that businesses be good citizens”. Philanthropic responsibilities refer to engaging in acts or programs to promote human welfare. These actions are not expected ethically or morally like ethical responsibilities, but the community desires them.
It is impossible to measure CSR, since CSR is not a variable. Corporate Social Performance (CSP) on the other hand is a manner of “making CSR applicable and putting it into practice”
(Maron, 2006). CSP, though also difficult to measure, can be translated into several measurable variables. In this study CSP is defined as a reference to a firm’s demonstration of several environmental and social measurable variables (Community, Corporate Governance, Diversity, Employee Relations, Environment, Human Rights and Product) as perceived by analysts of rating agency KLD.
2.2 The CSP-CFP Relationship
2.2.1 The shareholder approach and the stakeholder approach
Before testing the impact of CSP on CFP, the theoretical grounds that support a possible relation between CSP and CFP will be explained. In the field of CSR two different theories are addressed: the shareholder approach and the stakeholder approach.
The shareholder approach implies a negative relationship of the CSR-CFP Link. Quazi and O’Brien (2000) define the shareholder approach as the neo-classical view of CSR. The most famous supporter of the shareholder approach is Friedman (1970). He states that in a free society “there is one and only one social responsibility of business—to use its resources and engage in activities designed to increase its profits so long as it stays within the rules of the game, which is to say, engages in open and free competition without deception or fraud.”
According to Friedman, and for example Auperle et al. (1985), companies should therefore only focus on profit maximization. According to this idea engaging in CSR is “spending someone else’s money for general social interests” (Friedman, 1970). According to this approach governments are responsible for socially responsible activities, not the company.
They argue that if companies take money and resources from the company to address
socials problems, they spend money of the shareholder and that is not in his best interest.
An alternative view is the stakeholder approach. Freeman (1984) is regarded as the founder of this approach. The stakeholder approach states that an organization should not only take into account its shareholders but also be responsible for its stakeholders. Supporters of the stakeholder theory suggest that a “tension” exists between the company’s implicit costs to stakeholders (e.g.: environmental costs, product quality costs) and explicit costs (e.g.:
payment to bondholders). This theory argues that a company that tries to decrease its implicit costs by socially irresponsible actions will “as a result, incur higher explicit costs, resulting in competitive disadvantage” (Waddock and Graves, 1997).
To conclude, the neo-classical view advocates that any spending on CSR activities will lead to a competitive disadvantage and thus will result in a negative relationship between CSP and CFP. On the other hand, according to Ullman (1985) there is also a possibility that there is no CSR-CFP relationship, except maybe by chance. That is because there are so many variables that can possibly intervene the CSP-CFP relationship. Alternatively, the stakeholder theory provides a theoretical foundation for the CSP-CFP relationship: a higher CSP can lead to an improved relationship with stakeholders and therefore can lead to higher CFP. As stated however, engaging in CSR activities will lead to additional costs. Yet, these costs can be relatively small while the benefits that come with CSR activities can be potentially large (Moscowitz, 1972).
2.2.2 Mediators
Heal (2008) provides a comprehensive list of these benefits that work in the CSP-CFP relationship: (1) Risk Management, (2) Waste reduction, (3) Regulatory Protection, (4) Brand equity, (5) Employee productivity and (6) Cost of Capital.
First, one of the most important benefits that come with CSR is reducing the risks of
conflicts. Nowadays, companies cannot afford the risk of costly conflicts with stakeholders as
a consequence of their actions (e.g. NGO actions, bad press publications, boycotts or
lawsuits), as this will likely lead to a decrease of sales. For example, Shell suffered from a
customer boycott led by Greenpeace when it wanted to dispose the oil buoy Brent Spar in
the North Sea at the end of its life.
Second, by reducing waste certain companies can cut their costs and raise profits. For example, BP discovered that capturing associated gas, instead of releasing and burning it, facilitated its sale and transportation. This provided them with a new source of revenues that exceeded the costs of capture.
Third, a good relationship between company and an officialdom can be of significant importance. For instance, a reputation of being a “green” company can be of use for an oil company when negotiating for access to a potential oil field.
Fourth, for many companies “branding” is an important concept. Nowadays, with strong competition and small differences in differentiating products customers tend to buy products based on image or reputation. Waddock and Graves (1997) state that improvements in CSR can increase customer perceptions about the environmental awareness, quality of products and the reputation of the company. Because of the increased loyalty this will lead to reduced stakeholder management costs and improved sales.
Fifth, CSR can improve the relationship between a company and an employee, which may enhance employee productivity (Waddock and Graces, 1997). According to Heal (2008), people search for companies with a good and responsible image. As a result, companies with a good record have more success in attracting, retaining and motivating employees than companies with a poor record. For example, a survey of Montgomery and Rames (2003) of MBA attitudes toward potential employees noted that they are willing to receive less salary in order to work for companies with a more positive responsible image. In addition, Hill &
Knowlton’s Corporate Reputation Watch concluded in their survey that 88% of British companies think that in the future CSR will become more and more significant for recruiting and retaining employees (Adams, 2005).
