Amsterdam Business School
‘Is corporate social responsibility associated with earnings
management?’
Master thesis
Name Sjoerd Noordewier
Student number 10278915
Study MSc Accountancy & Control, Accountancy track First supervisor Dr. A. Sikalidis
Second supervisor Dr. G. Georgakopoulos Submission date 23th of June, 2014
Abstract
This study examines the relationship between corporate social responsibility and earnings management. I hypothesize that managers who engage in corporate social responsibility activities are less likely to engage in earnings management. This would mean that managers who actively participate in corporate social responsibility activities will act in an ethical manner and therefore provide stakeholders with reliable and transparent financial reports. Companies listed in the Standard & Poor’s 500 index between 2007 and 2011 are used to investigate the relationship described as above. The findings of this study provide support for the hypothesis and thus suggest that managers who engage in corporate social responsibility show truly ethical behaviour and provide reliable and more transparent financial reports than managers who participate less in corporate social responsibility activities.
Contents
1. Introduction ... 3
2. Literature review ... 6
2.1 CSR ... 6
2.2 Earnings Management ... 7
2.2.1 Earnings management through discretionary accruals... 8
2.2.2 Earnings management through real activities manipulations ... 8
2.3 Agency Theory ... 9
2.4 CSR and earnings management ... 10
3. Hypotheses development ... 11
4. Research design ... 13
4.1 Data sample ... 13
4.2 Measurement for CSR ... 13
4.3 Measurement of discretionary accruals ... 14
4.4 Measurement of real activities manipulation ... 14
4.5 Empirical model of discretionary accruals ... 15
4.6 Empirical model for real activities manipulation ... 15
4.7 Accounting and Auditing Enforcement Releases (AAERs) ... 17
4.8 Control variables ... 17
4.9 Empirical models ... 18
5. Results ... 20
5.1 Descriptive statistics ... 20
5.2 Main results ... 25
5.2.1 CSR and earnings manipulation through discretionary accruals ... 25
5.2.2 CSR and earnings management through real activities manipulation ... 27
5.2.3 Earnings management and individual CSR issue area’s ... 29
5.2.4. Earnings management and total CSR strengths and total CSR concerns ... 29
5.2.5. Earnings management and individual CSR components ... 32
6. Conclusion ... 37
1. Introduction
Corporate social responsibility (hereafter CSR) is a concept which has seen great development in the last decades. Now is the time for companies to improve stakeholder relationships by engaging CSR activities and show to society they do not only pursue profit (Lindgreen and Swaen, 2010). CSR is now becoming one of the main topics on the research agendas and even leads to the following discussion: ‘not only is doing good the right thing to do, it also leads to doing better’ (Sen, Bhattacharya and Korschun, 2006). An increase in transparent financial reporting and a significant change in attitude towards CSR has led to an important change in responsible behaviour by firms (Kim, Park and Wier, 2012). CSR is related to ethical issues and moral behaviour of a company and it addresses complexities such as the environmental problems, transparent financial reporting and it deal with the beliefs of many stakeholders (Prior, Surroca and Tribó, 2008). To address such complexities, organizations engage in CSR and communicate these activities through their websites, through their annual reports and through separate reports and thereby increase their reputation with several stakeholders (Sweeney and Coughlan, 2008).
The earning numbers of a firm are the most important statistics to several different stakeholder groups, according to Prior et al. (2008) the financial reports of a firm able them to distinguish themselves from the poor performers. Stakeholders base their financial decision making on the financial reports of a firm and this gives companies the ability to provide financial information to attract shareholders (Hong and Andersen, 2011). Managers have the ability to perform earnings management within the boundaries of generally accepted accounting principles and thus misleading stakeholders by consciously increasing or decreasing the performance of a company or to influence contractual outcomes (Healy and Wahlen, 1999). These actions could lead to an increase in reputation, if the activities are performed without the stakeholders’ knowledge, which seems rather unethical.
CSR behaviour can also result in a reputation increase, although these activities are seen as more ethical as society also benefits from this. Within this research I want to investigate whether managers who actively participate in CSR are truly ethical or if they entrench behind the positive image of CSR and do engage in earnings management. This results in the following research question:
To investigate the relationship between CSR and earnings management, I will follow the research model of Kim et al. (2012). Beforehand, I will conduct a literature review on scientific articles to obtain information about the different concepts in this study. Therefore, in my literature review I will provide an explanation of the concepts of CSR, earnings management through both discretionary accrual and earnings management through real activities manipulation. Then I will use evidence of existing literature to enhance about the relationship among CSR and earnings management to set my hypotheses. Due to the literature review and the research of Kim et al. (2012) I hypothesize there is a negative relationship between CSR and earnings management, this would suggest that firms who actively participate in CSR activities have ethical motivation and would thus engage less in earnings management. I have analysed the companies who are listed on the Standard & Poor’s 500 index (hereafter S&P 500) of 2007 until 2011. Necessary data to estimate discretionary accruals and real activities manipulation is gathered from Compustat. To estimate discretionary accruals the Modified Jones Model is used (Kim et al., 2012). To estimate real activities manipulation the abnormal levels of cash flows from operations were calculated, along with the abnormal level of production costs and the level of abnormal discretionary expenses. Finally, the three proxies for real activities manipulation formed one combined proxy, consistent with Cohen, Dey and Lys (2008).
To determine the CSR commitment of a firm the Kinder Lydenburg and Domini (hereafter KLD) database is used to obtain information. This database analyses firms and judges a firm on a certain set of strengths and concern on different criteria. In this study, five categories (community, diversity, employee relations, environment and product) were used to measure a firm’s CSR score. The total number of concerns was subtracted from the total number of strengths per category, while the net score of the five categories was added. A firm was ranked as a CSR firm if it received a positive net score, and a firm was ranked as a Non-CSR firm otherwise. As said, the research model of Kim et al. (2012) is used to analyse the data.
This study attempts to shed new light on the relationship between earnings management and CSR, as the findings of existing literature show different results. Furthermore, I want to add to existing literature by investigating the effects of individual CSR categories on constraining earnings management.
The most important results of this study suggest that, in this sample, firms who engage in CSR activities are less likely to engage in earnings management, through both discretionary accruals and through real activities manipulation. Although the findings for the latter means of
individual CSR categories indicate that high achievements in the CSR category community is more effective in constraining earnings management through real activities manipulation, although these results should be interpreted with cautiousness as the differences between these results are little.
My motivation for this study lies within several events in the past decades, as there are the major accounting scandals which caused society to revolt against the major accounting firms and its’ profession. Earnings management was of great interest during these scandals and is seen by some as the instigator of the financial crisis. Also, the recent evolvements with the Sarbanes-Oxley Act and the intervention to reduce earnings management, which might or might not have led to a reduction in earnings management. Or even to the reduction of one means of earnings management and an increase in the other means of earnings management. These events in the last decade are interesting developments in the accounting profession. Along with the evolution and the growing interest in CSR make it interesting to see whether there is a real ethical change or firms evolve along with the demands of society.
Existing literature shows evidence about the relationship between CSR, although conclusions about this relation differ. Several researchers conclude that CSR and earnings management are positively related (e.g. Prior et al., 2008), while others find evidence that there is a negative relationship between these two concepts (e.g. Kim et al., 2012). It is interesting to see whether managers and companies use CSR as an entrenchment tool to regain reputation lost because of earnings management or if firms are having true ethical believes and actively participate in CSR and thus make less use of earnings management. For this research, several issue areas from the KLD database are used to estimate a CSR measure per firm, which is done in prior research (e.g. Hong and Andersen, 2011; Kim et al., 2012). To contribute to the existing literature, I want to investigate what the effects of the individual CSR issue areas are on the subject of restraining earnings management. The effects of individual CSR issue areas on earnings management have not been investigated in existing literature. Hence it would add to the literature and it would benefit society by providing insight in what components of CSR constrain earnings management.
