Greenwashing and Earnings Management: Symptoms of the same problem?

(1)

Greenwashing and Earnings Management:

Symptoms of the same problem?

Abstract

In this thesis I have tested the relationship between greenwashing and earnings

management, through discretionary accruals or real earnings management, in US listed firms. Specifically, I examine whether greenwashing is being used to divert shareholder’s attention away from earnings management practices. Based on a sample consisting of 5198 firm-years, including 33 industries and 1394 companies, covering a time period ranging from 2010 to 2017; I find (1) no evidence that such a relationship exists between greenwashing and earnings management through discretionary accruals, (2) evidence which suggests that firms who make use of CSR reporting, regardless of whether they greenwash or not, are less likely to manipulate real operating activities. Secondary analysis into explanatory

mechanisms of greenwashing, show that while (3) real earnings management and

Sustainability compensation incentives increase the likelihood of greenwashing, (4) the use of the GRI reporting framework, as well as having a more gender diverse board reduces the likelihood for greenwashing.

University of Groningen G.L. Seekles

Faculty of Economics and Business s3490211

MSc Accountancy Aagje Dekenstraat 1

8023 BX Zwolle

Master Thesis g.l.seekles@student.rug.nl

Supervised by: Dr. T. Marra 06-31967368

Januari 21st_{, 2019} Word count: 13.150

(2)

1 Introduction

In recent years we have seen the annual reports of many firms being extended with information about non-financial sources being used or affected by the organization. These developments in corporate reporting are known as Corporate Social Reporting (CSR), Environmental, Social and Governance (ESG) Reporting, or Sustainability Reporting1_{. As a} result, the amount of publicly available corporate information has grown considerably in recent years.

As argued by Walker and Wan (2012), ESG performance is, as a construct, important for a company’s reputation, and yet consists of numerous separate issues, from greenhouse-gas-emissions and stakeholder engagement, to lifecycle analysis and water - and waste

management.

As a result, numerous companies as well as NGO’s and society now take an active role in environmental management, some going as far as lobbying slow-moving governments for greater environmental regulations. And perhaps even more important, various researchers amongst which Griggs et al. (2013, p 305) pressured governments and society to act because “…, the stable functioning of Earths systems, [..], is a prerequisite for a thriving global

society”. Their efforts were not in vain, as in Augustus 2015 193 countries within the United

Nations ratified the UN Sustainable Development Goals (SDGs), as a call for action by all countries, to promote prosperity and economic growth while protecting the planet and tackling climate change.

Consequently, corporate reporting initiatives such as those of the IIRC and the GRI have arisen and the SDGs have led to a growing global awareness combined with increased pressure from investors (IIGCC, 2018; HSBC, 2018; PRI, 2018). In a 2017 public letter to the CEOs of all public companies, Larry Fink, CEO of BlackRock, with $6.3 trillion in invested assets one of the world’s largest investors said that companies needed have a strategy for long-term value creation and financial performance. “To prosper over time, every company must not only deliver financial performance, but also show how it makes a positive

contribution to society. Companies must benefit all of their stakeholders, including shareholders, employees, customers, and the communities in which they operate.” In order to protect their reputation (Walker and Wan, 2012; Hooghiemstra, 2000) and to retain their investors, companies have acted and have expanded their corporate reporting to include more and more information on their actions with respect to the SDGs, and more specifically also on their environmental performance. This, however, presents a looming issue, called Greenwashing. Greenwashing is a strategy that companies adopt to engage in symbolic communications of environmental issues without substantially addressing them in actions (Laufer, 2003).

To provide an example, a recent study by technology website The Verge (2018) and a subsequent article in the Dutch media (NOS, 2018), illustrates that the amount of water needed to produce 1 liter of Coca-Cola is dependent upon how far back you trace the supply

1_{In this thesis the terms CSR Reporting, ESG Reporting and Social Reporting as well as related terms such as}

(5)

chain and subsequently attribute this water usage to Coca-Cola. Based upon the production process and the amount that goes in the bottle, 1 liter Coca-Cola requires 2,43 liters of water, this is as advertised by the company in 2010. If we, however, are to include the water needed to also produce the ingredients, such as sugar, the water usage increases to around 70 liters of water per 1 liter of Coca-Cola. And yet, in a 2016 full-page ad2_{published in The} New York Times, the company proclaimed, “For every drop we use, we give one back.”, the

question now becomes Do They? As with earnings, public perception matters. It is in fact, big business.

And yet the academic literature on what drives greenwashing is limited. I argue however, that we can draw from existing academic literature on earnings management and Corporate Social Responsibility (Kim et al., 2012; Jordaan et al., 2018; Hooghiemstra, 2000) to explain the underlying motivations why managers do engage in greenwashing.

Earnings management, which occurs “when managers use judgment in financial reporting

and in structuring transactions 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 reported outcomes.” (Healy and Wahlen, 1999 p. 6); is a

well-known method that is also used by managers in investor relations. Evidence from prior research indicates that managers’ concerns over current performance may motivate them to engage in manipulating current period earnings at the expense of future period earnings (Graham et al., 2005).

The presence or absence of (signals for) earnings management is relevant indicator for earnings quality. Higher quality earnings more faithfully represent the features of the firm’s fundamental earnings process that are relevant to a specific decision made by a specific decision-maker (Dechow, Ge & Schrand, 2010). It is not hard to see why this is relevant for investors, especially when looking at the longer-term value creation associated with sustainable investments (Clark, G.L., Feiner, A. & Viehs, M., 2015).

Empirical research into a direct association between greenwashing and earnings

management is limited. While this study is partly based upon prior work by Kim et al. (2012) and Jordaan et al. (2018), they do not measure greenwashing directly but focus purely on effect of CSR performance on earnings management. They do however refer to

greenwashing as a possible way to explain their results.

Carroll (1979) found that managers have an incentive to be ethical, honest and transparent in their financial reporting and to be socially and environmentally responsible in their activities. More recently however, Hemingway and Maclagan (2004) found evidence

suggesting that companies may voluntarily disclose information to obtaining the support of share- and stakeholders by diverting their attention, or as a means of symbolic action (Phillipe & Durand, 2011).

(6)

Evidence from prior research further suggests that managers may resort to (overly positive) CSR reporting to deflect unwanted attention away from earnings management activities (Prior et al., 2008; Grougiou et al, 2014), suggesting that greenwashing may prove to be an alternative measure for undesirable management behavior or corporate misconduct.

Considering the latter may be harmful to share- and stakeholder interests and impact future earnings, and the fact that shareholders make use of ESG information in their investment decisions (Amel-Zadeh & Serafeim, 2018); obtaining more understanding about the relationship between earnings management and greenwashing not only contributes the literature, but is also useful for investors and analysts during their forecasts and investment decision making process.

This study also contributes to the literature by testing the association between

greenwashing and earnings management, using a rigorous measure of greenwashing, which is based on assessment of a company’s reported performance versus their assessed

performance by a third party and that is not based on more subjective measures such as opinions or perception (Avcilar & Demirgünes, 2016).

The result of this study may also be relevant for auditors, especially those involved in ESG-assurance, and regulators, when it comes to determining audit risks and regulations with respect to disclosures and environmental law. If a relationship between Greenwashing and opportunistic earnings management is shown, auditors should be more aware of possible changes in accounting policies or real actions which influence the company’s current performance and may also present a risk to the company’s longer-term continuity. For regulatory institutions, such as governments and the UN, and for accounting standard setters, the results of this study may prompt new legislative actions in order to improve not only the quality of corporate reporting, but also towards more environmental laws and regulations in order to battle climate change and improve sustainable growth and prosperity on a global scale.