Sixth, the right choices with regard to CSR can decrease a company’s cost of capital through
the impact of Socially Responsible Investments (SRI). In the United States, 10% of funds
under professional management are now regarded as SRI (Heal, 2008).
To conclude, it is expected that CSP has a positive influence on CSP due to the above- mentioned benefits. All of these benefits are likely to increase relationships with different stakeholders.
2.2.3 Empirical Evidence
In the last three decades, a lot of research has been conducted on the relationship between corporate social performance (CSP) and corporate financial performance (CFP).
Nevertheless, the results are mixed: some authors found a negative relationship, others found a positive and some found no relationship or a mixed relationship (for a comprehensive review of studies see Chapter 3). Therefore some authors argue that evidence regarding the CSP-CFP is inconclusive (Donaldson, 1999; Margolis and Walsh 2001, McWilliams and Siegel, 2001 and Wu, 2006).
To resolve this inconclusiveness, Griffin and Mahon (1997) and Margolis and Walsh (2001) integrated and researched several studies on the CFP-CSR relationship with a so-called
“vote-counting” technique. They cautiously conclude that a positive link between CSP and CFP exists. However, Orlitzky et al. (2003) criticize this method, since it simply codes studies as showing significantly positive, negative or neutral results and therefore conclusions drawn from these studies are likely to be false.
Consequently, in their own often-cited study they suggest another technique: meta-analysis.
This method corrects for measurement- and sample error and is therefore a more applicable
method. Instead of counting the researches that concluded a positive, negative or no
relationship, meta-analyses statistically aggregate the outcomes of the different studies
(Rosenthal and DiMatteo, 2001). In their meta-analyses of 52 articles, resulting in a sample
size of 33,878 observations, they found that “Corporate social performance and financial
performance are generally positively related across a wide variety of industry and study
contexts.” This study also indicates reputation and learning as mediating effects. They
however also received criticism. Bird et al. (2007) note that Orlitzky et al. utilize more than
30 different measures of CFP and a broad variety of CSP measures. Therefore the validity of
these measurements is questionable. Moreover, only 18 of the articles were published after
1990.
Beurden and Gössling (2008) state in their research that it is important to include only research published after 1990, since “The Brandtland Report” was published in 1987. This report “can be seen as a turning point in the attention of CSR”. According to them it is not likely that the consequences as well as organizational reactions are included in academic research prior to 1990. This is echoed by Roman et al. (1999) who state that studies published before 1990 can be utilised as an argument, but should not be considered as empirical legitimacy. Consequently, in their own research Beurden and Gössling use 31 studies published after 1990. They found a significant positive relationship in 23 (68%) of the cases, a significant negative relationship in 2 cases (6%) and they found no significant relationship in 6 cases (26%). Yet, they also used the “vote-counting” technique.
Wu (2006) also conducted meta-analyses of 121 empirical studies to research the relationship between CSP, CFP and Size. He also revealed a positive relation in the CSP-CFP link. However, several studies before 1990 were incorporated.
To conclude, evidence regarding the CSR-CFP relationship is debatable and conclusions have been inconsistent. According to some, this relationship has been one of the least understood and one of the most researched relationships in the business literature (Wood and Jones, 1995; Griffin and Mahon, 1997). Ruf et al. (2001) provide several reasons for this inconsistency: (1) a lack of theoretical foundation, (2) a lack of proper methodology (3) a mismatch between social and financial variables, (4) a lack of comprehensive systematic measure of CSP and (5) sample size and composition limitations. However, despite these reasons, the majority of the meta-analyses lean to a positive relationship of the CSP-CFP link.
2.3 Differences in measurement 2.3.1 Measuring CSP
It is vital to have clear definitions when researching the relationship between two variables, in this case CSP and CFP (Ulmann, 1985).
As noted in the previous paragraph, many researchers state that measurement problems (of
both CSP and CFP) contribute considerably to the inconsistent and diverse results when
researching the CSP-CFP relationship. Specifically, without having a general accepted measure of CSR it is hard for scholars to duplicate the results of previous researchers (Hammond et al, 1996). Therefore, when researching the CSP-CFP relationship it is important to take into account the differences in measurement of CSP (and CFP), since they can influence the direction of this relationship (Orlitzky et al, 2003; Wu 2006).
As mentioned earlier, according to Carroll (1991) CSR constitutes of four kind of social responsibilities: economic, legal, ethical and philanthropic”. Hence, CSR is a multidimensional concept and therefore difficult to measure (Waddock and Graves, 1997).
This is echoed by Orlitzky et al. (2003), they state that CSP (and CFP) are broad meta- constructs that can create measurement problems.
Beurden and Gössling (2008) acknowledge this problem of inconsistency. In an attempt to minimize this problem and to compare the research conclusions they categorize the measurements of CSP that are commonly used into three categories: (1) measurement based on social disclosure, (2) measurement based on corporate actions and (3) measurement based on corporate reputations.