2. Literature review
In this chapter I will provide a background of the main aspects of this thesis, along with a review of prior literature about this subject. As mentioned in the introduction, this thesis consists of two main concepts. These concepts being CSR and earnings management. The agency theory is relevant for this study. Since the theory will have an impact on earnings management in a firm it will thus be explained. The background of this research will therefore consist of several paragraphs, in which the concepts of CSR, earnings management and agency theory will be explained. Furthermore, this chapter will elaborate about the existing relationships of the different concepts, based on existing literature. Starting with an explanation of CSR, followed by earnings management. Then the agency theory will be explained, which is followed by the relationship between CSR and earnings management.
2.1 CSR
As of today, the demand for CSR is increasing and reaches the top of the research agendas (Dhaliwal, Zhen, Tsang and Yang, 2011; Greenfield, 2004; Lindgreen and Swaen, 2010). The increasing demand of CSR reporting is met by including CSR reports in annual reports of companies (Sweeney and Coughlan, 2008), issue separate CSR reports (Neu, Warsame and Pedwell, 1998) or by reporting on CSR activities on the website of the company (Maignan and Ralston, 2002). Because of these recent events, CSR has become a reality instead of an idea and it should be a necessity for companies to carry out their role in current society to prove they behave ethically (Lindgreen and Swaen, 2010).
CSR is a very widely known concept nowadays, the origin of CSR however goes back to and has been evolving since the early 1930’s. Carroll (1979) is one of the first and more influential thinkers about the current view on CSR, defining a four-part framework which provides categories that society expects to be conducted by businesses. The definition by Carroll (1979) is as follows: “The social responsibility of business encompasses the economic, legal, ethical, and discretionary expectations that society has of organizations at a given point in time (Carroll 1979, p. 500).” This definition about CSR provides a broad statement about the behaviour of companies, as they should not only look at making profit, but also comply with laws and regulations, behave in an ethical way and satisfy a generally accepted view by society. According to Frederick (2008) the 1990’s and 2000’s became the most important era for global corporate citizenship, even though the early 2000’s endured some very crucial and
crisis was also triggered during this period. Despite the worldwide consequences of the crisis, the evolution of CSR kept going and significant progress was made during this period (Carrol and Shabana, 2010). Carrol and Shabana (2010) find that CSR has become a dominant case for the business community, as they want to rationalize and legitimize the activities they are performing. Bernstein and Pointon (2000) state that the public view provides strong support for companies to engage in CSR activities. The public seems to believe that companies should also take good care of their employees and add value to all stakeholders, along with their pursuit of profit, even if this means they will have to immolate some of these profits.
Porter and Kramer (2006) speak of ‘four prevailing justifications for CSR’. These are the moral obligation of a company, sustainability of the environment, a license to operate which is granted by stakeholders and acting socially responsible to increase the firms’ reputation. The first argument, the moral obligation, argues that companies should strive for good citizenship by providing reliable financial statements, comply with the law and make ethically accepted decisions. Sustainability refers to the triple bottom line of economic, social and environmental performance by demanding stewardship towards society and the environment. The license to operate has a more pragmatic approach than sustainability as companies need approval of stakeholders to continue their business. Finally, a firm needs to act socially responsible to increase its’ reputation. With these four arguments to engage CSR, Porter and Kramer (2006) argue that a company can improve its image and eventually even increase the value of its shares.
2.2 Earnings Management
External capital providers, suppliers and other stakeholders share a great interest in the accounting earnings of a company, as these are the most cited performance statistics with which companies can discern themselves from poor performers (Healy and Wahlen, 1999). Healy and Wahlen (1999) also state that managers have the ability to mislead stakeholders by adjusting their accounting earnings and thus their performance statistics, or in the same way are able to influence contractual outcomes. Managers are able to perform such activities without conducting illegal actions, as they have the ability to exercise some discretion in calculating their earnings without violating the generally accepted accounting principles (hereafter GAAP). This causes a distorted image of the performance statistics of a company, as the accounting earnings appear to be greater or less than they are in reality (Prior et al., 2008). Firms have the ability to engage in earnings management by two means, which are earnings management through discretionary accruals and earnings management through real activities manipulation
(Cohen et al., 2008). The differences between these two concepts are explained in the following two sections.
Kinney Jr, Palmrose and Scholz (2004) describe earnings management as actions which are conducted by management that reduce the quality of earnings in the financial statements of a company. Lev (2003) enhances this statement by arguing that it not only reduces the earnings quality, but it also lowers the predictive ability of future cash flows and earnings. Kaplan (2001) raises the discussion whether the engagement in earnings management is ethical or not. If the users of financial statements see earnings management as unethical, this behaviour may lead to reputation damage for managers and companies. Within this discussion, Kaplan (2001) provides evidence of two examples which have led to negative impact after the use of earnings management. The first example is one of Beneish (1997), who finds that firms who have infringed the GAAP earn negative returns for two subsequent years after their disclosure of the GAAP violation. Second, Dechow, Sloan and Sweeney (1996) found that allegations made by financial press or the Security and Exchange Commission (hereafter SEC) are associated with negative stock returns. To prevent a company from such consequences, executives might protect their damaged reputation by entrenching themselves behind CSR and distract stakeholders from potential earnings management (Cespa and Cestone, 2007).
2.2.1 Earnings management through discretionary accruals
As stated in the previous section, managing discretionary accruals is one form of earnings management. Engaging in discretionary accruals leads to discretion in earnings without cash flow consequences (Roychowdhury, 2006). Managers generally have some discretion over accruals, which can either be used positively or negatively. The discretion may be used positively by conveying private information to shareholders or negatively when discretionary accruals are used when earnings are opportunistically managed (Dechow, 1994). For instance, companies can engage in earnings management through discretionary accruals by changing their estimates of warranty liabilities thereby increasing the earnings of the firm (Hong & Andersen, 2011).
2.2.2 Earnings management through real activities manipulations
The second form of earnings management is real activities manipulation. The main difference between discretionary accruals and real activities manipulation is that the latter affects cash flows, while it sometimes also affects accruals (Roychowdhury, 2006). A definition of real activities manipulation is also given by Roychowdhury (2006): “departures from normal
operational practices, motivated by managers’ desire to mislead at least some stakeholders into believing certain financial reporting goals have been met in the normal course of operations.” This definition implies that managers of firms have incentives to perform certain discretion to meet certain targets and thereby mislead stakeholders. According to Graham, Harvey and Rajgopal (2005) managers are consciously manipulating earnings to meet targets such as last year’s earnings, forecasts set by analysts and beat the zero earnings benchmark even though it might harm the reputation and value of the firm.
Existing research about the consequences of real activities finds different evidence. For instance, Bhojraj, Hribar, Picconi and McInnis (2009) find that firms who engage in real activities manipulation and just beat the forecasts of analysts will perform worse in the subsequent three years compared to firms who not meet the forecasts of analysts and do not engage in earnings management. Gunny (2010) provides evidence for contrary findings, were firms who engage in real activities manipulation will outperform firms who did not engage in real activities manipulation and just miss out on the forecasts of analytics.