The research question will therefore be: “Is there a positive association between

greenwashing and earnings management, and does the amount of earnings management used dictate the degree of greenwashing?”

This thesis continues as follows. Section 2 provides the theoretical framework and the hypothesis development. Section 3 describes the research design used to measure the variables and to conduct the analysis. The results are subsequently presented in Section 4. Finally, in Section 5, the results are discussed, and attention is given to the implications of the results, the limitations of this study and recommendations for further research.

(7)

2 Theory

2.1 Theoretical Background

2.1.1 What is Greenwashing?

According to the Oxford English dictionary, the term Greenwashing (hereafter GW) originated in 1980s and is defined as follows: “Disinformation disseminated by an

organization so as to present an environmentally responsible public image.”

It is however import to realize that the academic community has yet to reach consensus on the exact definition of GW (Lyon & Maxwell, 2011; Lyon & Montgomery, 2015).In general, the term is used either when describing the practice of misleading customers about the environmental benefits or impact of a product through advertisements, or when, organizations spend more time and money on advertising and disclosing that they are “greener” than they actually are (Lyon & Montgomery, 2015; Delmas & Burbano, 2011; Laufer, 2003). This can be done by, for example, emphasizing the money spent on green innovations, whilst neglecting to mention the environmental impact of the organization’s normal operations.

A review of Lyon and Montgomery (2015) provides an overview of the multidisciplinary literature on greenwashing, with a sharp increase in articles since 2011. They also provide an extensive overview of different means to Greenwash, based up a review of 34 papers

published in ISI-ranked journals. Looking back at the example presented in the introduction as well as the overview presented by Lyon and Montgomery (2015), GW is concept which can take many forms, but the common factor is that they all rely on misdirection and disinformation in order to communicate an environmentally responsible public image. Another way to greenwash refers to a discrepancy between a firm’s internal activities and their outside projected image, for example the creation of a sustainability department, which may be understaffed or have little influence within the organization (Lyon & Montgomery, 2015 p.226).

Lastly, selective disclosure may also be considered a variety of greenwashing (Lyon & Montgomery, 2015; Cho and Patten, 2007; Patten, 2002; Clarkson et al., 2008). However, the literature finds conflicting results over whether cleaner firms issue more disclosure or not (Lyon & Montgomery, 2015).

For this thesis I will define greenwashing as: Disinformation presented by an organization, either through corporate ESG-reporting or advertisements, so as to present an overly positive environmentally responsible image.

(8)

2.1.2 Why does Greenwashing occur?

To explain why GW occurs, Walker and Wan (2012) refer to both institutional theory and signaling theory. According to signaling theory, appearing to conform to green norms can be effective at signaling to external stakeholders the firms’ values regarding green issues

(Ramus & Montiel, 2005), and it is easier than actual implementing these values with actions (Suchman, 1995).

The disclosure of non-financial information is also associated with clear financial benefits. A growing number of investors is taking non-financial information into account when making investment decisions (Amel-Zadeh & Serafeim, 2018). Prior research by Dhaliwal et al. (2015) shows a clear relation between voluntary disclosure of non-financial information and a reduction in firm’s cost of equity capital. They have also shown that firms that perform better (compared to their peers) on non-financial indicators enjoy easier access to dedicated investors when raising capital. Martínez-Ferrero et al. (2016) also found that CSR reporting reduces a company’s cost of capital. But more importantly, their results also show that the impact of CSR reporting on the cost of capital is larger for firms in which earnings

management was detected, thus giving managers that manage earnings a clear incentive to report on CSR.

The second motive that GW Walker and Wan (2012) mention is to attain legitimacy according to the institutional theory (Suchman, 1995). Institutional theory emphasizes the importance of regulatory, normative, and cognitive factors in shaping firms’ decisions to adopt specific organizational practices which conform to the demands of external parties such as stakeholders (DiMaggio & Powell, 1983). And yet the fact that this expected

conformity can hold advantages for firms may, according to Meyer and Rowan (1977), incite misbehavior and thus lead to organization managing their outside impressions by decoupling their internal actions and activities from their presented image.

According to Lyon and Montgomery (2015, p 235/236) most of the prior research has focused on the pressure placed on an organization by its external environment, suggesting that greenwashing is a reactive response. They have also identified a smaller body of research, which also aligns with this thesis, which suggests that greenwashing can also be used strategically to influence the external environment, by maintaining or preserving positions of power through rhetoric and symbolic action, cheap talk or unfulfilled promises. Evidence from research by Phillippe and Durand (2011) suggests that merely releasing a corporate sustainability report, without actually making improvements, improves a firm’s reputation. When the industry was environmentally sensitive however, the reputation could only be enhanced if substantive improvement where made. Bansal and Clelland (2004) find that firms with poor environmental reputation can improve investor relations and thus their reputation by making a public expression of commitment to an environment or a specific goal. Lastly Ramus and Montiel (2005) find that greenwashing occurs through unfulfilled promises, which appear to be more likely in certain industries then others.

(9)

Institutional theory is also used by Delmas and Burbano (2011) in their analysis into the drivers of greenwashing. Based upon their framework, see figure 1, Delmas and Burbano (2011, p 68-69) posit that “The regulatory context is a critical external institutional driver of

firm greenwashing.” as well as “Market external factors are important drivers of

greenwashing. Key firm characteristics, incentive structure and ethical climate, effectiveness of intra-firm communication, and organizational inertia play important roles in moderating a firm’s reaction to external drivers. In addition, individual-level psychological and cognitive factors influence managers’ decision-making processes and thus influence how external drivers translate into motivation for action.”

Looking back at the definition of GW and the work of Laufer (2003), these outcomes aren’t very surprising. A lack of regulations and good standards, together with strong managerial incentives, especially if it concerns their remuneration (Hosmer, 1987; Sims, 1991), that promote unethical behavior are important drivers of GW.

Figure 1: Drivers of Greenwashing (Delmas & Burbano, 2011)

Fortunately, the power of civil society to influence companies and challenge greenwashing has increased. According to Lyon and Montgomery (2015, p. 239) “several empirical papers

find that strong environmental groups may deter greenwash through a variety of activities including “naming and shaming” by individuals or civil society organizations.” This may

however result in less disclosure on environmental performance altogether (Lyon & Maxwell, 2011).

Various forms of (CSR) performance evaluation are believed to also contribute to addressing greenwashing, frequently through measures such as sustainability ratings (Parguel et al., 2011). I will elaborate on these ratings, and how they can help to measure GW in the following section.

(10)

9

Delmas and Burbano (2011), also offer recommendations to decrease greenwashing, of which I will discuss several that may be relevant for this thesis. Firstly, the alignment of employee (and management) incentives, which can reduce the likelihood of greenwashing. Instead of incentives that enable greenwashing, employees and management should receive incentives that encourage environmental performance or punish for the involvement in greenwashing incidents (Delmas & Burbano, 2011).

