The influence of tone at the top on earnings management : ethical leadership through an ethical organizational culture

The influence of tone at the top on earnings

management

Ethical leadership through an ethical organizational culture

Author: Johnny Chan

Student number: 10106944

Version and date: Master thesis final version, August 2014 Supervisor: Dr. Alexandros Sikalidis

MSc Accountancy & Control 2013-2014 Faculty of Economics and Business University of Amsterdam

Abstract

This thesis examines whether a positive ethical tone at the top contributes in reducing earnings management. Tone at the top refers to the ethical atmosphere that is created in the workplace by the organization's leadership. It acts as a mechanism based on a culture of ethics and trust to make proper ethical decisions which leads to less real earnings management. I predict that when top management sets an ethical organizational culture, firms are less likely to engage in earnings management. In sum, I document that (1) a general tone at the top is not related to earnings management, (2) an ethical tone at the top is negatively associated with real earnings management only, (3) firms have less real earnings management when top management has made a public commitment to ethics and when firms possess appropriate communication tools to improve ethics, (4) firms without an ethical tone at the top manage their earnings upward with real activities while firms that do possess an ethical tone at the top have weak downward real earnings management, (5) a code of ethics is a proper channel to communicate ethics and (6) tone at the top is not related to internal control deficiencies, suggesting that it works as an extra layer of protection next to internal controls. My results are partially robust when considering an alternative proxy for earnings management: the F-score for financial reporting aggressiveness. This study provides evidence of tone at the top as a mechanism against earnings management and therefore contributes to the literature, especially as there is barely any empirical research of tone at the top other than case studies, particularly in the field of earnings management.

Keywords: Tone at the top, ethics, organizational culture, earnings management, real activities management, financial reporting aggressiveness.

Table of contents

1 Introduction ... 5

2 Literature review and hypothesis development ... 7

2.1 What is tone at the top ... 7

2.2 Earnings management and its ethics ... 8

2.3 Influence of tone at the top elements on unethical behavior: earnings management ... 10

2.3.1 Organizational culture and ethics ... 10

2.3.2 Leadership ... 12

2.3.3 Communication ... 13

3 Research method ... 14

3.1 Sample ... 14

3.2 Dependent and independent variables of interest ... 16

3.3 Empirical models ... 21

3.4 Control variables ... 22

4 Results ... 25

4.1 Descriptive statistics and univariate analysis ... 25

4.2 The relation between Tone at the Top and Earnings Management ... 30

4.2 The relation between Tone at the Top and Internal Control Deficiencies ... 33

4.3 The relation between Code of Ethics and Earnings management ... 34

5 Sensitivity analysis... 36

The alternative measure F-Score for financial reporting aggressiveness ... 36

6 Conclusion ... 40

References ... 42

Appendix A: Examples of main DICTION variables from CEO letters ... 45

Table of Figures

Table 1: Sample selection ... 15

Table 2: Summary of main proxies ... 15

Table 3: List of variables ... 23

Table 4: Industry overview ... 25

Table 5: Full sample descriptive statistics ... 26

Table 6: Correlations among Tone at the Top, Earnings Management and other selected variables ... 29

Table 7: Difference tests (Independent two-sample t-tests) ... 30

Table 8: Multiple Regression of Earnings Management on Tone at the Top ... 32

Table 9: Difference tests using ICDs (Independent two-sample t-tests) ... 33

Table 10: Multiple regression of ICDs on Tone at the Top ... 33

Table 11: Difference tests for code of ethics (Independent two-sample t-tests) ... 34

Table 12: Multiple regression of Earnings Management on Code of Ethics ... 35

Table 13: Difference tests using the alternative F-Score for financial reporting aggressiveness ... 37

Table 14: Multiple regression of the alternative F-Score on Tone at the Top ... 38

1 Introduction

Managers are under pressure to perform well and meet expectations. To do this, some resort to opportunistic earnings management. An example of such behavior is the well-known WorldCom accounting scandal. In the 90’s, WorldCom had paid too much for acquisitions, was stuck with overcapacity and a bad integration between the acquired companies (Moberg & Romar, 2006). The market environment deteriorated due to heightened competition and decreasing demand, leading to price wars. Consequently, revenues and share prices fell. WorldCom’s CEO did not accept this and encouraged managers to do whatever necessary to bring in revenue (Kaplan & Kiron, 2007). As a result, the CFO decided to make use of earnings management. Even though the fraudulent reporting methods were relatively simple, the CFO was able to do this due to the corporate culture. WorldCom had an autocrative management style pressuring employees. Each department had its own rules and management style. A top down attitude ruled in which employees should not question their supervisors, but simply do what they are told. And when they had a complaint, there was no outlet for them to express their concerns. Several employees were not even aware of the existence of the internal audit department (Kaplan & Kiron, 2007). The problem in this case, besides a lack of internal controls, is a lack of a positive tone at the top which could have uncovered the fraud. Instead, employees were pressured to participate and keep quiet about the fraud.

In reaction to WorldCom and previous accounting scandals, regulation increased. In the U.S., the SOX was implemented. It was designed to reduce earnings management, strengthen internal controls and to protect shareholders, investors and employees (Coates, 2007). Despite these measures, opportunistic earnings management is still used. There have even been other issues such as banking scandals (e.g. Libor scandal). Regulation alone may not be sufficient. As Higson (2010, p.17) says: “One has to start to consider the possibility that organisations are becoming too large and too complex to be effectively controlled and monitored and that corporate governance may in fact be a modern fantasy.” Instead of focusing on more regulation to reduce earnings management, a possibility is to focus on the organizational culture and the tone at the top. This means a culture of ethics and trust fostered by a positive tone at the top to make ethical decisions.

An experiment on financial reporting decisions by D’Aquila and Bean (2011) demonstrates that employees tend to choose unethical behavior if they are put in a tone at the top that does not encourage ethical decisions. This suggests that a tone at the top is required to prevent unethical behavior such as earnings management. Such importance of tone at the top 5

has been acknowledged by the Committee of Sponsoring Organizations of the Treadway Commission (COSO). COSO sees tone at the top as part of the internal/control environment, which are included in its Internal Control and Enterprise Risk Management frameworks (COSO, 1992; 2004). To be clear, tone at the top refers to the ethical atmosphere that is created in the workplace by the organization's leadership (ACFE, 2013). It is the message, the attitude and the culture the board of directors disseminates throughout the organization. These are formed by the consistency between statements, assertions, explanations of the management and its actions (Bruinsma & Wemmenhove, 2009).

In this thesis the main research question is: ‘Does a positive ethical tone at the top contribute to reducing earnings management?’ The effects of tone at the top on lower leveled employees will be researched. To answer the research question I first look at what the existing literature says about tone at the top. Specifically, what it entails, how it affects earnings management through ethics and when these effects occur. Based on the literature, I have developed the hypotheses which are tested. To do this, regression models have been designed and data from American Fortune 500 firms are analyzed.

As there is little published research on tone at the top, especially in the field of earnings management, this thesis will fill that void. It also contributes by researching whether a positive ethical tone at the top really works against earnings management and how. This is interesting because tone at the top can be an alternative or complement to existing regulation in combating opportunistic earnings management. It is especially important considering that earnings management has led to the downfall of large companies such as WorldCom and Enron. Consequently, it could help firms in case they find standard internal controls to be insufficient.

The remainder of this thesis is organized as follows. In section 2, related literature are discussed and hypotheses are developed. Section 3 contains discussion on sample selection and research design. Results are presented in section 4. A sensitivity analysis is performed in section 5. Lastly, a summary of the findings and concluding remarks appear in section 6.

