Earnings management and Corporate Social Responsibility

(1)

Earnings management and Corporate

Social Responsibility

Abstract

This study examined whether a significant relation exists between earnings management (EM) and corporate social responsibility (CSR). Furthermore, the study identified whether investor protection moderates this relationship. Proxies for both accrual-based earning management (AEM) and real earnings management (REM) were used to test the hypotheses.

The contribution of this study to prior research is the extensive research period, 2002-2016. Another contribution is the inclusion of REM next to AEM in testing the moderating effect of investor protection. At last, where most of prior research used KLD data to measure CSR, this study contributes to prior research by making use of another CSR database, namely Thomson Reuters’ ASSET4 database.

Making use of multivariate regression analysis, this study found that CSR firms are less likely to be engaged in AEM and manipulation of operating cash flow, suggesting that firms make more discreet decisions when they are more concerned with socially responsible activities. However, opposite results were with regard to abnormal production costs, abnormal discretionary expense, and overall real activity manipulation. These findings align with the possibility that firms use CSR to mask their earnings management activities.

The obtained evidence with regard to the moderating effect of investor protection show that CSR firm, when located in countries with strong investor protection, are less likely to be engaged in AEM, but are more likely to be engaged in earnings management by manipulation abnormal discretionary expenses and overall real activity manipulation.

Patrick de Lange S3255891

MSc Accountancy

Supervisor: dr. T.A. Marra

Co-assessor: N. Hussain, PhD

Date: January 22, 2018

(2)

2

1. Introduction

Corporate Social Responsibility (CSR) has gained a lot of attention over the past few decades. Environmental issues are pushed into the mainstream because of global warming, emission trading and carbon tax issues (Yip, Van Straden, and Cahan, 2011). The election of Donald Trump as the president of the United States of America and his outspoken opinion on the climate issue for example, brought new fire to the subject. An article by B. Chakraborty (01 June 2017) posted on the website of FOX News about the withdrawal of the United States from the Paris Climate Agreement is just one example of this subject from a political point of view.

However, it is not just a political debate. Firms are also concerned about their reporting of CSR activity as stakeholders demand for greater transparency about the aspects of business (Kim, Park, and Wier, 2012). From the business point of view, an interesting article of M. Corbet in the Memphis Business Journal (27 April 2017) indicates the importance of CSR. This article mentions the suicide rate of Apple Inc.’s largest contract manufacturer and Foxconn employees falling from the company’s main production building in China, both events took place in 2010, which created some negative publicity for both companies. These types of events have contributed to the importance of CSR (Yip et al., 2011). Therefore, it may not be a surprise that reporting about CSR activity has gained the interest of companies. According to Jo and Kim (2008), more than half of the Fortune 1.000 companies regularly issue CSR reports. With the growing interest in CSR, independent agency such as Kinder Lydenberg and Domini, Bloomberg, and Thomson Reuters ASSET4 started to rate and rank organizations on their corporate social performance (CSP) (Ioannou and Serafeim, 2012).

With the rising importance of CSR, firms not only concern about their financial performance anymore. There are several reasons why firms spend more time and effort on CSR. For example, CSR might be used to lower the information asymmetry problem between managers and investors (Reverte, 2012). Another finding is that firms who have CSR disclosure are more transparent (Francis, LaFond, Olsson, and Schipper, 2005).

However, financial scandals in the US and Europe started a debate regarding the misuse of CSR strategies to cover up bad accounting practices (Gargouri, Shabou, and Francoeur, 2010). This statement is confirmed by other studies who found that certain firms try to compensate for poor organizational behavior (Prior, Surroca, and Tribó, 2008), or try to mask EM practices (Martinez-Ferrero, Banerjee, and García-Sánchez, 2016) with the help of CSR. Because of these findings, the relationship between CSR and EM became a topic of rising interest within academic research.

Prior research on the relationship between EM and CSR though also gave contrary findings and provided evidence that the risk of manipulating reported earnings is less likely when firms are CSR oriented (Kim et al., 2012; Scholtens and Kang, 2013). It is also mentioned that it is possible that a firm, in their CSR report, may only respond to the most powerful stakeholders (Buhr, 2002). Because of these findings it may be expected that a firm only focuses at the

(4)

4

aspects of CSR which are of interest of their most powerful stakeholders, instead of all dimensions. Finally, where prior research mostly focused on a one-way relationship between EM and CSR, Martinez-Ferrero, García-Sánchez, and Cuadrado-Ballesteros (2015) found evidence for a negative bidirectional relation when accrual measures are used for EM.

The inconsistency of findings with regard to the relationship between EM and CSR arise because of different EM proxies (Chih, Shen, and Kang 2008), different CSR dimensions (Trébucq and Russ, 2005), and other moderating variables. A possible important moderating variable has been mentioned by Leuz, Nanda, and Wysocki (2003) and is investor protection. They found that the use of EM differs between countries and that investor protection is an important factor in this. Countries with strong investor protection engage in less EM. This has also been found by Scholtens and Kang (2013) using Asian data.

This moderating variable has also been implemented in a study about the relationship between EM and CSR. In an international data research, Martinez-Ferrero et al. (2016) found that firms participate in CSR in order to avoid punishment because of their engagement in EM, particularly in countries with strong investor protection. In an earlier research, Martinez-Ferrero et al. (2015) found evidence for the existence of a negative bidirectional relationship between EM and CSR, which is even more significant in countries with strong investor protection due to investor pressure regarding CSR disclosure.

To extent the study by Martinez-Ferrero et al. (2015) and contribute to the knowledge about the relationship between EM and CSR and the moderating effect of the level of investor protection, the main research question of this paper will be:

To what extent does the level of investor protection moderate the associative relationship between EM and CSR?

1.1 Definitions

In order to increase the understandability of the variables and thereby the readability of this study, a few concepts will be defined in the next section. At first EM will be defined, followed up by CSR.

1.1.1 Earnings management

In his exploring article, Walker (2013) defines EM as follows:

‘The use of managerial discretion over (within GAAP) accounting choices, earnings reporting choices, and real economic decisions to influence how underlying economic events are reflected in one or more measures of earnings.’

This broad definition indicates that EM is not a bad thing in its origin, because the definition focuses on the legal possibility to make accounting choices and economic decisions. The possibility to manage earnings because of accounting principles’ flexibility is also mentioned

(5)

5

by Yip et al. (2011). Because of this, fraudulent accounting choices is not included in the definition. However, it has to be noted that fraud often is preceded by high levels of EM.

Another widely used definition of EM is formulated by Healy and Wahlen (1999). Following their definition ‘EM occurs when managers use judgement in financial reporting and in structuring transaction to alter financial reports to either mislead some stakeholders about the underlying economic performance of the company, or to influence contractual outcomes that depend on the reported accounting numbers.’

Similar to the definition by Walker (2013), Healy and Wahlen’s definition is also based on managerial decisions to alter financial reports and thereby misrepresent firm performance. Leuz et al. (2003) argue that a conflict of interest between insiders and outsiders of a firm give rise to the incentives to participate in EM.

As mentioned before, EM is not a bad thing by definition. In order to provide representative financial statements, managers are allowed to smooth out transitory components so accounting numbers provide a more realistic picture of the financial results of a firm. Danger arises when accounting results are managed intentionally for a managers’ own short-term benefits, rather than for the long-term interests of stakeholders (Healy and Wahlen, 1999). In this case earnings are managed to mislead stakeholders, which is considered to be unethical (Kaplan, 2001).

1.1.2 Corporate Social Responsibility and Corporate Social Performance

Defining CSR is not an easy task (Heltzer, 2011), and a lot of different definitions of CSR are being used in previous literature (Jo and Kim, 2008). This is mainly because an uncontested definition of CSR is not yet formulated. The first to define CSR has been Friedman (1970). The author defines CSR as conducting a business ‘in accordance with shareholders’ desires. These desires are being described as ‘making as much money as possible while conforming to the basic rules of society’. The most widely accepted definition is the one put forward by Carroll (1979) (Kim et al., 2012). In Carroll’s article, CSR is defined as; “the economic, legal, ethical, and discretionary expectations that society has of organizations at a given point in time”. The last definition of CSR which will be mentioned in this paper is the one of McWilliams and Siegel (2001). In their definition, CSR is about a firm’s actions to benefit the society, which are beyond the interests of the firm and are not required by any law. Based on these three definitions of CSR, and in accordance to many other researchers (e.g., Blowfield and Murray, 2008; Barnea and Rubin, 2005), it can be stated that CSR is every activity a firm undertakes and which go beyond what is legally required, in order to serve the society and environment as a whole.