Secondly, the board should provide ethical leadership and training that is designed to explain the risks of greenwashing. The board and senior management could institute ethical codes and lead by example, the diminish the likelihood of unethical behavior and

encouraging a culture of open communication and collaboration (Delmas & Burbano, 2011). Based on research by Neville et al. (2018), increased board independence is a commonly offered solution to limit the likelihood of corporate misconduct, activities or actions that organizational members engage in to deceive or mislead investors or other key stakeholders. Research by Al-Shaer and Zaman (2016) also suggests gender diverse board are associated with higher quality sustainability reports, suggesting that this may also reduce the likelihood of Greenwashing. This is based on the notion that a more diverse board will take a more balanced view and pay greater attention to stakeholder concerns (Al-Shaer & Zaman, 2016). And finally, improving the position and influence of for example the sustainability officer or department to a staff department would reduce the lack (of coherence) of communication between different divisions within a company and across countries, which should in turn lead to less greenwashing. The institutions of standards and requirements or frameworks such as included in the IIRC, ISO 14001 and/or the GRI can also provide a means to obtain and share more relevant information which may or may not be subject to greenwashing (Delmas & Burbano, 2011).

2.1.3 Measuring Greenwashing

As mentioned in the previous paragraph, prior research suggests that various forms of (CSR) performance evaluations can also contribute to addressing greenwashing, frequently

through measures such as sustainability ratings (Parguel et al., 2011). Through these ratings the public can obtain a measure of a company’s CSR performance compared to their

industry peers. Simultaneously, since the available ratings are maintained by different parties and measure a company’s performance using different methodologies, they also offer interesting opportunities for researchers.

For the purpose of this study I will measure greenwashing using the Asset4 and the Thompson Reuters ESG score metrics. I will use the environmental sub-pillar scores3_for Resource Use, Emission Reduction and Product Innovation from Asset4, which represent a company’s performance based on their corporate reporting and subtract those from the Thompson Reuters scores for those same sub-pillars, which represent a company’s performance as assessed by Thompson Reuters based on data in the public domain (thus including media).

If the resulting score is negative, I classify that company as greenwashing.

3_{I use the sub-pillars because, unlike Asset4, Thompson Reuters does not offer an overall Environmental Pillar}

(11)

2.1.4 What is Earnings Management?

According to Healy and Wahlen (1999, p. 6), “earnings management occurs when managers

use judgment in financial reporting and in structuring transactions 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 reported outcomes.” In other

words, earnings management (hereafter EM), is the intervention by, or the influence of choices by, managers in the external financial reporting process, with the intent to obtain company or private benefits.

The possibilities of managerial intervention in the reporting process can occur not only via accounting estimates and methods (discretionary accruals), but also through operational decisions or real activities management (Jones, 1991; Dechow et al., 1995; Zhang, 2012; Cohen et al., 2008). Roychowdhury (2006, p. 337) defines real activities management as “...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. These departures do not necessarily contribute to firm value even though they enable managers to meet reporting goals.” Some use-cases of EM may

benefit the company, where others might do harm. Simultaneously, acts of EM may also be either opportunistic or efficient dependent upon your point of view and the specific

situation.

2.1.5 What drives the use of Earnings Management?

When explaining earnings management behavior, agency theory is often used. Agency theory is about the relationship that exists between management and shareholders. Jensen and Meckling (1976, p. 6) define this relationship as "...a contract relationship where one or

more persons (the principal) engage another person (the agent) to perform some service on their behalf which involves delegating some decision-making authority to the agent”.

Because of this delegation of control, the fundamental concepts of the agency theory arise, firstly misalignment of interests. Given the diverging interests between the principal (share- and stakeholders) and the agent (management), the agent may not always act in the best interest of the principal. And because the decision-making authority is delegated, the shareholders rely on information, made available by the management to evaluate their performance. When relevant information in not communicated to the shareholders in full, this creates the secondary precondition for agency problems, known as information

asymmetry. Empirical research by Richardson (2000), found that a positive relationship

exists between the levels of information asymmetry and EM, suggesting that information

asymmetry is a necessary condition for EM.

Theory and evidence indicate that managers’ concerns over current performance may motivate them to engage in manipulating current period earnings at the expense of future period earnings (Graham et al., 2005). As main reasons, Roychowdhury (2006) and Kim & Sohn (2013) mention perceived private benefits to meeting the reporting goals, i.e. a manager’s remuneration, or to avoid debt covenant violations and credit risk (cost of capital). Since managers use real earnings management (REM) and accrual-based earnings management (AEM) as substitutes for each other, examining either type in isolation cannot lead to definitive conclusions (Zang, 2012).

(12)

Lastly, results from a study by Elias (2002) indicate a positive relationship between social responsibility, focus on long-term gains, idealism, and the ethical perception of EM and a negative relationship between focus on short-term gains, relativism and the ethical

perception of EM. This suggests that, as with greenwashing, ethics related to management behavior is an important driver for EM, depending on reasons for managers to use EM. 2.1.6 Measuring Accruals-based and Real Earnings Management

In this section I will briefly address the models used to calculate the discretionary accruals (as proxy for Accrual-based earnings management) and Roychowdhury’s real activities manipulation model (as proxy for Real earnings management). Both these models are widely used in academic literature (Graham et al. ,2005; Roychowdhury, 2006; Cohen et al., 2008; Zhang, 2012; Kim et al., 2012; Jordaan et al., 2018).

2.1.6.1 Modified Jones Model for discretionary accruals

Following Dechow et al. (1995) the discretionary accruals are calculated by measuring the non-discretionary accruals as a portion of the total accruals in the Modified Jones model. The total accruals are defined as equal to the change in accrual in year t as given by equation A1 in the appendix. The non-discretionary accruals are equal to a linear function of Gross value of Plant, property and equipment in the current period and the change in Revenues – the change in Receivables in the current period, both over lagged assets. To estimate the non-discretionary accruals, I run the following cross-section OLS regression for every industry and year4_:

𝑇𝐴𝐶𝐶_𝑡 𝐴𝑡−1 = 𝛼1 1 𝐴𝑡−1 + 𝛼2 (∆𝑅𝐸𝑉_𝑡− ∆𝑅𝐸𝐶_𝑡) 𝐴𝑡−1 + 𝛼3 𝑃𝑃𝐸_𝑡 𝐴𝑡−1 + 𝜀𝑡 (Eq. 1)

For every firm-year, the discretionary accruals are equal to the total accruals minus the non-discretionary accruals using the estimated coefficients from the corresponding year-industry model: 𝑁𝐷𝐴𝐶𝐶𝑡 𝐴_𝑡−1 = 𝛼̂1 1 𝐴_𝑡−1+ 𝛼̂2 (∆𝑅𝐸𝑉𝑡− ∆𝑅𝐸𝐶𝑡 ) 𝐴_𝑡−1 + 𝛼̂3 𝑃𝑃𝐸𝑡 𝐴_𝑡−1 (Eq. 2)

To simplify however, the discretionary accruals are equal to the error term in Eq.1 (Jordaan

et al., 2018). More information on this models and variables used can be found in the

appendix.

(13)

2.1.6.2 Roychowdhury’s real activities manipulation Model

Based upon Roychowdhury (2006) and Cohen et al. (2008), I use Roychowdhury’s combined measure as a proxy for real earnings management (see Eq. 3).

𝐶𝑂𝑀_𝑅𝐸𝑀 = 𝐴𝐵𝑁_𝐶𝐹𝑂− 𝐴𝐵𝑁_{𝑃𝑅𝑂𝐷}+ 𝐴𝐵𝑁_{𝐷𝐼𝑆𝐸𝑋𝑃}

(Eq. 3) The variables abnormal cashflow for operations (𝐴𝐵𝑁𝐶𝐹𝑂), abnormal production costs (𝐴𝐵𝑁𝑃𝑅𝑂𝐷) and abnormal discretionary expenses (𝐴𝐵𝑁𝐷𝐼𝑆𝐸𝑋𝑃) are each calculated based on separate cross-section OLS regressions for every industry and year. More information on this model and the variables used can be found in the appendix.