2.3 Agency Theory
Jensen and Meckling (1976) define the relationship between a principal and an agent as follows: “a contract under which one or more persons (the principal(s)) engage another person (the agent) to perform some service on their behalf which involves delegating some decision making authority to the agent” (p. 308). An example of this, could be a shareholder (principal) hiring a manager (agent) to perform managerial duties in the best interest of the company. As both the principal and the agent have different interests, information asymmetry could arise. Therefore, it is very difficult to align their behaviour and their interests and the manager could act in such a way that it is only beneficiary to himself.
To overcome certain problems, the shareholder could provide the agent with equity or other incentives, to align the interests of both the principal (shareholder) and the agent (manager). As the manager is, however, able to utilize his own profit by engaging in earnings management and receive certain incentives, which is not an uncommon practice (Davidson III, Jiraporn, Kim and Nemec, 2004). This might lead to earnings management being an agency cost, as the manager will release financial statements which may not be consistent with the real economic state of the firm. Eventually, stakeholders possibly make investment decisions which are not optimal, because they are based on these financial statements
2.4 CSR and earnings management
Prior et al. (2008) argue that earnings management activities conducted by management do not only affect its shareholders, it also has an impact on other stakeholders. This is due to the misleading consequences of earnings management, meaning that untrue values of firm’s assets or financial positions are reflected in the financial statements. This could lead to serious consequences for shareholders, employees, creditors and the whole society (Zahra, Priem and Rasheed, 2005). Some of the consequences of these actions could be a decrease in value on the stock market, overstated loans by banks based upon boosted earnings and managers might even reduce the earnings during labor union contract negotiations to reduce the loan costs (Prior et al., 2008). The media has the ability to intensify the effect of actions undertaken by management and is therefore able to reduce such actions. The effect, or even threat, of negative media attention has two consequences for managerial practices (Bansal, 2005). The first being pressure from negative publicity to engage sustainable development. Another consequence is encouraged stakeholders who provoke governments and lobby organizations. This negative image could tempt stakeholders into punishing management (Rowley and Berman, 2000).
The circumstances above could lead to a severe reputation damage to the CEO and if CEOs want to protect their job, and reputation, they might want to participate in a wide range of activities to improve their relationship with corporate stakeholders and environmental activists, thus leading to CSR activities (Prior et al., 2008). The relationship above is written from the starting point of earnings management and has been severely tested, in contrast with the effects of CSR on earnings management, which might lead to another conclusion. This relation will be elaborated in the following paragraph.
Although less than the other way around, the effects of CSR on earnings management have been tested in prior studies. By investigating this relationship researchers have found several conclusions, such as increasing CSR performances by firms over the years, which might have a positive effect on the agency problem (Liu and Lu, 2007). Furthermore, Jones (1995) suggests that agency costs could be the result of undertaking CSR activities that have a positive effect on stakeholder relations. Reasoning for the former two conclusions is found in the article written by Chih, Shen and Kang (2008), as the authors state that a socially responsible firm will not conduct unethical actions and thus will not use its ability to manage earnings. CSR is associated with transparency, which is in contrast with earnings management. Hence, companies who actively participate in CSR should not perform earnings management, as these actions are seen as unethical.
In the light of the evidence provided by Chih et al. (2008), the research of Kim et al. (2012) confirms this statement. Kim et al. (2012) find that firms who are actively participating in CSR activities are less likely to engage in earnings management through both discretionary accruals and earnings management.
3. Hypotheses development
Existing literature suggests that CSR and earnings management know two different connections. The most investigated connection is where CSR is used as an entrenchment mechanism to increase the reputation of a firm or manager when their reputation was damaged due to earnings management (Hemingway and Maclagan, 2004). If managers will engage in CSR activities due to opportunistic behaviour with earnings managements, a positive relation between CSR and earnings management would be expected.
More recent ideas and articles look at the other connection, from the point of view of CSR, thus looking at an ethical concept. Lichtenstein, Drumwright and Braig (2004) argue that CSR has become reality and companies have to accept that certain social standards are set by society to continue their operations. Furthermore, as said in the literature review, Porter and Kramer (2006) provide in their paper four different justifications for a firm to participate in CSR activities. One of those justifications is the moral obligation to pursue good citizenship and provide reliable financial statements. These statements suggest that companies have an ethical obligation to provide stakeholders with transparent and reliable financial statements. As earnings management is still seen as unethical, a negative relationship between CSR and earnings management is expected, leading to the following hypothesis:
Ethical hypothesis A firm with a higher ranking of CSR is less likely to engage in earnings management.
On the contrary, firms might engage in CSR activities, while they also participate in earnings management, albeit through discretionary accruals or through real activities manipulation. This is suggested in the paper of Prior et al. (2008) as they state that CSR is a powerful tool to increase the reputation of the firm with the stakeholders of the company and gain support. Through this way managers are able to restore their faith with for instance the board or other stakeholders after they have engaged in earnings management. If managers would engage in such activities described as above, they would use CSR as entrenchment. Hemingway and
Maclagan (2004) support this reasoning, by arguing that firms would engage in CSR activities to cover up misconduct which has harmed the firm. They also state that CSR will restore their reputation and they can continue operations and earnings management. Although they would seemingly operate in a transparent way, they in fact entrench behind the positive image of CSR and still engage in earnings management.
Actions performed by management described as above would suggest that managers have an incentive to engage in CSR while they are also actively participating in earnings management. Therefore, a positive relationship between earnings management and CSR is expected, which leads to the following hypothesis:
Entrenchment hypothesis A firm with a higher ranking of CSR is more likely to engage in earnings management.
According to Kim et al. (2012) real activities manipulation and discretionary accruals are used as substitutes, so they use both means of earnings management as an alternative for each other. This evidence is supported by the research of Cohen et al. (2008). A relationship as such, would suggest that earnings management through discretionary accruals and earnings management through real activities manipulation have a negative relationship. Therefore, the following hypothesis is formulated:
Substitution hypothesis There is a negative relationship between earnings management through discretionary accruals and earnings management through real activities manipulation.
4. Research design
In this chapter I will describe the research methodology that I have used to test my hypotheses. First, the sample selection is clarified. Then I will provide a description of the CSR scores and KLD database. Then the model to calculate discretionary accruals is explained, followed by the empirical model of real activities manipulation. Then the Accounting and Auditing Enforcement Releases (AAERs) are clarified, followed by the control variables.
4.1 Data sample
This research is restricted to companies that are listed on the Standard & Poor’s 500 and to companies of which data was available on Compustat from 2007 until 2011. Financial institutions (SIC codes 6000 – 6999) are left out of the sample as characteristics to calculate accruals are different for these firms. Companies with missing values were removed and outliers were winsorized. The data of the KLD database was gathered of the remaining firms. Data on equity offerings was hand collected from the Securities Data Company’s New Issue Database (now: Thomson One). Furthermore, firm age was collected from the CRSP database and enforcement data was hand collected from the SEC website. My final sample consists of 261 firms and 1.104 observations.
4.2 Measurement for CSR
To classify firms into certain CSR categories, the Kinder Lydenburg and Domini database (KLD) is heavily used for other articles to gather information and is argued to be one of the best sources available for CSR (Hillman and Keim, 2001; Hong and Andersen, 2011; Kim et al., 2012). Hillman and Keim (2001) even argue that the KLD database is the best source to obtain CSR information. This database has covered the Standard & Poor’s 500 index since 1991 and in 2002 the database was extended by including the 3000 largest U.S. publicly traded firms. The companies in the KLD database are reviewed by independent research analysts. They use several different sources such as financial statements, academic articles and surveys in different categories to conclude whether a company is acting socially responsible or not. For this research I will use the same CSR categories as Kim et al. (2012), which are the strengths and concerns of community, diversity, employee relations, environment and product to construct the CSR measure and the strengths and concerns of corporate governance will be used as a control variable. Corporate governance is not included in the construct for CSR, as
corporate governance is seen as a different construct from CSR, although it is included as a control variable.