Despite the lack of an uncontested definition, the definitions put forward in previous literature show corresponding elements and are basically based on three pillars (Martinez-Ferrero et al., 2015):

- Social responsible practices aiming to resolve the conflict of interest between shareholders and stakeholders;

- which go over and beyond legal requirements, and; - take the ethical aspect of ‘doing right’ in mind.

(6)

6

Based on the many different but corresponding definitions and the three pillars described, CSR research is mostly build on three main categories which are impacted trough CSR activities; social, economic and environment (Adams and Zutshi, 2004).

Where CSR mainly refers to the socially responsible activities and reporting by organizations, CSP refers to the individual agencies who rate and rank-order organizations on their CSR activities. In order to define CSP, this study follows Wood (1991) who defines CSP as; ‘a business organization's configuration of principles of social responsibility, processes of social responsiveness, policies, programs, and observable outcomes as they relate to the firm's societal relationships’. CSP therefore is the social performance resulting from the CSR activities undertaken by organizations.

2. Academic contribution

Due to the financial scandals in the US and Europe in the late 90’s and early 00’s, the society started to lose confidence in companies’ accounting activities. In order to try and regain the trust of stakeholders, companies started to disclose more and more information about their business behavior on a voluntary basis. This gave an impulse to what nowadays is known as CSR reporting (Martinez-Ferrero et al., 2015). These scandals also provided new areas for academic research.

For example studies started looking into, and found evidence for, the association between CSR and EM (Chih et al., 2008; Prior et al., 2008). Studies found a positive relationship between diverse socially responsible activities and the quality of accounting results (Hong and Andersen, 2011; Kim et al. 2012; Scholtens and Kang 2013; Schleifer, 2004; Shen and Chih, 2005). However, whether CSR has a positive or a negative effect on EM, and vice versa, is still inconclusive based on empirical research (Chih et al., 2008).

Previous studies in the area of CSR and EM showed results about the higher rate of transparency and lower risk of manipulating earnings in socially responsible firms (Kim et al. 2012). The research by Kim et al. (2012) also examined the relation between CSR subscores and financial reporting behavior, making use of KLD data as a proxy for CSR. Previous studies also found that environment respecting firms do not engage in more or less EM, an increase in environmental strengths does not change this relationship (Heltzer, 2011). However, an increase of EM is found within firms that disrespect the environment (Heltzer, 2011). Another area that has been part of research is the misuse of CSR in relation to EM (Muttakin, Khan, and Azim, 2015). In this case firms try to cover up their bad accounting practices by reporting about CSR. Finally, Martinez-Ferrero et al. (2015) found evidence for a negative bidirectional relation when accrual measures are used for EM. Firms tend to be more transparent if they have greater social commitment. On the other hand, opportunistic behavior by management does not make them participate in CSR practices, which is contrary to findings by Prior et al. (2008).

(7)

7

In addition to the findings mentioned above, Martinez-Ferrero et al. (2015) also considered investor protection as a moderating effect and found that the relationship between EM and CSR was stronger in countries with strong investor protection.

This study will contribute on prior research by extending the findings by Martinez-Ferrero et al. (2015). Their sample consisted out of international data from 26 countries in the period 2002-2010. The CSR data used in this research was extracted from the EIRIS database. It has been mentioned that the relationship between EM and CSR is very much dependent of the measurement of the variables. Where most of prior research used KLD data to measure CSR (Kim et al. 2012), this study contributes to prior research by making use of another CSR database, namely Thomson Reuters’ ASSET4 database. This database has been widely used by investors who want to integrate CSR in their investment analysis (Ioannou and Serafeim, 2012). Also, this study will use an extended research period compared to the study by Martinez-Ferrero et al (2015), namely the period 2002-2016, in order to obtain more generalizable evidence. At last, this study will involve real earnings management next to accrual earnings management, where Martinez-Ferrero et al. (2015) only included accrual earnings management in their research.

3. Literature review

3.1 Earnings management

Gargouri et al. (2010) provided several motivations for managers to participate in EM. The motivations mentioned are; the positive accounting theory, income smoothing, initial public offering (IPO), changes in control, labor negotiations and tax minimization. Also it is argued that from a rational point of view there would be no EM without the expectation to gain private benefit (Prior et al. 2008).

Despite these motivations, managerial choices to indulge in EM for a long period of time in order to satisfy their own interests can possibly affect a firm in a negative way (Prior et al. 2008) and damage the corporate reputation (Fombrun, Gardberg and Barnett, 2000; Martinez-Ferrero et al., 2016; Roychowdhury, 2006). Prior research provided reasoning for these negative effects from different theoretical points of view.

In the legitimacy theory, EM is an undesired practice and a latent threat. When EM practices go public because they are signaled by different sources, they could potentially lead to devastating effects (Dechow and Skinner, 2000). These effects could be negative press coverage (Chen, Patten, and Roberts, 2008; Dedoulis, 2006; Moerman and Van Der Laan, 2005) or even litigation proceedings (Dechow, Ge, and Schrand, 2010).

The social norm theory argues that the beliefs of the community determine economical behavior (Romer, 1984). It is argued that the acceptance of questionable reporting practices, like EM, is

(8)

8

significantly lower when economic attitudes adhere to accepted codes of corporate behavior (Leventis, Hasan, and Dedoulis, 2013).

The beliefs of the community mentioned above, lead to the stakeholder theory. Firms are confronted with divers stakeholder pressures which gives rise to managerial incentives to indulge in EM to influence the perceptions of their stakeholders (Bowen, Johnson, Shevlin, and Shores, 1992). These incentives are increasing when there is a multitude of stakeholder groups who have insufficient resources, incentives or access to information (Richardson, 2000), which causes information asymmetry (Jensen, 2001).

The last theory to be discussed is the agency theory. This theory has been the standard approach in studies about EM, and the relationship with EM has been established (e.g., Davidson III, Jiraporn, Kim, and Nemec, 2004). The agency theory sees the firm as a nexus of contracts (Walker, 2013). The two contracts that matter the most in the agency approach are the reward contracts and the financial contracts. Reward contracts play a role in the relationship between managers and shareholders, financial contracts are important in the relationship between shareholders and debtholders. Potential conflicts of interest arise between those parties when they start a business relationship. The most important factor which plays a role in these conflicts of interest is information asymmetry between insiders and outsiders of a firm and the divergence of interest. Information asymmetry causes difficulties for the various outside parties to observe and control insiders’ actions and decisions, thereby creating possibilities for opportunistic behavior by managers at the expense of share- and stakeholders.

3.2 Corporate Social Responsibility

The information asymmetry which comes with the agency theory can be reduced with the help of CSR. CSR reports help to communicate information to the firms’ stakeholders, thereby providing more transparent information about how results were obtained (Gray, Kouhy, and Lavers, 1995; Brady and Honey, 2007; Dragomir and Cristina, 2009).

As for EM, various theories have been used to research CSR. Because of the complexity of CSR, opinions on the theories vary among researchers. However, the theories have in common that they value social and environmental information disclosure to promulgate the reputation, identity and legitimacy of a firm (Hooghiemstra, 2000).