2.2 Hypothesis development

The drivers of GW and EM, as identified by Lyon and Montgomery (2015), Delmas and

Burbano (2011) and by Roychowdhury (2006) and Graham et al. (2005) respectively, are very similar yet motivated from different theoretical approaches.

As evidenced by Elias (2002), Hosmer (1983) and Sims (1991) social and personal values also play an important role. The problem is now, how to reconcile the different theories into a logical framework on which hypothesis can be built. Jordaan et al. (2018), while conducting a similar study, use an approach based upon Kim et al. (2012), which I would argue is also applicable for this study, that focuses an ethical and an opportunistic approach towards management behavior.

2.2.1 Opportunistic management behavior

The ethical hypothesis states that managers have an incentive to be ethical, honest and transparent in their financial reporting and to be socially and environmentally responsible in their activities (Carroll, 1979), thus not negatively impact a company’s market value through activities such as earnings management. This doesn’t mean that using earnings management is necessarily a bad thing, especially if used efficiently, for example when it comes to making sure that terms within debt contracts are met through the management of accruals as is allowed within GAAP. Research by Kim et al. (2012) found support for this hypothesis in their investigation of a sample of companies in the United States. They found that companies with higher CSR performance scores delivered more transparent financial information to

investors.

Given that greenwashing is also driven by ethics and morality (Hosmer, 1983; Sims 1991), this would suggest that the ethical argument may also be true for Greenwashing. Combined the findings of Martinez-Ferrero et al. (2013), who argue that managers who have an

incentive to reduce information asymmetry will minimize earnings management, the results from Kim et al. (2012) may suggest that managers with incentives to report more

transparently and more responsibly, have less incentive to resort to earnings management. It is also possible that managers may have an incentive to act opportunistically in their decisions on whether to be socially and environmentally responsible (Jensen & Meckling, 1976; Carroll,1979), suggesting that ESG reporting may be linked to manager’s self-interest or personal benefit over that of the firm or her stake- and shareholders. From an agency perspective, prior research by McWilliams et al. (2006) also finds that managers can use ESG reporting to advance their careers or other personal agendas.

(14)

The disclosure of non-financial information is also associated with clear financial benefits. A growing number of investors is taking non-financial information into account when making investment decisions (Amel-Zadeh & Serafeim, 2018). Hemingway and Maclagan (2004), as well as Grougiou et al. (2014), found evidence that companies may voluntarily disclose information, such as a CSR report, to cover up corporate misconduct, thus obtaining the support of stakeholders by diverting their attention. Prior et al. (2008) have examined whether firms use CSR reporting strategically to disguise earnings management and found a significant, positive relationship between earnings management and CSR, but for regulated firms only. When taken together, the results suggest that CSR Reporting can be used as reputation insurance or window-dressing, for management to hide behind.

In summary, studies that examine the relationship between CSR reporting and earnings management are few and they provide mixed results, such that it is difficult to draw a definitive conclusion about the nature of the relationship (either ethical or opportunistic). It is thus possible, that when managers manage earnings, they may resort to CSR reporting, which can be overly positive and therefore subject to greenwashing, in order to divert shareholder’s attention, suggesting the following hypothesis:

H1: There is positive association between greenwashing and earnings management.

2.2.2 Integrated Reporting and Compensation incentives

Prior research on the influence of several governance factors on greenwashing, posits that both the use integrated reporting (such as the GRI framework) and the use of compensation incentives can help to can help to curb the effects of corporate misconduct, i.e.

greenwashing and earnings management (Delmas & Burbano, 2011).

The key principle behind integrated reporting is that it reduces the existing information asymmetry between the principals and the agents, by integrating more information both financial and non-financial into one annual report. Prior research (Kim et al. (2012), Jordaan

et al. (2018)) have shown that CSR disclosures, which are still largely voluntary are, are

associated with earnings management, thereby implying that greenwashing (as the difference between the companies reported and actual performance) is as well. Martinez-Ferrero et al. (2013) argue that, based on their findings, managers who have incentives to reduce information asymmetry will minimize earnings management and disclosure more CSR information.

Based on the key principle of integrated reporting and the result of Martinez-Ferrero et al. (2013), as well as the theoretical argumentation by Delmas and Burbano (2011) and to lesser extent Lyon and Montgomery (2011), it is feasible to expect that the use of GRI (as a proxy for integrated reporting) and the use of Sustainability Compensation Incentives for

Executives help reduce earnings management, leading to the following hypotheses:

H2A: There is a positive association between the use of GRI and earnings management. H2B: There is a positive association between the use of Sustainability Compensation Incentives for Executives and earnings management.

And looking at the ESG scores more closely, it is feasible to expect that the use of GRI would reduce greenwashing (i.e. the difference between the ESG scores), whereas the use

(15)

Sustainability Compensation Incentives for Executives is expected to increase the difference because of the associated personal benefit with better reporting scores, thus implying the following two hypotheses:

H3A: There is a negative association between the use of GRI and the difference in ESG scores.

H3B: There is a positive association between the use of Sustainability Compensation Incentives for Executives and the difference in ESG scores.

(16)

3 Research design and methodology

3.1 Data collection and sample

I restrict my research to a period of 2010-2017, to coincide with the introduction of GRI and match the period where research in greenwashing has started to gain academic attention. I focus my research on US listed firms because firstly, in a recent study, Amel-Zadeh and Serafeim (2018) have shown that a growing number of investors, is taking non-financial information into account when making investment decisions. Their results indicate that a significant portion of investors (84% in the EU vs 75% in the US) makes use of ESG

information in their investment decisions, signifying its importance. Secondly, in focusing solely on the US, I can example a large sample of companies, without having to control for various cross-country differences in legislations, and in corporate or organizational culture. Thirdly, the US stock markets deal with the world’s largest companies, as well as some of the world’s most influential investors like Warren Buffet and Larry Fink, the latter of which is very outspoken about the importance of ESG and sustainability.

Following Dechow et al. (1995) and Roychowdhury (2006) I use annual data, which I obtain from the US COMPUSTAT database, and require sufficient data to calculate the

COMPUSTAT-based variables in the section 6.1 for every firm-year and I also eliminate firms in regulated industries (SIC codes between 4400 and 5000) and banks and financial

institutions (SIC codes between 6000 and 6500) and lastly, following Dechow et al. (1995) and Roychowdhury (2006) I also require at least 15 observations per year-industry grouping. I use this data to run the regression analysis needed to estimate my proxies for independent variables, AEM and REM.

Data on ESG reporting scores is available from DATASTREAM as a timeseries for the period 2010-2017 for which I require the data on the Asset4 and Thompson Reuters (hereafter TRESG) variables mentioned in section 3.2. I require both Asset4 and TRESG data per firm-year to compute the dependent variable, (the level of) greenwashing.

The data needed for the control variables is also obtained from COMPUSTAT and the Eikon Thompson Reuters databases.

The dataset used to calculate the earnings management proxies consists of 39.362 firm-year observations, covering 5485 firms and 99 industries. After dropping missing values and regulated industries, the discretionary accruals are calculated over a sample of 17.696 firm-years, and the abnormal cashflow from operations, abnormal production costs and

abnormal discretionary expenses are calculated over samples that have 19.030, 16.566, and 20.030 firm-years respectively.

After combining this with the data from the control variables and ESG data, and again dropping for any missing values, the full sample remains, consisting of 5198 firm-year observations, covering 33 industries and 1394 firms.

(17)

3.2 Estimation models and variable description

In this section the estimation models used to test the hypotheses will be presented and the variables will be explained.