2 Literature review and hypothesis development

This section discusses related literature on tone at the top, ethics and earnings management. It is structured in three subsections. The first subsections explains what tone at the top is and presents its elements. In subsection 2.2 earnings management and its ethics are discussed. Subsection 2.3 discusses what the influence of the tone at the top elements are on (un)ethical behavior such as earnings management and also presents the hypotheses.

2.1 What is tone at the top

There are different concepts and frameworks of tone at the top, three of these are presented. Firstly, COSO (1992) created an internal control framework with the purpose of achieving, among others, reliability of financial statements and to comply with applicable laws and regulations. Earnings management does not fit into this. The framework consists of five components of which the first, control environment, interests this research. This is because tone at the top is part of the control environment. The control environment serves as the foundation for internal control as it sets the tone of an organization which has influence on the control consciousness of its employees. To be specific, the upper management establishes the tone of which the intent is to convey a commitment to integrity and high ethical standards in the execution of all activities throughout the organization, translating into an ethical organizational culture (COSO, 1992). According to COSO (1992), such an ethical organizational culture contributes to the effectiveness of policies and internal controls. It also helps influence behavior that is not even subject to the most advanced controls. Policies are what management wants, but the organizational culture determines what really happens; which policies are obeyed or ignored.

Secondly, Swinkels (2003, p.12) defines tone at the top as “management control that consists of all actions taken deliberately, with which the management influences employee behavior to such an extent that the proper strategy is implemented in the organization”. He adds that employees their perception of the tone at the top is positively influenced if the top provides guidance, clear assertions together with the right example. Swinkels (2003) separates tone at the top into three elements: leadership, communication and culture. He emphasises the consistency between these elements and guarding it. In that, when management makes assertions and policies their actions should be consistent with their own

words. The leaders should act as role models and their words should be communicated top – down, creating a culture of ethical values and norms (Swinkels, 2013).

Lastly, as there are multiple concepts of tone at the top Bruinsma and Wemmenhove (2009) have examined what experts, such as auditors and professors, think of the subject. They find a consensus of four elements: management, communication, culture and structure. Management is about demonstrating inspiring leadership and setting the right example. Management forms the dominant culture. This is communicated from top down and bottom up through channels such as symbols, rituals and assumptions in verbal, visual and digital form. An example is implementing a code of ethics. After communicating, the culture is affected. Employees should be encouraged to take ethically correct decisions and be rewarded for it. If something goes wrong, a policy should be present to protect whistleblowers (Bruinsma & Wemmenhove, 2009). Lastly, the corporate structure facilitates the three elements explained above. Organizations should have independent functions with responsibilities, such as an independent supervisory board (Bruinsma & Wemmenhove, 2009).

All authors above have three elements in common that form the tone at the top. These are leadership, communication and culture. I will continue this thesis based on these three elements, emphasizing leadership and (un)ethical culture.

2.2 Earnings management and its ethics

What is earnings management

“Earnings management is the choice by a manager of accounting policies, or actions affecting earnings, so as to achieve some specific reported objective” (Scott, 2011, p. 403). Healy and Wahlen (1999, p. 368) say: “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 accounting numbers.” Levitt (1998) says earnings management is the gray area between legitimacy and outright fraud, where where earnings reports reflect the desires of management rather than the underlying financial performance of the company. He calls it poison to the financial reporting process. The highest degree of earnings management is fraudulent financial reporting (Dechow & Skinner, 2000). This is misleading users of financial reports by recording sales before they are realizable, recording fictitious sales, backdating sales invoices and overstating inventory by recording fictitious inventory (Dechow & Skinner, 2000).

The ethics of earnings management

In general, ethics is the concept of right and wrong, good and evil, justice and crime. Unethical behavior at the workplace is defined as behavior in and by organizations that violate generally accepted moral norms of behavior (Jones, 1991). Earnings management is a form of unethical behavior. The reason for that is because earnings management is used selfishly in the interest of managers or companies and not in the best interest of its stakeholders. As seen in the previous paragraph, earnings management misleads stakeholders about the underlying economic performance of the company (Healy & Wahlen, 1999) with some authors even calling it poison (e.g. Levitt, 1998).

Management can also force its employees to participate in earnings management practices (e.g. WorldCom, see introduction). When they submit, even though they are forced, they become accomplices of earnings management which is also unethical. Being aware of earnings management but ignoring it is negligence and also unethical. Employees should have the responsibility to report malpractice. Whistleblowing policies facilitate this by allowing employees to take up normative responsibility and act out a ‘duty owed’ (Tsahuridu & Vanderkerckhove, 2008).

As Levitt (1998) mentions, earnings management is a gray area between legitimacy and outright fraud. This makes it difficult to judge whether certain earnings management practices are unethical or not. But nonetheless, to reduce earnings management ethics should be paid attention to. When Elias (2002) researched determinants of earnings management ethics such as moral philosophies and social responsibility. He found that “individuals who believed in corporate social responsibility and long-term gains rated earnings management as more unethical while individuals who believed in short-term gains rated them as more ethical. These findings suggest that attention should be paid to an individual’s moral philosophy and time horizon if the objective is to reduce earnings management occurrence” (Elias, 2002, p. 43). Elias (2004) continued by researching the impact of corporate ethical values on perceptions of earnings management. He found that certified public accountants with high ethical standards were more likely to view earnings management activities as unethical compared to certified public accountants with low standards. Once again, this suggests attention should be paid to ethics to reduce earnings management.

2.3 Influence of tone at the top elements on unethical behavior: earnings management

This section focuses mainly on how the top can influence lower level employees through the organizational culture and ethics, leadership and communication. As there is very little research on tone at the top, especially in the field of earnings management, I don’t look at the influence on specifically earnings management. Instead I look at a broader unethical behavior that can lead to earnings management and fraudulent financial reporting (e.g. Hunton, Hoitash and Thibodeau, 2010). This paragraph is structured in three subsections based on the three elements of tone at the top: organizational culture and ethics, leadership and communication.

2.3.1 Organizational culture and ethics

The organizational culture consists of shared assumptions, beliefs and values of an organisation. This culture forms employees their attachment, commitment, identification and feelings for the company. When employees make decisions they take into account their organisation culture and goals (Almhelm, 1992). The organizational culture helps employees understand what they do and why by providing a justification for their actions. Because of this, the organizational culture can influence the good/bad or ethical/unethical behavior of employees by showing what the correct manner is (Watson, 2003).

Past literature mostly looks at unethical or deviant behavior at the workplace. As mentioned before, opportunistic earnings management is a form of unethical behavior. Kaptein (2011) describes six dimensions of ethics which can influence ethical/unethical behavior such as earnings management. These are clarity of ethical standards, ethical role modeling of management and supervisors, capability to behave ethically, commitment to behave ethically, visibility of (un)ethical behavior, openness to discuss ethical issues and reinforcement of ethical behavior. These will be discussed shortly.

First of all, unethical behavior can occur due to an ambiguous ethical situation. Some situations don’t scream out the difference between ethical and unethical. As Kaptein (2011) says, general intuitive moral norms are insufficient to make a distinction between ethical and unethical behavior. That’s why organizations should clarify the ethical standards, explaining what behavior is desirable. Greater clarification communicates the importance of ethical standards rendering employees and managers less likely to violate them (Kaptein, 2011). So far clarity focuses on preventing unethical behavior. But it can also help detect financial statement fraud. By providing ethics training and explaining the societal and economic 10

consequences of financial statement fraud, employees are more likely to report and therefore whistleblow on financial fraud (Robinson, Robertson and Curtis, 2012).