A widely used theory in CSR research is the stakeholder theory. Managing the parties influenced by and / or who influence the organization (stakeholders) is the key element of concern of the theory (Freeman, 1984; Mitchell, Agle, and Wood, 1997). Aligning firm and stakeholder objectives and expectations is seen as the aim of the theory. Gray et al. (1995) mentioned the possibilities for CSR activities to be part of the “dialogue between the company and its stakeholders”. They also call CSR activities a “successful means of negotiating these relationships”. Titman (1984) and Banerjee, Dasgupta, and Kim (2008) mention the importance of aligning company objectives with the objectives of their stakeholders because it creates a valuable intangible asset which helps in achieving stability and future growth opportunities for

(9)

9

the company. It has also been argued that firms who succeed in meeting stakeholders’ expectations will report more transparent information and present a more truthful and realistic image of the company (Salewski and Zulch, 2012).

The legitimacy theory, which is the most used theory in studying sustainable disclosure, is based on the stakeholder theory but is a broader concept. Following this theory, a business’s activities are ‘legitimate’ if they are accepted within the broad society (Castello and Lozano, 2011; Dawkins and Fraas, 2011; Patten, 2002; Woodward, Edwards, and Birkin, 1996). A firm has an implied social contract which connects the interests of society with the interests of the business (Martinez-Ferrero et al., 2015). A legitimate activity is not only focused on the needs of stakeholders, nor does it only focus on the alignment of interest. The legitimacy theory takes a broader perspective in a way that it also focuses on the principles in accordance with social norms, values and beliefs (Gray et al., 1995).

In order to obtain legitimacy, a firm can use different strategies such as changing their objectives, methods and output or changing stakeholders’ perception of these factors (Dowling and Pfeffer, 1975; Lindblom, 1993). One way to affect or change stakeholders’ perception is by providing CSR disclosure.

It has been argued that CSR disclosures are perceived to be useful by stakeholders (Dierkes and Antal 1985; Gray et al., 1995) and are appreciated because of their voluntary character (Basu and Palazzo 2008). Stakeholders can use these disclosures to judge if corporate strategies and activities are in line with their desires (Suchman, 1995). Also, good CSR disclosure is positively related to good CSR performance (Gelb and Strawser, 2001) which suggest that CSR reports do present a truthful image of a firms’ social activities. Companies who promote CSR activities are considered to be financially more stable, less vulnerable to activism and to have a long-term plan in maintaining and building loyalty and community relationships (Martinez-Ferrero et al., 2016). These findings are consistent with the believes of the stakeholder- and legitimacy theory.

In contrast to EM, stakeholders do value CSR strategy positively and a positive effect on corporate reputation has been shown (Martinez-Ferrero et al., 2016). Previous literature provides information about the important role of CSR in corporate strategy. Research has shown that CSR decreases shareholder equity (McWilliams and Siegel, 2001) but it is a crucial strategic element for success (Sen and Bhattacharya, 2001). However, the reasons for companies to engage in CSR remain unclear (Martinez-Ferrero et al., 2016).

The challenge for organizations is managing all of the different stakeholder groups. Corporate financial reporting is inevitably influenced by the tensions created by divers and competing stakeholder groups (Bowen et al., 1992; Freeman, Harrison, Wicks, Parmar, and De Colle, 2010). The same thing applies to CSR reporting which often is used to manage or manipulate the information provided to various powerful stakeholders in order to get their support and to be able to survive as a firm (Buhr, 2002; Gray et al., 1995). If a firm succeed in satisfying the demands and desires of the various stakeholders, it can be considered as a successful firm.

(10)

10

Besides comes the incentive for managers to focus on coping with CSR norms in their disclosures in order to obtain legitimacy (Grougiou, Leventis, Dedoulis, and Owusu-Ansah, 2014). Prior research also found evidence with regard to managers using CSR for personal benefit rather than promoting sustainable growth beyond their own welfare (Handelman and Arnold, 1999; Banerjee, 2007) or they use CSR to secure their jobs (Cespa and Cestone, 2007). From this point of view, CSR disclosures could be seen as a strategy to manage legitimacy (Patten, 1991; Deegan and Rankin, 1996; Campbell, 2000; Hutchings and Taylor, 2000; Woodward et al., 1996), and to increase corporate reputation (Fombrun and Shanley 1990; Verschoor, 2005; Linthicum, Reitenga, and Sanchez, 2010). Stakeholders do not take these actions lightly and a response may be to punish the managers, by boycotts or lobbying (Baron, 2001; Feddersen and Gilligan, 2001; John and Klein, 2003), in order to try and change this behavior (Rowley and Berman, 2000).

CSR and more specifically the rating of CSR activities has had some criticism and skepticism during the time. The CSR ratings developed and provided by small and private organizations for instance. These organizations provide little insights in the evaluation process regarding the ratings (Delmas, Etzion, and Nairn-Birch, 2013). However, even the most widely used CSR ratings in the US, the KLD ratings, are not free of criticism (Chatterji, Levine, and Toffel, 2009). It is found that these rating do not make optimal use of public available data. Also, it plays to the strength of large companies that there is a heavy emphasis on CSR scores in order to measure sustainable activities. These scores can easily be influenced by large companies, even if they have no real care about sustainability.

3.3 Relationship EM - CSR:

EM practices could harm a firm and its stakeholders. Adopting CSR is one way in which managers try to satisfy stakeholder interests and avoid the negative effects of EM. To avoid negative reactions or publicity, it has been found that CSR is often used as a defense system (Cespa and Cestone, 2007).

The association between EM, which has often been used as proxy for financial reporting quality, and CSR has been established in previous studies by for example Chih et al. (2008) and Prior et al., (2008). However, these studies show divergent results when it comes to the direction of the relationship between EM and CSR and whether there is a positive or a negative effect. The different EM proxies play a role in these divergent results, as shown in the research by Chih et al. (2008). They found, using multinational data, that firms who participate in CSR show more aggressive accrual management, but show less activity in earnings loss avoidance and earnings smoothing.

The divergent results are also shown by Trébucq and Russ (2005) who studied the relationship between CSR and EM. In their research no significant association was found when using a net CSR score. When observing other specification, they found a negative relation. Also different CSR dimensions gave inconsistent results.

(11)

11

The relationship between EM and CSR is also dependent on other, moderating, variables. Prior et al. (2008) for instance distinguished regulated and unregulated firms in their research whether firms mask EM by making strategical use of CSR. The results showed a positive relationship for regulated firms, for unregulated firms the results where statistically insignificant. However, regulated firms do not have the same discretion in accounting choices as unregulated firms have, therefore these results are less compelling (Kim et al., 2012).

Other important aspects in determining the relationship between EM and CSR are difference in accounting standard across countries, difference in investor protection and whether CSR is required by law (Reinhardt, Stavins, and Vietor, 2008). Next to that, it has been argued that the level of earnings management practices differ across countries (Leuz et al., 2003). The inconsistent results by Chih et al. (2008) mentioned earlier for example could be a result of these country differences instead of different CSR activities.

Studies that found a clear positive relationship between CSR and accounting quality (e.g., Prior et al., 2008, Gargouri et al., 2010), show that managers are more likely to be concerned with ethical and social activities if they make corporate decisions with discretion. By making discreet decisions, managers try to acquire stakeholders’ support and reducing negative effects of EM such as dismissal, or a bad reputation. This is consistent with the studies of Gelb and Strawser (2001), and Shen and Chih (2005) who found that social responsible and ethical firms provide more transparent information and tend to highlight their good behavior. These firms are found to be less inclined to engage in EM (Kim and Venkatachalam, 2011). In this vein Choi and Pae (2011) found that firms are less likely to engage in EM when they show high commitment towards ethical behavior.

As for the concept of EM and CSR apart, the relationship between the two factors are being explained by different theories as for instance agency theory, legitimacy theory and stakeholder theory.

From an agency theory perspective it has been suggested that CSR might possibly help managers to act in a self-serving manner and to advance their careers (e.g., Jensen and Meckling, 1976; Carroll, 1979, McWilliams, Siegel, and Wright, 2006). Results by Petrovits (2006) and Prior et al. (2008) are consistent with these suggestions. The possibility for managers to use CSR to mask opportunistic reporting behavior exists because of the voluntary nature of CSR reporting. This makes stakeholders skeptical with regard to corporate disclosures about CSR (Bakan, 2005; Barnett, 2007).