3.2.1 Variable description 3.2.1.1 Dependent variable(s)

Based on Kim et al. (2012), Jordaan et al. (2018) and Zhang (2012), I use two measures of earnings management, one for accrual-based earnings management based on discretionary accruals, as calculated with the model explained in section 2.1.6.1; and one that measures real earnings management, as calculated with the model explained in section 2.1.6.2. Since managers may use either or both methods to manage earnings (Cohen et al., 2008) and Zhang (2012) finds that both methods may substitute each other, examining either in isolation cannot lead to definitive conclusions.

Discretionary accruals

Based on Jordaan et al. (2018) and Hooghiemstra et al. (2019) and in line with prior literature, I use the absolute value of the discretionary accruals, calculated based on the modified Jones models (Dechow et al., 1995). If the results are consistent with ethical

(opportunistic) hypothesis, I expect a negative (positive) association between greenwashing

and earnings management through discretionary accruals.

Combined real earnings management

Following Roychowdhury’s Model (2006) and Cohen et al. (2008), the main proxy for real earnings management in this study is calculated by aggregating the residuals from the regression on the abnormal cashflow from operations, abnormal productions and abnormal discretionary expenses.

Lower cashflows from operation than the industry average is associated with earnings

management, since they may be the result of managers offering higher discounts or offering more lenient credit terms leading to higher sales that may never result into cash.

Higher than average production costs may relate to overproduction of inventories which

may also indicate earnings management. This is mainly due to the fact the overproduction may reduce the cost of goods sold as the fixed costs per unit are smaller.

Lower levels of discretionary expenses are also associated with more earnings management,

since these costs (R&D, advertising) may be reduced in order to meet short term targets at the expense of future performance.

Since the abnormal production costs are deducted when calculating the aggregate proxy,

(18)

3.2.1.2 Independent variable

As described in section 2.1.3 greenwashing can be measured as deviation between different ESG ratings which measure either a firm’s reported performance or a firm’s real

performance based on a third-party assessment.

As explained in section 2.1.3, I use the Asset4 ESG score as a measure of a firms reported performance, and I use the Thompson Reuters ESG scores as a measure of a firm’s actual performance. A measure of GW (Independent variable) is thus given by subtracting the Asset4 sub-pillar scores from the Thompson Reuters sub-pillar scores.

If the remain score is negative, the dummy variable (dGW) will take a score of 1, indicated greenwashing, and will be zero otherwise. I also retain the ‘real’ difference (realGW).

Prior research also suggests that renumeration incentives may drive both greenwashing and earnings management, thus I also control for executive compensation incentives on

sustainability targets using a dummy (dSUSComp) taking 1 if such incentives are present and 0 otherwise.

Lastly, the use of integrated reporting is expected to curb the effects of corporate misconduct, since it decreases the opportunity for managers to make use of information asymmetry and mislead or misdirect stake- and shareholders. In order to control for this effect, I add a dummy variable for GRIUse (as a proxy for the use of integrated reporting) as a control variable, taking 1 if GRI is used and zero otherwise.

3.2.1.3 Controls

Lastly, several control variables are included in the estimation models: SIZE (measured by the logarithm of the firm’s total assets), RDIntensity (R&D Intensity, measured by the research and development expenses scaled by total revenue), NI (Net Income (Loss)), LEVERAGE (measured by the total long-term debt scaled by shareholder’s equity), Board Independence (BoardInd) and Board Gender Diversity (BoardGDiv). Prior studies show that these variables are associated with Greenwashing and Earnings Management (Kim et al., 2012; Jordaan et al., 2018; Hooghiemstra et al., 2019).

Since Greenwashing and Earnings Management are also related to investor relations, I also add EPS (Earnings per share) as a measure for performance as well as Tobin’s Q as a measure for shareholder pressure.

(19)

3.2.2 Regression models to test the hypotheses

The regression models are based upon prior work by Kim et al. (2012) and Jordaan et al. (2018). However, they do not use the difference between the Asset4 ESG scores and the Thompson Reuters ESG scores as a metric for Greenwashing.

Model A: Accrual-based earnings management (AEM)

𝑎𝑏𝑠𝐷𝐴𝐶𝐶_𝑡= 𝛼₀+ 𝛼₁𝑑𝐺𝑊_𝑡+ 𝛼₂𝐶𝑂𝑀_𝑅𝐸𝑀 + 𝛼₃𝑆𝐼𝑍𝐸_𝑡 + 𝛼₄𝑅𝐷𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦_𝑡+ 𝛼₅𝑁𝐼_𝑡 + 𝛼₆𝐿𝐸𝑉𝐸𝑅𝐴𝐺𝐸_𝑡+ 𝛼₇𝐸𝑃𝑆_𝑡+ 𝛼₈𝑇𝑂𝐵𝐼𝑁𝑄_𝑡

+ 𝛼₉BoardInd_t+ 𝛼₁₀BoardGDiv_𝑡+ 𝛼₁₁𝑑𝑆𝑈𝑆𝐶𝑜𝑚𝑝𝐼𝑛𝑐_𝑡+ 𝛼₁₂𝑑𝐺𝑅𝐼_𝑡+ 𝜀_𝑡

Model B: Real earnings management (REM)

𝐶𝑂𝑀_𝑅𝐸𝑀_𝑡= 𝛼₀+ 𝛼₁𝑑𝐺𝑊_𝑡+ 𝑎𝑏𝑠𝐷𝐴𝐶𝐶 + 𝛼₃𝑆𝐼𝑍𝐸_𝑡 + 𝛼₄𝑅𝐷𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦_𝑡+ 𝛼₅𝑁𝐼_𝑡 + 𝛼6𝐿𝐸𝑉𝐸𝑅𝐴𝐺𝐸𝑡+ 𝛼7𝐸𝑃𝑆𝑡+ 𝛼8𝑇𝑂𝐵𝐼𝑁𝑄𝑡

+ 𝛼9BoardIndt+ 𝛼10BoardGDiv𝑡+ 𝛼11𝑑𝑆𝑈𝑆𝐶𝑜𝑚𝑝𝐼𝑛𝑐𝑡+ 𝛼12𝑑𝐺𝑅𝐼𝑡+ 𝜀𝑡 In both models, 𝛼1 is the coefficient of interest, since it tests the association between the earnings management through AEM and REM respectively and Greenwashing. The REM and AEM measures, as well as the ESG scores from which the GW variable is derived per industry and year. To test the hypothesis, I will use panel data regression models, using the company as a grouping variable.

To test the third hypothesis and a as control for the results of both earnings management models, I use the following model:

Model C: Greenwashing and integrated reporting

𝑟𝑒𝑎𝑙𝐺𝑊𝑡 = 𝛼0+ 𝛼1𝑑𝐺𝑅𝐼𝑡+𝛼2𝑑𝑆𝑈𝑆𝐶𝑜𝑚𝑝𝑡+ 𝛼3𝑆𝐼𝑍𝐸𝑡 + 𝛼4𝑅𝐷𝐼𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦𝑡+ 𝛼5𝑁𝐼𝑡 + 𝛼6𝐿𝐸𝑉𝐸𝑅𝐴𝐺𝐸𝑡+ 𝛼7𝐸𝑃𝑆𝑡+ 𝛼8𝑇𝑂𝐵𝐼𝑁𝑄𝑡

+ 𝛼9BoardIndt+ 𝛼10BoardGDiv𝑡+ 𝛼11𝑎𝑏𝑠𝐷𝐴𝐶𝐶𝑡+ 𝛼12𝐶𝑂𝑀_𝑅𝐸𝑀𝑡+ 𝜀𝑡 In this model, 𝛼1 and 𝛼2 are the coefficients of interest with respect to hypothesis H3A and H3B, where the last two can be used to control the results of the primarily analysis. Again, a panel data regression model is used, with the company as a grouping variable.