The five categories that complete the construct of CSR, community, diversity, employee relations, environment and product are measured by subtracting the total concerns from the total strengths, which will lead to a net score per category. Eventually, all category net scores are added and this total net score will determine whether a company is rated as a CSR firm, when achieving a positive score, or as a Non-CSR firm otherwise.
4.3 Measurement of discretionary accruals
Many studies use discretionary accruals as a measurement for either audit quality or earnings management (Kim et al., 2012; Cohen et al., 2008; Hong and Andersen, 2011). In this study I will use two proxies for earnings management, of which discretionary accruals is the first measure. A higher amount of discretionary accruals is associated with lower earnings quality and thus with firms who engage in earnings management. To estimate the level of discretionary accruals the Modified Jones Model is used and by expanding the model with ROAt-1 as in Kothari, Leone and Wasley (2005) the reliability is enhanced.
A negative relation between CSR and discretionary accruals is expected, which would support the Ethical hypothesis. Discretionary accruals are denoted as ResDA, as it is the residual value of the model. Absolute values of discretionary accruals are used, as earnings management can be expressed through income-increasing or income-decreasing accruals (Kim et al., 2012).
4.4 Measurement of real activities manipulation
As said, the second proxy for earnings management is real activities manipulation. Existing research mainly focuses on three different proxies to capture real activities manipulation (Roychowdhury, 2006; Kim et al., 2012; Cohen et al., 2008). These three proxies are; abnormal levels of operating cash flows, denoted ResCFO, abnormal production costs, denoted ResPROD, and abnormal discretionary expenses, denoted ResEXP. Furthermore, as in Cohen et al. (2008) I will use a combined proxy that consists of the three real activities manipulations measurements mentioned before, although this proxy will be calculated as in the paper of Kim et al. (2012). Therefore the following formula is used for the combined proxy, or CRAM; ResCFO – ResPROD + ResEXP.
A positive relation between CSR and ResCFO, ResEXP and CRAM is expected, while a negative relation between CSR and ResPROD is expected, as this would be consistent with the
research of Kim et al. (2012) and Cohen et al. (2008). This would also provide evidence to support the Ethical hypothesis.
4.5 Empirical model of discretionary accruals
To estimate both empirical models, the research of Kim et al. (2012) is used. For the discretionary accruals, the residual is used of a cross-sectional regression for industry-years with at least 10 observations. To improve reliability, the Modified Jones Model is expanded by including ROAt-1 (Kothari et al., 2005). This leads to the following regression:
TAit/Ait-1 = 0 (1/Ait-1) + 1 (REVit – RECit)/Ait-1 + 2 PPEit/Ait + 3 IBXIit-1/Ait-1 + it
Where:
TAit = total accruals for a firm i in year t;
REVit = change in net revenues in year t from year t-1; RECit = change in net receivables in year t from year t-1; PPEit = gross property, plant and equipment for firm i in year t; IBXIit-1 = income before extraordinary items in year t-1;
Ait-1 = lagged total assets;
it = residual of firm i in year t;
4.6 Empirical model for real activities manipulation
Real activities manipulation is earnings management with a cash flow effect by pushing forward or delaying economic activities. There are three proxies that will be used to estimate real activities manipulation. The model of Roychowdhury (2006) is used to estimate the normal level of operating cash flows or the acceleration of the timing of sales through increased price discounts or more lenient credit terms. This leads to the following model:
CFOt/At-1 = 0 + 1 (1/At-1) + 1 (St + At-1) + 2 (St/At-1) + t Where:
CFOt = cash flow from operation in year t; At-1 = lagged total assets;
S = change in net sales in year t from year t-1; t = residual of firm i in year t;
The abnormal cash flow from operations or ResCFO is the residual from the mode.
The second proxy for real activities manipulation is reporting of lower cost of goods sold through increased production or overproduction. In Kim et al. (2012) production costs is defined as the sum of cost of goods sold, denoted COGS, and change in inventory during the year, denoted INV. The model for COGS is:
COGSt/At-1 = 0 + 1 (1/At-1) + (St/At-1) + t
In this model COGSt is the cost of goods sold in year t. A somewhat similar model is used for the growth in inventory:
INVt/At-1 = 0 + 1 (1/At-1) + 1 (St/At-1) + 2 (St-1/At-1) + t
Consistent with Roychowdhury (2006) and Kim et al. (2012) production costs is defined as the sum of the two equations above, leading to PRODt = COGSt + INVt. This leads to the following
model:
PRODt/At-1 = 0 + 1 (1/At-1) + 1 (S/At-1) + 2 (S/At-1) + 3 (St-1/At-1) + t The abnormal production cost or ResPROD is the residual from the model.
The third and last proxy for real activities manipulation is the abnormal discretionary expenses, or the decrease in discretionary expenses that include advertising expenses, research and development (denoted R&D) expenses and selling, general and administrative (denoted SG&A) expenses. Consistent with Roychowdhury (2006) and Kim et al. (2012) the following equation is used:
where EXPt is the discretionary expense in year t, calculated by the sum of advertising expense, R&D expense and SG&A expense. The residual of the model, ResEXP, is the abnormal discretionary expenditure. Following Roychowdhury (2006), when R&D and advertising expenses are unavailable and SG&A expenses are available, the others are set to 0.
Finally, a combined proxy of real activities manipulation is constructed. This is done by the following sum; ResCFO – ResPROD + ResEXP. This is consistent with Kim et al. (2012) and differs from Cohen et al. (2008) as they sum all proxies to estimate the combined proxy.
4.7 Accounting and Auditing Enforcement Releases (AAERs)
Dechow et al. (1996) argues that firms may be subject to AAERs due to earnings management. AAER is identified as a firm is sanctioned to Section 13 (a) of the Securities Exchange Act of 1934. AAER is hand collected from the SEC website. AAER is set as a dummy, were 1 means the firm is subject to enforcement in that fiscal year and 0 otherwise. A negative relationship between AAER and CSR is expected.
4.8 Control variables
To control for potential affection of the discretionary accruals and real activities manipulation by other determinants, several control variables are included. Consistent with the research of Kim et al. (2012) the following control variables are added:
SIZE Natural logarithm of the market value of equity (MVE)
MB Market-to-book equity ratio, measured as MVE/BVE, where BVE is the book value of equity
ADJROA Industry mean-adjusted ROA in the previous year, where ROA is measured as income before extraordinary items, scaled by lagged total assets
AU An indicator variable that takes a value of 1 if the firm is audited by a Big 4 auditor and 0 otherwise
LEV Long-term debt scaled by total assets
EO An indicator variable that takes on a value of 1 if the firm has equity offerings in the following year and 0 otherwise
SGAint SG&A intensity for the two-digit SIC code industry for the year
CGnet Net score of KLD ratings in the corporate governance area, total strengths minus total concerns
Admired an indicator variable that takes a value of 1 if the firm is listed in the Fortune’s America’s Most Admired 50 Companies, and 0 otherwise
In this study, ten control variables are used. Roychowdhury (2006) argues that growth opportunity and size of a firm explain variation in earnings management. Therefore, the first two control variables, SIZE and MB are included. These two proxies are correlated with CSR performance of firms (Roychowdhury, 2006). To control for financial performance of a firm, the industry adjusted return on assets is also included (AJDROA). As there could be differences in earnings management activities between firms audited by Big 4 auditors and firms audited by other firms, AU is included as a dummy variable.