First, measurement based on social disclosure entails “the extent of social disclosure about matters of social concern” (e.g. content analyses of annual reports). Second, measurement based on corporate actions. These actions refer to “concrete observable CSR processes and outcomes”. These actions involve social programs, philanthropy and pollution control. Third and last, measurement based on corporate reputation provided by rating agencies as KLD and Fortune. “These reputation ratings assume that CSP reputations are good reflections of underlying CSR values and behaviours.”
In general, in the first studies researchers have used a single measure of CSP (e.g. pollution) whereas it is now more standard to use a multidimensional measure such as ratings provided by KLD or fortune (Waddock and Graves, 1997).
In the studies of Beurden and Gössling (2008) 6% used social disclosure as a measurement
method, 35% used corporate actions and 59% used corporate reputation.
These measurement methods have several drawbacks. Measurement based on social disclosure (e.g. pollution) is, as being said, one-dimensional. Since CSP is a multidimensional concept, one might not be able to capture the whole concept by using this measurement.
Moreover, it is said that this kind of measurement can be biased through inclusion or omission, since social disclosures provide information about public relations and a company’s social orientation rather than corporate actions (Waddock and Graves, 1997).
Measurement based on corporate reputation is the most frequently used method.
According to Orlitzky (2003), CSP reputation indices are also more highly correlated with CFP than are other indicators of CSP. However, some researchers think that agencies that measure CSP based on corporate reputation do not offer a large enough sample for one industry. Moreover, some aspects of the assessment (e.g. Fortune) of companies are based on subjective elements, such as the opinions of senior executives. It is argued that measures as KLD are quite complete, but are affected by the subjective weighting of several sub- aspects (Griffin and Mahon, 1997).
Despite these arguments, Wood and Jones (1995) state that the KLD database is the “best- researched and most comprehensive” database for CSP. Moreover, Ruf et al. (2001) state that the KLD data are suitable for several reasons. First, these data reflect the worries traditionally held by social investors and consist of all aspects held important according to surveys among social fund managers for the time period investigated. Second, the KLD data distinguish different dimensions of CFP that results in a composite measure of CFP. Third, organizations are rated over time allowing researchers to investigate change in social performance. In their research regarding how well KLD measures the environmental variable of CSR, Chatterji et al. (2009) found that KLD’s “concern” ratings are quite good indicators of past environmental performance. In later years, companies with more KLD concerns receive more pollution and regulatory compliance violations. They did, however, find that KLD
“strengths” do not precisely predict compliance violations or pollution levels. In addition, in his research regarding the construct validity of the KLD rating, Sharfman (1996) argues that researchers studying CFP can be confident to use the KLD data. Furthermore, the KLD database is described as the “largest multidimensional CSP database available to the public”
(Deckop, 2006) and as the “de facto CSP research standard at the moment” (Waddock,
2003). Not only scholars use the data of KLD, according to KLD 15 of the top 25 institutional financial managers utilize KLD research (www.kld.com). To conclude, we can state that data provided by KLD is the most accurate data to measure CSP.
2.3.2 Measuring CFP
Previously conducted research shows that there are two different measurements of corporate financial performance: market-based measures and accounting-based measures (Orlitzky et al., 2003; Wu, 2006; Beurden and Gössling, 2008).
Market-based measures consist of market return, stock performance, share price appreciation, market value to book value and other market-based measures.
“Stock market participants determine a firm’s stock price and consequent market value, and then base their decisions on their perception of past, current, and future stock returns”
(Beurden and Gössling, 2008). This can be influenced by CSP.
Accounting-based measures refer to profitability measures, asset utilization, such as return asset (ROA) and asset turnover, return on sales (ROS), return on Equity (ROE), earnings per share (EPS) and growth (Wu, 2006). These measures use data from the income statement and balance sheet (book values). These measures refer to historical performance and reflect the internal efficiency of a company, which can be influenced by CSP.
Both measures have their own advantages and therefore authors are divided on which one
to use. According to Orlitzky et al. (2003) and Wu (2006), CSP seems to have a higher
correlation with CFP measured by accounting based measures (such as asset utilization,
profitability measurements and growth) than CFP measured my market-based measures
(such as market return, stock performance and share price appreciation). Therefore, Wu
(2006) states that CFP measured by accounting measures are a better predictor of CSP than
CFP measured by market-based measurements. Nevertheless, these measures can be
subject to different accounting treatments and managerial manipulation. Moreover, these
measures focus on internal decision-making rather than responses from stakeholders to the
activities of the company. Therefore, it can be difficult to compare results across different
companies and studies (Moore, 2001).
Davidson and Worrell (1990) rather use market-based measurements. According to them it is almost not feasible to separate activities related to CSR. Moreover, these measurements are more connected to shareholders’ wealth. Shareholders are only interested in accounting-based measures when it actually has influence on the shareholders’ wealth.