(20)

4 Results

4.1 Descriptive statistics

Tables 1 presents the descriptive statistics for the sample which has values for full sample, covering 5198 firm-years, including 33 industries and 1394 companies.

The absolute discretionary accruals (absDACC) has a mean of 0.055 of lagged assets, while the REM (COM_REM) has a mean of 0.037 of lagged assets. This suggests that on average, companies in the sample manage earnings more through discretionary accruals then through REM. This is consistent with prior research by Zhang (2012) and Jordaan et al. (2018), who also find similar results. We can also see that dummy variable indicating greenwashing (dGW) has a mean of .3941, indicating that in 39.41% of the firm-years observations, the company is greenwashing. Lastly, we can also see that, perhaps

surprisingly, only in 25,3% firm-year observations, the company makes use of integrated reporting (dGRI). Sustainability compensation incentives are used in 17.4% of the

observations.

Table 1: Descriptive statistics for the full sample

Variables N Mean Median Std. Deviation

Dependent variables absDACC 5198 .0550325 .0294168 .12410 ABNCFO 5198 .0314337 .0229841 .22471 ABNPROD 5198 -.0183574 -.0120737 .22176 ABNDISEXP 5198 -.0123084 -.0374839 .68997 COM_REM 5198 .0374828 -.0234664 .85301 Test variables absGW 5198 9.88272 13.075 49.79849 dGW 5198 .3941901 0 .48872 dGRI 5198 .2529819 0 .43476 dSUSCompInc 5198 .1742978 0 .3794019 Control variables SIZE 5198 8.249581 8.212306 1.629449 RDIntensity 5189 .044272 .008443 .0961645 NI 5198 740.5397 149.586 2868.789 LEVERAGE 5198 1.287112 .477392 3.499311 EPS 5198 1.958867 1.588629 5.488099 TOBINQ 5198 2.413665 1.878704 1.895502 BoardInd 5198 _.7701381 .8182 _.1648894 BoardGDiv 5198 _.1565428 .1429 _.1101915

(21)

Table 2: Breakdown per industry, sorted by greenwashing

Table 2 presents the breakdown of firm-year observations in the sample per industry, sorted by whether they are greenwashing or not. What we can see from this table, is that

greenwashing occurs in many different industries, from utilities to production, to business services and retail. Demonstrating again that greenwashing can be a problem for all sorts of investors and stakeholders.

dGW = 0

dGW

= 1 Totals

Industry (based on 2-digit SIC code) N N N %

Metal Mining 117 35 152 2,92%

Oil and Gas Extraction 289 79 368 7,08%

Heamy Construction, Except Building Construction, Contractor 3 23 26 0,50%

Food and Kindred Products 71 131 202 3,89%

Apparel, Finished Products from Fabrics & Similar Materials 32 34 66 1,27%

Furniture and Fixtures 27 26 53 1,02%

Paper and Allied Products 6 69 75 1,44%

Printing, Publishing and Allied Industries 44 4 48 0,92%

Chemicals and Allied Products 414 313 727 13,99%

Petroleum Refining and Related Industries 65 51 116 2,23% Rubber and Miscellaneous Plastic Products 10 41 51 0,98% Stone, Clay, Glass, and Concrete Products 12 48 60 1,15%

Primary Metal Industries 73 39 112 2,15%

Fabricated Metal Products 32 67 99 1,90%

Industrial and Commercial Machinery and Computer Equipment 93 251 344 6,62% Electronic & Other Electrical Equipment & Components 127 320 447 8,60%

Transportation Equipment 47 161 208 4,00%

Measuring, Photographic, Medical, & Optical Goods, & Clocks 201 154 355 6,83% Miscellaneous Manufacturing Industries 8 4 12 0,23%

Motor Freight Transportation 18 21 39 0,75%

Wholesale Trade - Durable Goods 77 37 114 2,19%

Wholesale Trade - Nondurable Goods 64 14 78 1,50% Automotive Dealers and Gasoline Service Stations 60 0 60 1,15%

Apparel and Accessory Stores 87 2 89 1,71%

Eating and Drinking Places 77 15 92 1,77%

Miscellaneous Retail 96 9 105 2,02%

Real Estate 16 4 20 0,38%

Holding and Other Investment Offices 69 31 100 1,92%

Business Services 649 51 700 13,47%

Amusement and Recreation Services 57 6 63 1,21%

Health Services 90 0 90 1,73%

Educational Services 32 0 32 0,62%

Engineering, Accounting, Research, and Management Services 86 9 95 1,83%

(22)

I tested the all variables for multicollinearity and found that the variance inflation factors (VIF) ranged between 1.00 and 1.96 for both sets of models. Based upon the Pearson Correlation matrixes (see tables A1 and A2 in the appendix), neither of the variables significantly correlate with any other variable at values above (-)0.5. The strongest correlation, in both models, exists between dGRI and SIZE (at 0.492), but based upon the corresponding VIF factor (1.96) this is not enough to raise concerns regarding

multicollinearity.

Since I am using panel data, and both the Pearson correlation matrix as well as the VIF estimates are primarily used for normal regression models, I also test(untabulated) the correlation of the coefficients (VCE) in the panel data regression and find that none of the coefficients correlate with each other at values above (-)0.5, again indicating that

multicollinearity isn’t a large concern in my models. 4.2 Regression results

Before presenting the results, I first need to test the panel regression models for endogeneity or, the correlation between a variable and the error term in the model.

This is important, since endogeneity problems may introduce a bias in both the coefficients and the errors due to misrepresentation of the model variables. The Hausman Test

(Hausman, 1978) can be performed to detect endogenous variables in a regression model and based on the outcome either a fixed effect or random effect panel regression will be used. The result from the Hausman tests for both models are tabulated in table 3 and 4. Table 3: Hausman Test for Accrual based Earnings Management (Model A)

Coefficients (b) (B) (b-B) sqrt(diag(V_b-V_B)) fe re Difference S.E. dGW -.0125935 .0049338 -.0175273 .0072639 dGRI .0039596 .0110591 -.0070996 .0085672 COM_REM -.008178 .0063278 -.0145058 .0012882 SIZE -.0120547 -.0025718 -.0094829 .0079848 RDIntensity -.2933159 .0479825 -.3412983 .0695581 NI 7.38e-09 1.99e-07 -1.91e-07 1.17e-06 LEVERAGE .0000144 .0000101 4.28e-06 .0000175 EPS -.0002226 -.0001728 -.0000497 .0003201 TOBINQ .0095251 .0044583 .0050668 .0020814 BoardInd -.0162651 -.0166474 .0003823 .0291219 BoardGDiv -.0133072 .0200527 -.03336 .0345356 dSUSComp .0179838 .0178333 .0001505 .0034876 chi2 164,41 Prod>chi2 0,0000***

b = consistent under Ho and Ha; B = inconsistent under Ha, efficient under Ho *** = significant at P<0.001

From the results we conclude following for model A. Since the p-value of the test criteria in model A (see table 3) is less than 0.001, we find sufficient evidence to reject the

null-hypothesis, meaning using the fixed effect model yields more appropriate results for the Accrual based Earnings Management model (model A). Using the real difference in ESG scores (realGW) instead, yields similar test results (chi2 163.04, p<0.000).