Similar to the research of Kim et al. (2012) leverage and equity offerings in the following year are included to control for incentives for earnings management. The measure for equity offerings are gathered from the Thomson One database. SG&A intensity in the industry is positively associated with CSR and Earnings (Roychowdhury, 2006). As corporate governance is a separate construct from CSR, while both can affect the financial reporting behaviour of a firm, this is included as a control variable (CGnet). Due to the changing conditions when a firm matures, for instance financial reporting and CSR activity could change over time, AGE is included a control variable for development of these factors. Finally, Admired is included as firm reputation, since reputation and earnings management are positively related (Musteen, Datta and Kemmerer, 2009) and Kim et al. (2012) suggest that KLD’s evaluation of a firm might also be compromised. In the research of Kim et al. (2012) the complete Fortune’s America’s Most Admired Companies list is used, while for my research only the top fifty is included, as my sample is significantly smaller. Furthermore, as a relative big share of the companies on the S&P 500 list are also listed on the Fortune’s America’s Most Admired Companies list, this control variable would lose its value.
4.9 Empirical models
Consistent with the research of Kim et al. (2012), the following empirical model is used to measure the relationship between earnings management and CSR:
ResDAt = 0 + 1 CSRt +2 CRAMt + 3 SIZEt-1 + 4 MBt-1 + 5 ADJROAt-1 + 6 AUt + 7 LEVt-1 + 8 EOt + SGAintt + 10 CGnett + 11 AGEt + 12 Admired + t
CRAMt = 0 + 1 CSRt +2 ResDAt + 3 SIZEt-1 + 4 MBt-1 + 5 ADJROAt-1 + 6 AUt + 7 LEVt-1 + 8 EOt + SGAintt + 10 CGnett + 11 AGEt + 12 Admired + t
The first regression, ResDA is the dependent variable, which is the absolute level of discretionary accruals. Furthermore, the positive discretionary accruals (PosDA) and negative discretionary accruals (NegDA) will also be used as a dependent variable. The same goes for the second regression, were CRAM as a dependent variable will be replaced by ResCFO, ResPROD and ResEXP. This is consistent with the research of Kim et al. (2012).
To contribute to the existing literature, I want to extend the research which was performed by Kim et al. (2012). I want to investigate whether the individual CSR components generated by KLD have more or less effect on restraining earnings management than the other individual CSR components. This means that in the regressions the combined and aggregated CSR measure will be replaced by a net score per CSR component. For instance, the community will be calculated by subtracting total community concerns from total community strengths and this net score will replace the aggregated CSR measure.
5. Results
In this section of the paper, the results of my research will be analyzed. The first paragraph will consist of an explanation of the descriptive statistics and will be followed by an analysis of the main results of the paper.
5.1 Descriptive statistics
Table 1 presents the distribution of firm-year observations, which shows how many firms represent each industry that is present in the sample selection. The industry with the most firms is the chemicals and allied products industry, with 195 firms which represents 17.66 percent of the sample (SIC code 28), followed by the Electric and Other Electric Equipment industry (132 observations which represent 13.32 percent of the sample, SIC code 36) and Industrial Machinery and Computer Equipment industry (132 firm observations which represent 11.96 percent, SIC code 35). The industry that represents the least firms is the communication industry, which represents only 76 observations or 6.88 percent of the sample (SIC code 48).
Table 2 of the descriptive statistics shows three distinct tables. The first table, Panel A, reports the descriptive statistics of this study. Panel B will also report on the descriptive
Industry Two-Digit S IC # of obs % of sample Cummulative Percent
Oil and Gas 13 126 11.41 11.41
Food, BeverAGE 20 93 8.42 19.84
Chemicals and Allied Products 28 195 17.66 37.50
Industrial M achinery and Computer Equipment 35 132 11.96 49.46 Electronic and Other Electric Equipment 36 147 13.32 62.77
Transportation Equipment 37 68 6.16 68.93
Instruments and Related Products 38 101 9.15 78.08
Communication 48 76 6.88 84.96
General M erchandise Store 53 43 3.89 88.86
Business Services 73 123 11.14 100.00
1.104 100.00% TABLE 1
S ample Description
statistics, although in this panel a distinction is made between CSR firms and Non-CSR firms. The third table, Panel C, presents a Pearson correlation coefficients table of the variables that are used in this thesis. The continuous variables that are used in this paper are all winsorized at the first and ninety-ninth percentile.
As can be seen in Panel A of table 2, ResCFO, ResPROD, ResEXP and CRAM all show a mean value of approximately 0.000. This suggests that on average the firms in this sample do not participate in earnings management through real activities management. The absolute value of discretionary accruals, ResDA, shows a mean value of 0.035, fairly lower than in the study of Kim et al. (2012), who find a value of 0.200, which indicates that there is less earnings management through discretionary accruals in this sample.
Panel A - Full sample N Mean Median S td. Dev.
25th Percentile 75th Percentile Dependent Variables ResDA 1104 0.035 0.025 0.036 0.012 0.044 PosDA 616 0.032 0.025 0.029 0.011 0.045 NegDA 488 - 0.037 - 0.025 0.043 - 0.041 - 0.012 ResCFO 1104 - 0.000 - 0.003 0.060 - 0.037 0.033 ResPROD 1104 0.000 - 0.002 0.117 - 0.070 0.069 ResEXP 1104 0.000 - 0.009 0.131 - 0.076 0.058 CRAM 1104 0.000 - 0.005 0.258 - 0.162 0.136 Variables of Interest CSR 1104 1.633 1.000 4.072 - 1.000 4.000 Control Variables SIZE 1104 9.292 9.137 1.215 8.505 10.040 M B 1104 0.384 0.367 1.546 0.229 0.588 AdjROA 1104 0.074 0.076 0.093 0.037 0.116 AU 1104 0.997 1.000 0.052 1.000 1.000 LEV 1104 0.219 0.193 0.160 0.116 0.286 EO 1104 0.034 0.000 0.182 0.000 0.000 SGAint 1104 0.287 0.299 0.124 0.232 0.402 CGnet 1104 - 0.559 - 1.000 0.902 - 1.000 0.000 AGE 1104 35.290 31.000 23.942 16.000 48.000 ADM IRED 1104 0.078 0.000 0.268 0.000 0.000 TABLE 2 Descriptive statistics
The variable of interest in this study, CSR, has a mean of 1.633. This suggests that, on average, firms listed on the S&P 500 between 2007 and 2011 achieve a positive score in CSR activities according to the KLD rankings. Furthermore, the control variables in this research suggest that firms in this sample are more profitable than their peers as AdjROA is 0.074. More than 99 percent of the sample firms are audited by a Big 4 Audit firm. A little over 3 percent offers equity in the following year, as EO is 0.034. Although corporate governance was not included in the CSR construct, it was included as a control variable and the mean of CGnet suggests that the sample firms have a weak corporate governance, on average. The average age of the firms in this sample is over 35 years and almost 8 percent of the sample is represented the Fortune’s America’s most admired 50 companies list.