According to Griffin et al. (1997) however, market-based measures may hold more than just
a company’s financial performance. For that reason, accounting-based measures may be
more applicable than market-based measures. In our research accounting-based measures
will be used.
3. Literature analysis
3.1 Analysis CSP-CFP literature
As noted, the literature regarding the CSP-CFP relationship is extensive and inconclusive, therefore in this section we provide a literature analysis of all the published literature after 1990. This is a collection of the studies used in the researches of the earlier mentioned studies: Griffin and Mahon (1997), Margolis and Walsh (2001), Orlitzky et al. (2003) Wu (2006), Beurden and Gössling (2008) and we added additional literature published from 2008 till 2011. In addition to the analyses of the (meta-) studies in previous paragraphs, our analyses will result in an up-to-date overview of all published literature regarding the CSP- CFP relationship. This is needed because the latest literature review of Beurden and Gössling (2008) does not incorporate all available relevant studies. Furthermore, literature of the last three years is not included.
Despite its drawbacks a “vote-counting” technique will be used to analyse these studies.
Meta-analysis, like Orlitzky (2003), that statistically aggregates results of prior researches, is beyond this paper (since its primary goal is to indicate differences in the CSP-CFP relationship among industries).
Note that some conditions will be met. Only literature that was published after 1990 is added. Furthermore, only studies that focus on US companies will be used, since this research will only focus on the S&P 500. Thereby, case studies were excluded. Also, a few studies have not been incorporated since they were not available in Ebsco Host.
First, to find relevant literature published after 2008 different search strings in Ebsco Host
have been entered: CSP, CSR, Corporate Social Performance and Corporate Social
Responsibility; CFP, Corporate Financial Performance and Financial Performance. Second,
the literature lists of the found studies were manually examined in order to find more
appropriate literature.
3.1 Results relationship
Between 1990 and 2010, 93 studies meet the previously described conditions and empirically study the CSP-CFP relationship. 51 Studies (55%) found a positive CSP-CRP relationship, whereas 20 studies (21%) point to a non-significant relationship. Note that several studies in fact found a positive link but that this link was considered not significant due to methodological issues. Only 8 (9%) significant negative relationships were reported, while 14 (15%) studies found a mixed relationship. Figure 2 show these results. Appendix I show the results with the different articles.
Figure 2: CSP-CFP Relationship
These results are in line with the results of the latest published analysis of Beurden and Gössling (2008) (68% a significant positive relationship; 26% no significant relationship and 6% a significant negative relationship). However, we found less studies with a positive relation. This could be due to the better methodology used nowadays and for example due to the better systematic measure of CSP. The data of KLD to measure CSP is far more used nowadays than in the early nineties. Moreover, Beurden and Gössling also did not find or add mixed relationships.
To conclude, these results (still) show an inconclusive CSP-CFP relationship. Nevertheless a majority of 55% of the studies show a positive significant relationship. Note that, like the
55%
21%
9%
15%
CSP-CFP Relationship
Positive relationship
Non-significant relationship
Negative Relationship
Mixed Relationship
research of Beurden and Gössling (2008), more studies did find a positive relationship, but that this was not significant.
3.2 Results CSP and CFP
We already noted in paragraph 2.3 that researchers use different measures to examine both CSP and CFP. This leads to different measurement problems. In this paragraph an analysis regarding the different CFP and CSP combinations is provided. Table 1 shows the number of CSP and CFP combinations found in the literature analysis (In appendix II we show the same information but sorted by sort of relationship).
Table 1: Number of CSP and CFP combinations in 93 studies focused on US and published after 1990
CFP 1 CFP 2 CFP 3 CFP 1 and 2 Total
CSP 1 4 1 0 3 8
CSP 2 10 7 2 2 21
CSP 3 31 14 0 19 64
Total 45 22 2 24 93
CSP 1: Based on social disclosure; CSP 2: based on corporate actions and CSP 3: based on corporate reputations. CFP 1: market-based measurements; CFP 2: accounting-based measurements and CFP 3:
other measurements (categorization adopted from Beurden and Gössling, 2008).
As can be seen from the table, the category CSP 3 (measurement based on corporate reputations) is used the most, in 64 of 93 studies (70%). From the CFP categories the majority of 45 studies make use of the category CFP 1 (market-based measurements) (48%).
However a combination of the two categories (CFP 1 and CFP 2) is also often used (26%).
This also applies to the CFP 2 category (accounting-based measurements)(24%). This analysis
is in line as we had previously concluded: there seems to be no conclusive manner of
measuring CFP (see 2.3.2). However, with regard to CSP it can be concluded that the
majority of the researchers nowadays (especially in the last few years) choose for the CSP 3
category to measure CSP. Not surprisingly, since we already concluded earlier that KLD Data
is the best suitable data to measure CSP. In contrast with for example Orlitzky’s (2003) meta-
analyses, the CSP 3 category were far less often used to measure CSP. This is probably
because of the fact that Orlitzky’s (2003) meta-analysis incorporated studies prior to 1990
and KLD only started publishing its data since 1991. To conclude, even if we categorize CSP
scholars concerning the methodology that should be used to study the CSP-CFP relationship
and the conceptualization of these variables.