In contrast to this, it has been found that CSR is used to reduce the agency problem (Jensen, 2001; Calton and Payne, 2003; Scherer, Palazzo, and Baumann, 2006; Harjoto and Jo, 2011). The suggestion is that conflicts of interest can be mitigated by CSR and as a consequence earnings management activities may be reduced. Also, Jo and Kim (2008) found evidence for a negative relation between corporate disclosure and EM, and vice versa, since disclosure reduces information asymmetry.

(12)

12

Legitimacy theory predicts a positive impact of EM on CSR because managers are incentivized to report nonrealistic financial profits which do not reflect the true economic situation out of their own interests (Scholtens and Kang, 2013). It is found that these managers, in order to protect their position and to distract attention from their reporting practices, are more likely to engage in CSR (Prior et al., 2008).

It has been mentioned that following the stakeholder theory, it is important for organizations to meet stakeholder expectation and let stakeholders participate in corporate decision making. By doing this, social legitimacy will be reinforced, board of directors’ involvement increases and top managers are held to a higher standard of performance (Prior et al., 2008). However, previous studies found opposite results for CSR to have a positive impact on EM and suggests that managers who engage in CSR to negotiate the interests of their stakeholders, are also participating in questionable reporting practices (Grougiou et al., 2014).

From both theories, legitimacy and stakeholder theory, it can be concluded that managers believe they can reduce the probability of an investigation into EM practices if they satisfy stakeholders’ interests (Prior et al., 2008). Therefore managers are incentivized to show their concern and awareness for the environment and society.

To predict the relationship between EM and CSR, Chih et al. (2008) presented four hypothesis which are widely used in previous studies. The first hypothesis is the myopia avoidance hypothesis which predicts a negative correlation between Corporate Social Performance (CSP) and EM. The second and third hypothesis are the predictable earnings hypothesis and the multiple objectives hypothesis according to whom there is a positive relation. At last they formulated the institutional hypothesis which stipulates a neutral relationship.

The multiple objectives hypothesis (Chih et al., 2008) is a frequently used hypothesis in studies about EM and CSR. The hypothesis argues that managers are being motivated to engage in EM because CSR intensifies the agency problem. The reasoning behind this is that managers have different goals compared to stakeholders and use information to make decisions in their own interests. They use CSR to disguise this opportunistic behavior (see also; Hemingway and Maclagan, 2004).

Evaluating managerial performance becomes even more difficult when firms try to satisfy a wide group of stakeholders because a lack of performance criteria arises. The lack of criteria worsens agency problems and creates possibilities for managers to participate in EM (Gargouri et al., 2010). These findings are in line with the multiple objectives hypothesis.

Gargouri et al. (2010) investigated the relation between CSP and EM and found a significant and positive association between the environmental aspect of CSP and EM. In their research this association is explained with the reasoning that it is expensive to participate in environmental activities, which has a negative effect on financial performance. Therefore managers are incentivized to manage earnings. In this same research a positive association

(13)

13

between EM and the employee aspect of CSP has been found which suggests collusion of managers and employees in order to mask opportunistic behavior.

Most of the prior studies mentioned above use managerial opportunism to explain the relationship between CSR and EM. Kim et al. (2012) used a different point of view and studied how financial reporting practices are affected by motives for CSR in general. Another distinctive feature of this study is that it takes into account real earnings management (REM) next to accrual-based earnings management (AEM). According to the study of Kim et al. (2012) it appears that CSR firms do not use discretionary accruals in the same amount as non-CSR firms. It is also less likely that CSR firms will participate in REM since this affects real operations of a firm. Their last finding argues that it is less likely that managers of CSR firms will be subject of an investigation. Therefore they suggest that CSR is an important factor in reducing EM activities. CSR firms seem to make more conservative corporate decisions and their financial information is more transparent.

Where prior research mostly focused on a one-way relationship between EM and CSR, Martinez-Ferrero et al. (2015) found evidence for a negative bidirectional relation when accrual measures are used for EM. The authors presented findings that more social responsible firms are less inclined to engage in EM, are more transparent and provide more financial information. In the same study they also showed evidence for EM to negatively affect CSR, which confirms that high quality reporting leads to high levels of social commitment and that opportunistic managerial behavior is not always being covered up by CSR practices.

Based on the above conflicting remarks the following hypothesis is formulated:

H1: EM is significantly associated to CSR.

(Null hypotheses: there is no significant association between EM and CSR.)

3.4 Investor protection

Investor protection is the protection of shareholders by legal systems. With the help of these legal systems, insiders’ private control benefits are being limited and they are granted rights to discipline insiders (e.g., La Porta, Lopez-de-Silanes, Shleifer, and Vishny, 1998; Nenova, 2000; Claessens and Fan, 2002; Dyck and Zingales, 2004). It has been found that investor protection is an important factor in the international differences in corporate earnings management and that manipulative activities occur less frequently in countries with strong investor protection (Leuz et al., 2003; Chih et al., 2008; Scholtens and Kang, 2013).

However, Prado-Lorenzo, García-Sánchez, and Gallego-Álvarez (2012) found that in countries with stronger investor protection, managers concern less about CSR in their decision making which suggest that they may be less committed to CSR development (Simnett, Vanstraelen, and Wai Fong, 2009).

(14)

14

In contrast to this, Martinez-Ferrero et al. (2015) found that the negative bidirectional relationship between EM and CSR is more significant in countries with stronger investor protection. They found that in these countries companies are less likely to engage in manipulative actions and thereby do not need to use CSR actions to disguise EM activities. In another research Martinez-Ferrero et al. (2016) found a favorable effect of CSR on EM in stronger investor protected countries.

Based on these findings the following hypotheses is proposed:

H2: The level of investor protection does moderate the relationship between EM and CSR. (Null hypotheses: The level of investor protection does not moderate the relationship between EM and CSR.)

4. Data & Methodology

4.1 Population and sample

The sample used in this study consists of international data. Financial data is obtained from the Compustat Global and Compustat North America database. CSR data is obtained from the ASSET4 database. The research period of this paper is from fiscal year 2002 to 2016. The start date of the research if restricted to 2002 because ASSET4 started measuring CSR since fiscal year 2002.

In order to calculate earnings management, companies in regulated industries (SIC codes 4400 – 5000) and financial institutions (SIC codes 6000 - 6500) are being eliminated from the sample because of their specific characteristics. This is in line with Roychowdhury (2006). Also, following sample requirements of among others Bozzolan, Fabrizi, Mallin, and Michelon (2015), at least 10 observations for every country-year grouping are required.

After calculating the proxies for EM, the data has been merged with the CSR data and the control variable data. This resulted in a dataset with 23,108 firm-year observations, from 54 countries.

For testing the second hypothesis, the dataset for testing hypothesis 1 has been merged with investor protection data from La Porta et al. (1998). This resulted in a dataset with 21,820 firm-year observations from 37 countries.

4.2 Measures of EM

4.2.1 Accrual earnings management

EM will be divided into accrual earnings management (AEM) and real earnings management (REM). AEM stands for the discretionary component of accruals adjustment and measures accounting manipulation (Christensen, Kent, and Stewart, 2010; Navarro-García and

(15)

Madrid-15

Guijarro 2014; Wilson 2011). The discretionary component must be separated from the non-discretionary component. In this way the existence and extent of EM can be determined.

In order to calculate the discretionary component first total accruals adjustment (TACC) will be calculated following Jones (1991):

𝑇𝐴𝐶𝐶_𝑡= ∆𝐶𝐴_𝑡− ∆𝐶𝐴𝑆𝐻 − ∆𝐶𝐿_𝑡+ ∆𝐷𝐶𝐿_𝑡− 𝐷𝐸𝑃_𝑡 (1A)

where, ∆CAt is the change in the current assets in year t; ∆CASHt is the change in cash and cash equivalents in year t; ∆CLt reflects the change in current liabilities in year t; ∆DCLt is the change in short-term debt included in current liabilities in year t; and DEPt is the depreciation and amortization expense in year t.