Panel B of table 2 reports on the descriptive statistics of the variables, although in this case a distinction is made between CSR firms and Non-CSR firms. A firm is classified as a
Panel B
N Mean Median N Mean Median t-test
Wilcoxon test Dependent Variables ResDA 764 0.033 0.025 340 0.037 0.025 0.091 0.281 PosDA 415 0.031 0.024 201 0.035 0.025 0.097 0.537 NegDA 349 - 0.036 - 0.025 139 - 0.040 - 0.027 0.340 0.313 ResCFO 764 - 0.006 0.001 340 - 0.014 - 0.012 < 0.001 < 0.001 ResPROD 764 - 0.015 - 0.015 340 0.033 0.021 < 0.001 < 0.001 ResEXP 764 0.017 0.005 340 - 0.038 - 0.025 < 0.001 < 0.001 CRAM 764 0.038 0.035 340 - 0.085 - 0.062 < 0.001 < 0.001 Variables of Interest CSR 764 3.445 2.000 340 - 2.438 - 2.000 < 0.001 < 0.001 Control Variables SIZE 764 9.390 9.229 340 9.071 9.000 < 0.001 < 0.001 M B 764 0.296 0.333 340 0.582 0.453 < 0.001 < 0.001 AdjROA 764 0.078 0.081 340 0.064 0.066 0.017 < 0.001 AU 764 0.999 1.000 340 0.994 1.000 0.294 0.178 LEV 764 0.220 0.193 340 0.216 0.194 0.731 0.799 EO 764 0.039 0.000 340 0.024 0.000 0.146 0.186 SGAint 764 0.311 0.299 340 0.234 0.232 < 0.001 < 0.001 CGnet 764 - 0.437 - 1.000 340 - 0.832 - 1.000 < 0.001 < 0.001 AGE 764 35.736 27.000 340 34.288 35.000 0.358 0.128 ADM IRED 764 0.094 0.000 340 0.0412 0.000 < 0.001 0.002
CS R Firms Non-CS R firms
Panel B Descriptive statistics CS R firms versus Non-CS R firms
CSR firm when it has a positive net score based on the total strengths and total concerns reported in the five issue areas of the KLD database. If firms achieve a negative net score when the total concerns are subtracted from the total strengths, the firm is classified as a Non-CSR firm.
The mean and median values of ResDA are slightly higher for Non-CSR firms than for CSR firms, although the mean between these two groups is statistically significant at the 0.1 level, this suggests CSR firms are less likely to engage in earnings management through discretionary accruals. This is in contrast with the measure for the CRAM proxy and for the real activities manipulation proxies itself, as the results between these groups are significant at the 0.01 level and show bigger differences amongst (Non-) CSR firms. ResCFO, ResEXP and CRAM have higher mean and median values for CSR firms, while the mean and median of ResPROD are higher for Non-CSR firms. Furthermore, the results suggest that CSR firms seem to be bigger in size, have a better corporate governance, although both results are negative (-0.437 against -0.832) and have a higher rate of firms which are listed in the Fortune’s 50 America’s most admired companies (10 percent against 4 percent).
The results of Panel B of Table 2 suggest that firms which are socially responsible are less likely to participate in earnings management. The mean and median value of ResDA also support this, but this result is only slightly significant. The proxies for real activities manipulation are statistically significant and therefore both findings show support the Ethical hypothesis, as CSR firms seem less likely to participate in earnings management.
The third Panel (C) of Table 2 presents the Pearson correlation coefficients between the variables. The most important correlations for this paper are the correlations between CSR and ResDA, ResCFO, ResPROD, ResEXP and CRAM. CSR is negatively correlated with ResDA (p < 0.01), which suggests that CSR firms are less likely to engage in earnings management through discretionary accruals. CSR is also significantly and negatively correlated with ResPROD and positively correlated with ResCFO, ResEXP and CRAM. These findings also suggest socially responsible firms are less likely to participate in earnings management than Non-CSR firms. Therefore, both findings support the Ethical hypothesis.
The most important findings in Table 2 are that the sample firms are CSR responsible on average. Panel B shows a higher mean value for ResCFO, ResEXP and CRAM for CSR firms and a lower mean value for ResPROD. Panel C correlations provide the same results and both therefore suggest that CSR firms are less likely to engage in earnings management. Hence, Table 2 provides support for the Ethical hypothesis.
Panel C
CSR Resda Rescfo ResPROD ResEXP CRAM SIZE M B AdjROA AU LEV EO SGAint Cgnet AGE ADM IRED
CSR 1.000 ResDA - 0.0974 1.0000 0.0012 ResCFO 0.1814 - 0.1524 1.0000 0.0000 0.0000 ResPROD - 0.2612 0.0911 - 0.5349 1.0000 0.0000 0.0025 0.0000 ResEXP 0.3190 0.0707 0.2008 - 0.6897 1.0000 0.0000 0.0188 0.0000 0.0000 CRAM 0.3234 - 0.0404 0.5759 - 0.9301 0.8704 1.0000 0.000 0.1797 0.0000 0.0000 0.0000 SIZE 0.2878 - 0.1641 0.3641 - 0.3040 0.1974 0.3229 1.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 M B - 0.0826 0.0338 - 0.0850 0.0680 - 0.0498 - 0.0759 0.0271 1.0000 0.0060 0.2612 0.0047 0.0239 0.0981 0.0116 0.3684 AdjROA 0.1112 0.3855 0.5298 - 0.3448 0.0325 0.2954 0.2876 0.0229 1.0000 0.0002 0.0000 0.0000 0.0000 0.2801 0.0000 0.0000 0.4473 AU 0.0380 - 0.0008 0.0343 - 0.0751 0.0554 0.0704 0.0669 0.0023 0.0089 1.0000 0.2065 0.9777 0.2554 0.0126 0.0655 0.0194 0.0261 0.9381 0.7687 LEV - 0.0392 0.0241 - 0.0562 0.0253 - 0.0681 - 0.0592 - 0.1347 - 0.0531 - 0.1743 - 0.0341 1.0000 0.1926 0.4247 0.0621 0.4017 0.0237 0.0494 0.0000 0.0779 0.0000 0.2571 EO 0.0353 0.0151 0.0470 - 0.0172 - 0.0040 0.0166 0.0172 0.0047 0.0460 0.0099 0.0212 1.0000 0.2408 0.6168 0.1190 0.5673 0.8946 0.5815 0.5672 0.8756 0.1265 0.7436 0.4811 SGAint 0.2996 - 0.0152 - 0.0000 - 0.0000 0.0000 0.0000 0.0323 - 0.0187 0.0843 - 0.0742 - 0.0194 0.0323 1.0000 0.0000 0.6137 1.0000 1.0000 1.0000 1.0000 0.2840 0.5359 0.0051 0.0136 0.5198 0.2842 CGnet 0.3548 - 0.0648 0.0637 - 0.0949 0.0649 0.0910 0.1335 - 0.0212 0.1144 0.0255 - 0.0251 - 0.0317 0.0599 1.0000 0.0000 0.0312 0.0344 0.0016 0.0309 0.0025 0.0000 0.4814 0.0001 0.3968 0.4055 0.2920 0.0468 AGE 0.0966 - 0.0555 - 0.0366 0.0576 - 0.0501 - 0.602 0.2908 - 0.0594 0.0178 - 0.0415 0.0396 - 0.0326 - 0.0212 0.1624 1.0000 0.0013 0.0653 0.2242 0.0557 0.0958 0.0454 0.0000 0.0485 0.5537 0.1678 0.2792 0.2792 0.4822 0.0000 ADM IRED 0.2787 - 0.0850 0.1276 - 0.1124 0.0956 0.1293 0.4728 - 0.0156 0.0907 0.0152 - 0.0645 0.0378 0.0378 0.0789 1.0000
5.2 Main results
The main results section provides seven tables about the results of the regression analysis of the relationship between earnings management and CSR. The first table (Table 3) will present the results of the multivariate regression analyses of the discretionary accrual proxies. Table 4 will report the regression analyses of the real activities manipulation proxies, while Table 5 presents the relationship between earnings management and the net scores of the five CSR issue areas. Table 6 reports on the relationship between earnings management and total CSR strengths and total CSR concerns. The last three tables report on the relationship between earnings management and individual CSR components. Table 7 shows a correlation table of the individual CSR components, Table 8 reports on its relation with earnings management through discretionary accruals and Table 9 presents the regression analysis of the separate CSR issue areas and its’ effect on real activities manipulation.