4. Hypotheses
4.1 CSP-CFP
Our literature analysis shows that 55% of the investigated papers found a positive relation between CSP and CFP. Therefore, we expect a positive relation between CSP and CFP.
Hypothesis 1: There is a positive relationship between CSP and CFP.
This hypothesis is tested against the zero hypothesis of no effect of CSP on CFP. We will first use a composite measure of CSP, but we will also test the relation between separate CSP aspects with CFP.
4.2 Industry effects
In the CSP-CFP relationship, Industry has often been explained as a moderating variable.
Industries vary in the way they handle their environment. Chand (2006) states that industries operate in various contexts and have to deal with diverse financial, social and environmental issues.
Many studies investigating the CSP-CFP link have focused on multi-industry data (hypothesis 1). For example, of the 51 analysed researches Griffin and Mahon (1997) found 40 studies that focused on multiple industries. This is “despite numerous suggestions that future work on this topic needs to be conducted within specific industries” (Wokutch and Spencer, 1987).
Although researchers stress the importance of the research of CSR on an industry level, most
researches use industry only to control for. Griffin and Mahon (1997) state that "industries
exhibit special uniqueness in that the internal competencies or external pressures inherent
in the industry create a "specialization of social interests"". By researching the CSP-CFP link
on multiple industries, results may not incorporate individual differences based on the
specific context of an industry. Already in the seventies Sturdivant and Ginter (1977) noticed
the importance of the industry when researching CSR. This is in accordance with Cottrill
(1990), who found that important differences exist in CSR between industries and suggests
that future research should focus on different industries. In their study of the banking
industry, Simpson and Kohers (2002) state that differences between industries regarding
CSR are so enormous that research in the CSR area should focus on one industry. This is also
accuracy and validity. Moreover, Moore (2001), who focuses on the supermarket industry, states that comparing the CSP from an oil company, where employee security and environmental issues are more important, with a high street retailer makes no sense.
Furthermore, Amato and Amato (2007) state that industry characteristics are a key determinant for the amount of corporate giving.
To conclude, Griffin and Mahon (1997) give several reasons for focusing on specific industries: (1) the uniqueness of the internal competencies or the external forces that are inherent in an industry, (2) the degree of public visibility and (3) the different stakeholders each industries has and their different reactions to particular issues.
When studying the CSP-CFP relationship different results are to be expected among various industries. Before testing whether the relationship between CSP and CFP differs among industries, we first test whether there are any differences between the levels of CSR among industries. This leads to the following hypothesis:
Hypothesis 2: There are differences in the CSP ratings among industries.
This second hypothesis will be tested against the zero hypothesis that there are no differences in the CSP ratings among industries.
Though some research suggests that there are differences among industries, research that empirically investigates differences in the CSP-CFP relationship among industries is lacking.
Therefore, the next hypothesis is stated as:
Hypothesis 3: The relation between CSP and CFP differs among industries.
Also this hypothesis will be tested against the zero hypothesis that there are no differences
in the relationship between CSP and CFP in industries.
5. Results
5.1 Inconclusive results
As previously described, research on the CSP-CFP relationship is inconclusive. Ruf et al.
(2001) provide several reasons for this inconsistency: (1) a lack of theoretical foundation, (2) a lack of proper methodology, (3) a mismatch between social and financial variables, (4) a lack of comprehensive systematic measure of CSP and (5) sample size and composition limitations. Our methodology will try to take into account all 5 aspects. The first issue is addressed by providing a theoretical foundation of the CSP-CFP relation in the literature section, the stakeholder theory. All four other aspects will be taken into account and/or will be further explained in this methodology section.
5.2 Variables: CSP
As shown in section 2.3.1, there are a different number of measures how to measure CSP. To measure CSP we already concluded in section 2.3.1 that data of KLD is the best data around.
Therefore, for our research we use the data of KLD that falls in the CSP 3 category. Our sample includes CSP data of S&P 500 companies of the year 2002. Unfortunately for us, due to the high price newer KLD data was not available. This is a drawback since CSR is a modern theme and stakeholders nowadays could react differently on CSR now than nine years ago.
KLD is a rating agency that provides “Socrates”. This is a “research database, measuring the social and environmental performance of corporations. This database “contains social, environmental, and governance research on more than 4,000 companies in 50+ global markets” (http://www.kld.com). It includes quantitative measures of over 90 environmental and social indicators that are clustered into eight aspects: Community, Corporate Governance, Diversity, Employee Relations, Environment, Human Rights, Product and Exclusionary Screens (Alcohol, Gambling, Firearms, Military, Nuclear Power and Tobacco).
Each of these categories relates to one of six main stakeholders. With community, employee
relations and environment each representing one of these stakeholders, Corporate
Governance corresponds to shareholders, Product to customers and Human Rights and
Diversity to society and employees.