Secondly, the capability to behave ethically refers to “the extent to which managers and employees believe they have sufficient time, budgets, equipment, information and authority at their disposal to fulfill their ethical responsibilities” (Kaptein, 2011, p. 849). When they are incapable of doing this through legitimate means, they resort to unethical behavior. Especially in high-pressure cultures set by top management, employees might prioritize other matters, such as meeting financial targets through earnings management, over complying with ethical standards (Treviño, 1986).

Thirdly, a lack of commitment to ethical standards increase the risk of unethical behavior. These risks exist in cultures consisting of demotivation, mistrust and dissatisfaction (Kaptein, 2011). Instead, encouraging employees to identify with the organizational values motivate them intrinsically to comply with them (Tyler & Blader, 2005).

Fourthly, “studies emphasize the importance of perceived visibility because of its potential to expose unethical behavior and for acting as deterrent owing to the increased likelihood of getting caught” (Kaptein, 2011, p. 850).

Fifthly, an ethical culture is one in which employees are open to discuss ethical dilemmas and alleged unethical behavior. Such open communication engages people to share and learn from each other (Kaptein, 2011). According to Bird and Waters (1989) moral issues go unnoticed and unacknowledged if they are not openly talked about. This leads to higher moral stress and lower authority of ethical standards. Such an open culture can also encourage whistleblowers to report fraud (Robinson, Robertson and Curtis, 2012). For these reasons, the openness of ethics can be a good predictor of the frequency of unethical behavior (Treviño et al., 1999).

Lastly, ethical behavior should be reinforced by rewarding employees for its ethical behavior and punishing its unethical behavior. This works as a message that unethical behavior is unacceptable (Kaptein, 2011). When the punishment outweighs the potential reward employees avoid such unacceptable behavior.

As top management sets the tone of ethics in the organization through the six dimensions explained above, top management has the ability to encourage or discourage lower leveled employees to behave unethically which can lead to more or less earnings management. Therefore I hypothesize that:

Hypothesis 1: a positive ethical tone at the top is negatively related to earnings management. Thus, a firm with a positive ethical tone at the top is less likely to engage in earnings management

2.3.2 Leadership

Brown, Treviño and Harrison (2005, p.120) define ethical leadership as “the demonstration of normatively appropriate conduct [...], and the promotion of such conduct to followers through two-way communication, reinforcement, and decision-making”. Employees observe their leaders’ appropriate conduct and adopt it, according to the social learning theory (Bandura, 1986). Leaders therefore set the norm, if leaders act ethically so will the employees. This is also called the trickle-down mechanism, in which traits of ethical leadership at the top influences ethical conduct at lower levels management (Schaubroeck et al., 2012).

Consequently, when employees feel are treated properly, they reciprocate this with beneficial work behavior (i.e. more ethical behavior) and refrain from engaging in destructive behavior (i.e. deviant behavior: earnings management) (Brown & Mitchell, 2010). Ethical leadership is also positively related to employees’ job satisfaction, organizational commitment, willingness to report problems to supervisors and willingness to put in extra effort on the job (Brown & Mitchell, 2010).

On the other hand, unethical leadership is destructive and harmful. Brown and Mitchell (2010) name a few of such leaderships: abusive supervision, supervisor undermining, toxic leadership, and tyrannical leadership. Such leaders are oppressive, abusive, manipulative and undermining (Tepper, 2007). These characteristics can also be seen in the WorldCom case. Unethical leadership affects employees’ work attitude. They engage in deviant work behavior, exert less effort and retaliate (Brown & Mitchell, 2010). A consequence is engaging in more opportunistic earnings management or even fraud. So far this has been in accordance with the previous section: organizational culture and ethics.

Aside from the ethical perspective of tone at the top on earnings management, there is also an internal control perspective. For example Ashbaugh-Skaife et al. (2008, p. 221) say: “Weak internal control in the form of inadequate segregation of duties can allow the misappropriation of assets and alteration of recorded amounts by an employee that is not detected because the company has inadequate staff for monitoring, or lack of action by top management because of a lax control environment. In addition, misstatements can be introduced into the financial reporting process through ‘selective oversights or omissions’ in

accumulating segment and subsidiary information for consolidated reports, as well as through management emphasizing earnings targets in instructions to employees”. This suggests that the top facilitates internal controls by creating a tone which emphasizes a more strict or lax control environment resulting in less or more earnings management. As a positive tone at the top establishes and maintains a strong control environment (COSO, 1992), I hypothesize that:

Hypothesis 2: A positive tone at the top is negatively related to internal control deficiencies. Thus, a firm with a positive tone at the top will less likely have deficient internal controls.

2.3.3 Communication

Communication helps strengthen the norms and values within an organisation. It has an indirect effect on (un)ethical behavior because it communicates the ethical culture and leadership, which should help in preventing and detecting earnings management. There are many channels of communication in the form of verbal, visual and digital form. This paragraph focuses on code of ethics.

A code of ethics is a formal document set by the top containing the accepted and expected standards of ethical behavior. They are aimed at reducing ambiguity and increasing consistent behavior (Stevens, 1999). The advantage is that such codes are mass-orientated and the top can convey its intended tone to a large range of people with the same message (Swinkels, 2003). The disadvantage is that it’s a one-way communication, top-down. Because of this the extent of implementation cannot be measured. Also the success of code of ethics depends largely on other influences, as other organizational systems, informal norms and values can overshadow code of ethics (Weaver, Treviño and Cochran, 1999). But according to the theory code of ethics should have a positive influence. Therefore I hypothesize that:

Hypothesis 3: A positive tone at the top communicated through a code of ethics is negatively related to earnings management. Thus, a firm with a code of ethics is less likely to engage in earnings management.

3 Research method

3.1 Sample

For the sample of firms I have two requirements. Firstly, these firms must be mid to large sized with distinct upper and mid/lower level management. This is necessary because I measure the influence of the top on lower levels. Secondly, English must be their first language. This is required due to the nature of one of my proxies for tone at the top. This proxy is created by analyzing CEO letters with a text-analyzing software called DICTION. This software only works with English text. Additionally, English must be company’s first language and not their second or third language. This is because psychological research shows that when people speak and write in a language different from their mother tongue, they think more rationally (Keysar, Hayakawa & An, 2012). However, tone at the top is something intuitive which is expressed through the management’s words and actions. Such intuitive words and actions are better captured when CEO’s think and write in their mother tongue: English as their first language.

Due to these two requirements I use the Fortune 500 companies from 2008. The Fortune 500 lists the 500 largest companies in the US as measured by their revenues. As a result of their size, they all have a distinct upper and lower level management, thus fulfilling the first requirement. Additionally, these American companies, all speak English as their first language, thereby fulfilling the second requirement.

Data is acquired from four sources: the ASSET4 ESG database, the Compustat database, the AuditAnalytics database and CEO letters from the firms’ annual reports as reported on their websites. These will be from fiscal years 2008 and 2009, the beginning of the global economic crisis. The reason for this is because during tough economic challenges and uncertainty CEOs are forced to fully exploit their leadership skills, tone at the top becomes critical and incentives to manage earnings are higher (Patelli & Pedrini, 2013).

After matching data from the four sources, I obtain an initial sample of 1,000 firm-year observations. Observations missing necessary data are excluded from the sample. Financial firms are also excluded due to their deviating accrual characteristics (Kim, Park and Wier, 2012). I end up with a main sample of 641 firm-year observations and a subsample using CEO letters of 532 firm-year observations. A summary of the sample selection and main proxies are provided on the next page.