After running equation (A), the Modified Jones Model as proposed by Dechow, Sloan, and Sweeney (1995) will be used to calculate the abnormal accruals (DACC). However, before being able to calculate DACC, the following model is estimated for each country-year grouping by making use of an ordinary least squares regression (OLS):

𝑇𝐴𝐶𝐶_𝑡 𝐴_𝑡−1

= 𝛼

1 1 𝐴_𝑡−1

+ 𝛼

2 (∆𝑅𝐸𝑉_𝑡−∆𝑅𝐸𝐶_𝑡) 𝐴_𝑡−1

+ 𝛼

3 𝑃𝑃𝐸_𝑡 𝐴_𝑡−1

+ 𝜀

𝑡 (1B)

where At-1 represents the lagged total assets; ∆REVt is the change in revenues; ∆RECt is the change in receivables; and PPEt is the property, plant and equipment.

DACC is the difference between total accrual and nondiscretionary accruals. Therefore the next step is to calculate the nondiscretionary accruals (NDACC) using the following equation:

𝑁𝐷𝐴𝐶𝐶𝑡 𝐴𝑡−1

= 𝛼̂

1 1 𝐴𝑡−1

+ 𝛼̂

2 (∆𝑅𝐸𝑉𝑡−∆𝑅𝐸𝐶𝑡 ) 𝐴𝑡−1

+ 𝛼̂

3 𝑃𝑃𝐸𝑡 𝐴𝑡−1 (1C)

At last the discretionary accruals, which will be used as a measure for EM, can be calculated with the help of the following formula:

𝐷𝐴𝐶𝐶

_𝑡

=

𝑇𝐴𝐶𝐶𝑡

𝐴𝑡−1

−

𝑁𝐷𝐴𝐶𝐶𝑡

𝐴𝑡−1 (1D)

4.2.2 Real earnings management

Real earnings management (REM) is the actions of a firms’ management that deviates from the regular actions, in order to meet or beat earnings thresholds (Roychowdhury, 2006). To measure REM, this study will make use of the method as put forward by Cohen, Dey, and Lys (2008) and Roychowdhury (2006).

To detect real earnings activities, this study will rely on prior studies and use four different measures (e.g., Roychowdhury, 2006; Cohen et al., 2008; Cohen and Zarowin, 2010; Badertscher, 2011; Kim et al., 2012; Zang 2012). These measures are; abnormal levels of

(16)

16

operating cash flows (ABN_CFO), abnormal production costs (ABN_PROD), abnormal discretionary expenses (ABN_DISEXP), and a combined measure of real activities manipulation (COMBINED_REM). This combined measure is calculated as ABN_CFO -/- ABN_PROD + ABN_DISEXP (Kim et al., 2012).

ABN_CFO is calculated as the actual CFO minus the normal CFO by using estimated coefficients from the corresponding country-year model (equation 2A) and the firm-year’s sales and lagged assets.

The estimated country-year model is:

𝐶𝐹𝑂_𝑡 𝐴_𝑡−1

= 𝛽

0

+ 𝛽

1 1 𝐴_𝑡−1

+ 𝛽

2 𝑆_𝑡 𝐴_𝑡−1

+ 𝛽

3 ∆𝑆_𝑡 𝐴_𝑡−1

+ 𝜀

𝑡 (2A) where CFOt is cash flow from operations; At-1 represents the lagged total assets; St is total sales; ∆St is the change in sales.

ABN_CFO is then calculated by using the following formula:

𝐴𝐵𝑁_𝐶𝐹𝑂

_𝑡

=

𝐶𝐹𝑂𝑡 𝐴𝑡−1

− 𝛽̂

0

− 𝛽̂

1 1 𝐴𝑡−1

− 𝛽̂

2 𝑆𝑡 𝐴𝑡−1

− 𝛽̂

3 ∆𝑆𝑡 𝐴𝑡−1 (2B)

ABN_PROD is calculated as the actual production costs minus the normal production costs by using estimated coefficients from the corresponding country-year model (equation 3B) and the firm-year’s sales and lagged assets. The first step in calculating ABN_PROD is calculating the production costs (PROD) with the variables cost of goods sold (COGS) and the change in inventories (∆INV), using the following formula:

𝑃𝑅𝑂𝐷_𝑡 = 𝐶𝑂𝐺𝑆_𝑡+ ∆𝐼𝑁𝑉_𝑡 (3A)

The next step is estimating the model:

𝑃𝑅𝑂𝐷𝑡 𝐴𝑡−1

= 𝛽

0

+ 𝛽

1 1 𝐴𝑡−1

+ 𝛽

3 ∆𝑆𝑡 𝐴𝑡−1

+ 𝛽

4 ∆𝑆𝑡−1 𝐴𝑡−1

+ 𝜀

𝑡 (3B)

where At-1 represents the lagged total assets; St is total sales; ∆St is the change in sales; ∆St-1 is sales in period -1 minus sales in period -2.

At last ABN_PROD is calculated with the following formula:

𝐴𝐵𝑁_𝑃𝑅𝑂𝐷

_𝑡

=

𝑃𝑅𝑂𝐷𝑡 𝐴𝑡−1

− 𝛽̂

0

− 𝛽̂

1 1 𝐴𝑡−1

− 𝛽̂

3 ∆𝑆𝑡 𝐴𝑡−1

− 𝛽̂

4 ∆𝑆𝑡−1 𝐴𝑡−1 (3C)

(17)

17

The third measure of REM is ABN_DISEXP and is calculated as the actual discretionary expenses minus the normal discretionary expenses by using estimated coefficients from the corresponding country-year model (equation 4B) and the firm-year’s sales and lagged assets. The first step in calculating ABN_DISEXP is calculating the discretionary expenses (DISEXP) with the variables R&D (R&D), advertising (ADV) and selling, general and administrative expenses (SG&A), using the following formula:

𝐷𝐼𝑆𝐸𝑋𝑃𝑡 = 𝑅&𝐷𝑡+ 𝐴𝐷𝑉𝑡+ 𝑆𝐺&𝐴𝑡 (4A)

After DISEXP has been calculated, the following model will be estimated:

𝐷𝐼𝑆𝐸𝑋𝑃𝑡 𝐴𝑡−1

= 𝛽

0

+ 𝛽

1 1 𝐴𝑡−1

+ 𝛽

+ 𝜀

𝑡 (4B)

where At-1 represents the lagged total assets and St is total sales.

After estimating the model, the last step is to calculate ABN_DISEXP by using the formula:

𝐴𝐵𝑁_𝐷𝐼𝑆𝐸𝑋𝑃

_𝑡

=

𝐷𝐼𝑆𝐸𝑋𝑃𝑡 𝐴𝑡−1

− 𝛽̂

0

− 𝛽̂

1 1 𝐴𝑡−1

− 𝛽̂

2 𝑆_𝑡 𝐴𝑡−1 (4C) 4.3 Measures of CSR

Following Ioannou and Serafeim (2012), the CSR data used in this study is obtained from the Thomson Reuters ASSET4 global database. This database has collected CSR data since fiscal year 2002. Firms are being rated by research analysts who collect 900 evaluation points per firm. The data collected by the analysts is transformed from qualitative to quantitative data in order to allow quantitative analyses. The 900 evaluation points are used as input to calculate 250 key performance indicators (KPIs). These KPIs are organized within 4 main pillars, namely: environmental performance score, social performance score, corporate governance score, and economic performance score. These four pillars each got their own individual score, ranging from 0 to 100, which is an average of the KPI scores within that pillar.

In constructing the CSR measure, this study followed Ioannou and Serafeim (2012). The CSR measures used to test the hypothesis is the equal-weighted rating of the environmental score, the social score, and the corporate governance score. Further, a robustness test will be performed whereby the CSR measure is the equal-weighted rating of only environmental score and the social score. Corporate governance is excluded from the CSR measure in this regression, because the relation between corporate governance and financial reporting practices has widely been investigated in prior research (Kim et al., 2012). In this regression, the corporate governance score is added as control variable.