5.2.1 CSR and earnings manipulation through discretionary accruals
Table 3 reports on the analyses of discretionary accruals’ proxies by performing a multivariate regression, in line with the research performed by Kim et al. (2012). In this particular paper the test statistics and significance levels are based on the standard errors adjusted by a two-dimensional cluster, since residuals can be correlated.
For the CSR measure only the relation with ResDA and PosDA are significant (p < 0.1 and p < 0.05 respectively). These relations are both negative which provides support for the Ethical hypothesis, as a firm with high CSR achievements is less likely to engage in earnings management through discretionary accruals. The relation between NegDA and CSR is positive but is statistically insignificant. These results are consistent with the findings of Kim et al. (2012).
For the real activities manipulation proxy, CRAM, I find one result that is in contrast with the research of Cohen et al. (2008) and Kim et al. (2012). As the findings of the other papers suggest that real activities manipulation and discretionary accruals are substitutes of each other, the multivariate regression in Table 3 gives some indication that these two means of earnings management are too some extent complementary. The relation between CRAM and PosDA is negative and significant (p < 0.1) and thus suggests companies who participate in real earnings manipulation are more likely to engage in discretionary accruals. As the coefficient of this relation is fairly small and it is only slightly significant, this result should be interpreted with cautiousness, although it provides evidence to reject the Substitution hypothesis.
NegDA and ResDA do not show any significant relationship with CRAM. Interesting to note is the relationship between SIZE and ResDA and PosDA, as these results are significant (p < 0.05 and p < 0.01 respectively) and suggest that bigger firms are less likely to engage in earnings management through discretionary accruals.
Res DA Positive DA Negative DA
Coefficient Coefficient Coefficient
(t-stat) (t-stat) (t-stat)
CSR - 0.001 - 0.001 0.001 - (1.88)* - (2.14)** (0.85) CRAM 0.0034 - 0.009 - 0.013 (0.73) - (1.74)* - (1.56) SIZE - 0.003 - 0.003 0.001 - (2.50)** - (2.63)*** (0.72) M B 0.002 - 0.005 - 0.005 (0.348) - (1.87)* - (1.50) AdjROA - 0.007 0.007 0.027 - (0.57) (0.49) (1.28) AU 0.008 - 0.006 - 0.036 (0.38) - (0.27) - (0.82) LEV 0.003 - 0.009 - 0.011 (0.37) - (1.08) - (0.99) EO 0.003 0.006 - 0.001 (0.52) (1.01) - (0.12) SGAint 0.013 0.006 - 0.014 (1.41) (0.64) - (0.84) CGnet - 0.001 - 0.001 0.001 - (0.78) - (0.87) (0.32) AGE - 0.000 - 0.000 0.000 - (0.32) - (0.31) (0.38) ADM IRED - 0.002 0.002 0.009 - (0.46) (0.33) (1.03) Adj. R2 0.026 0.054 0.031 N 1,104 616 488
* Significant at the 10% level. ** Significant at the 5% level. *** Significant at the 1% level.
TABLE 3
5.2.2 CSR and earnings management through real activities manipulation
Table 4 is an analysis of the multiple regression of the real activities manipulation proxies and CSR. The test statistics and significance levels are based on the standard errors adjusted by a two-dimensional cluster, similar to Table 3.
The relation between CSR and the several proxies for real activities manipulation are all statistically significant. ResCFO, ResEXP and CRAM are positively related to CSR, while ResPROD is negatively related to CSR. All of these results in the regression suggest that CSR firms are less likely to participate in earnings management through real activities manipulation as higher level of abnormal operating cash flows, abnormal expenses and overall real activities manipulation and lower abnormal production are all indicators of conservative decisions made by a firms’ executives. These findings therefore support the Ethical hypothesis, which is consistent with the articles of Kim et al. (2012) and Cohen et al. (2008).
The relations between the proxies for real activities manipulation and ResDA show somewhat different results than in the research performed by Kim et al. (2012) and Cohen et al. (2008), as ResCFO and ResDA are significantly and negatively related. This would again, as is the case in Table 3, suggest that there is some evidence that real activities manipulation and discretionary accruals are complementary. In contrast, the relationship between ResEXP and ResDA are consistent with Kim et al. (2012), since this result is statistically significant and positive and thus suggests that both means of earnings management are substitutes. Table 4 therefore provides evidence to both support and reject the Substitution hypothesis.
In addition, the fourth column of Table 4 suggests that bigger and better performing are positively and significantly associated with CRAM (p < 0.01). This indicates that these firms are less likely to participate in earnings management through real activities manipulation. This is partly in contrast with the findings of Kim et al. (2012), who find that bigger firms are more likely to engage in earnings management through real activities manipulation. Furthermore, AGE and SGAint are significantly and negatively related with CRAM (p < 0.01), indicating that older firms and firm with higher SG&A intensity are more likely to participate in earnings management through real activities manipulation.
ResCFO ResPROD ResEXP CRAM
Coefficient Coefficient Coefficient Coefficient (t-stat) (t-stat) (t-stat) (t-stat)
CSR 0.002 - 0.006 0.011 0.019 (3.37)*** - (6.61)*** (10.47)*** (9.37)*** ResDA - 0.166 0.125 0.422 0.142 - (3.85)*** (1.39) (4.20)*** (0.73) SIZE 0.011 - 0.020 0.016 0.047 (6.72)*** - (6.09)*** (4.09)*** (6.49)*** M B - 0.004 0.011 - 0.013 - 0.027 - (1.64) (2.12)** - (2.19)** - (2.51)** AdjROA 0.230 - 0.251 - 0.020 0.460 (13.47)*** - (7.05)*** - (0.49) (6.00)*** AU 0.002 0.088 0.050 0.139 (0.07) (1.42) (0.70) (1.04) LEV 0.008 - 0.024 - 0.040 - 0.007 (0.87) - (1.19) - (1.71) - (0.16) EO 0.013 - 0.004 - 0.013 0.003 (1.52) - (0.22) - (0.66) (0.09) SGAint - 0.024 0.075 - 0.121 - 0.221 - (1.87)* (2.77)*** - (3.87)*** - (3.75)*** CGnet - 0.001 0.000 - 0.007 - 0.008 - (0.52) (0.05) - (1.64) - (1.01) AGE - 0.000 0.001 - 0.001 - 0.002 - (4.28)*** (4.71)*** - (3.37)*** - (4.92)*** ADM IRED - 0.005 0.017 - 0.016 - 0.038 - (0.77) (1.20) - (1.03) - (1.28) Adj. R2 0.267 0.183 0.148 0.213 N 1.104 1.104 1.104 1.104
* Significant at the 10% level. ** Significant at the 5% level. *** Significant at the 1% level.