Table 1: Sample selection

2008 2009 Total firm-years

Fortune 500 US 500 500 1000

Less: missing from Compustat (52) (57) (109)

Less: missing from Asset4 ESG (69) (64) (133)

Less: financial firms (SIC 6000-6999) (54) (54) (108)

Less: missing data (5) (4) (9)

Main sample 320 321 641

Less: missing CEO letters (54) (55) (109)

Subsample using CEO letters 266 266 532

Tone at the top

Specification 1 (Archival data)

Capture whether tone at the top emphasizes

ethics based on whether:

• The top has made a public commitment to business ethics;

• The company has processes in place to improve business ethics;

• Appropriate communication tools are in place to improve business ethics.

Specification 2

Content analysis of CEO letters -Understand how tone at the top is enacted linguistically and reflected in CEO letters. -Frequency measures and 5 main categories

compared to the norm

Earnings management

Archival data -Discretionary accruals

(Modified Jones model) -Real Earnings Management

(Abnormal cash flow, abnormal production and abnormal

discretionary expenses) Table 2: Summary of main proxies

3.2 Dependent and independent variables of interest

Accrual-based Earnings management

Many studies on earnings management use discretionary accruals as a proxy for earnings quality and earnings management. I too use discretionary accruals based on the modified cross-sectional Jones model as in DeFond and Subramanyan (1998) because of its exceptional specification while requiring less restrictive data. This data is acquired from Compustat. Following Kothari et al. (2005), return on assets (ROA) from the previous year is added as a regressor to control for the effect of performance on measured discretionary accruals. It avoids misspecification, thereby enhancing the reliability of inferences from discretionary accrual estimates (Kim, Park and Wier, 2012). The modified Jones model is estimated for each two-digit SIC firm-year grouping and the residuals from this model are the discretionary accruals (DACC): 𝑇𝑇𝑇𝑇𝑖𝑖𝑖𝑖 𝑇𝑇𝑖𝑖𝑖𝑖−1

= 𝛼𝛼

+ 𝛼𝛼

+ 𝛼𝛼

+ 𝛼𝛼

+ 𝜀𝜀

TAit = total accruals for firm i at year t, calculated as income before extraordinary items1 minus operating cash flow;

∆REVit = change in net revenues in year t from t-1; ∆RECit = change in net receivables;

PPEit = gross property, plant and equipment;

IBXIit-1 = income before extraordinary items at year1 t-1; Ait-1 = lagged total assets.

Real Earnings Management

Firms likely use a mix of accrual-based earnings management and real earnings/activities management. Alternatively they substitute each other using the technique that is less costly (Cohen, Dey and Lys, 2008; Zang, 2012) or when managers are constrained in their ability to manage accruals when constrained by higher quality auditors (Chi, Lisic and Pevzner, 2011). To cover the substitutive nature of these two earnings management methods I include REM as

1 _{Income before extraordinary items is taken from the cash flow statement instead of the balance sheet as Hribar}

and Collins (2002) argue that estimating accruals using a balance sheet approach are potentially contaminated with measurement errors.

16

a proxy for real earnings management (e.g. Cohen, Dey and Lys, 2008). REM is based on the abnormal level of cash flow from operations, abnormal level of production costs from cost of goods sold and change in inventories and the abnormal level of discretionary expenses from R&D, advertising and SG&A expenses.

Sales can be manipulated through price discounts and lenient credit terms (Cohen, Dey and Lys, 2008). In the case of upward real earnings management, cash flow in the current period will decrease. To estimate such abnormal cash flow from operations I follow Roychowdhury’s (2006) model: 𝑅𝑅𝐶𝐶𝐶𝐶𝑖𝑖 𝑇𝑇𝑖𝑖−1

= 𝛼𝛼

+ 𝛼𝛼

+ 𝛼𝛼

+ 𝛼𝛼

+ 𝜀𝜀

CFOt = cash flow from operations in year t; A = total assets;

S = net sales;

∆S = change in net sales.

This model is estimated for each two-digit SIC firm-year grouping and the residuals are the abnormal cash flows from operations. Consistent with Zang (2012) I multiply the abnormal cash flows by negative one, so that the higher the amount, the more likely that the firm is engaging in sales manipulations upwards and vice versa. This results in variable REM_CFO.

The second measure of real earnings/activities management is abnormal production costs. Managers can increase production more than necessary in order to increase earnings (Cohen, Dey & Lys, 2008). This way fixed overhead costs are spread over a larger number of units, thus reporting lower cost of goods sold (COGS) and higher operating margins. Roychowdhury (2006), Cohen, Dey and Lys (2008) define production costs defined as the sum of COGS and change in inventory, and express them as a linear function of contemporaneous sales . Following their studies, I estimate the following model for normal COGS: 𝑅𝑅𝐶𝐶𝐶𝐶𝑆𝑆𝑖𝑖 𝑇𝑇𝑖𝑖−1

= 𝛼𝛼

+ 𝛼𝛼

+ 𝛼𝛼

+ 𝜀𝜀

Next, I estimate the model for normal inventory growth: ∆𝐼𝐼𝐼𝐼𝑅𝑅𝑖𝑖 𝑇𝑇𝑖𝑖−1

= 𝛼𝛼

+ 𝛼𝛼

+ 𝛼𝛼

+ 𝛼𝛼

+ 𝜀𝜀

Since production costs (PROD) = COGS + ∆INV, I use equation (3) and (4) to estimate the normal level of production costs:

𝑃𝑃𝑅𝑅𝐶𝐶𝑃𝑃𝑖𝑖 𝑇𝑇𝑖𝑖−1

= 𝛼𝛼

+ 𝛼𝛼

+ 𝛼𝛼

+ 𝛼𝛼

+ 𝛼𝛼

+ 𝜀𝜀

Abnormal production costs are the residuals (REM_PROD). The residuals do not need to be multiplied by negative one like the other residuals, because higher production costs are indicative of overproduction to reduce cost of goods sold. So more abnormal production costs means more upward real earnings management.

The third measure of real earnings management is abnormal discretionary expenses. Reducing discretionary expenses, such as R&D, advertising and SG&A expenses, will boost current period earnings (Cohen, Dey and Lys, 2008). I estimate the normal level of discretionary expenses using the following equation:

𝑃𝑃𝐼𝐼𝑆𝑆𝑅𝑅𝐼𝐼𝑃𝑃𝑖𝑖 𝑇𝑇𝑖𝑖−1

= 𝛼𝛼

+ 𝛼𝛼

+ 𝛼𝛼

+ 𝜀𝜀

Where DISEXP = discretionary expenses defined as the sum of R&D, advertising and SG&A expenses. For every firm-year, abnormal discretionary expenses is the residual. This is then multiplied by negative one to create REM_DISEXP, so that the higher the amount the more likely it is that the firm is cutting discretionary expenses to manage reported earnings upwards and vice versa. REM_CFO, REM_PROD and REM_DISEXP are then aggregated to obtain the combined measure for real earnings/activities management REM.

Specification 1: Ethical Tone at the Top

Tone at the top is measured with two proxies. The first proxy is a self-devised method and the second proxy is based on an existing method (e.g. Amernic, Craig and Tourish, 2010; Patelli & Pedrini, 2013).