(18)

18

4.4 Measures of investor protection

This study followed the study of Martinez-Ferrero et al. (2015) in computing the level of investor protection. Three sub-indices are created, which are based on La Porta et al. (1998) their country-level governance indices. These sub-indices are DCL, DAR, and DEF. DCL indicates whether a firm is located in a common law country (1) or whether a firm is located in a civil law country (0). DAR indicates whether a firm is located in stronger anti-director rights country (1) compared to a median country or whether it is located in a lower anti-director rights country (0). At last, DEF, indicates whether a firm is located in a higher law enforcement index country (1) compared to the median country, or 0 otherwise. DEF is the sum of two indices (La Porta et al., 1998), namely; the efficiency of the judicial system, and law and order (Martinez-Ferrero et al., 2015).

The sum of these three indices will result in a dummy variable which will proxy for effective investor protection (IP). IP indicates whether a firm is located in a country with above-average investor protection (1) or below-average investor protection (0).

4.5 Control variables

The study will make use of control variables which might affect EM and CSR. Following Martinez-Ferrero et al. (2015) this study uses company size, financial resources, growth opportunities, and R&D investment as control variables. Also leverage (Kim et al. 2012) and return on assets (Scholtens and Kang, 2013) will be used as control variables.

Company size (SIZE) will be measured by making use of the total asset natural logarithm. The firms’ financial resources (FIN_RES) will be measured by working capital divided by total assets, working capital is calculated as current assets minus current liabilities. GROWTH indicates the growth opportunities of the firm and will be measured using the ratio of the market to book value. The ratio of R&D expenditure to total revenue will measure R&D investment (RDINVEST). Leverage (LEV) will be computed as the liabilities to total assets ratio, and financial performance is measured by return on assets (ROA).

Furthermore, it is likely that firms use a combination of AEM and REM in order to manage their reported earnings (Kim et al., 2012). It has also been found that firms choose one of the methods which is the less costly to the organization (Cohen et al., 2008; Zang, 2012). Following prior research by Cohen et al. (2008) and Kim et al. (2012), this study uses COMB_REM as a control variable in the equation of AEM, and DACC as a control variable in the equations for REM, to control for this substitute effect.

(19)

19

4.6 Method of analysis

To test the first hypotheses, the following model has been estimated by multiple regression (Kim et al., 2012):

DACCt = ø + ø1CSRt + ø2COMB_REMt + ø3SIZEt + ø4LEVt + ø5ROAt + ø6GROWTHt + ø7RDINVESTt + ø8FIN_RES + ηi + μit

[Model1.1]

REMt = ø + ø1CSRt + ø2DACCt + ø3SIZEt + ø4LEVt + ø5ROAt + ø6GROWTHt + ø7RDINVESTt + ø8FIN_RES + ηi + μit

[Model1.2]

where REM is one of the REM proxies (ABN_CFO, ABN_PROD, ABN_DISEXP, or COMB_REM) as defined above.

The second hypothesis includes the interaction effect of investor protection on CSR (CSR*IP). Therefore a further regression model will be estimated to test this hypothesis. The equation is as follows:

DACCt = ø + ø1CSR∗IPt + ø2CSRt + ø3IPt + ø4COMB_REMt + ø5SIZEt + ø6LEVit + ø7ROAt + ø8GROWTHt + ø9RDINVESTt + ø10FIN_RES + ηi + μit

[Model2.1]

REMt = ø + ø1CSR∗IPt + ø2CSRt + ø3IPt + ø4DACCt + ø5SIZEt + ø6LEVit + ø7ROAt + ø8GROWTHt + ø9RDINVESTt + ø10FIN_RES + ηi + μit

[Model2.2]

In order to test the robustness of the CSR measure, additional analysis will be estimated whereby CSR represents the equal-weighted rating of the environmental score and the social score (CSR_exCG). In this regression the corporate governance score (CG_SCORE) is included as control variable. In order to test the robustness, the following models will be estimated:

EMt = ø + ø1CSR_exCGt + ø2COMB_REMt + ø3CG_SCOREt + ø4SIZEt + ø5LEVt + ø6ROAt + ø7GROWTHt + ø8RDINVESTt + ø9FIN_RES + ηi + μit

[Model3.1]

EMt = ø + ø1CSR_exCG∗IPt + ø2CSR_exCGt + ø3IPt + ø4COMB_REMt + ø5CG_SCOREt + ø6SIZEt + ø7LEVit + ø8ROAt + ø9GROWTHt + ø10RDINVESTt + ø11FIN_RES + ηi + μit

[Model3.2]

where EM is one of the five EM proxies as defined above (DACC, ABN_CFO, ABN_PROD, AN_DISEXP, and COMB_REM).

(20)

20

At last, this study will contain additional analysis for the possible bidirectional relationship between EM and CSR, as proposed and found by Martinez-Ferrero et al. (2015). The models to be estimated are:

CSRt = ø + ø1EMt + ø2SIZEt + ø3LEVt + ø4ROAt + ø5GROWTHt + ø6RDINVESTt + ø7FIN_RES + ηi + μit

[Model4.1]

CSRt = ø + ø1EM∗IPt + ø2DACCt + ø3IPt + ø4SIZEt + ø5LEVit + ø6ROAt + ø7GROWTHt + ø8RDINVESTt + ø9FIN_RES + ηi + μit

[Model4.2]

where EM is one of the five EM proxies as defined above (DACC, ABN_CFO, ABN_PROD, AN_DISEXP, and COMB_REM).

5. Results

5.1 Descriptive statistics and correlations

The descriptive statistics of the main variables used to test hypothesis 1 are presented in table 1a and the descriptive statistics of the main variables used to test the second hypothesis are presented in table 1b. Following Kim et al. (2012), the variables presented in table 1a and table 1b are all winsorized at the top and bottom 1 percent level.

Table 1a

Descriptive statistics of data hypothesis 1

Mean St. Dev. Min Max

Dependent variables DACC -0.2201 0.6575 -3.7148 0.5251 ABN_CFO 0.1344 0.2763 -1.1755 0.8377 ABN_PROD -0.0743 0.2524 -0.8332 0.6847 ABN_DISEXP -0.2811 0.5246 -3.2596 0.8955 COMB_REM -0.0723 0.6271 -4.7419 2.4881 Variables of interest CSR 0.5149 0.2451 0.0460 0.9777 CSR_exCG 0.5175 0.2973 0.0623 0.9809 Control variables SIZE 9.4255 2.6002 4.5667 16.511 LEV 0.5296 0.2076 0.0590 1.1534 ROA 0.0469 0.0934 -0.4153 0.3017 GROWTH 2.9145 3.5165 -5.5800 23.340 RDINVEST 0.0291 0.0651 0 0.4304 FIN_RES 0.1591 0.1786 -0.2292 0.6937

(21)

21

Table 1a and table 1b summarize the descriptive statistics for the variables used in the research models. For example, the mean values of discretionary accruals of the samples is 0.2201 and -0.2236 with standard deviations of 0.6575 and 0.6694. Also, the companies in the samples have an average CSR performance as shown by the mean values of CSR (0.5149, 0.5241) and CSR without CG_SCORE (0.5175, 0.5267).

Table 1b

Descriptive statistics of data hypothesis 2

Mean St. Dev. Min Max

Dependent variables DACC -0.2236 0.6694 -3.7869 0.3516 ABN_CFO 0.1548 0.2538 -1.1755 0.8377 ABN_PROD -0.0741 0.2519 -0.8332 0.6847 ABN_DISEXP -0.3195 0.5025 -3.2596 0.8955 COMB_REM -0.0895 0.6231 -4.7419 2.4881 Variables of interest CSR 0.5241 0.2431 0.0460 0.9777 CSR_exCG 0.5267 0.2974 0.0628 0.9809

Moderator & interaction variable

IP 0.9227 0.2671 0 1 CSR*IP 0.4831 0.2722 0 0.9777 CSR_exCG*IP 0.4783 0.3172 0 0.9809 Control variables SIZE 9.4059 2.6405 4.5667 16.511 LEV 0.5318 0.2082 0.0590 1.1534 ROA 0.0462 0.0933 -0.4153 0.3017 GROWTH 2.9227 3.5345 -5.5800 23.340 RDINVEST 0.0299 0.0662 0 0.4304 FIN_RES 0.1598 0.1777 -0.2223 0.6952

The frequencies of the moderating dummy variable investor protection have been summarized in table 2. As defined before, the values of this dummy variable is 0 if a country has below-average investor protection and 1 if a country has above-below-average investor protection. Table 2 shows that the largest part of the sample, 92.27% of the firm-year observations, have above-average investor protection.