TABLE 4
5.2.3 Earnings management and individual CSR issue area’s
Table 5 reports on the relationship between earnings management through both discretionary accruals and real activities manipulation on the distinct CSR issue areas. Prior studies (Kim et al., 2012; Mattingly and Berman, 2006) also perform an analysis about the relationship between both combined and distinguished CSR scores and earnings management, therefore sub scores are constructed for each CSR issue area. This was done by subtracting the total concerns from total strengths, leading to a net score for Community, Diversity, Employee Relations (hereafter Employee), Environment and Product. Afterwards, the relation between these net score of the CSR issue areas and earnings management was then analyzed in Table 5. The first column narrates the relationship between discretionary accruals and the five CSR issue areas. Only two results are statistically significant (p < 0.01) and show a negative relation with ResDA. Both Community and Environment show a negative relationship with discretionary accruals, indicating that these CSR components are likely to constrain earnings management through discretionary accruals. The other three categories do not show any significant results in this regression, although Diversity and Employee show a negative relation and Product shows a positive relationship.
The regression analysis between CRAM and the separate CSR categories is shown in the second column of Table 5. Except for the Product category, which is negative and insignificant, all the issue areas are positively and significantly (p < 0.01) related with earnings management through real activities manipulation. This indicates that positive CSR scores in Community, Diversity, Employee and Environment will constrain earnings management through real activities manipulation. Both findings are consistent with the research performed by Kim et al. (2012) and thus provides support for the Ethical hypothesis. Interesting to note is that Product seems to be the opposite of the other CSR components in relation to both CRAM and ResDA, although these relations are insignificant.
5.2.4. Earnings management and total CSR strengths and total CSR concerns
According to the research of Kim et al. (2012) earnings management should not only be regressed with CSR as an aggregated measure, but also by a distinct measure were total strengths and total concerns are separately analyzed. Mattingly and Berman (2006) argue that this is due to possible cloaking of important differences, as a firm with many strengths and
Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff.
(t-stat) (t-stat) (t-stat) (t-stat) (t-stat) (t-stat) (t-stat) (t-stat) (t-stat) (t-stat)
CS R Community - 0.005 0.070 - (3.64)*** (8.20)*** Diversity - 0.000 0.032 - (0.67) (8.18)*** Employee - 0.000 0.051 - (0.20) (8.98)*** Environment - 0.003 0.033 - (3.86)*** (6.82)*** Product 0.002 - 0.007 (1.41) - (0.91) CRAM - 0.002 - 0.005 - 0.006 - 0.002 - 0.006 - (0.43) - (1.14) - (1.35) - (0.53) - (1.38) ResDA - 0.090 - 0.240 - 0.271 - 0.114 - 0.298 - (0.43) - (1.14) - (1.35) - (0.53) - (1.38) Adjusted R2 0.012 0.000 0.000 0.013 0.002 0.058 0.057 0.068 0.040 0.001
* Significant at the 10% level. ** Significant at the 5% level. *** Significant at the 1% level.
ResDA CRAM
TABLE 5
ResDA CRAM AAER ResDA CRAM AAER
Coeff. Coeff. Coeff. Coeff. Coeff. Coeff.
(t-stat) (t-stat) (t-stat) (t-stat) (t-stat) (t-stat)
CS R Strengths - 0.001 0.015 0.001 - (3.74)*** (8.03)*** (3.42)*** Concerns - 0.001 - 0.016 - 0.000 - (1.35) - (5.00)*** - (0.29) CRAM - 0.002 - 0.002 - 0.006 0.003 - (0.42) - (0.26) - (1.35) (0.51) ResDA - 0.089 - 0.024 - 0.327 - 0.042 - (0.42) - (0.55) - (1.53) - (0.95) Adjusted R2 0.012 0.055 0.009 0.002 0.022 0.002
* Significant at the 10% level. ** Significant at the 5% level. *** Significant at the 1% level.
TABLE 6
Relationship between earnings management and total strengths and concers of CSR rating
equally many concerns is different from a firm with a single strength a single concern. Therefore, Table 6 will report on the regression between earnings management through both discretionary accruals and real activities manipulation and CSR measured by total strengths and total concerns.
The first column shows a negative and statistically significant (p < 0.01) relation between total CSR strengths and ResDA. The relation between CRAM and total CSR strengths is statistically significant and positive, these results both suggest that firms with a high achievement in CSR strengths are less likely to engage in earnings management through both discretionary accruals and real activities manipulation.
Therefore, these findings provide support for the Ethical hypothesis. As in the research of Kim et al. (2012) AAER is also taken into account in this regression and it does show a significant relation, but since n=3 and the regulation enforcements can only be found at CSR firms, these results shall not be taken into account.
The second three columns of Table 6 reports on the relationship between total CSR concerns and earnings management. ResDA shows an insignificant result in Table 6, while CRAM is showing a negative and statistically significant (p < 0.01) result. This result suggests that firms with higher CSR concerns are more likely to engage in earnings management through real activities manipulation. Concerns are therefore associated with more aggressive and opportunistic accounting.
5.2.5. Earnings management and individual CSR components
To expand the research of Kim et al. (2012) I will investigate the effect of the individual CSR components on earnings management through both discretionary accruals and real activities manipulation. This is done by performing a multiple regression as was done in Table 3 and Table 4. The difference with those results is that CSR was used as a combined result in this regression, while in Table 8 and Table 9 the individual CSR components are used. Before presenting the regression with the individual CSR components, Table 7 presents the correlation results amongst the individual CSR components.
Table 7 reports on the correlations between the separate CSR components, as Table 8 and Table 9 will present the regression of the effects of the individual CSR components on both earnings management measurements. As can be seen in table 7, most separate CSR components are to a large extent positively and significantly related. Only Product and Community (p < 0.1) and Product and Diversity (p < 0.01) are negative and significantly related. The other singular
CSR components are positively and statistically significant (p < 0.01) related. When carrying out the results of the individual regression of Table 8 and Table 9, the results may contain some bias. For instance, a firm with high results in Community is also expected to achieve a higher ranking in Environment, as Table 7 shows a correlation of 0.4284 amongst the two individual CSR components.
Table 8 reports on the multiple regression using the proxy for earnings management through discretionary accruals or ResDA. The results of this regression differ from the results of Table 3 as this regression uses the net score of a firm on the individual components of CSR instead of the combined and aggregated CSR score of a firm. The first five rows of the table show that three out of five individual CSR components are negatively related with ResDA, as expected as these observations have also been noted in the complete regression as in Table 3. Although only Community, Environment and Product are statistically significant (p < 0.01), these are also the CSR components that are negatively related to ResDA. Therefore, Table 8 provides support for the Ethical hypothesis, as these findings suggest that firms with high achievements on these three CSR components are less likely to engage in earnings management through discretionary accruals.
Important to note is that bigger firms seem to engage less in earnings management through discretionary accruals which is consistent with Table 3, as all carried out results come back negative and statistically significant. Furthermore, as in Table 3 there is no significant relationship between CRAM and ResDA.
Community Diversity
Employee
Relations Environment Product
Community 1.000 Diversity 0.4210 1.0000 0.0000 Employee Relations 0.2517 0.2124 1.0000 0.0000 0.0000 Environment 0.4284 0.2941 0.2458 1.0000 0.0000 0.0000 0.0000 Product - 0.0519 - 0.1672 0.0844 0.1324 1.0000 0.0849 0.0000 0.0050 0.0000
Relations between CS R issue area's TABLE 7