The first tone at the top proxy (TONE_TOP_ETHICS) is specifically aimed at measuring the ethical tone at the top. This proxy is a self-devised never-used method and will be compared with proxy 2 for every test to be careful. Data is acquired from the ASSET4 ESG database. This database contains data about environmental, social and corporate governance related information, also known as extra-financial information. TONE_TOP_ETHICS is a summary indicator variable equal to 1 if the top emphasizes business ethics, 0 otherwise. It consists of the following elements:

- The firm’s management or board has made a public commitment to business ethics; 18

- The company describes, claims to have or mentions processes in place to improve business ethics;

- Appropriate communication tools are in place to improve business ethics (e.g. whistleblower, ombudsman, suggestion box, hotline, newsletter, website);

If the answer is YES to any of these elements then TONE_TOP_ETHICS equals 1, and 0 otherwise. These elements represent the dimensions of ethics which can influence ethical/unethical behavior, as mentioned in the literature review. To be specific they represent commitment to behave ethically, reinforcement of ethical behavior and openness to discuss ethical issues.

Specification 2: General Tone at the Top

The second proxy (TONE_GOOD_GNRL & TONE_BAD_GNRL) measures the general tone at the top, assessed by analyzing the tone of CEO letters of annual reports from the firm’s websites. Amernic, Craig and Tourish (2010, pp. v-vi) say that “CEO letters to shareholders in annual reports are a potentially important vehicle for communicating the attitudes, values and behaviours of those in senior leadership roles. The significance of such letters and what they demonstrate about the culture of organisations, and the personalities of CEOs, is assessed”. The authors continue by saying: “The letter has the capacity to reflect the CEO’s priorities, mindset, implicit ideologies and perceived charisma and greatness” (Amernic, Craig and Tourish, 2010, p.31). By auditing the tone at the top it is better understood what the motives and rationales are for the underlying accounting policies and procedures. It reveals whether the tone at the top will lead to a highly conservative accounting, adventurous accounting, or an accounting that is contemptuous of conventional norms, regulations and proprieties.

To perform the content analysis of CEO letters, a software called DICTION (v.7) is used. DICTION is a computer-aided text analysis program for determining the tone of a verbal message. DICTION analyzes the text using 33 dictionaries to search certain text. It categorizes the use of language in 35 subcategories and produces frequency measures and scores. Subsequently DICTION standardizes each score against its normative database of corporate financial reports producing five main thematic categories: activity, optimism, certainty, realism and commonality. Use of certain categories can indicate the kind of tone at the top established. DICTION also interprets patterns and looks for material changes in use of language.

Based on the five main thematic categories, I create TONE_GOOD_GNRL as the average of realism + commonality and TONE_BAD_GNRL as the average of activity + optimism + certainty. I now continue with the explanation of every category. Examples of these categories in CEO letters can also be found in appendix A.

Activity is a measure of movement, change and the implementation of ideas and the avoidance of inertia (Hart, 2000). Language characterized by activity emphasizes accomplishments and conveys narcissistic self-confidence (Patelli & Pedrini, 2013). Brown and Treviño point (2006) out that leaders who promote transformational change are unethical when driven by self-confidence. Additionally, Schrand and Zechman (2012) find overconfidence of CEOs to be correlated with financial restatements. Hence, activity indicates a negative tone at the top.

Optimism is a measure of language endorsing some person, group, concept or event or highlighting their positive entailments (Hart, 2000). In corporate narratives optimism is considered as a form of impression management (Hooghiemstra, 2000). It is a rhetorical manipulation aimed at distorting the interpretation of financial results, influencing investors’ expectations to depict a more favorable situation (Patelli & Pedrini, 2013). Hence, optimism indicates a negative tone at the top.

Certainty is a measure of language indicating resoluteness, inflexibility, and completeness and a tendency to speak ex cathedra (Hart, 2000). Certainty depends on the context. During an economic context of financial instability, as in the case of this sample, resoluteness and a sense of certainty is inappropriate. Instead, a flexible language appears to be more legitimate to seek understanding rather than approval (Yuthas et al., 2002). Hence, certainty indicates a negative tone at the top.

Realism is a measure of language describing tangible, immediate, recognizable matters that affect people’s everyday lives (Hart, 2000). Realism captures the ease of reading and decreases with complex language. Communication should be comprehensible and transparent to improve mutual understanding and ethical leadership. Lack of such matter can signal ethical lapses in leadership, negatively affecting financial reporting practices (Schaubroeck et al. 2012). Hence, realism indicates a positive tone at the top.

Commonality is a measure of language highlighting the agreed-upon values of a group and rejecting idiosyncratic modes of engagement (Hart, 2012). “Such language seeks to establish mutual understanding by underlining common values. Conversely, commonality decreases as communication rejects social conventions and stresses differences from norms. Therefore, commonality in CEO letters is aimed at creating a community and stimulating 20

commitment toward common goals” (Patelli & Pedrini, 2013). Hence, commonality indicates a positive tone at the top.

Internal control deficiencies

I follow Ashbaugh-Skaife et al. (2008) and acquire data regarding internal control deficiencies (ICDs) from AuditAnalytics. An internal control deficiency or material weakness is captured if this has been reported by the company’s management or its auditor according to SOX 302 and 404. This is reflected in my data as an indicator equal to 1 if an internal control deficiency has been detected according to SOX 302 and 404, and 0 if there are no ICDs.

Code of ethics and ethics commitment

As a proxy for code of ethics I use CODE_ETHICS which equals 1 if the firm describes in the code of conduct that it strives to maintain the highest level of general business ethics, and 0 otherwise. To measure whether a firm is committed to ethics I use ETHICS_COMMITMENT. This proxy is equal to 1 if a firm’s management or board has made a public commitment to business ethics, and 0 otherwise. This is only used in combination with proxy 2 of tone at the top as proxy 2 already incorporated it. I expect both ETHICS_COMMITMENT and CODE_ETHICS to be related to earnings management. In that, if equal to 0 there will be more earnings management.

3.3 Empirical models

To capture the relation between earnings management and tone at the top I estimate the following models:

H1:

DACC or REM = α₀+ α₁ TONE_TOP or α₁TONE_TOP_ETHICS+

α2ETHICS_COMMITMENT+ α3SIZE+ α4MB+ α5ROA+

α6LEV+ α7GOVERNANCE+ ε

H2:

ICD = α₀+ 𝛼𝛼₁TONE_TOP or α₁TONE_TOP_ETHICS+

α2ETHICS_COMMITMENT+ α3SIZE+ α4MB+ α5ROA+

α6LEV+ α7GOVERNANCE+ 𝜀𝜀

H3:

DACC or REM = α₀+ α₁CODE_ETHICS+ α₂ETHICS_COMMITMENT+ α₃SIZE+ α4MB+ α5ROA+ α6LEV+ α7GOVERNANCE+ ε

3.4 Control variables

To improve the model and to avoid correlated omitted variables I add control variables based on Kim, Park and Wier (2012). These control variables affect financial reporting behavior and tone at the top performance.

Firm-specific growth opportunity and the size of a firm can explain some variation in earnings management. To control for this I add Market-to-book value (MB) and SIZE which is the natural logarithm of market value of equity. To isolate the effect of the ethical perspective of tone at the top on earnings management I add return on assets (ROA) to the regression. Furthermore, earnings management is known to be related to leverage and equity-offering related incentives, therefore I add LEV (Kim and Park, 2005). As corporate governance can affect firms’ financial reporting behaviors, I include ASSET4 ESG’s corporate governance score (GOVERNANCE) in the regression models.

Table 3: List of variables

Variable Description

Dependent Variables

DACC Discretionary accruals defined as income before extraordinary items minus operating cash flow, estimated by the cross-sectional modified Jones model including lagged ROA as a regressor (Compustat);

REM Sum of real earnings/activities manipulation proxies, measured as the sum of

REM_CFO, REM_PROD and REM_EXP (Compustat); REM_CFO Level of abnormal cash flows from operations;

REM_PROD level of abnormal production costs, where production costs are defined as the sum of cost of goods sold and the change in inventories;

REM_DISEXP level of abnormal discretionary expenses, where discretionary expenses are the sum of R&D expenses, advertising expenses, and SG&A expenses;

ICD An indicator variable equal to 1 if an internal control deficiency has been detected according to SOX 302&404 and 0 if there are no ICDs

(AuditAnalytics).