Table 2

Frequencies

Investor protection Absolute Relative

0 1,687 7.73

1 20,133 92.27

(22)

22

In table 3a and table 3b, the results of the Pearson correlation tests are presented. Table 3a shows the correlation coefficients between the variables used to test hypothesis 1 and table 3b shows the correlation coefficient between the variables used to test hypothesis 2. With these coefficient, the bivariate correlation between the variables can be analyzed. In table 3a, DACC is negatively correlated with CSR (CSR and CSR_exCG). This suggests that CSR firms are less engaged in EM by manipulating accruals. When the interaction effect of investor protection is taken into account, DACC is also negatively correlated to CSR. However, when the corporate governance score is excluded from the CSR measure, the coefficient shows a positive correlation.

With regard to the REM variables, ABN_CFO and ABN_PROD are positively correlated and ABN_DISEXP and COMB_REM are negatively correlated to CSR scores. When the corporate governance scores are excluded from the CSR scores, ABN_CFO and COMB_REM are negatively correlated and ABN_PROD and ABN_DISEXP are positively correlated to CSR scores. This suggests that CSR firms are more (less) likely to be engaged in EM by production cost, discretionary expense and overall REM (operating cash flow) manipulating activities. When the corporate governance score is excluded from the CSR measure, the matrix shows different results with regard to ABN_CFO and ABN_DISEXP. In this case higher (lower) levels of operating cash flow (discretionary expenses) manipulation are correlated with higher CSR scores. The same findings apply when the interaction effect between investor protection and the CSR scores are added in the correlation matrix, table 3b.

Also the VIF ratios have been analyzed. The values are all lower than the suggested maximum of 10 (Neter, Wasserman and Kunter, 1990), therefore the variables can simultaneously be used in multivariate analysis.

Table 3a

Pearson correlation data hypothesis 1

1 2 3 4 5 6 7 DACC (1) 1.0000 ABN_CFO (2) -0.0095 1.0000 ABN_PROD (3) -0.0525 -0.2941 1.0000 ABN_DISEXP (4) 0.0608 -0.2345 -0.0190 1.0000 COMB_REM (5) 0.0678 0.3627 -0.5479 0.7408 1.0000 CSR (6) -0.1327 0.1113 0.0692 -0.1079 -0.0690 1.0000 CSR_exCG) (7) -0.1209 -0.0397 0.1452 0.0098 -0.0677 0.9139 1.0000 SIZE (8) 0.0963 -0.2048 0.1560 0.1814 -0.0013 0.0631 0.3207 LEV (9) -0.0618 0.0041 0.1599 -0.1631 -0.1990 0.2086 0.1922 ROA (10) -0.0381 0.2471 -0.2810 -0.0465 0.1830 0.0646 0.0648 GROWTH (11) -0.0109 0.1563 -0.2409 -0.0504 0.1237 0.0135 -0.0351 RDINVEST (12) 0.0843 -0.0013 -0.1888 0.0785 0.1411 -0.0190 -0.0240 FIN_RES (13) 0.1662 0.0194 -0.1503 0.0405 0.1030 -0.1222 -0.1070

(23)

23 Table 3a (continued) 8 9 10 11 12 13 SIZE (8) 1.0000 LEV (9) 0.0577 1.0000 ROA (10) 0.0415 -0.1242 1.0000 GROWTH (11) -0.1642 0.1128 0.2021 1.0000 RDINVEST (12) -0.0753 -0.1513 -0.1696 0.1479 1.0000 FIN_RES (13) -0.1022 -0.4978 0.0888 0.0310 0.3097 1.0000 Table 3b

Pearson correlation data hypothesis 2

1 2 3 4 5 6 7 DACC (1) 1.0000 ABN_CFO (2) -0.0190 1.0000 ABN_PROD (3) -0.0587 -0.3368 1.0000 ABN_DISEXP (4) 0.0806 -0.1379 0.0001 1.0000 COMB_REM (5) 0.0810 0.4323 -0.5414 0.7502 1.0000 CSR (6) -0.1385 0.0609 0.0665 -0.0576 -0.0486 1.0000 IP (7) 0.2133 0.0575 -0.0907 -0.0470 0.0222 -0.0076 1.0000 CSR*IP (8) -0.0057 0.0821 0.0056 -0.0793 -0.0327 0.8199 0.5138 SIZE (9) 0.0943 -0.2256 0.1583 0.1927 -0.0005 0.0654 -0.2944 LEV (10) -0.0633 0.0001 0.1522 -0.1579 -0.1888 0.2110 0.0337 ROA (11) -0.0411 0.2978 -0.2764 -0.0679 0.1783 0.0807 -0.0827 GROWTH (12) -0.0103 0.1840 -0.2376 -0.0606 0.1222 0.0197 -0.0303 RDINVEST (13) 0.0841 -0.0164 -0.1914 0.0970 0.1489 -0.0281 0.0929 FIN_RES (14) 0.1679 0.0282 -0.1476 0.0352 0.0995 -0.1248 0.0503 Table 3b (continued) 8 9 10 11 12 13 14 CSR*IP (8) 1.0000 SIZE (9) -0.0911 1.0000 LEV (10) 0.2023 0.0558 1.0000 ROA (11) 0.0300 0.0424 -0.1122 1.0000 GROWTH (12) 0.0040 -0.1635 0.1127 0.2000 1.0000 RDINVEST (13) 0.0282 -0.0749 -0.1561 -0.1743 0.1470 1.0000 FIN_RES (14) -0.0817 -0.0995 -0.4983 0.0763 0.0301 0.3137 1.0000

5.2 The relation between earnings management and corporate social responsibility

The aim of hypothesis 1 is to examine the associative relationship between EM and CSR. Table 4 presents the results for the estimated multivariate regressions of model 1.1 and 1.2. Each column represents a different measure of EM used as dependent variable.

(24)

24

Table 4

(1) (2) (3) (4) (5)

Models 1.1 & 1.2 DACC ABN_CFO ABN_PROD ABN_DISEXP COMB_REM

CSR -0.328*** 0.123*** 0.0540*** -0.175*** -0.107*** (0.0176) (0.00719) (0.00633) (0.0140) (0.0166) COMB_REM 0.0605*** (0.00703) DACC 0.0160*** -0.0194*** 0.0174*** 0.0528*** (0.00267) (0.00234) (0.00519) (0.00614) SIZE 0.0312*** -0.0229*** 0.0123*** 0.0398*** 0.00464*** (0.00165) (0.000675) (0.000593) (0.00131) (0.00155) LEV 0.181*** 0.0164* 0.120*** -0.464*** -0.568*** (0.0245) (0.00981) (0.00863) (0.0191) (0.0226) ROA -0.352*** 0.721*** -0.765*** -0.310*** 1.175*** (0.0490) (0.0196) (0.0172) (0.0382) (0.0451) GROWTH -0.000397 0.00524*** -0.0106*** 0.00100 0.0168*** (0.00129) (0.000520) (0.000457) (0.00101) (0.00120) RDINVEST 0.274*** 0.0884*** -0.736*** 0.571*** 1.396*** (0.0714) (0.0287) (0.0253) (0.0559) (0.0661) FIN_RES 0.672*** -0.0304*** 0.0213** -0.181*** -0.233*** (0.0284) (0.0116) (0.0102) (0.0226) (0.0268) Constant -0.533*** 0.235*** -0.201*** -0.293*** 0.143*** (0.0232) (0.00950) (0.00836) (0.0185) (0.0219) Observations 23,108 23,108 23,108 23,108 23,108 R-squared 0.063 0.126 0.191 0.081 0.101

Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

5.2.1 Accrual-based earnings management and CSR

The first column of table 4 presents the findings with regard to accrual-based earnings management. A negative and significant effect (p<0.01) of CSR has been found with regard to DACC, which indicates that CSR firms are less likely to manage earnings through accruals. Further, the combined measure of REM (COMB_REM), has a positive and statistically significant effect on DACC. This indicates that firms who choose to manage earnings through accruals are less likely to also manage earnings through real activity manipulation. These findings are in line with the findings presented by Kim et al. (2012). Consistent with Cohen et al. (2008) and Zang (2012) this would imply that firms choose between one of the methods to manage earnings, AEM or REM. In the DACC regression, it is found that SIZE, LEV, RDINVEST, and FIN_RES are positively and significantly related to DACC. This suggests that larger firms, with higher leverage, higher R&D investments and more financial resources, are more likely to be engaged in accrual-based earnings management. ROA is negative and significant related to DACC, suggesting that firms with higher return on assets engage less in earnings management through accruals.