Each of these categories is reviewed per firm and each of these firms is assigned a “concern”
(e.g.: “high compensation” is grouped under corporate governance) or a “strength” (e.g.:
“strong union relations” under employee relations) on each of these categories.
To calculate the CSP score for each category, a score of -1 will be added for each “concern”
and a score of +1 for each “strength”, which will be summed up and will result in a CSP Score. A calculated weighted average of this CSP score of the seven aspects will be combined and this will result in an aggregate average for each company to represent its level of CSP.
Instead of assigning each category the same weighting, the weightings will be used that resemble with the one developed by Waddock and Graves (1997). They use evaluations provided by an expert CSR panel. The following weightings will be used:
Table 2: Weightings of CSP Score Weight Aspects
0,132 Community
0,118 Corporate Governance 0,120 Diversity
0,152 Employee Relations 0,126 Environment 0,130 Human rights 0,138 Product
0,084 Exclusionary Screens
This will eventually result in one definite aggregate CSP score per company. These aggregate CSP scores will be used since this will facilitate the comparison of the different outcomes per industry. If the categories would be separately linked with CFP than this would have resulted in too much outcomes and would be too difficult to compare.
Since the data of KLD will be used, one of the lacks Ruf et al (2001) pointed out, the lack of comprehensive systematic measure, will be taken care of. KLD data is secondary data from an independent and qualified organization and focuses on a wide spread of categories.
Moreover, it is tested on validity by Sharfman (1996).
5.3 Variables: CFP
As described in section 2.3.2 there are several ways to measure CFP: Market-based
measures and accounting-based measures. As previously mentioned, each of these
measures has various advantages and disadvantages. For this research accounting-based
measures will be used: (1) return on assets (ROA) and (2) return on equity (ROE). This falls in the CFP 2 category.
ROA is a ratio that focuses on the profitability of company assets in realising revenue. The ROA is calculated as net Income divided by total assets.
ROE is a ratio that focuses on the profitability of a company compared to its equity. The ROE is calculated as net income divided by total equity.
An issue indicated by Ruf (2001) with respect to measuring CFP is the timing difference between the introduction of a CSR practice and its possible financial benefit. The stakeholder theory provides a foundation with regard to investigating the relationships between the company and society, and thereby provides a theoretical understanding of the CSP-CFP relationship, but it does not take into account this timing difference. For this study it is assumed that the introduction of improvements of CSR take a certain time to have effect on CFP. As indicated CSP data of year 2002 will be used and therefore a lack of 1 year financial data of the following year 2003 will be used. This is in line with research of Waddock and Graves (1997). The financial data with regard to ROA and ROE of the year 2003 will be extracted from Bloomberg.
5.4 Control variables
Waddocks & Graves (1997) explain in their research that there are three control variables that need to be taken into account when examining the CSP-CFP relationship: size, industry and risk.
Meta-studies of Orlitzky et al. (2003), Wu (2006) and Beurden and Gössling (2008) show that size is the most confounding factor in the CSP-CFP relationship. Wood and Jones (1995) state in their research that large companies give more money than small companies (e.g.
corporate philanthropy). Orlitzky (2001) however did not find empirical support for this
relationship, but half of the investigated studies of Beurden and Gössling (2008) did find a
significant effect of size on CSP. Yet, it is unclear what the effect of size on the CSP-CFP link
truly is. Anyway, for this study we will use size as a control variable, measured by total sales
and total number of employees of the year 2002.
The companies’ risk tolerance can be important since it controls the way the company handles CSR activities that have the potential to (1) incur present or future costs (e.g. to help to avoid fines pollution control equipment can be introduced), (2) elicit savings (e.g. a waste reduction or recycling reduction) and (3) “build (environmentally friendly firm) or destroy (perceived as unfriendly to certain types of people) markets” (Waddock and Graves, 1997).
Thus we will incorporate risk as a control variable, measured by the long term debt to total assets ratio (leverage) of the year 2002.
We already stated in the Hypotheses section that the variable industry could have an influence on the level of CSR and the CSP-CFP relationship. Adding the control variable industry would take these differences into account. However for obvious reasons, this research will not use industry as a control variable, as this research regards several industries as different samples.
By controlling for extraneous variables the methodological rigor, one of the lacks of Ruf et al.
(2001), will be improved. The data of the control variables will be extracted from Bloomburg.