Independent Variables of Interest

TONE_TOP_ETHICS Indicator variable for an ethical tone at the top emphasizing ethics. 1 if the firm’s management or board has made a public commitment to business ethics, the firm has processes in place to improve business ethics and appropriate communication tools are in place to improve business ethics, 0 otherwise (ASSET4 ESG);

ETHICS_ COMMITMENT

An indicator variable equal to 1 if the firm’s management or board has made a public commitment to business ethics, 0 otherwise (ASSET4 ESG);

ETHICS PROCESSES An indicator equal to 1 if the firm describes, claims to have or mentions processes in place to improve business ethics, 0 otherwise (ASSET4 ESG);

ETHICS

COMMUNICATION

An indicator equal to 1 if appropriate communication tools are in place to improve business ethics (e.g. whistleblower, ombudsman, suggestion box, hotline, newsletter, website), 0 otherwise (ASSET4 ESG);

TONE_GOOD_GNRL DICTION’s standardized scores based on frequency and the use of language in CEO letters in categories of realism + commonality (Annual reports +

DICTION);

TONE_BAD_GNRL DICTION’s standardized scores based on frequency and the use of language in CEO letters in categories of activity + certainty + optimism (Annual

reports + DICTION);

Variable Description

Activity A measure of movement, change and the implementation of ideas and the avoidance of inertia. Expresses narcissistic self-confidence and thus indicates a negative tone at the top (Annual reports + DICTION);

Optimism A measure of language endorsing some person, group, concept or event or highlighting their positive entailments. Expresses a form of impression management aimed at distorting the interpretation of financial results and thus indicates a negative tone at the top (Annual reports + DICTION);

Certainty A measure of language indicating resoluteness, inflexibility, and

completeness and a tendency to speak ex cathedra. Certainty is inappropriate in the context of financial instability, instead, flexible language is more legitimate. Hence, certainty indicates a negative tone at the top (Annual

reports + DICTION);

Realism A measure of language describing tangible, immediate, recognizable matters that affect people’s everyday lives. It expresses comprehensibility and transparency to improve mutual understanding and ethical leadership. Hence, realism indicates a positive tone at the top. (Annual reports + DICTION);

Commonality A measure of language highlighting the agreed-upon values of a group and rejecting idiosyncratic modes of engagement. It is aimed at creating a community by establishing mutual understanding and stimulating

commitment toward common goals. Hence, commonality indicates a positive tone at the top. (Annual reports + DICTION);

CODE_ETHICS An indicator variable equal to 1 if the firm describes in the code of conduct that it strives to maintain the highest level of business ethics, 0 otherwise.

(ASSET4 ESG).

Control Variables

SIZE Natural logarithm of the market value of equity (Compustat);

MB Market-to-book equity ratio, measured as MVE/BVE, where BVE is the book value of equity (Compustat);

ROA ROA in the previous year, measured as income before extraordinary items scaled by lagged total assets (Compustat);

LEV Long-term debt scaled by total assets (Compustat);

GOVERNANCE ASSET4 ESG’s score reflecting a company's capacity, through its use of best management practices, to direct and control its rights and

responsibilities through the creation of incentives, as well as checks and balances in order to generate long term shareholder value (ASSET4 ESG). = indicates a main variable used in regressions

= indicates an element of the main variable above

4 Results

4.1 Descriptive statistics and univariate analysis

Table 4: Industry overview

Industry Two-digit

SIC

# of obs. % of sample Cumulative

%

Agricultural Production - Crops 01 2 0.31% 0.31%

Metal Mining, Ores 10 4 0.62% 0.94%

Oil & Gas extraction 13 21 3.28% 4.21%

General Building Contractors 15 8 1.25% 5.46%

Heavy Construction, Except Building 16 6 0.94% 6.40%

Food, Beverage 20 37 5.77% 12.17%

Food & Kindred Products 21 4 0.62% 12.79%

Lumber & Wood Products 24 2 0.31% 13.10%

Furniture & Fixtures 25 6 0.94% 14.04%

Paper and Allied Products 26 12 1.87% 15.91%

Printing & Publishing 27 4 0.62% 16.54%

Chemicals & Allied Products 28 53 8.27% 24.80%

Petroleum 29 16 2.50% 27.30%

Rubber 30 6 0.94% 28.24%

Stone, Clay, & Glass Products 32 6 0.94% 29.17%

Primary Metal Industries 33 10 1.56% 30.73%

Fabricated Metal Products 34 8 1.25% 31.98%

Industrial Machinery & Computer Equipment

35 49 7.64% 39.63%

Electronic & Other Electric Equipment 36 26 4.06% 43.68%

Transportation Equipment 37 29 4.52% 48.21%

Instruments & Related Products 38 22 3.43% 51.64% Miscellaneous Manufacturing Industries 39 4 0.62% 52.26%

Railroad Transportation 40 7 1.09% 53.35%

Trucking & Warehousing 42 2 0.31% 53.67%

Air Transportation 45 8 1.25% 54.91%

Transportation Services 47 4 0.62% 55.54%

Communication 48 26 4.06% 59.59%

Electric, Gas & Sanitary Services 49 70 10.92% 70.51%

Wholesale—Durable Goods 50 19 2.96% 73.48%

Wholesale—Non-Durable Goods 51 10 1.56% 75.04% Building Materials & Gardening Supplies 52 4 0.62% 75.66%

General Merchandise Store 53 24 3.74% 79.41%

Food Stores 54 8 1.25% 80.66%

Automotive Dealers & Service Stations 55 10 1.56% 82.22% Apparel & Accessory Stores 56 12 1.87% 84.09% Furniture & Home furnishings Stores 57 6 0.94% 85.02%

Eating & Drinking 58 8 1.25% 86.27%

Miscellaneous Retail 59 21 3.28% 89.55%

Hotels & Other Lodging Places 70 4 0.62% 90.17%

Business Services 73 38 5.93% 96.10%

Auto Repair, Services, & Parking 75 6 0.94% 97.04%

Health Services 80 12 1.87% 98.91%

Engineering & Management Services 87 3 0.47% 99.38% Non-Classifiable Establishments 99 4 0.62% 100.00%

Total 641 100.00%

Table 5: Full sample descriptive statistics

N Mean Median St. Dev. Min. p25 p75 Max.