(25)

25

5.2.2 Real earnings management and CSR

In order to be able to correctly interpret the results of the regressions, it is important to understand the directions of the REM measurements. In accordance with Kim et al. (2012), higher levels of abnormal operating cash flows, abnormal discretionary expenses, and the combined REM measure, indicate less earnings management by manipulating real operating activities. The level of abnormal production costs has to be interpreted the other way around. In this case a lower level of abnormal production costs mean less earnings management activities.

In columns two till five the results of the REM regressions are presented. The results show that CSR is positive and significant associated to ABN_CFO and ABN_PROD. Also, CSR is negative and significant associated with ABN_DISEXP and COMB_REM. This means that CSR firms are more (less) likely to be engaged in real activity manipulation by abnormal production costs, abnormal discretionary expense, and overall REM (abnormal operating cash flow). These findings indicate that CSR firms are more likely to manage earnings through real activity manipulation, except for operating cash flow manipulation.

DACC has been used as control variable in the REM regressions. The results show a positive (negative) effect for DACC on ABN_CFO, ABN_DISEXP, COMB_REM (ABN_PROD). As in the DACC regression, this suggests that firms use REM and AEM as substitutes of each other. With regard to the control variables, the regression of COMB_REM shows positive and significant coefficients for SIZE, ROA, GROWTH, and RDINVEST, suggesting that bigger firms with higher return on assets, higher market to book values and higher R&D investments are less likely to manage earnings through overall real activity manipulation. LEV and FIN_RES are negatively and significantly associated with COMB_REM, which indicates that firms with higher leverage and more financial resources are more likely to be engaged in overall activity manipulation.

The significant results found in both the AEM and REM regressions lead to rejecting the null hypothesis 1. A significant association between EM and CSR has been found.

5.3 Moderating effect of investor protection

The second hypothesis of this study aims to examine to moderating effect of investor protection on the relation between EM and CSR. In order to be able to examine this effect, the dummy variable DINV_PROT and the interaction effect of DINV_PROT with CSR is added to the regression. The variable INT_CSR_IP represents this interaction effects of investor protection with CSR. Table 5 shows the results of the estimated models 2.1 and 2.2.

(26)

26

5.3.1 Accrual-based earnings management

The first column of table 5 shows that the interaction effect of investor protection with CSR is negatively and significantly (P<0.05) associated with DACC, indicating that CSR firms, when located in countries with strong investor protection, are less likely to be engaged in AEM. As in the regression of DACC for testing hypothesis 1, the overall REM measure (COMB_REM) is positive and significant associated with DACC, indicating that firms substitute between AEM and REM. With regard to the control variables it is found that SIZE, LEV, GROWTH and FIN_RES are positively and significantly related to DACC. A negative and significant relation has been found for ROA.

Table 5

(1) (2) (3) (4) (5)

Models 2.1 & 2.2 DACC ABN_CFO ABN_PROD ABN_DISEXP COM_REM

CSR*IP -0.139** 0.0224 -0.0169 -0.224*** -0.185*** (0.0653) (0.0243) (0.0236) (0.0501) (0.0617) CSR -0.228*** 0.0291 0.0728*** 0.155*** 0.111* (0.0626) (0.0233) (0.0226) (0.0480) (0.0591) IP 0.718*** -0.000745 -0.0511*** 0.116*** 0.166*** (0.0386) (0.0145) (0.0141) (0.0299) (0.0367) COMB_REM 0.0672*** (0.00716) DACC 0.00825*** -0.0156*** 0.0359*** 0.0598*** (0.00252) (0.00244) (0.00518) (0.00638) SIZE 0.0507*** -0.0219*** 0.0105*** 0.0387*** 0.00631*** (0.00173) (0.000655) (0.000636) (0.00135) (0.00166) LEV 0.112*** 0.0296*** 0.117*** -0.459*** -0.546*** (0.0248) (0.00912) (0.00886) (0.0188) (0.0231) ROA -0.302*** 0.799*** -0.771*** -0.387*** 1.183*** (0.0496) (0.0182) (0.0177) (0.0375) (0.0462) GROWTH 0.00458*** 0.00600*** -0.0106*** -0.000155 0.0165*** (0.00130) (0.000482) (0.000468) (0.000993) (0.00122) RDINVEST 0.0339 0.0396 -0.721*** 0.675*** 1.435*** (0.0715) (0.0263) (0.0256) (0.0543) (0.0668) FIN_RES 0.644*** -0.0126 0.0198* -0.208*** -0.240*** (0.0289) (0.0109) (0.0105) (0.0224) (0.0275) Constant -1.333*** 0.268*** -0.136*** -0.478*** -0.0746* (0.0449) (0.0171) (0.0166) (0.0352) (0.0433) Observations 21,820 21,820 21,820 21,820 21,820 R-squared 0.125 0.157 0.193 0.086 0.101

Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

(27)

27

5.3.2 Real earnings management

The results with regard to the REM regression, presented in columns two till five, show negative and significant effect (P<0.01) of CSR*IP on ABN_DISEXP and COMB_REM. These findings suggest that firms located in countries with stronger investor protection and are committed to CSR activities, are more likely to be engaged in earnings management by manipulating abnormal discretionary expenses and overall real activity manipulation. The findings with regard to abnormal operating cash flow and abnormal production costs manipulation are not significant.

As found in the regression to test hypothesis 1, DACC as control variable is positive (negative) associated with ABN_CFO, ABN_DISEXP, and COMB_REM (ABN_PROD). Turning to the control variables, the regression of COMB_REM shows positive and significant effects with regard to SIZE, ROA, GROWTH, and RDINVEST. LEV and FIN_RES are negatively and significantly associated with COM_REM. These results are equal to the results obtained from the regressions to test hypotheses 1.

The results discussed above lead to reject null hypotheses 2 for both AEM and REM, since it has been found that investor protection moderates the relationship between CSR and EM.

5.4 Additional analysis

5.4.1 CSR without corporate governance score

In order to test hypothesis 1 and 2, the CSR measures used as independent variable consisted of the equal-weighted rating of environmental score, social score, and corporate governance score. However, the relation between corporate governance and financial reporting practices has widely been investigated in prior research (e.g. Klein, 2002; Bergstresser and Philippon, 2006). Kim et al. (2012) therefore excluded corporate governance from their CSR measure. Following Kim et al. (2012), additional analysis will examine the effect of excluding corporate governance from the CSR score. The corporate governance score will be included as a control variable in regression model 3.1, presented in table 6. Table 7 presents the results of this same regression, but here the moderating role of investor protection on CSR is added, resulting in model 3.2.

The results in table 6 show that CSR is negatively and significantly (p<0.01) associated with DACC, which is in line with the findings with regard to hypothesis 1. This result indicates that CSR firms are less likely to be engaged in accrual-based earnings management. The results of hypotheses 1 hold when corporate governance is excluded from the CSR score. However, when the moderating role of investor protection is added to the model, no significant relation has been found for DACC. This is in contrast to model 2.1, where the CSR measure includes corporate governance. In other words, the results of hypothesis 2 with regard to DACC do not hold when the CSR measure does not contain corporate governance.

Earnings management and Corporate Social Responsibility