5.5 Sample and industries
To investigate the CSP-CFP relationship the KLD CSP data of the year 2002 and the financial
data of the S&P500 companies of the year 2003 will be examined. The S&P 500 is an US
index of the prices of the 500 large-cap common stocks that are actively traded. KLD
provides data of the S&P 500 and focuses primarily on the US market, thus this could be
easily matched with the financial data. To analyse differences among industries, this sample
will be split up into different industries (Hypotheses 2 and 3). For this research The Global
Industry Classification Standard (GICS) will be used. This standard was developed by Morgan
Stanley Capital International (MSCI) and Standard and Poor’s (S&P) 500 in 1999 to provide
an accurate and standardized industry definition for the global financial community. The
GICS distinguishes 10 industries: Energy, Materials, Industrials, Consumer Discretionary,
Consumer Staples, Healthcare, Financials, Information Technology, Telecommunication
Services and Utilities. The industries speak mainly for themselves, except for the Consumer
Discretionary and Consumer Staples. These two industries both contain consumer services
and products. However, Consumer Discretionary is a cyclical industry. This means that this industry is sensitive to a business cycle, thus revenues are impacted by economic prosperity (e.g.: restaurants, hotels and automobile manufacturers). On the other hand, Consumer Staples is non-cyclical. The included companies do not produce services and goods that are deemed necessities and are not affected by an economic downturn.
The sample size will be determined by taking all matching companies from the KLD
“Socrates” database and the financial data of Bloomberg of the year 2002. For methodological sake, only industries that have more than 10 companies will be used.
One of the issues provided by Ruf et al. (2002), sample size and composition limitations, is now addressed by utilizing data of a fairly large number of companies and different industries. For this research we incorporate multiple industries to enhance external validity (Hypothesis 1). Then we focus on single industries for internal validity (Hypothesis 2 and 3).
5.6 Methods
The hypotheses are tested by running different linear regression analyses and analyses of variance (ANOVA) using the program SPSS 17.0. For our first hypothesis, where we test whether there is a positive relation between CSP and CFP, we execute a linear regression analysis. We also execute different linear regression analyses to unveil what aspects (Community, Corporate Governance, Diversity, Employee Relations, Environment, Human Rights, Product and Exclusionary Screens) of CSP have a possible positive relation with CFP.
For our second hypothesis another ANOVA analysis is executed and we test whether the CSP
scores differ among industries. Last, for our third hypothesis, we need to test whether the
CSP-CFP relation differs among industries. In other words, we assume that there is an
interaction effect between CSP and Industry. We categorize CSP into a three category
variable (low CSP, medium CSP and high CSP). We perform another ANOVA analysis where
we regard the industry*CSP (cat) interaction as independent variable and CFP as the
dependent variable. For all the models we will take into account the different control
variables.
6. Results
6.1 Sample and variables
A total of 9 industries and 302 companies remain in the sample after eliminating companies with missing financial data. The industry telecommunication services was excluded from the sample since there were only four companies included in that industry.
We use two variables to measure the dependent variable CFP: Return on Assets (ROA) and return on equity (ROE). These variables are uncorrelated (r= -.02, ns) and will be analyzed separately.
Our independent variables are CSP, Industry and CSP*Industry (interaction). For the dependent variable CFP we use ROA and ROE. We use different control variables where NE is the control variable number of employees, TS the control variable Total Sales (in million dollars) and LEV the control variable long term debt divided by total assets.
6.2 Descriptives
The table below shows the averages and standard deviations for all the variables in our model. Notable elements in this table are the large SD of ROE compared to ROA and the large SD of TS. The latter shows us the importance of incorporating this variable as a control variable.
Table 3: Descriptives for all variables
Mean Std
Deviation Minimum Maximum Median Valid N
CSP (2002) -.02 .40 -1.31 1.08 .00 302
ROA (2003) .05 .07 -.48 -038 -.04 302
ROE (2003) .08 .71 -11.34 1.59 .13 302
LEV (2002) 18.7 14.1 .00 67.31 17.03 302
TS (2002) 13.809 24.019 404.22 229616.00 6416.62 302
NE (2002) 48.76 99.15 740.00 1400000.00 22850.00 302
6.3 Relationship between CSP and CFP
To investigate whether there is a positive relationship between CSP and CFP we execute a linear regression model. The following equation is assumed:
= + + + + +
Where refers to company .
6.3.1 Results
The results shown in table 4 shows us that we indeed found a positive effect of CSP on ROA ( =.028, p<.01). The first row indicates the , whereas the second row is the significance level. We do not find this effect when regarding ROE as dependent variable. We do find a significant negative effect of leverage on ROE ( =-.007, p<.01).