Dependent variables DACC 641 -0.005 0.000 0.049 -0.322 -0.020 0.015 0.196 ABS. DACC 641 0.029 0.018 0.040 0.000 0.005 0.037 0.322 Positive DACC 329 0.024 0.014 0.030 0.000 0.003 0.034 0.196 Negative DACC 312 -0.035 -0.021 0.048 -0.322 -0.041 -0.006 -0.000 REM 640 -0.008 0.000 0.233 -1.042 -0.087 0.091 0.764 REM_CFO 641 -0.002 0.000 0.042 -0.157 -0.024 0.021 0.165 REM_PROD 640 -0.002 0.000 0.115 -0.593 -0.044 0.041 0.440 REM_DISEXP 641 -0.003 0.000 0.113 -0.496 -0.037 0.035 0.403 ABS. REM 640 0.152 0.088 0.177 0.000 0.017 0.226 1.042 Positive REM 351 0.132 0.070 0.159 0.000 0.010 0.197 0.764 Negative REM 289 -0.177 -0.109 0.194 -1.042 -0.275 -0.030 -0.000 Variables of interest TONE_TOP_ETHICS 641 0.980 1.000 0.141 0.000 1.000 1.000 1.000 ETHICS_COMMITMENT 641 0.626 1.000 0.484 0.000 0.000 1.000 1.000 ETHICS PROCESSES 641 0.836 1.000 0.370 0.000 1.000 1.000 1.000 ETHICS COMMUNICATION 641 0.945 1.000 0.227 0.000 1.000 1.000 1.000 TONE_GOOD_GNRL 532 51.97 51.97 1.715 43.56 50.96 52.94 62.64 TONE_BAD_GNRL 532 51.28 51.35 1.781 41.12 50.40 52.32 57.86 Activity 532 49.80 49.87 2.771 26.67 48.75 51.07 76.33 Optimism 532 56.29 55.73 3.994 46.04 53.64 58.27 80.36 Certainty 532 47.74 48.62 4.451 13.53 46.43 50.46 59.44 Realism 532 54.73 54.58 2.813 45.46 53.13 56.12 73.79 Commonality 532 49.21 49.28 2.038 35.68 48.26 50.38 54.38 ICD 641 0.083 0.000 0.276 0.000 0.000 0.000 1.000 CODE_ETHICS 641 0.967 1.000 0.178 0.000 1.000 1.000 1.000 Control variables SIZE 632 9.088 9.017 1.301 4.266 8.154 9.879 12.89 MB 635 1.382 1.884 32.74 -688.5 1.244 3.035 368.7 ROA 641 0.045 0.047 0.081 -0.418 0.015 0.087 0.346 LEV 639 0.245 0.230 0.164 0.000 0.132 0.337 1.203 GOVERNANCE 641 80.17 83.91 14.73 1.57 74.10 90.95 96.67 = indicates a main variable used in regressions

= indicates an element of the main variable above or an extra descriptive statistic

Table 4 presents the sample distribution by the two-digit SIC code industry. The most heavily represented industries are Electric, Gas & Sanitary services (10.92%, SIC code 49) followed by Chemicals & Allied products (8.27%, SIC code 28). These industries are likewise most often represented in the Fortune 500 of US companies in 2008 and 2009.

Table 5 reports the full sample descriptive statistics. All continuous variables are winsorized at the top and bottom 1 percent of their distributions. The table shows a mean value of -0.005 for discretionary accruals (DACC), which is 0.5 percent of lagged total assets.

This number is similar to Kim, Park and Wier (2012) and Cohen, Dey and Lys (2008) who also report 0.5 percent. However the mean value of absolute discretionary accruals (ABS_DACC) is only 0.029, almost ten times smaller than Kim, Park and Wier (2012). These results suggest that the behavior of upwards and downwards accrual-based earnings management is similar to ‘normal’ times before the global economic crisis. However, during the global economic crisis of 2008 and 2009 the large Fortune 500 firms do not manage earnings as much as ‘normal’ times. Furthermore, negative discretionary accruals (Negative DACC: 0.034) are, on average, slightly larger in magnitude than positive discretionary accruals (Positive DACC: 0.024). This suggests that the examined companies lean slightly more on downward earnings management during the global economic crisis, possibly to show a small but certain improvement in revenues next year.

Continuing with the real earnings management proxies (REM, REM_CFO, REM_PROD, REM_EXP), their mean values are -0.008, -0.002, -0.002, -0.003 respectively. They are all negative, suggesting that on average the Fortune 500 firms manage their earnings downwards with real activities. This is reaffirmed as Negative REM shows a higher absolute value of -0.177 (17.7%) while Positive REM is 0.132 (13.2%). However, the average magnitudes are small with an average of 0.8 percent of lagged assets.

The tone at the top variables are standardized z-scores with 50 as the average normal score. On average, the sample firms score within the normal boundaries (TONE_GOOD_GNRL: 51.97 & TONE_BAD_GNRL: 51.28). Interesting to see though is that the general tone at the top is more optimistic (56.29) and realistic (54.73) however not certain (47.74). This has to do with the global economic crisis in 2008 and 2009. Most CEOs write that shareholders have to be realistic of possible negative performance in the future but that they are optimistic to overcome the economic crisis, however, they are not able to guarantee this. Regarding the tone at the top emphasizing ethics, on average the sample firms score high (0.98). 63 percent of the top has made a public commitment to ethics, 84 percent already has or is planning to implement processes to improve ethics and 94.5 percent has appropriate communication tools in place to improve ethics. Additionally, 96.7 percent strive to maintain the highest level of ethics in their code of ethics. Lastly, only 8 percent of the sample firms have internal control deficiencies (ICDs), meaning most firms passed the internal control audits according to SOX 302 and 404.

With respect to the control variables, the sample firms have a fair market-to-book ratio (MB) of 1.4 and a return on assets (ROA) of 5% during tough economic times of 2008 and 2009. As LEV is only 0.245, it appears the firms do not possess much long-term debt relative 27

to total assets. Lastly, the companies are identified with an average score of 80 for GOVERNANCE. As this is a standardized z-score, an average score is 50. The sample firms therefore score 25 times the standard deviation higher than average according to the ASSET4 ESG database. This high score reflects the firms’ high capacity, through its use of best management practices, to direct and control its rights and responsibilities through the creation of incentives, as well as checks and balances in order to generate long term shareholder value.

Table 6 on the next page presents the Pearson correlation coefficients for selected variables. Ethical tone at the top (TONE_TOP_ETHICS) and general tone at the top (TONE_GOOD_GNRL, TONE_BAD_GNRL) are not significantly correlated to discretionary accruals (DACC). General tone at the top is also not significantly correlated to real earnings management (REM). However, ethical tone at the top is significantly and negatively correlated with REM and its component REM_DISEXP (p < 0.01). Similarly, the components of ethical tone at the top (ETHICS_COMMITMENT, ETHICS_PROCESSES, ETHICS_ COMMUNICATION) are also correlated with REM or one of its components (REM_CFO, REM_PROD, REM_DISEXP). These results suggest that general tone at the top is not related to earnings management, while on the other hand, ethical tone at the top is negatively related to real earnings management only, and not accrual-based earnings management. This might be the result of accrual-based earnings management being substituted by real earnings management.