(23)

Table 4: Hausman Test for Real Earnings Management (Model B) Coefficients (b) (B) (b-B) sqrt(diag(V_b-V_B)) fe re Difference S.E. dGW -.0794456 .0709677 -.1504133 .0481733 dGRI .0022504 .0225638 -.0203134 .0566164 absDACC -.3688704 .2241843 -.5930548 .05439 SIZE -.2484101 .0172505 -.2656606 .0534027 RDIntensity -.3615917 -.5171268 .1555351 .4670164 NI 7.33e-06 .0000106 -3.31e-06 7.74e-06 LEVERAGE .0001279 -.0000592 .0001871 .000112 EPS .0042409 .0032628 .0009781 .0020853 TOBINQ .0361743 .0786315 -.0424572 .0139255 BoardInd -.2470706 -.0267923 -.2202783 .1941455 BoardGDiv -.1503381 .113445 -.2637831 .2296508 dSUSComp .1426807 .1385286 .0041522 .0225728 chi2 174,87 Prod>chi2 0,0000***

b = consistent under Ho and Ha; B = inconsistent under Ha, efficient under Ho *** = significant at P<0.001

For model B, (see table 4) we find a p-value of less than 0.001 from which we infer a significant difference. We therefore reject Ho (p<0.000), meaning that according to the Hausman test, the fixed effects model yields more appropriate results with respect to real earnings management model (model B). Using the real difference in ESG scores (realGW), the Hausman test yields similar results (chi2 197.89, p<0.000).

What this test cannot cover however, is a second problem which also associated with endogeneity, known as simultaneity, meaning a direct causal relation between two variables. In case of my study, based on my hypothesis, this problem may still exist. However, based on the multicollinearity analysis, and the use of random or fixed effects models to better estimate the regression models, the remaining bias in my models should be limited.

4.2.1 Accrual based earnings management and Greenwashing (model A)

The results depicted in table 5 suggest that there is no associated between earnings

management through discretionary accruals (absDACC) and greenwashing(dGW), although contrary to the expectation the coefficient is negative. Based on these results I am unable to find support for hypothesis H1.

The interaction between AEM and REM as the coefficient is negative and strongly significant (p-value of less than 0.001), suggesting that AEM and REM are used as trade-offs for each other, which is consistent with prior research (Roychowdhury, 2006; Cohen et al., 2008; Zhang, 2012).

Looking at the results regarding hypotheses H2A and H2B I conclude that, in line with my expectations, the coefficients for GRI use (dGRI) as well as Sustainability Compensation Incentives for Executives (dSUSComp) are positive. But since the coefficient for dGRI is not significant and the coefficient for dSUSComp is only significant at p<0.10, I find only weak support for hypothesis H2B, implying that Sustainability Compensation Incentives for

(24)

Executives does reduce earnings management through discretionary accruals. This result is in line with prior research by Jordaan et al. (2018).

Table 5: Earnings management and Greenwashing Accrual based earnings management

(Model A) Real earnings management (Model B) Dependent variable

(absDACC) Coefficient

Dependent variable

(COM_REM) Coefficient

Test variables Test variables

dGW -.01259345 dGW -.07944561

dGRI .00395956 dGRI .00225042

dSUSComp .16063033 * dSUSComp .14268071 ***

Control variables Control variables

COM_REM -.008178 *** absDACC -.36887041 *** SIZE -.01205473 SIZE -.24841008 *** RDIntensity -.29331589 *** RDIntensity -.36159175 NI 7,38E-06 NI 7,33E-03 LEVERAGE .00001443 LEVERAGE .00012792 EPS -.00022257 EPS .0042409 TOBINQ .0095251 *** TOBINQ .03617426 * BoardInd -.01626508 BoardInd -.24707059 BoardGDiv -.01330722 BoardGDiv -.15033805 constant .16063033 * constant 2.241549 *** R2 0.33588969 R2 0.36597384 Adj-R2 0.08982561 Adj-R2 0.13105645

*, indicates significance, two-tailed at the 10% level; **, indicates significance, two-tailed, at the 5% level; ***, indicates significance, two-tailed, at the 1% level

4.2.2 Real earnings management and Greenwashing (model B)

The results depicted in table5suggest thatthere is no associated between real earnings management (COM_REM) and greenwashing (dGW), although contrary tothe expectation the coefficient is negative.Based on these results I am unable to find support forhypothesis H1.

The interaction between REM and AEM as the coefficient is negative and strongly

significant (p-value of less than 0.001), suggesting that REM and AEM are used as trade-offs for each other, which is consistent with prior research(Roychowdhury, 2006; Cohenet al.,

2008; Zhang, 2012).

Looking at the results regarding hypotheses H2A and H2B I conclude that, in line with my expectations,the coefficients for GRI use (dGRI) as well asSustainability Compensation Incentives for Executives (dSUSComp) are positive. I again find insufficient evidence for hypothesis H2A since thecoefficient is not significant, I can however support H2B, sincethe coefficient is highly significant (p-value of less than 0.001). And since higher levels of COM_REM are associated with real earnings management, I can conclude that based on these results, Sustainability Compensation Incentives for Executives do indeed reduce real earnings management.This result is inline with prior research by Jordaanet al.(2018).

(25)

4.2.3 Earnings management and the level of greenwashing

Instead of using a dummy variable for greenwashing (dGW), I can also use the real difference in ESG scores (realGW), thereby assuming that there is a casual relationship between

earnings management and greenwashing. To test this, I adapt the regression models A and B mentioned in 3.2.2 to include the real difference in ESG scores (realGW) instead of

greenwashing dummy (dGW) and rerun the regression.

Table 6: Earnings management and level of greenwashing (real difference) Accrual based earnings management

(Model A) Real earnings management (Model B) Dependent variable

(absDACC) Coefficient

Dependent variable

realGW .00018923 realGW .00374053 ***

dGRI .0050694 dGRI .02838001

dSUSComp .01796698 ** dSUSComp .13569654 ***

COM_REM -.00835183 *** absDACC -.37496242 *** SIZE -.01251478 SIZE -.25469914 *** RDIntensity -.29463985 *** RDIntensity -.39379871 NI 6,59E-05 NI 8,15E-03 LEVERAGE .00001458 LEVERAGE .0001333 EPS -.00023652 EPS .00401843 TOBINQ .00954218 *** TOBINQ .03492797 * BoardInd -.01795237 BoardInd -.26228159 BoardGDiv -.01258034 BoardGDiv -.06501593 constant .1585068 * constant 2.2227046 *** R2 0.33586819 R2 0.36889698 Adj-R2 0.08979615 Adj-R2 0.13506266

*, indicates significance, two-tailed, at the 10% level; **, indicates significance, two-tailed, at the 5% level; ***, indicates significance, two-tailed, at the 1% level; Accrual based earnings management

The results depicted in table 6, when looking at the Accrual based earnings management (model A) are similar to the results in table 5. Although the coefficient for greenwashing (realGW) is now positive, it is still not significant. This means that, yet again, I find

insufficient evidence to conclude in favor of hypothesis H1 when looking at relation between greenwashing and earnings management through accruals.