Table 4: The regression coefficients of the independent variables for ROA and ROE for 2003
aa) Significance levels between the parentheses 6.3.2 Relationship between aspects of CSP and CFP
For our previous analysis we used an aggregate measure for CSP, but it might be useful to know if also separate aspects of CSP have an influence on CFP. As pointed out, KLD distinguishes CSP in different aspects: Community, Corporate Governance, Diversity, Employee Relations, Environment, Human Rights, Product and Exclusionary Screens (Alcohol, Gambling, Firearms, Military, Nuclear Power and Tobacco). We use these eight different categories and execute eight linear regression analyses, regarding ROA as the dependent variable. We find two aspects that have a positive significant effect on ROA:
Diversity ( =.007, p<.05) and Environment ( =.010, p<.05). We do not find a significant
ROA ROE
CSP (2002) .028 -.084
(.008) (.426)
LEV (2002) .000 -.007
(.411) (.026)
TS (2002) .000 .000
(.603) (.811)
NE (2002) .000 .000
(.742) (.684)
Table 5: The regression coefficients of the independent variables for ROA for 2003
aCOM CGOV DIV EMP ENV HUM PRO EXC
CSP (2002) .001 .001 .007 .002 .010 -.003 .005 .011
(.781) (.873) (.016) (.582) (.027) (.797) (.264) (.329)
LEV (2002) .000 .000 .000 .000 .000 .000 .000 .000
(.208) (.206) (.264) (.227) (.535) (.201) (.219) (.331)
TS (2002) .000 .000 .000 .000 .000 .000 .000 .000
(.815) (.728) (.998) (.805) (.293) (.797) (.536) (.748)
NE (2002) .000 .000 .000 .000 .000 .000 .000 .000
(.708) (.744) (.733) (.692) (.794) (.711) (.805) (.742)
a) Significance levels between the parentheses
6.4 Relationship between industry and CSP score
The CSP scores range from -1.31 to 1.08, with an average of CSP -.02. The energy and utilities industries have relatively low average CSP scores; information technology has a relatively high average CSP score. The standard deviations of the industries lie more or less between .30 and .45.
Table 6: Descriptives for CSP Scores per industry
Mean Std
Deviation Minimum Maximum Median Valid N
Information Technology .17 .37 -.51 1.08 .14 48
Health Care .09 .29 -.41 .90 .02 33
Consumer Staples .07 .46 -.78 1.02 .11 24
Financials .07 .28 -.51 .69 .01 45
Consumer Discretionary .03 .29 -.83 .75 .00 48
Materials -.09 .34 -.78 .53 .00 23
Industrials -.15 .46 -1.03 .83 -.09 41
Utilities -.34 .41 -1.31 .30 -.28 23
Energy -.4 .45 -1.25 .43 -.39 17
Total -.02 .40 -1.31 1.08 0.00 302
We see three ‘clusters’ of CSP-scores in: (1) Energy and Utilities score lower than all others,
(2) Materials and Industrials have medium CSP and (3) IT, Health Care, Consumer Staples,
Financials and Consumer Discretionary have relatively high CSP.
To test whether average CSP differ significantly between industries, we use univariate analysis of variance (ANOVA) with industry as the independent variable, and number of employees, total sales and leverage as control variables. CSP is the dependent variable.
6.4.1 Results
The results of the ANOVA analysis show a significant effect of industry on CSP (p<.001). This indicates that the average CSP are different for different industries. These results also suggest that it is relevant to distinguish different industries when examining the CSP-CFP relationship. Table 7 shows the summary of the outcomes of the ANOVA analysis. In appendix III we show the complete table.
Table 7: Sum of Squares for the independent variables with CSP
aCSP
Corrected Model 8.939
(.000)
Intercept .051
(.538)
LEV (2002) .002
(.897)
TS (2002) .041
(.581)
NE (2002) .099
(.389)
Industry 6.632
(.000)
Error 38.781
Corrected Total 47.720
a) Significance levels between the parentheses
6.5 Relationship between CSP, CFP and Industry
To investigate whether the relationship between average CSP-scores and CFP-scores differs between industries, we use two univariate analyses of variance (ANOVA) with ROE and ROA as dependent variables, and industry, CSP and the industry*CSP (cat) interaction as independent variables. The number of employees, total sales and leverage are used as control variables.
Hypothesis 3 is based on the presence of an interaction effect between CSP (continuous
transform our independent scale variable CSP into a three category variable: low CSP (lower than -.15), medium CSP (between -.15 and -.13) and high CSP (higher than -.13). The category boundaries are chosen to create an equal distribution of companies over categories; each category contains 33% of our companies.
Table 8: Distribution of companies by industry and CSP (categorical)
low medium high Total
Consumer Discretionary 13 21 14 48
Consumer Staples 8 4 12 24
Energy 13 2 2 17
Financials 9 19 17 45
Health Care 9 11 13 33
Industrials 18 14 9 41
Information Technology 7 16 25 48
Materials 7 9 7 23
Utilities 16 5 2 23
100 101 101 302
6.5.1 CSP and ROA/ROE
In this paragraph we describe an ANOVA with ROA and ROE as dependent variables, industry, CSP and the industry*CSP (cat) interaction as independent variables, and number of employees, total sales and leverage as control variables.
6.5.2 Results
The analysis, shown in table 9, with ROA as a dependent variable shows a significant effect of both industry (p<.001) and, in line with our regression analysis, CSP (cat) (p<.01). We also find an almost significant interaction effect between industry and CSP (cat) on ROA (p=.08).
The analysis with ROE as a dependent variable shows a significant effect of industry (p<.05) and the control variable Leverage (2002) (p<.001) but no significant effect of CSP (cat), and no significant interaction effect between industry and CSP on ROE 1 . Table 9 shows the summary of the outcomes from both the ANOVA for ROA and ROE. In appendix III we show the complete tables.
1