Table 6: Correlations among Tone at the Top, Earnings Management and other selected variables 1 2 3 4 5 6 7 8 9 10 11 12 1 TONE_TOP_ETHICS 1 2 TONE_GOOD_GNRL 0.018 1 3 TONE_BAD_GNRL -0.016 0.092** 1 4 ETHICS_COMMITMENT 0.186*** -0.012 -0.055 1 5 ETHICS_PROCESSES _{0.325*** -0.047 -0.148*** 0.25***} 1 6 ETHICS_COMMUNICATION 0.599*** 0.021 -0.063 0.155*** 0.209*** 1 7 ICD 0.043 0.052 -0.032 0.092** 0.026 0.047 1 8 CODE_ETHICS 0.533*** 0.005 -0.096* 0.184*** 0.132*** 0.689*** 0.055 1 9 DACC 0.005 0.058 0.041 0.024 0.028 0.009 0.054 0.026 1 10 REM -0.106*** 0.02 -0.06 -0.086** -0.059 -0.105*** -0.03 -0.089** 0.057 1 11 REM_CFO -0.043 0.03 0.02 -0.088** -0.097** -0.057 0.032 -0.014 0.201*** 0.59*** 1 12 REM_PROD -0.05 0.031 -0.07 -0.088** -0.049 -0.064 -0.037 -0.045 -0.019 0.931*** 0.518*** 1 13 REM_DISEXP _{-0.15*** 0 -0.06} -0.056 -0.036 -0.13*** -0.037 -0.133*** 0.061 0.892*** 0.315*** 0.706*** 14 SIZE 0.083** 0.094** 0.032 0.202** 0.158*** 0.04 -0.069* 0.028 0.008 -0.269*** -0.305*** -0.253*** 15 MB 0.001 -0.034 0.003 0.01 0.113*** 0.144*** 0.02 -0.004 0.029 -0.033 -0.034 -0.044 16 ROA -0.049 0.05 0.061 0.092** 0.008 0 -0.033 -0.051 0.414*** -0.256** -0.329*** -0.259** 17 LEV -0.059 0.011 0.075* -0.077* -0.091** -0.053 -0.006 0.017 -0.024 0.089** 0.157*** 0.052 18 GOVERNANCE 0.185*** 0.132*** 0.018 0.281*** 0.193*** 0.129*** 0.018 0.131*** 0 -0.112*** -0.125*** -0.084** 13 14 15 16 17 18 13 REM_DISEXP 1 14 SIZE -0.184*** 1 15 MB -0.01 0.038 1 16 ROA -0.142*** 0.475*** 0.041 1 17 LEV 0.072* -0.265*** -0.046 -0.346*** 1 18 GOVERNANCE -0.099** 0.319*** 0.023 0.089** -0.232*** 1

*, **, *** Indicate statistical significance at the 10 percent, 5 percent and 1 percent levels respectively

4.2 The relation between Tone at the Top and Earnings Management

First I perform difference tests with independent two-sample t-tests between firms with and without a tone at the top that emphasizes ethics. Both accrual-based and real earnings management are used. This test is to find out whether there is a difference in earnings management between a company with and without an ethical tone at the top. The results are presented in table 7. Note that because Tone Top Ethics is a summary proxy for ethical tone at the top, its components are also included to provide more detail of which instrument of an ethical tone is of influence on earnings management.

Table 7: Difference tests (Independent two-sample t-tests)

Panel A: Using Accrual-based Earnings Management

Group Mean Significance t-value

Tone Top Ethics Positive Ethics -0.0046 0.896 0.130 No ethics -0.0064

Public Ethics Commitment Yes -0.0037 0.548 0.601

No -0.0061

Ethics Improvement Processes Yes -0.0040 0.480 0.707

No -0.0077

Ethics Communication Improvement Yes -0.0045 0.826 0.219

No -0.0064

Ethics sought in Code Yes -0.0044 0.513 0.654

No -0.0115

General Ethics Policy Yes -0.0047 0.623 0.491

No 0.0034

Panel B: Using Real Earnings Management

Group Mean Significance t-value

Tone Top Ethics Positive Ethics -0.01118 0.008*** 2.681 No ethics 0.16327

Public Ethics Commitment Yes -0.02320 0.029** 2.192

No 0.01847

Ethics Improvement Processes Yes -0.01365 0.139 1.482

No 0.02336

Ethics Communication Improvement Yes -0.01354 0.008*** 2.672

No 0.09432

Ethics sought in Code Yes -0.01148 0.024** 2.269

No 0.10560

General Ethics Policy Yes -0.01004 0.029** 2.187

No 0.16079

**, *** Indicate statistical significance at the 5 percent and 1 percent levels respectively

Analysis of differences in mean values of earnings management

Panel A of table 7 shows no statistical significance between the accrual-based earnings management of firms who do and do not have an ethical tone at the top. This was not expected, however not unforeseen. It is known that companies, especially after the implementation of SOX, switched from accrual-based to real earnings management (Cohen, Dey and Lys, 2008). Additionally, all firms in the sample, with the exception of only a handful of them, are audited by high quality auditors of the Big 4. Chi, Lisic and Pevzner (2011) argue that firms resort to real earnings management when their ability to manage accruals are constrained by high quality auditors. This means that there should be significant results for real earnings management instead of accrual-based earnings management. Panel B confirms this. This panel reports a significant difference in real earnings management between firms that do and do not have an ethical tone at the top. Firms with a tone that emphasize ethics have smaller magnitudes of real earnings management downwards, while firms that do not emphasize ethics have larger magnitudes of real earnings management upwards. This is only the case when firms are accompanied with a public commitment to ethics from the top, when the highest level of ethics is endeavored as described in the code of conduct, when firms have a general ethics policy and particularly when firms possess appropriate communication tools to improve ethics, such as a suggestion box or a whistleblower.

Regression analyses of earnings management on tone at the top

Results of the multivariate regression analyses of earnings management on tone at the top are reported in table 8. These regressions are performed four times using two different proxies/specifications of tone at the top for both accrual-based and real earnings management.

Starting off with discretionary accruals, both tone at the top proxies are not significantly associated with discretionary accruals. This suggests that there is no relation between tone at the top and accrual-based earnings management. However this was expected as the previous analysis showed only a difference in means for real earnings management. This is possibly due to the fact that firms substitute accrual-based with real earnings management, especially after SOX and when constrained by high quality auditors.

In the case of real earnings management, general tone at the top (TONE_GOOD_GNRL, TONE_BAD_GNRL) is not statistically significant. However, the second proxy, ethical tone at the top, is statistically significant. I find a significant and negative association between ethical tone at the top (TONE_TOP_ETHICS) and real earnings management (REM). This suggests that that a tone at the top which emphasizes ethics more leads to less earnings management through real activities.

*, **, *** Indicate statistical significance at the 10 percent, 5 percent and 1 percent levels respectively

Table 8: Multiple Regression of Earnings Management on Tone at the Top

Panel A: Using Accrual-based Earnings Management

Panel B: Using Real Earnings Management Specification 1

Ethical Tone at the Top Coefficient (t-stat) Specification 2 General Tone at the Top Coefficient (t-stat)

Intercept 0.5106 (5.26)*** Intercept 0.3463 (0.88)* TONE_TOP_ETHICS -0.0528 TONE_GOOD_GNRL 0.0070 (-1.03) (1.21) SIZE 0.0041 TONE_BAD_GNRL -0.0077 (0.65) (-1.38) MB 0.0002 ETHICS_COMMITMENT -0.0177 (1.1) (-0.83) ROA 0.3763 SIZE -0.0291 (3.72)*** (-3.17)*** LEV -0.0775 MB -0.0001 (-1.63) (-0.49) GOVERNANCE 0.0001 ROA -0.6320 (0.24) (-4.23)*** LEV -0.0376 (-0.52) GOVERNANCE -0.0002 (-0.22) Adj. R² 0.0469 Adj. R² 0.0924 Specification 1

Ethical Tone at the Top

Coefficient (t-stat)

Specification 2

General Tone at the Top

Coefficient (t-stat) Intercept 0.0274 (1.43) Intercept -0.0319 (-0.43) TONE_TOP_ETHICS 0.0192 TONE_GOOD_GNRL 0.001361 (1.48) (1.27) SIZE -0.0094 TONE_BAD_GNRL 0.000261 (-5.84)*** (0.25) MB 3.02E-05 ETHICS_COMMITMENT 0.004955 (0.57) (1.24) ROA 0.3511 SIZE -0.01006 (13.73)*** (-5.9)*** LEV 0.0398 MB 3.54E-05 (3.31)*** (0.7) GOVERNANCE 0.0001 ROA 0.297241 (0.87) (10.71)*** LEV 0.020407 (1.51) GOVERNANCE 0.000184 (1.16) Adj. R² 0.2260 Adj. R² 0.1863 32