With respect to the use of GRI (dGRI) and Sustainability Compensation Incentives for

Executives (dSUSComp), the results are also similar although the coefficient for dSUSComp is now significant at the 5% level, meaning that I can now also conclude in support of H2B when it concerns earnings management through discretionary accruals. This result is in line with the theoretical argument presented by Delmas & Burbano (2011), as well as the findings of Martinez-Ferrero et al. (2013) and Jordaan et al. (2018), who argue that managers with incentives to reduce information asymmetry will minimize earnings management, especially since the coefficient for dSUSComp in the real earnings

(26)

Real earnings management

When looking at the earnings management model (model B), I find significantly different results when looking at the association between the level of greenwashing and real earnings management, not only is the coefficient positive, it is also highly significant (p-value of less than 0.001). This result is consistent with my expectations, suggesting that greenwashing is indeed used as a substitute for real earnings management, which would imply a conclusion in favor of hypothesis H1 when related to real earnings management.

This result lends some support for the opportunistic approach, where managers, based on prior research by Grougiou et al. (2014) and Hemingway and Maclagan (2004), turn to CSR reporting to divert shareholder’s attention. This result, however, is also somewhat

consistent with prior research by Bozzolan et al. (2015) who found that CSR activities act as a constraint on REM but not AEM, although this may also be explained by that fact that

companies with better CSR performance are more likely to give up REM than AEM because of negative effects of REM on future performance (Bozzolan et al., 2015).

The results of Bozzolan et al. (2015) suggest that the positive result in table 6 may not necessarily be due to greenwashing, but simply due to fact that CSR reporting coincides with an actual reduction in earnings management practices, thereby implying support for the ethical argument, where managers that are more transparent in their corporate reporting are less likely to manage earnings.

Table 7: Real earnings management and level of greenwashing (real difference), by dGW Real earnings management (Model B),

non-greenwashing

Real earnings management (Model B), greenwashing

Dependent variable

realGW .00460964 *** realGW .00364229 *

dGRI .11393104 dGRI -.04308495

dSUSComp .10026539 * dSUSComp .17620867 *

absDACC -.36579902 * absDACC -.55242111 ** SIZE -.13735188 * SIZE -.29887091 * RDIntensity -.13131904 RDIntensity -3,0286510 NI -9,55E-03 NI .00001349 LEVERAGE .00007081 LEVERAGE .00016449 EPS .00106963 EPS .01426884 TOBINQ .03935671 * TOBINQ .03649856 BoardInd -.79362387 *** BoardInd .18470476 BoardGDiv .1845064 BoardGDiv -.26818688 constant 1.372798 ** constant 2.7540283 * N 3149 N 2049 R2 0.52150872 R2 0.28044576 Adj-R2 0.27825082 Adj-R2 0.00697637

*, indicates significance, two-tailed, at the 10% level; **, indicates significance, two-tailed, at the 5% level; ***, indicates significance, two-tailed, at the 1% level;

(27)

This thus warrants further review, for which I split my sample intosubsamples containing greenwashing and non-greenwashing companies (using dGW) and rerun the regressions for the real earnings management model (model B).

The results of this analysis are presented in table 7, and theyseem to support the conclusion thatin non-greenwashing companies, CSR reportingdoes indeed lead to a reduction in real earnings management since the coefficient (realGW) is again positive and highly significant (p-value of less than 0.001). The coefficient in the subsample of companies that do

greenwash, is also positive, but only significant at the 10% level.

Overall the results, based upon tables 6 and 7, would lead me to conclude, consistent with prior research (Kimet al.,2012;Bozzolanet al.,2015) andwith respect to real earnings management,that even thoughthereisapositive relationship betweenlevel of

greenwashing and real earnings managementin both subsamples,the effect is far stronger in companies that do not greenwash. Thiswould imply thatCSR reporting in general, while it may be subject to some amount of greenwashing, reduces the likelihood that managers engage in real earnings management, and in fact does not support a conclusion that greenwashing is asubstitute for realearnings management.

Since it is clear that a difference in ESG scores, regardless ofwhether that difference would constitute greenwashing, has impact on the use of real earnings management, this brings to mind a matter upon I have touched briefly before at the start ofsection 4.2, simultaneity or causality.One might argue that insteadof being independent measures, the interplay between earnings management and greenwashing as well as the fact that the different measures of earningsmanagement are usedas substitutes for each other, may very well mean that there is in fact a causal relationship between earnings management and

greenwashing.While I will leave further study of this possible causality for a future study, I will add that,based upon the results in table 7,I cannot simply state that this causal relationship doesn’t exist, meaning that I will emphasize some caution when using the results of this study.

4.2.4 Greenwashing and Integrated reporting

The results in table 8 contain the regressionresult for model C,with which Iaim to investigate the impact of integrated (dGRI) and the use ExecutiveSustainability

Compensation Incentives (dSUSComp) on the greenwashing (dGW), and on the difference between the ESG scores (realGW).

The firstthingI notice is the very high explanatory power of model fits (R2 of 0.831 and 0.931 respectively), indicating that predictive power of both models isvery good. Looking at the model for greenwashing (dGW), we can see that again the coefficient for GRI use is positive, as expected, but notsignificant. This suggests that the useofGRI, contrary to expectation isnot an important factor when predicting greenwashing, which is consistent withmy earlier analysis.

As expected, the coefficient for Executive SustainabilityCompensation Incentives is negative and significant, suggesting that Sustainability incentivesdoincrease the likelihood for

greenwashing.

(28)

Zaman (2016) who find that (more) gender diverse board are associated with higher quality sustainability reports, suggesting that this may also reduce the likelihood of Greenwashing. This is based on the notion that a more diverse board will take a more balanced view and pay greater attention to stakeholder concerns (Al-Shaer & Zaman, 2016).

Lastly, neither measure for earnings management is a significant predictor, which is consistent with my earlier findings.

Table 8: Integrated reporting and sustainability incentives Greenwashing and integrated reporting

(Model C)

Greenwashing and integrated reporting (Model C)

Dependent variable

(dGW) Coefficient

Dependent variable

(realGW) Coefficient

dGRI .02474897 dGRI -7.5115255 ***

dSUSComp -.03522182 ** dSUSComp 2.435927 **

absDACC -.04964167 absDACC 3.1327892 COM_REM -.00694296 COM_REM 1.3793442 *** SIZE .00972498 SIZE 1.7819105 RDIntensity -.04645168 RDIntensity 10.037992 NI -1,79E-03 NI -.0001905 LEVERAGE .0000176 LEVERAGE -.00196677 EPS .00042829 EPS .04519491 TOBINQ -.00830622 TOBINQ .46420102 BoardInd .1000346 BoardInd 2.2562142 BoardGDiv .41356997 *** BoardGDiv -31.368228 *** constant .19762056 constant -1.90291 N 5198 N 5198 R2 0.83120163 R2 0.93171593 Adj-R2 0.76865898 Adj-R2 0.90641553

Looking at the difference in ESG scores (real GW), table 8 shows results in line with the hypotheses H3A and H3B, namely a highly significant negative coefficient for the use of GRI (dGRI) and a significant positive coefficient for Executive Sustainability Compensation Incentives (dSUSComp) such that I conclude that I can accept both hypothesis H3A and H3B. Based on these results we can thus state that the use GRI decreases the difference in ESG scores, thereby reducing the likelihood of greenwashing, whereas the opposite is true for the use of Sustainability compensation incentives for Executes which increases the difference in scores and thus increases the likelihood of greenwashing.

The coefficient for the real earnings management proxy (COM_REM) is also positive and highly significant, which is consistent with my earlier findings. Lastly, the coefficient for Board Gender Diversity is strongly significant and negative, which is consistent with prior research by Al-Shaer & Zaman (2016). I find no support for significant influence of Board Independence, as one might expect based on prior research by Neville et al. (2018).

Greenwashing and Earnings Management: Symptoms of the same problem?